The South Carolina Epidemiological Studies of Epilepsy & Seizure Disorders (SCESESD) Final Report THE MEDICAL UNIVERSITY OF SOUTH CAROLINA In Partnership with The SC Budget and Control Board The Office of Research & Statistics PRINCIPAL INVESTIGATOR Anbesaw Selassie, DrPH Co-Principal Investigator Braxton Wannamaker, MD Investigators Elisabeth Pickelsimer, DA Robert Turner, MD, MSCR Gigi Smith, RN, MSN, CPNP Walter “Pete” Bailey, MPH Mary Tyrell, PhD Epidemiologist & Research Manager Pamela Ferguson, PhD Research Associate Lee Lineberry, BS, (PhD Student) Statistician & Programmer Ja Kook Gu, MSPH CDC Technical Advisor David Thurman, MD, MPH Project Office: (843) 876-1100 Executive Summary The South Carolina Epidemiological Studies of Epilepsy and Seizure Disorders (SCESESD) in the Department of Biostatistics, Bioinformatics, and Epidemiology at the Medical University of South Carolina (MUSC) proposed to conduct a population-based study to determine the prevalence and incidence of epilepsy in South Carolina. The study targeted the general population of South Carolina. According to the 2002 census data, the total population of the state is 4,100,000 and it is the 26th most populated state in the union. The population is 30% black, 67% white, and 3% other races. The lifetime prevalence of epilepsy has been determined to be 2%. Persons with epilepsy and seizure disorders were identified from four major data sources: Statewide Hospital Discharge (HD), Emergency Department (ED) Visits, Physician Outpatient Visits (POV), and the epilepsy module of the SC Behavioral Risk Factor Surveillance System (BRFSS). The first three data sources supplied information on persons with epilepsy and seizure disorders that have been clinically evaluated. Persons were identified with the diagnosis codes (International Classification of Disease 9th Revision Clinical Modification, ICD-9-CM) of 345.x (epilepsy and status epilepticus), 780.31 (febrile seizures), 780.39 (seizure unspecified), 780.2 (syncope), and 293.0 (delirium). Case identification through BRFSS relied on self-report using random-digit dialing telephone survey of the adult population of the state. Broader case ascertainment codes were used to identify individuals that could be misclassified. Detailed clinical information was abstracted from medical charts of representative sample of 3,983 persons (5.4%) with these conditions clinically evaluated in 2001 and 2002. By the end of the third year, a total of 21,480 individuals had completed the BRFSS telephone survey making SCESESD among the largest population-based epidemiological study of epilepsy in the United States. SCESESD personnel abstracted medical record information on 3,881 randomly selected persons with epilepsy and seizure disorder. Case-level uniform billing abstracted data were acquired on 73,955 individuals using the expanded case ascertainment codes and a database has been developed. The database includes one primary and up to 9 secondary diagnoses, demographics, number of visits made during the year, payer information, CPT codes, and unique identifiers tied to SSN. The database also includes extensive and detailed information from chart abstraction on ii 3,881 individuals. Case-level data with the appropriate weights on 21,480 respondents has been generated. SCESESD will make these data public domain after major findings have been published. This is expected to be no later than October 31, 2008. Among the major activities completed by the project, determination of prevalence, incidence, venues of epilepsy care, and measures of data validity are proud accomplishments. Within such a short time, SCESESD impacted public policy debate, and enhanced social activism and awareness about epilepsy and seizure disorders. A steering committee comprised of persons and families with epilepsy, key stakeholders, and community leaders has been formed to promote further research on epilepsy. Data generated by the SCESESD has been used to formulate legislative initiative for social support and waiver programs for persons with epilepsy. Governor Mark Sanford declared November as epilepsy awareness month and more than three major public campaigns have been initiated to promote support groups and raise funds. BRFSS-based prevalence estimate in SC indicate a rate of 2.0% (95% CI: 1.8-2.3) for ever having epilepsy and 1.1% (95% CI: 0.9-1.2) for active epilepsy among individuals age 18 and older. For the two year period, the annualized estimate of prevalence derived from healthcare encounters for 2001 and 2002 ranges from 0.8% to 0.9% and appears to be in total agreement with the lower range estimate of active epilepsy derived from BRFSS. We further generated a model-based prevalence rate that provided a conservative estimate of 0.8% using information from the case-level data on 70,000 individuals in a loglinear model. The model shows potential to be tested elsewhere and uses uniformly available variables from administrative data systems. The estimated annual incidence rate of epilepsy for the general population of the state solely based on HD and ED data is 3 per 10,000 (0.03%) based on a follow-up period of 3 years. This suggests that there are at least 1,200 individuals who develop epilepsy every year. Assuming a reduction due to the force mortality at 1.68%, a rate 2.5 times higher than the mortality rate in the general population of the state, SC could have about 120,000 persons living with epilepsy by 2011 if all other contributing causes of seizure are held constant. Our incidence estimate also shows that the incidence rate of epilepsy is at least 30 times higher among persons with TBI. Our TBI Registry data indicate that the incidence of epilepsy after TBI is 2.6/100 person-year within the first iii three years of follow-up. With increased incidence of TBI due to various causes, the influx of persons who develop epilepsy may be higher than the current estimate our analysis provides. Our study showed that the healthcare data sources we used can be used to monitor the prevalence of epilepsy and seizure disorders. The predictive value positive (PVP) of HD, ED, and POV suggest usefulness of these data sources for surveillance activities. Our findings indicate that the PVP of 345.x code is remarkably high for all the sources ranging between 93% and 96%. However, the PVP of 780.3x is very low ranging between 14% and 20%. It can be assumed that about 80% of those coded with seizure unspecified code of 780.3x are epilepsy cases based on 3,983 records we reviewed. Given that 70-80% of the persons coded with780.3x are misclassified as seizure and the probability of a case of epilepsy being coded as 780.39 is 11 times higher than being coded as 345.x, it is imperative to develop a surveillance guideline that takes into account this prolific use to generate a reliable estimate. Our experienced team is more than willing to work with the CDC in assisting the development of such a guideline. Finally, we are proud to declare that SCESESD met all of its objectives according to the stipulation of the award. We have developed a strong, multifaceted surveillance system with diversified data sources that are complementary to each other. A decision analysis tool with an exquisite algorithm addressing each set of diagnosis codes has been developed and validated. Our team and partners are well prepared and ready for more productive research to better the lives of persons with epilepsy. iv Table of Contents Page Executive Summary…………………………………………………………………. i Prevalence rates……………………………………………………………………... 1 Behavioral Risk Factor Surveillance System data……………………………… 1 Chart abstraction data…………………………………………………………... 3 Inpatient and emergency department (HD/ED)………………………. 3 Physician office visits (POV)………………………………………… 5 Prevalence combining HD/ED/POV data……….........................…… 7 Model-based approach to estimate prevalence………………………. 7 Incidence rates……………………………………………………………………… 12 Posttraumatic epilepsy………………………………………………………. 12 Incidence of epilepsy utilizing HD/ ED data………..............………………. 22 Causes (triggers) of seizures and etiologies of epilepsy…………………………. 28 Target populations for intervention……………………………………………… 35 Medication use - HD/ED data…………............…………………………… 35 Medication use - POV data……………….........................……………….. 38 Posttraumatic epilepsy……………………………………………………… 40 Behavioral Risk Factor Surveillance System data………………………….. 40 Focus groups data…………………………………………………………… 41 Severity and subtypes of epilepsy…………………………………………………. 42 HD/ED data…………........…………………………………………………. 42 POV data…………………………..........................………………………… 45 Behavioral Risk Factor Surveillance System data…………………………… 48 v Venues and levels of care………………………………………………………….. 48 Behavioral Risk Factor Surveillance System data………………………….. 48 POV data…………………….........................…………………………….. . 49 Focus groups data……………………………………………………..…….. 50 Traumatic Brain Injury among persons with seizure disorder………....…..…... 50 Data quality………………………………………………………………..…….….. 56 Population and sample information………………………………..………... 56 HD/ED data…………...............................................……....…….….. 56 POV data……….........................…………………………..………... 58 Predictive Value Positive and Sensitivity of seizure and epilepsy codes…..... 60 HD and ED Data.........................................……………………….….60 POV Data...........................…………………………………………. 61 Estimates of PVP and Sensitivity for HD/ED.data………….……… 62 Estimates of PVP and Sensitivity for POV data....…………….…… 63 Algorithm to identify epilepsy patients from administrative datasets….…… 63 Data quality measures……………………………………………………….. 68 Dissemination of results…………………………………………….……………… 72 Purpose and Plan……………………………………………………….……. 73 Reports, Websites, Presentations, and Publications…………………….…… 73 Appendices……………………………………………..…………………………… 76 A. Abstraction manual for HD/ED charts….............................................….. 77 B. Abstraction manual for POV charts……….........................………….… 108 vi C. Comparison of HD/ED sample to population........……………………… 122 D. Comparison of POV sample to population….........................…………. 125 E. Model-based case-level likelihood of an epilepsy diagnosis…………… 127 F. Posttraumatic epilepsy incidence analysis……………………………… 133 G. Sampling plan for HD/ED charts……..............................................……. 136 H. Sampling plan for POV charts………..........................…………………… 138 Acknowledgement..……………………………………..…………………………… 139 vii I. Prevalence Rates Aim—Provide reliable and stable population-based estimates on the incidence and prevalence rates of epilepsy and other seizure disorders. Prevalence rates were generated using a multifaceted approach. The state-based Behavioral Risk Factor Surveillance System (BRFSS) was used to generate self-reported epilepsy covering three years (2003-2005) of surveillance. A total of 21,480 individuals age 18 and older responded to the survey questions. Existing data sources pertaining to 2001 and 2002 Hospital Discharges (HD), Emergency Department (ED) and Physician Office Visits (POV) provided statewide information on persons with seizure disorders. Each of these is described below. Prevalence rates were generated from both state-based BRFSS and data collected from the clinical encounters A. BRFSS is a state-based, random-digit–dialed telephone survey of the noninstitutionalized, U.S. civilian population aged >18 years. The South Carolina Behavioral Risk Factor Surveillance System surveys for 2003 through 2005 included questions on epilepsy listed in Table 1. The survey included questions regarding history of epilepsy and number of seizures experienced during the past three months. Respondents were considered to have active epilepsy if they 1) reported ever having been told by a doctor that they had a seizure disorder or epilepsy and 2) either were currently taking medicine to control epilepsy or had had one or more episodes of seizure during the preceding 3 months. Active epilepsy was categorized further by whether the respondent had had one or more seizures during the preceding 3 months. Data was weighted by sex, race and age to adjust for differences between the survey population and the South Carolina population. 2003 and 2004 survey results were reported earlier (MMWR, October 28, 2005 / 54(42);1080-1082). By the end of the third year, a total of 21,480 individuals had completed the survey—5,926 in 2003, 7,114 in 2004 and 8,440 in 2005—for response rates of 41.6%, 43.8% and 59.1% respectively. The first question was considered as measuring the lifetime prevalence of epilepsy, and had a response rate of 92.0%. Respondents were considered to have active epilepsy if they 1) reported ever having been told by a doctor that they had a seizure disorder or epilepsy and 2) either were currently taking medicine to control it or had had one or more episodes of seizure in the preceding 1 3 months. Active epilepsy was further categorized as controlled or uncontrolled based on whether the respondent had had a seizure in the preceding 3 months. Condensed results for the five questions, with active, non-active, controlled, and non-controlled are shown in Table 2, with 95% confidence intervals (CI). Table 1. Survey questions included in the SC BRFSS epilepsy module 1. Have you ever been told by a doctor that you have a seizure disorder or epilepsy? 1. Yes 2. No† 7. Don’t know/Not sure† 9. Refused† 2. Are you currently taking any medicine to control your seizure disorder or epilepsy? 1. Yes 2. No 7. Don’t know/Not sure 9. Refused 3. How many seizures have you had in the last three months? 1. None 2. One 3. More than one 4. No longer have epilepsy or seizure disorder† 7. Don’t know/not sure 9. Refused 4. During the past 30 days, to what extent has epilepsy or its treatment interfered with your normal activities like working, school, or socializing with family or friends? Would you say... 1. Not at all 2. Slightly 3. Moderately 4. Quite a bit 5. Extremely 7. Don’t know/Not sure 9. Refused 5. In the past year have you seen a neurologist or epilepsy specialist for you epilepsy or seizure disorder?* 1. Yes 2. No 7. Don’t know/Not sure 9. Refused † If these responses were given, interviewer skipped the rest of the epilepsy questions. *Included years 2004-2005 only. 2 Table 2. Weighted summary of the SC BRFSS survey, 2003-2005 Total Epilepsy Status Do not have epilepsy Have epilepsy N % (95% CI) 19,390 379 98.0 (97.7-98.2) 2.0 (1.8-2.3) Taking medicine 190 46.1 (39.6-52.5) Not taking medicine 189 53.9 (47.5-60.4) 24.5 (18.4-30.6) 84 Had seizure in prev. 3 mos. 72.9 (66.7-79.1) 277 No seizures in prev. 3 mos. 2.6 (1.0-4.2) 11 No longer have epilepsy Epilepsy interfered 92 28.2 (21.7-34.7) Epilepsy did not interfere 272 71.8 (65.3-78.3) Epilepsy, nonactivea 172 1.0 (0.8-1.1) 207 1.1 (0.9-1.2) Epilepsy, activeb c Active, controlled 117 52.8(43.7-62.0) 84 47.2(38.0-56.3) Active, uncontrolledd Seen neurologist in past year Yes 88 37.8(29.4-46.3) No 160 62.2(53.7-70.6) a nonactive=’yes’ to question 1, but not taking medication and no seizure in previous 3 months b active=taking medication or seizure in previous 3 months c active, controlled=taking medication and no seizure in previous 3 months d active, uncontrolled=seizure in previous 3 months The results show prevalence rate of 2.0% (95% CI: 1.8-2.3) for ever having epilepsy and 1.1% (95% CI: 0.9-1.2) for active epilepsy among individuals age 18 and older. Chart Abstraction Data—following are data sources with abstracted information: B. Inpatient and Emergency Department— The SC Budget and Control Board, Office of Research and Statistics (ORS), is the entitled by state law to serve as the repository of data from all nonfederal hospitals and EDs. Data are dumped 90 days after the end of the calendar quarter and the format of data submission is based on the Uniform Billing, 1992 layout, often referred to as UB-92. We obtained data on 70,955 unduplicated individuals from all 62 nonfederal hospitals and 64 emergency departments (EDs) across the state for the calendar year 2001 and 2002 to capture persons with a diagnosis of epilepsy, seizure disorders, syncope, and delirium. Data were unduplicated using the patients’ unique identifiers. A representative sample of 3,881 unduplicated records (5.5%) were abstracted to collect additional data (please see Appendix A–Abstraction manual for 3 inpatient/ED charts for information collected). Table 3 compares patient characteristics by abstraction status. Table 3. Characteristics of Patients by Abstraction Status Total Abstracted Characteristics N=70,955 (%)† n1=3,881 (%)‡ Year: 2001 30,948 (43.6) 2,428 (7.9) 2002 40,007 (56.4) 1,453 (3.6) Age group: 0-9 5,233 ( 7.4) 439 (8.4) 10-19 5,715 ( 8.1) 256 (4.5) 20-39 14,944 (21.1) 925 (6.2) 40-59 19,210 (27.1) 1,207 (6.3) 60-79 17,425 (24.6) 789 (4.5) 80-89 6,980 ( 9.8) 226 (3.2) 90+ 1,448 ( 2.0) 39 (2.7) Primary Payer: Private 20,528 (28.9) 892 (4.4) Medicaid 11,865 (16.7) 963 (8.1) Medicare 28,429 (40.1) 1,458 (5.1) Uninsured 10,133 (14.3) 568 (5.6) UB-92 diagnosis: 345.x 2,954 ( 4.2) 1,186 (40.2) 7803 32,892 (46.4) 2,566 ( 7.8) 7802 33,094 (46.6) 112 ( 0.3) 2930 2,015 ( 2.8) 17 (0.8) † Column percent ‡ Row percent Not abstracted n2=67,074 (%)‡ 28,520 (92.1) 38,554 (96.4) 4,794 (91.6) 5,459 (95.5) 14,019 (93.8) 18,003 (93.7) 16,636 (95.5) 6,754 (96.8) 1,409 (97.3) 19,636 (95.7) 10,902 (91.9) 26,971 (94.9) 9,565 (94.4) 1,768 (59.9) 30,326 (922) 32,982 (99.7) 1,409 (97.2) Cases were selected after stratification by diagnosis category using ICD-9-CM codes of epilepsy (including status epilepticus) (345.x), convulsions (780.3), syncope and collapse (780.2), and acute delirium (293.0). In our evaluation to determine the prevalence of epilepsy from these sources, the following general principles were observed: 1. All characteristics of the sample (sex, race, age, or payer) are comparable with the referent population from which the sample is per diagnosis strata (345.x or 780.3) and calendar year (2001 or 2002). (Appendix C for actual comparisons). A total of 3,983 charts were sampled and reviewed. The accuracy of the charts reviewed is summarized in Table 4. 2. ICD codes of 780.2 and 293.0 are listed as independent diagnosis when there is neither a 345.x nor 780.3 diagnoses. 3. If there is more than one seizure-related code, the code utilized is assigned according to the following hierarchy: 345.x > 780.3. 4. Data were combined for 2001 and 2002 and annualized. 4 5. For duplicate observations in 2001 and 2002, individuals present in both years were unduplicated. If they had both a 780.3 and a 345.x code, the 345.x visit was retained. Abstracted records were reviewed by trained neurology nurse practitioner and nurses. When information extracted required a second opinion, an adult epileptologist (BW) and a pediatric epileptologist (RT) provided their opinion. Following is the summary of the abstracted HD and ED charts comparing diagnosis assigned in UB-92 and clinical reviewers. Table 4. Accuracy of abstracted HD and ED charts, SC 2001-02 Diagnosis after clinician review Diagnosis UB-92 Epilepsy Status Seizure Syncope 345.x 1016 (85.2%) 61 (5.1%) 41 (3.4%) 0 780.3 2040 (73.1%) 94 (3.4%) 430 (15.4%) 3 (0.1%) All 3056 (76.7%) 155 (3.9%) 471 (11.8%) 3 (0.0%) Inadequate information 74 (6.2%) 224 (8.0%) 298 (7.5%) Total 1192 (29.9%) 2791 (70.1%) 3983 (100%) Based on the findings of Table 4, 85.2% of 345.x codes, and 73.1% of 780.3 codes, have evidence of epilepsy. When the proportion of clinically confirmed epilepsy is applied to the referent population, the following annualized frequency is obtained: Diagnosis† Proportion Estimated # of Group Frequency after review epilepsy cases 345.x: 3177 85.2% 3177*.852 = 2,707 780.3: 31,695 73.1% 31,695*.731 = 23,169 Estimated number of HD and ED visits with epilepsy = 25,876 † There is 4.8% difference in diagnoses when compared to sample population . If we make the assumption that the inconclusive cases due to inadequate information would follow the same distribution as those records that furnished adequate information and ignore them, the proportion of epilepsy will be 90.9% and 79.5% respectively for 345.x and 780.3 codes. This latter proportion yields 28,086 persons with epilepsy and could be interpreted as upper range of the estimate. C. Physician Office Visit—For data obtained from POV, the sample of records reviewed were 302 records, 9.3% of 3,253 unduplicated encounters. Although the sampled of proportion is nearly twice that of HD/ED sample, the smaller number of patients in the POV setting makes it difficult to conduct subset analysis by various attributes due to 5 scanty distribution in some of the cells. Hence, categories were collapsed for payer status and only epilepsy (345x) and seizure (780.3) were identified. Detailed distribution by abstraction status with 95% confidence limits is presented in Appendix D. Unlike the HD/ED data, POV data were not identified by year excepting Medicare data which came from 2001 visits. The POV patients have had no ED/HD encounters for the years in mention. The diagnosis codes were hierarchically ranked in the following order: 345.0, 345.1, 345.4, 345.5, 345.6, 345.7, 345.8, 345.9 > 345.2, 345.3 > 780.3. Table 5 shows the distribution of billing codes after review of the records by the clinical nurse practitioner and nurses, and input of the two epileptologists. Table 5. Accuracy of abstracted POV charts, SC 2001-02 Diagnosis after clinician review Diagnosis Inadequate UB-92 Epilepsy Status Syncope information 345.x 202 (91.4%) 1 (0.5%) 10 (4.5%) 8 (3.6%) 780.3 74 (91.4%) 0 4 (4.9%) 3 (3.7%) All 276 (91.4%) 1 (0.3%) 14 (4.6%) 11 (3.6%) Total 221 (29.9%) 81 (70.1%) 302 (100%) For both 345.x and 780.3 codes, the proportion of the records determined to be epilepsy cases was 91.4%. When these proportions were applied as weighting factor, of the 1,157 patients with 345.x, 1,057 and of the 2,096 patients coded with 780.3, 1,916, total 2,973 were deemed to be epilepsy cases. If the 11 records were ignored, the total number would be 3,086. More detailed sample information is found on page 57. The data analyzed from POV is not generalizable to the general population since the data sources are limited to selected practices and do not include children under the age of 2 years. However, epilepsy in this age group is expected to be small and negligible. To approximate the POV data to the general population of epilepsy patients from physician offices through out the state, we assumed that the prevalence of epilepsy among patients encountered in physician offices elsewhere is approximately comparable to the distribution noted in the sampled records among Medicare, Medicaid and SHP insured patients. Based on state demographic data, we assumed that the sample area coverage is approximately 39% of the state. If these values are extrapolated to the rest of the state, the sample analyzed represents a low of 8,494 (2,973/0.35) and a high of 8,817 (3086/0.35) patients with epilepsy. These numbers are most likely an underestimate until we can obtain the number of SC Medicare patients with a 345.x or 780.3 diagnosis 6 code in 2002 who were not seen as an HD/ED patient during 2001 or 2002. Following is a step-by-step calculation of prevalence based on the three data sources. Prevalence for Combined HD/ED/POV Data • Average population of the state for 2001-2002= 4,081,794 (Source: US Census Bureau) • Individuals with epilepsy in SC: o Lower range: 25,876 HD/ED + 8,494 POV = 34,370 o Upper range: 28,086 HD/ED + 8,817 POV = 36,903 • Annualized prevalence estimate of clinically attended (~Active) epilepsy in SC for 2001 and 2002 o Lower range: 34,370/4,081,794= 0.84% o Upper range: 36,903/4,081,794= 0.90% D. Model-based Approach to Estimate Prevalence:—this approach utilized the 2001 and 2002 ED/HD data with 70,955 unduplicated observations shown in Table 3. The main research effort of this modeling approach is to determine the utility of routinely available surveillance variables in correctly predicting epilepsy using the case ascertainment differential diagnosis codes of 345.x, 780.3, 780.2, and 293.0 and covariates. The covariates selected for this modeling approach were seven variables: Demographics (age, sex, race, and payer status) and clinical (UB-92 diagnosis, number of visits during the two years, comorbid conditions frequently associated with seizure disorders). The response variable was dichotomous level of clinically confirmed epilepsy (epilepsy vs. no epilepsy) based on the adjudication of the clinical neurology nurse practitioner and nurses and the two epileptologists on the 3,983 charts reviewed (Table 4). The final analysis relied on 3,881 records after deleting 102 records that were abstracted and reviewed in each year. The covariates selected were first evaluated for their bivariate association with the response and only those with p<0.10 were included in the multivariable logistic regression. Only sex was excluded from the multivariable model due to lack of association. The final model included the variables listed in Table 6 along with the various levels of effect. The beta-coefficients from the model were applied to the observations that were not reviewed (n=67,074) to identify the predicated probability of 7 epilepsy for each patient conditional on the covariate values each patient satisfied. We used the following formulae to calculate the probability of epilepsy and its 99% confidence interval for each case: (1) P=exp(α+β*Χ)/(1+exp exp(α+β*Χ)) (2) 95%CI= exp(α+β*Χ)/(1+ exp(α+β*Χ)) +/- 1.96*sd. where: α – the intercept β = the vector of mean beta-coefficients from the 100 estimation sets X = the vector of explanatory variables sd = the standard deviation of the distribution of mean predicted probabilities To assess the predictive power of the model, a Receiver Operating Characteristic (ROC) curve was constructed for the validation data (n=3,881). A ROC curve is a graphical representation of the trade off between false negative and false positive rates for every possible probability cut off (for example, the tradeoff if only those with a probability of 6% or higher are defined as likely to have epilepsy). Equivalently, the ROC curve is the representation of the tradeoffs between sensitivity and specificity. The curve shows sensitivity on the Y-axis and one minus specificity on the X-axis. A ROC curve that climbs rapidly toward the upper left hand corner of the graph indicates that the true positive rate is high while the false negative rate is low. When the ROC curve follows a diagonal path from the lower left hand corner to the upper right hand corner, it means that every improvement in false positive rate is matched by a corresponding decline in the false negative rate. The Area Under the Curve (AUC) is a representation of the model’s ability to correctly discriminate a pair of true epilepsy and non-epilepsy patients—the larger the AUC, the higher the ability of the model to correctly discriminate those who have epilepsy from those who do not. Generally, AUC 0.90-1.00 is considered outstanding discriminatory power, 0.80-0.89 as excellent, 0.70-0.79 as very good, 0.600.69 as good, and values less than 0.50 are worse than chance 1. As shown in Figure 1, the sensitivity analysis of our model shows an AUC of 0.75, suggesting very good discriminatory power with the covariates identified (Figure 1). The model fit and predictive power is also strong. Given that the model utilized the most parsimonious set 1 Hosmer DW, Lemeshow S. Assessing the fit of the model. In: Applied Logistic Regression. New York: John Wiley & Sons; 2000:160-164. 8 of covariates that are routinely available in administrative datasets, the potential usefulness of the model warrants further consideration in other settings. Table 5. Logistic Regression Parameter Estimates Wald Confidence Interval for Parameters Parameter Level of effect Estimate 99% Confidence Limits Intercept β0 Baseline -1.5612 -2.0566 -1.0658 Agegp β10 0-9 Agegp β11 10-19 1.4528 0.9692 1.9364 Agegp β12 20-39 2.0098 1.6200 2.3996 Agegp β13 40-59 1.8267 1.4561 2.1972 Agegp β14 60-79 1.3783 0.9287 1.8278 Agegp β15 80-89 1.3879 0.7920 1.9839 Agegp Β16 90+ 1.0141 -0.00274 2.0309 Racegp β20 White Racegp β21 Black 0.2144 -0.0106 0.4394 Racegp Β22 Other -0.4160 -1.1813 0.3493 Comorbgp β30 Lo/No risk Reference Reference Reference † Comorbgp β31 Hi-risk 0.4744 0.1215 0.8274 Visit β40 1 Visit β41 2-5 1.0356 0.7980 1.2732 Visit β42 >=6 1.7164 1.0489 2.3838 Esdgp (UB-92) β50 Syncope/Delirium Esdgp (UB-92) β51 Epilepsy 1.0584 0.5893 1.5276 Esdgp (UB-92) β52 SeizureNos 0.2243 -0.1683 0.6168 Payer β60 Private Payer β61 Medicaid 0.5911 0.2679 0.9143 Payer β62 Medicare 0.5299 0.1769 0.8830 Payer β63 Uninsured 0.1860 -0.1702 0.5422 Reference Reference Reference † Hi-risk comorbid conditions include the presence of mental retardation, psychiatric problems, depression, substance abuse, paralysis, and anemia in the secondary diagnosis field. We calculated the probability of epilepsy for each case in the 2001-2002 ED/HD surveillance dataset by applying the parameter estimates. For individual cases, the probability could be written in the following manner. P(D) = {1+ exp – ( β0 + β10 + β11 + …….+ β62 + β63 )}-1 9 To illustrate how the model parameters work in assessing the probability of epilepsy based on the covariates, the following three examples are provided. Example 1: Patient A is 27 years-old, black, with right hemiplegia, had 6 visits during the year, UB-92 diagnosis was 345.1, is Medicaid insured. The probability that this would turn out to be true epilepsy is: {1+exp-(-1.5612+2.0098+0.2144+0.4744+1.7164+1.0584+0.5911)}-1 = 0.9890 (98.9%) Example 2: Patient B is 17 years-old, native American, with no comorbid condition, had no previous visit during the year, UB-92 diagnosis was 780.2, is privately insured. The probability that this would turn out to be true epilepsy is: {1+exp-(-1.5612+1.4528+0.0+0.0+0.0+0.0+0.0)}-1 = 0.1216 (12.2%) Example 3: Patient C is 74 years-old, Hispanic, with alcoholism and anemia recorded as comorbidities, had eight visits during the year with seizure episode, UB-92 diagnosis was 780.39, is Medicare insured. The probability that this would turn out to be true epilepsy is: {1+exp-(-1. 5612+1.3783+(-0.4160) +0.4744+1.7164+0.3522+0.2166)}-1 = 0.8967 (89.7%) Based on the estimated probabilities and a cutoff point of 47%, patients A and C are likely to be epilepsy cases while patient B is less likely to be an epilepsy case. Figure 1. Sensitivity analysis for predicting Epilepsy from Administrative Datasets Legend: 1. AUC= Area under the curve--ability of the model to correctly discriminate those with epilepsy from those without. It is the reflection of the C statistics from the logistic output. 2. PoI= Point of intersection (0.695)--the point at which sensitivity and specificity are equal, i.e., a cutoff point where false positives and false negatives are balanced. 10 To determine the appropriate cutoff point for the predicted epilepsy, we used the 50th percentile (41.85) of the true positives cases as the cutoff for probable epilepsy and the 40th percentile (33.48) as the cutoff for possible cases. While it is customary to use the mean and 1 standard deviation below the mean as the upper and lower range of the cutoff points, the choice of the median is a good substitute when data are not normally distributed, which is the case in our distribution. Based on these cutoff levels the proportions of cases that are highly likely and likely to be epilepsy are shown in Table 7. Table 6. Cutoff points and model-based distribution of epilepsy among HD &ED visits, 2001-2002 Epilepsy Location Cutoff Level Frequency Percent Cumulative Frequency Cumulative Percent 1)Probable ≥ 50th% of T.P 0.4185-1.000 21,046 29.66 21,046 29.66 2)Possible 40th % of T.P 0.3348-0.4185 9,875 13.92 30,921 43.58 3)Unlikely <40th % of T.P 0.0000-0.3347 40,034 56.42 70,955 100.00 Any Epilepsy ≥ 40th % of T.P 03348-1.000 30,921 Rate = (30,921/4,100,000) =0.7500% For individual prediction of epilepsy, we utilized the model parameters as described earlier and calculated the probabilities across the variables reflected as risk characteristics for each person. Appendix E shows a 0.1% randomly selected sample of records to which such an estimate is demonstrated. Overall, when the UB-92 diagnosis is 345.x, the model’s prediction agrees with the clinically confirmed diagnosis in 8 out of 10 cases. Conversely, when the diagnosis is 780.3, the model’s prediction agrees with the expert decision in 4 out of 10 cases. This suggests that the prediction has higher sensitivity with a tradeoff on specificity. However, for the purposes of surveillance the current cutoff appears to be adequate. Finally, it is important to note that our effort in developing this model as an alternative to time/labor-intensive and costly review of records is an important step towards improving epilepsy surveillance in the US. We feel there is plenty of room to improve the model and test it in various settings. The global estimate of epilepsy noted from our model is comparable with the approach presented earlier. 11 II. Incidence rates Aim—Determine the incidence of epilepsy in SC from existing data sources A. Posttraumatic epilepsy—we utilized the SC Traumatic Brain Injury Follow-up Registry (SCTBIFR) data to estimate the incidence of epilepsy after TBI. The cohort is comprised of persons aged 15 years and older with TBI randomly selected from the SC statewide non-federal hospital discharge data set over four years (January 1, 1999 through December 31, 2002) and recruited to participate in the follow-up telephone interviews. TBI was defined as any discharge with a primary or secondary diagnosis of trauma to the head associated with decreased consciousness, amnesia, other neurological or neuropsychological abnormalities, skull fracture, or intracranial lesion, in accordance with the CDC case definition of TBI. During the recruitment period 4,519 persons were discharged alive. At time of first interview, 3,746 persons (82.9%) were alive and eligible to participate. 2,118 (56.5%) of these were both able to be located and agreed to participate for the first interview one year after their discharge. This cohort was used to examine the relationship between TBI and epilepsy after injury. During the interviews, individuals were asked about the presence of seizures or epilepsy, both before and after their TBI. As detailed in Table 8, initially the questions were less sensitive to the presence of epilepsy after TBI, since an individual with epilepsy, but no recent seizures, might respond negatively. The questions were later changed to be more sensitive and a few were added to gain additional information. Table 8. Epilepsy-related questions used for initial identification of epilepsy cases. 1: Before your injury, did a doctor ever tell you that you had a seizure disorder or epilepsy? 1=Yes 2=No Used for first 1972 interviews: 2: During the past 4 weeks, have you had seizures or epilepsy? 1=Yes 2=No Used from interview number 1973 onward in place of #2 above: 2a: Since your injury [since the last time we talked to you], has a doctor told you that you had developed a seizure disorder or epilepsy? 1=Yes 2=No 2b: Are you currently taking any medicines to control your seizure disorder or epilepsy? 1=Yes 2=No All questions have response options of ‘refused’, ‘don’t know’, and ‘not applicable’. 12 If an individual had an ICD-9-CM discharge diagnosis code for epilepsy (345.x) or seizure (780.39) at discharge, or if they responded positively to the epilepsy-related questions during one of their interviews, their TBI charts were re-abstracted for additional information pertaining to seizures or epilepsy. Information gathered from chart re-abstraction included evidence of previous history of seizures or epilepsy, antiepileptic drugs (AEDs) prescribed prior to admission and those prescribed at discharge, whether the TBI was a result of a seizure, whether a seizure occurred after the TBI and if so when. The chart re-abstractions were completed prior to the completion of all followup interviews, however, and 30 additional potential cases of posttraumatic epilepsy (PTE) were identified after interview completion. These cases were not re-abstracted for additional information, but were evaluated in regard to seizures based on only the original abstraction and interview information (Figure 2). Due to changes in questions and skip patterns, we were not always able to determine the year of epilepsy onset with certainty. Such cases were assigned the year of epilepsy 13 diagnosis. In analyzing incidence, we used epilepsy onset at anytime during the three years following TBI discharge as our outcome. Both interview and medical record information were used to determine the presence of epilepsy prior to and after TBI (see Figure 2). Responses to the questions in Table 8, and re-abstraction information (when available), were used to determine initial categorization. After such determination, all cases with some indication of seizures or epilepsy had all available information reviewed individually by an epidemiologist and by a certified pediatric nurse practitioner who specializes in epilepsy to determine final epilepsy categorization. Any questionable cases were discussed by the entire team, including one or both of the epileptologists. Individuals with pre-existing epilepsy and those with uncertain status were excluded from all analyses of PTE. The incidence rate of PTE during the three years following TBI was calculated based on aggregate data taking into account those individuals not completing three years of interviews, and confidence intervals (CI) were calculated using tabulated values for a Poisson-distributed variable. Incidence of PTE was also calculated stratified by head injury severity using Abbreviated Injury Scale scores, categorized into mild, moderate, and severe. Possible factors involved in developing PTE were analyzed using Poisson regression, with a scale parameter estimated by the square root of Pearson’s chi-square divided by the degrees of freedom. The independent variables were derived from initial chart abstraction, chart re-abstraction, and first interview. Independent variables that showed no significant association with the dependent variable on chi-square analyses (p>0.10) were excluded from the initial regression model. Fisher’s Exact Test was used in place of chi-square when appropriate. Outcome variables were PTE and no epilepsy. Unlike the incidence rate calculation above, since this analysis involved examining individual characteristics related to outcome, a conservative approach was used and study participants included in the analyses who did not complete all three years of interviews, and who did not report epilepsy prior to their last interview, were assumed to have an outcome of no epilepsy. Table 9 shows information on the independent variables. The Cochran-Armitage test for trend was used to look for trends in outcome across ordinal variables. 14 Diagnostics were used to examine the appropriateness of the model. Variance inflation factors of the independent variables were calculated in an equivalent linear regression model to check for multicollinearity. The age category 55+ was inflated, but since this category was necessary in the model it was retained. Variables suspected of interaction were checked, and none was found. Deviance and Pearson chi-square statistics showed adequate fit with no evidence of overdispersion. Since there was attrition in the cohort, those completing three interviews were compared by chi-square to those who did not, to look for differences. Of 3,746 eligible persons, 2,118 (57%) were both located and participated in the first year interview. Of first year participants, 1,536 (72.5%) participated in the 2nd year interview, and 1,173 (55.4%) participated in the 3rd year interview. ICD-9-CM codes and interview responses were used to identify 325 potential cases with seizures or epilepsy, and 241 (74%) of their charts were re-abstracted. A total of 115 individuals were determined to have developed PTE in the three years following discharge for TBI. The incidence of PTE in the three years was 0.077 per person-year, which is equivalent to 2.6 cases per 100 person-years (95% CI 2.1, 3.1). Incidence by head injury severity was 1.2 cases per 100 person-years (95% CI 0.7, 1.8) for mild, 2.2 per 100 person-years (95% CI 1.3, 3.6) for moderate, and 3.9 per 100 person-years (95% CI 3.1, 4.9) for severe. See Appendix F (posttraumatic epilepsy incidence analysis) for more detail on the analysis. Table 10 shows the characteristics of the cohort by epilepsy status. In this analysis, 834 individuals who did not complete all three annual interviews, if they did not report epilepsy prior to withdrawal, were assumed to have an outcome of no epilepsy. Compared to individuals who did not develop PTE, the group with PTE has higher proportions of individuals who are middle-aged, male, Medicaid recipients, have severe TBI, early PTS, have sustained their injuries from violence, have no other injuries, have three or more comorbid conditions, and have a history of a previous head injury, stroke, or depression. Variables in which no significant difference was seen included race, trauma level status of hospital, and pre-TBI education, income, and substance abuse. It should be noted that preTBI income was unknown for 7.8% of the cohort, but did not differ between the outcome groups. 15 Table 9. Independent variables evaluated for the regression model. Variable and Source Age – Discharge dataset Sex - Discharge dataset Ethnicity - Discharge dataset Education – Interview report Income – Interview report Insurance - Discharge dataset Severity of TBI – Discharge dataset Early posttraumatic seizure – Re-abstraction information and any information that individuals may have volunteered during interview Etiology of TBI – Discharge dataset and original abstraction Number of concomitant injuries – Discharge dataset Number of comorbid conditions – Discharge dataset Trauma level status of hospital – Discharge dataset Previously knockedout/unconscious – Original abstraction and interview History of stroke – Interview report History of depression – Interview report and discharge dataset History of substance abuse – Discharge dataset, original abstraction, and interview Additional information on variable At time of injury Prior to TBI In the year prior to TBI Insurance status at time of discharge was grouped into the following categories: uninsured, Medicare, Medicaid plus other indigent programs, and private insurance, which included commercial insurance, Champus, Worker’s Compensation (WC), other agencies, and unknown. Commercial insurance made up 87% of the private insurance category. Unknown was 1% of the total cohort. Based on AIS (Association for the Advancement of Automotive Medicine, 1990) score for the head. Assigned by ICDMAP-90 software (Center for Injury Research Policy of the Johns Hopkins University School of Public Health, 1997) based on ICD-9-CM codes. Defined as any seizures within the first month, or if time was unknown, during the acute hospitalization. Violence category includes both with and without the use of weapons. Sports/other/unknown category is 54% ‘other’ and 18% ‘unknown’. Based on ICD-9-CM codes. Based on ICD-9-CM codes and Elixhauser et al’s categories.(Elixhauser et al. 1998) Interview report of prior episodes of being knocked out or unconsciousness and/or abstraction information on previous TBI. Having been told by a doctor that they had a stroke prior to their TBI. In less than 1% of cases were previous head injury or stroke unknown, and these were grouped in with those categorized as negative for those conditions. Prior to, or at time of discharge from, TBI. Interview report (“Before your injury, did a doctor ever tell you that you had depression?”) and/or ICD-9-CM code 296.2, 296.3, 300.4, or 311 (the latter code, for depressive disorder, was the most common of the 4 codes, representing 72% of individuals with one of these codes). 16% had an ICD-9-CM code and 84% were self-report only. A total of 95% asserted they had been told by a doctor before their injury that they had depression. Prior to, or at time of discharge from, TBI. Based on ICD-9-CM codes 303 through 305 (excluding 305.1), information in chart, and interview questions on drug and alcohol use. 16 Table 10. Comparison of characteristics in cases with and without posttraumatic epilepsy. Characteristic Age : 15-29 30-54 55+ Gender : Female Male Race: Nonwhite White Education: <HS grad HS grad >HS grad Income: <$20,000 $20-34,000 $35,000+ Unknown Insurance: Uninsured Medicaid/indigent Private/Other Medicare TBI Severity: Mild (AIS=2) Moderate (AIS=3) Severe (AIS=4,5) Early PTS: None known Yes Mechanism: Transportation Fall Violence Other/Unknown Multi Trauma: None 1-2 >=3 No. of comorbidity: None 1-2 >=3 Trauma level: 1 2 3 Undesignated Hx. unconsciousess: No Yes Hx. of stroke: No Yes Hx. of depression: No Yes Hx. of substance abuse: No Yes Posttraumatic epilepsy (N=115) 27 (23.5%) 55 (47.8%) 33 (28.7%) 36 (31.3%) 79 (68.7%) 32 (27.8%) 83 (72.2%) 36 (31.3%) 43 (37.4%) 36 (31.3%) 57 (49.6%) 34 (29.6%) 17 (14.8%) 7 (6.1%) 13 (11.3%) 28 (24.4%) 52 (45.2%) 22 (19.1%) 21 (18.3%) 16 (13.9%) 78 (67.8%) 93 (80.9%) 22 (19.1%) 51 (44.4%) 39 (33.9%) 19 (16.5%) 6 (5.2%) 69 (60.0%) 20 (17.4%) 26 (22.6%) 51 (44.4%) 46 (40.0%) 18 (15.7%) 64 (55.7%) 14 (12.2%) 25 (21.7%) 12 (10.4%) 76 (66.1%) 39 (33.9%) 102 (88.7%) 13 (11.3%) 77 (67.0%) 38 (33.0%) 49 (42.6%) 66 (57.4%) 17 No known epilepsy (N=1846) 642 (34.8%) 600 (32.5%) 604 (32.7%) 740 (40.1%) 1106 (59.9%) 430 (23.3%) 1416 (76.7%) 613 (33.4%) 608 (33.1%) 614 (33.5%) 943 (51.1%) 441 (23.9%) 316 (17.1%) 146 (7.9%) 210 (11.4%) 242 (13.1%) 949 (51.4%) 445 (24.1%) 744 (40.3%) 306 (16.6%) 796 (43.1%) 1809 (98.0%) 37 (2.0%) 1028 (55.7%) 522 (28.3%) 144 (7.8%) 152 (8.2%) 837 (45.3%) 470 (25.5%) 539 (29.2%) 1069 (57.9%) 683 (37.0%) 94 (5.1%) 900 (48.8%) 273 (14.8%) 430 (23.3%) 243 (13.2%) 1387 (75.1%) 459 (24.9%) 1735 (94.0%) 111 (6.0%) 1479 (80.1%) 367 (19.9%) 895 (48.5%) 951 (51.5%) Chi-square p-value .002 .062 .267 .643 .513 .008 <.001 <.001 .002 .009 <.001 .515 .031 .024 .001 .221 Table 11 shows the risk ratios of the independent variables and their 95% confidence intervals from the multivariable Poisson Regression. Results show that after adjusting for all the other variables in the model, individuals more likely to develop PTE had an early PTS, a severe head injury, three or more comorbid conditions, depression prior to or at the time of their TBI, and/or Medicaid health care coverage at time of TBI compared to individuals with the respective referent characteristics. In those variables mentioned above with multiple categories, the other categories did not show a significant impact on risk of PTE. For instance, moderate head injury, being insured by private/Worker’s Compensation(WC)/Champus/other/unknown insurance, being uninsured, and having one to two comorbid conditions were not significantly related to risk of PTE when compared to the reference categories. The model was also run with private/WC/Champus/other/unknown as the reference category for insurance. Individuals with Medicaid or other indigent insurance showed 2.31 (95% CI 1.38, 3.89) greater risk than those with private insurance. There were significant trends of increased risk of PTE with increasing head injury severity (p>.001) and an increasing number of comorbid conditions (p<.001). There was no significant association noted for the remaining variables. Chi-square analyses compared length of participation in the interviews by characteristics to calculate differential rates of attrition. A larger proportion of individuals developing PTE participated all three years (p<.001), with 55% of individuals not developing epilepsy completing all three years, and 70% of those with PTE completing all years. However, chi-square analyses showed no differences between length of participation when compared to severity of head injury (p=.891), or any other of the independent variables except insurance status (p<.001) and education (p=.047). Individuals participating all three years were more likely to have private insurance and to have post high school education. The risk of PTE could be falsely inflated if individuals who were more likely to develop PTE were more likely to participate all three years. However, since there was no difference in participation by severity or most of the other variables, it would appear more likely that the longer individuals remained in the cohort, the more likely we were to identify PTE. This, together with all individuals not 18 developing PTE prior to their leaving the cohort early being given an outcome of no epilepsy, might have reduced the strength of the actual risk relationships. Table 11. Risk Ratios of PTE in first 3 years after TBI for 1961 individuals Risk Ratio Characteristics (Reference level) (95% CI) Age (Ref=15-29 years) 30 – 54 1.63 (0.96, 2.77) 55+ 1.19 (0.54, 2.63) Gender (Ref=Female) Male 1.36 (0.85, 2.17) Insurance (Ref=Medicare) Medicaid or other indigent 3.52 (1.50, 8.25) Uninsured 1.98 (0.75, 5.19) Private/WC/Champus/other/unknown 1.52 (0.71, 3.26) Severity of TBI (Ref=Mild, AIS=2) Severe, AIS = 4 or 5 2.41 (1.39, 4.17) Moderate, AIS = 3 1.70 (0.82, 3.50) Early posttraumatic seizure (Ref=None known) Yes 6.52 (3.81, 11.17) Mechanism of injury (Ref=Transportation) Violence 1.67 (0.89, 3,13) Fall 1.05 (0.61, 1.83) Other (incl sports)/unknown 0.61 (0.24, 1.60) Number of concomitant injuries (Ref =3+) None 1.26 (0.73, 2.18) 1-2 0.82 (0.42, 1.59) Number of comorbid conditions (Ref=0) 3+ 3.14 (1.57, 6.28) 1-2 1.22 (0.76, 1.96) Previously knocked-out/unconscious (Ref=No) Yes 1.30 (0.83, 2.03) History of stroke (Ref=No) Yes 1.83 (0.88, 3.80) History of depression (Ref=No) Yes 1.86 (1.17, 2.96) Most studies have focused on clinical factors directly related to the TBI to determine the risk of PTE. This study differs by including limited clinical factors related to the brain injury, but a number of demographic, socioeconomic, and clinical factors occurring prior to the TBI. In concordance with much of the literature, there was an increased risk of PTE with increased severity. 19 Our study showed early PTS increased the probability of epilepsy more than any other factors. Our definition of early was liberalized to include seizures within a month of injury, or if time was unknown, anytime during the acute hospitalization. Since there was no specific interview question concerning early seizures, and not all charts were reabstracted for additional seizure and epilepsy information, it is possible that we did not identify all cases of early seizures. Choice of charts for re-abstraction was based on the presence of epilepsy or seizure discharge codes, or a positive reply to pre- or post-TBI epilepsy, with most of the early PTS identified through re-abstraction information. The 780.39 code had been used most often in our cohort for pre-existing epilepsy and/or for early PTS. If there were a large number of cases in which 780.39 was not used for early PTS, there could be a bias toward identifying early PTS in those developing PTE, resulting in an inflated association between PTE and early PTS. Unfortunately, while we know that 780.39 was often utilized for seizures other than early PTS (ie, pre-existing epilepsy or seizures that caused the TBI) we do not know how often early PTS was not given a 780.39 code. Transportation-related injuries were the most common mechanism in our TBI cohort. In our cohort, falls and violence had the highest proportions of severe head injuries, and transportation had the greatest proportion of mild head injuries. Mechanism was not significantly related to risk of PTE. In our analysis there was no association between the development of PTE and reporting prior episodes of being ‘knocked out’ or unconscious, or prior substance abuse. Since alcoholism and the prior occurrence of TBI are more prevalent in individuals with TBI, and both alcoholism and TBI are related to epilepsy, it could be argued that including individuals with these characteristics in our cohort may have increased the incidence of PTE. Ultimately we found no significant relationship between the development of PTE and either characteristic. We were limited in the amount of information we collected on previous substance abuse. Substance abuse included both alcohol and illicit drugs, and was based on ICD-9-CM codes at TBI discharge, self-report and abstraction information on alcohol or drug treatment, self-report on illicit drug use, and self-report on current drinking with comparison to pre-TBI drinking (using the COMBI definition of alcohol use (http://www.tbims.org/combi/subst/index.html)). 20 Since there was no specific measure of alcohol use prior to TBI, some individuals with prior alcohol abuse may have been missed. Individuals identified as having depression in our cohort were almost twice as likely to develop PTE. Although the majority of these individuals were identified solely through self-report of history of depression (84%), this provides further evidence of depression as a risk factor not only for epilepsy in general, but also for epilepsy after head injury. Individuals in our study showed both an increased risk of PTE with three or more comorbid conditions at discharge, and a trend of increasing proportion of PTE with increased number of comorbid conditions. Since stroke can be a predecessor to epilepsy, especially in older people, we included it as a separate variable. However, people reporting that they had been told they had had a stroke prior to their head injury did not show an increased risk of developing PTE. In addition, neither number of concomitant injuries nor trauma level status of the hospital showed an association with risk of PTE. In our cohort of TBI patients, no significant difference in risk of PTE between age categories was found. It must be noted that among the 4,519 people with TBI eligible for the study, 382 (8.5%) died prior to their first year anniversary of discharge. Of the deceased, 80% were 55 years and older, while that age group made up only 33% of the eligible population. Older persons were more likely to die in the first year, likely removing more seriously injured individuals who would have been at higher risk for PTE, and thus possibly lowering the reported risk of PTE in that age group and in the cohort. Our study found no difference in risk of PTE by race, either before or after adjusting for income, as well as no difference in risk of PTE by gender. There was no difference in the development of PTE by pre-TBI income or education. Interestingly, individuals with Medicaid or other indigent assistance on discharge from their TBI had a significantly increased risk of developing PTE over individuals with Medicare or some form of private health insurance. It is not known whether SES, quality or continuity of health care, or some other factor is responsible for this relationship. The relationship to insurance status remained significant when adjusted for both pre-TBI income and trauma level status of the acute care hospital. It is possible that those 21 individuals requiring long-term hospitalization became Medicaid recipients prior to their discharge. Some limitations of this study have been previously mentioned. Most important is the reliance on self-report of seizures or epilepsy status, as well as pre-TBI clinical variables. Individuals could have erroneously reported the occurrence of epilepsy, stroke, depression, substance abuse, or previous TBI, or have underreported them, since we had corroborating clinical information in only some cases. It is possible that individuals who only answered ‘yes’ to the interview question regarding taking medicine for epilepsy were on medication prophylactically after their TBI. However, this would seem unlikely because of the fairly long span of time between discharge and the beginning of interviews. While recognizing these limitations, our study has the advantages of being population-based, including large numbers, demographic heterogeneity, three years of follow-up, and representing all degrees of severity seen among hospitalized TBI. It also identifies characteristics present either prior to or concurrent with injury associated with the later development of PTE. While confirming some risk factors established by other studies, such as early PTS and severity of head injury, our study has identified other associations with PTE—especially depression, the presence of three or more comorbid conditions, and Medicaid insurance—that are less well established and warrant further research. Information on such associations can be used to better predict those at increased risk of PTE, and may eventually enable early interventions to reduce this risk or its consequences. B. Incidence of epilepsy utilizing ED and HD data—A main difficulty in determining epilepsy incidence is determining the onset of epilepsy. Rarely can a first seizure indicate epilepsy. In general, this only occurs when an individual has a type of epilepsy with a distinctive EEG pattern (for example, infantile spasms where the seizure and EEG findings (hypsarrythmia pattern) are specific to the epilepsy syndrome, or absence epilepsy where the seizure and EEG findings (3 per second spike and wave pattern) are specific to the epilepsy syndrome). We decided to take all cases in which the clinician review indicated a new seizure or new epilepsy case or questionable diagnosis, or if it 22 was a new case but the final diagnosis was questionable, and ask ORS to follow them forward in time, looking for any additional diagnoses of 780.3 or 345.x. Below are the numbers of new cases. Inpatient/ED charts abstracted: 2001 N=2530, 2002 N=1453 • New onset cases, inpt/ED, epilepsy or seizure: 2001 n=307, 2002 n=124 • New onset cases, inpt/ED, questionable final dx: 2001 n=8, 2002 n=1 Physician office visit (POV) charts abstracted (2001 & 2002): N=302 • New onset cases, POV, inpt/ED: 2001 n=8, 2002 n=3 • New onset cases, POV, questionable final diagnosis: none A total of 451 cases were identified as new onset cases. Seven of the POV cases were Medicare, and ORS was unable to follow them forward since they do not have any other years of data for Medicare. The other four cases would have involved enlisting two other personnel (one from State Health Plan and one from Medicaid), and it was deemed that the time and expense in tracking just these four would be prohibitive. However, it does not seem unlikely to assume that most individuals with a new onset seizure would likely be seen in an ED or as an inpatient, rather than in a physician’s office. In the case of more than one diagnosis code after clinician review, we assigned a code hierarchically: 345.x > 780.31 > 345.2, 345.3 > 780.3 > 780.2 > 293.0. The following 440 cases were considered new onset after clinician review: Dx. After review ? ? Epilepsy Febrile Febrile Seizure Seizure Status Status Year 2001 2002 2001 2001 2002 2001 2002 2001 2002 Frequency 8 1 4 51 24 192 78 60 22 The above cases were sent to ORS to follow them forward in time for later 780.3 or 345.x diagnoses. Of note, there were an additional 298 cases in the database in which new onset could not be determined. Most of these (66%) also had questionable final diagnoses. 23 For the purposes of determining epilepsy incidence, a case was considered epilepsy if they had two diagnoses of 780.39 and/or 345.x, with onset considered the initial seizure. There were 4 cases in which the clinical reviewers had previously assigned or confirmed a diagnosis of new onset epilepsy. Three of these four cases showed subsequent seizures. The one without any subsequent seizures recorded at ORS was removed as a new onset epilepsy case for this analysis. Febrile seizures (780.31) were not counted as seizures. If, however, an individual with a 780.31 diagnosis had at least two subsequent non-febrile seizure diagnoses, they were counted as a new onset case of epilepsy, with the onset as the year of the first febrile seizure. There were six such cases. The following analysis includes only inpatient and ED cases. 1. ALL INCIDENT CASES FROM SAMPLE By diagnosis assigned after clinician review, and year: Year=2001 ASSIGNED CODE EPILEPSY SEIZURE STATUS Frequency 4 83 31 Percent 3.39 70.34 26.27 Cumulative Frequency 4 87 118 Cumulative Percent 3.39 73.73 100.00 Frequency 2 37 15 Percent 3.70 68.52 27.78 Cumulative Frequency 2 39 54 Cumulative Percent 3.70 72.22 100.00 Year=2002 ASSIGNED CODE FEBRILE SEIZURE STATUS By original diagnosis and year: Year=2001 ORIGINAL CODE EPILEPSY SEIZURE STATUS Frequency 12 97 9 Percent 10.17 82.20 7.63 Cumulative Frequency 12 109 118 Cumulative Percent 10.17 92.37 100.00 Frequency 10 38 6 Percent 18.52 70.37 11.11 Cumulative Frequency 10 48 54 Cumulative Percent 18.52 88.89 100.00 Year=2001 ORIGINAL CODE EPILEPSY SEIZURE STATUS 24 2a. TOTAL EPILEPSY SAMPLE PER ABSTRACTION DATA By original diagnosis and year: Year=2001 ORIGINAL CODE EPILEPSY FEBRILE SEIZURE STATUS Frequency 628 110 1656 136 Percent 24.82 4.35 65.45 5.38 Cumulative Frequency 628 738 2394 2530 Cumulative Percent 24.82 29.17 94.62 100.00 Frequency 331 37 989 96 Percent 22.78 2.55 68.07 6.61 Cumulative Frequency 331 368 1357 1453 Cumulative Percent 22.78 25.33 93.39 100.00 Year=2002 ORIGINAL CODE EPILEPSY FEBRILE SEIZURE STATUS 2b. TOTAL EPILEPSY SAMPLE PER ORS (included status as epilepsy, and febrile as seizure) By code and year: Year=2001 CODE 2930 345X 7802 7803 Frequency 12 762 66 1690 Cumulative Frequency 12 774 840 2530 Percent 0.47 30.12 2.61 66.80 Cumulative Percent 0.47 30.59 33.20 100.00 Year=2002 CODE 2930 345X 7802 7803 Frequency 5 488 47 913 Percent 0.34 33.59 3.23 62.84 Cumulative Frequency 5 493 540 1453 Cumulative Percent 0.34 33.93 37.16 100.00 3. TOTAL EPILEPSY POPULATION (included status as epilepsy, and febrile as seizure) By code and year: Year=2001 CODE 2930 345X 7802 7803 Frequency 1020 1723 16574 18749 Percent 2.68 4.53 43.54 49.25 Cumulative Frequency 1020 2743 19317 38066 Cumulative Percent 2.68 7.21 50.75 100.00 Percent 2.64 3.97 44.36 49.02 Cumulative Frequency 1057 2646 20395 40007 Cumulative Percent 2.64 6.61 50.98 100.00 Year=2002 CODE 2930 345X 7802 7803 Frequency 1057 1589 17749 19612 25 The original plan was to choose 35% of 345.x codes (epilepsy), 5% of 780.3 codes (convulsions), 1% of 780.2 codes (syncope & collapse), and 5% of 293.0 codes (acute delirium) for 2001. In 2002 we wanted to have a total abstraction of approximately 1500 charts, and inflated that number upward to adjust for expected 15% unlocated charts. As seen further on, these percents were not exact, possibly due to what charts were able to be located, change in personnel, as well as the need to determine final percents on one diagnosis rather than the 2 or more that were sometimes present. In pulling the sample, ORS allowed those with more than one seizure-related diagnosis (ie, 780.3 and 780.2) to go into more than one group for pulling the sample. When sending us the population from which the sample was pulled, we requested ORS assign one diagnosis to each case, using the hierarchy 345.x>780.3>780.2>293.0, which is the same one used in assigning diagnoses to the abstracted cases. I believe that this most likely inflated the proportions of the ‘higher’ diagnoses (345.x and 780.3) since they would have priority in labeling. In 2001, 2.1% of the cases had 2 seizure-related diagnoses, and in 2002, 3.0% had 2 seizure-related diagnoses. While ORS was able to match the individuals back to the appropriate individuals in each year, the personnel putting together the population dataset was different from that which pulled the original sample, and they had difficulty matching cases to the exact visit, resulting in some variation in diagnoses and payers. Also of note, in addition to the 11 physician office visits that could not be followed forward, there were 7 HD/ED cases that ORS personnel could not match back to the database so they was unable to follow them forward. Also, there were an additional 3 HD/ED cases in which the diagnoses from ORS did not match the original diagnoses sent in the abstraction sample, so those were not followed forward. Together, that is a total of 21 cases which were new onset but were not able to be followed to determine whether they had additional seizures. Finally, individuals may have moved out of SC or received ED or inpatient care outside of SC between the year of the abstraction and 2005, and thus additional episodes of seizure could have been missed. If any of these cases became epilepsy, it would mean our estimate is an under estimate of the true incidence of epilepsy. 26 Of the 10 HD/ED cases that could not be matched, all had original diagnoses of 780.31. Eight of them were considered after clinical review to be correct, and two were reassigned codes of 780.39 since their ages were 10 and 74 years. Eight were 2001 cases and 2 were 2002 cases. The incorrectly coded cases were both from 2001. Of the 11 physician office visit cases, there were 3 cases from 2001 all of which were originally coded as seizure, and were considered to be correct after clinical review. The other eight cases, five from 2001 and three from 2002, were all originally coded as epilepsy, but were considered to be seizure cases after clinical review. There were slightly fewer epilepsy diagnoses in 2002 than in 2001 (1589/1723=92%), and there were slightly more seizure-related diagnoses in 2002 than in 2001 (19612/18749=105%). Because the sample reflects the population of cases with epilepsy and seizure diagnoses, we did not weight numbers when applying the sample incidence to the population. Therefore, total new cases of epilepsy in 2001 are 47 with 345.x diagnoses, and 1030 with 780.3 diagnoses. Total new cases of epilepsy in 2002 are 60 with 345.x diagnoses, and 726 with 780.3 diagnoses. In 2001, 2.7% of 345.x cases and 5.5% of 780.3 cases became epilepsy within the next 4 years. In 2002, 3.8% of 345.x cases and 3.7% of 780.3 cases became epilepsy within the next 3 years. Population information was acquired from SC Statistical Abstract (Table 23), 1990, 2000 and US Census Bureau, State Population Estimates 2, 3. Using 345.x (epilepsy & status) and 780.3 (seizures, including febrile) categories: • Inpt/ED population, 2001: 345.x = 1723; 780.3 = 18,749 • Inpt/ED sample per abstraction, 2001: 345.x = 764; 780.3 = 1766 [Inpt/ED sample per ORS, 2001: 345.x = 762; 780.3 = 1690] • Inpt/ED incident cases, 2001: 345.x = 21; 780.3 = 97 *2001 cases have at most 4 years of follow-up (through 2005) • 2001 sample was 44% of 345.x cases & 9% of 780.3 cases. Inflating incident cases to population results in 47 cases 345.x and 1030 cases 780.3. • Inpt/ED population, 2002: 345.x = 1589; 780.3 = 19,612 • Inpt/ED sample per abstraction, 2002: 345.x = 427; 780.3 = 1026 2 3 http://www.ors2.state.sc.us/abstract/chapter14/pop23.asp http://www.census.gov/popest/states/asrh/SC-EST2003-02.html 27 • • • [Inpt/ED sample per ORS, 2002: 345.x = 488; 780.3 = 913] Inpt/ED incident cases, 2002: 345.x = 16; 780.3 = 38 *2002 cases have at most 3 years of follow-up (through 2005) 2002 sample was 27% of 345.x cases & 5% of 780.3 cases. Inflating incident cases to total population results in 60 cases of 345.x and 726 cases 780.3. If NO adjustment for age, sex, or race, using the 2000 SC Census population, result would be the following: • 2001: 1077/4,012,000 = 0.00027 = 0.27 per 1000 people incident cases of epilepsy • 2002: 786/4,012,000 = 0.00020 = 0.20 per 1000 people incident cases of epilepsy III. Causes (triggers) of seizures and etiologies of epilepsy Aim—Determine the underlying causes and etiologies of epilepsy in South Carolina. In abstracting the data from charts, we asked the abstractors to note any information on what caused seizures (i.e., ‘triggers’) and the etiology of epilepsy. As part of the skill building instruction to abstractors, differences between these two concepts were presented, however we had to acknowledge that this distinction can be difficult for someone without a clinical background in epilepsy, and that there can be an overlap, such as the person who experiences a head injury and an immediate seizure, and then goes onto develop epilepsy, or the person with alcoholism and epilepsy. As shown in Appendices A&B, we had specific causes individually listed (injury, illness, fever, lack of sleep, pregnancy, eclampsia, alcohol use, drug use, change in medication, weight gain in children, noncompliance with medication) to try to prevent them from missing any that might be mentioned in the chart. We provided a place for narrative for the abstractors to write any information on causes and etiologies (ie, ‘Brief description of current circumstances that might have contributed to this seizure or seizure-like episode (what provoked this seizure?)’, and ‘Any past illness, condition, or injury in the patient’s history that initially caused the seizure disorder/epilepsy’). The initial results for cause and etiology had a high proportion of cases in which there was no information. For instance, in the inpatient/ED abstracted data, only about 20% of cases had an etiology listed under the actual ‘etiology’ variable. However, in reviewing the results of the chart abstractions, we noted that there were occasions when the 28 abstractors appeared to mix up cause and etiology, and times when a cause or etiology was noted among the text in other variables. In an attempt to use all abstracted information, we have had our pediatric nurse practitioner (GS) who specializes in epilepsy begin reviewing each case individually. If there are triggers that are mentioned in the text but not checked off in the categorical variables, then that is corrected. In addition, we developed a list of categories of etiologies, and she is categorizing text from the etiology text variable, as well as any possible etiology mentioned elsewhere in the abstraction. Since there are over 4,000 cases, this has involved a large time commitment. At present writing, she has completed reviewing 2,000 (50.2%) of the inpatient/ED cases, and intends to complete all cases as time allows, including physician office visit cases. The following analyses on the inpatient/ED data are from the 2000 completed cases, and are limited to those cases determined to be epilepsy after clinician review (n=1,476). When characteristics of age group, sex, race, and primary payer are compared, the records reviewed to date (nearly 50%) are reflective of the total group of epilepsy cases. In reviewing the abstracted data, information that could be used to infer possible etiologic/causal factors were noted to be in multiple areas. Diagnoses were used to assist in understanding causes of the seizure for that visit as well as etiologies for epilepsy, but it should be noted that not all diagnoses were always coded in that section. Instead, diagnostic information could also be found in chart histories and exams that were quoted by the abstractors as additional information deemed as important. The information given often was not complete enough to note if a condition was the etiology or a co-morbid condition of the epilepsy. (The etiologies list was modified and developed from the Sander’s article on “The epidemiology of epilepsy revisited.” Please see Table 12 below.) If such information was present, it was taken into consideration. For example, occasionally a chart would note pre-existing diagnosis of epilepsy but a new onset cerebrovascular accident (CVA) occurring at the hospital visit being reviewed. If this occurred, care was taken to avoid identifying the CVA as the etiology of the epilepsy. However, if CVA was merely listed without any temporal description, an inclusive approach was taken, and it was included under ‘CVA’. There are known reasons for the development of secondary epilepsy like Alzheimer’s, substance abuse (although debated), head trauma, brain tumors, AIDS, secondary metastasis to the brain, lupus, etc. 29 There are also diseases/disorders where the development of secondary epilepsy appears to occur frequently whether as part of the disease process or treatment, for example, diabetes, asthma, autoimmune disorders, other cancers, etc. There is also new research proposed that may identify the increased occurrence of a co-morbid condition, like depression or anxiety, with epilepsy due to similar pathophysiology. Therefore, all of these conditions were noted as possible etiologies to be considered. If not specifically noted as a separate etiology, they were coded as ‘other disease processes’. Table 12. Potential Etiologies of Epilepsy per Sander†. A. Childhood, adolescence and early adulthood 1. Congenital 2. Developmental 3. Genetic 4. Prenatal/perinatal/postnatal factors 5. Febrile seizures 6. Neonatal seizures (not assigned as data do not identify age <12 months) B. At any time 7. Head trauma 8. Central nervous system infections 9. Tumors (brain) 10. Substance abuse 11. Other disease process (Alzheimer’s, AIDS, autoimmune, other cancer, cancer metastases, diabetes, asthma, depression/anxiety renal, etc.) C. Elderly 12. CVA D. Certain environments 13. Endemic infections (malaria, neurocysticerocosis, paragonomiasis & toxicariasis) E. Family 14. Relative with epilepsy F. Unknown/not applicable 15. Unknown etiology 16. No identified epilepsy † Based upon Sander, JW. (2003). The epidemiology of epilepsy revisited. Current Opinion in Neurology. 16, 165-70. 30 Other concerns in using the modified epilepsy etiologies list from Sander include identifying neonatal seizures and febrile seizures. The data from ORS did not identify the specific age of an infant between 0 and 11 months. This meant neonatal seizures could not be identified. Also, this may skew the febrile seizure data, as per our protocol all children under 6 years of age were considered to have febrile seizure disorder unless other data was presented which identified another disorder that would rule out febrile seizures, such as cerebral palsy, infantile spasms, etc. Those children younger 6 months were not identified so all children listed as 0 years were given a febrile seizure diagnosis unless other information is included that disputed otherwise. Based on our chart review, 1,430 unduplicated individuals that were determined to have epilepsy had been assessed to identify the etiology of epilepsy. In some cases more than one etiology may be listed for an individual making the total over 100%. Most commonly, the etiology is either unknown, or is not stated in the chart (see Figure 3 below). The next most common, is ‘other disease processes’. As mentioned above, etiology is loosely defined, allowing what could be comorbid conditions to be stated as etiology. UNKNOWN/NOT STATED 37.5% OTHER DISEASE PROCESS 28.5% 13.9% SUBSTANCE ABUSE CVA 7.6% HEAD TRAUMA 6.7% DEVELOPMENTAL 4.8% NATAL FACTORS 3.8% 2.8% RELATIVE W/EPILPESY CONGENITAL 2.4% TUMORS(BRAIN) 2.2% CNS INFECTIONS GENETIC 0.8% 0.8% FEBRILE SEIZURES 0.6% 0.0% 5.0% 10.0% 15.0% 20.0% 25.0% 30.0% 35.0% 40.0% Figure 3. Etiology categories of inpatient/ED epilepsy cases Any possible causes of seizures for a visit were identified from the narrative, diagnoses, and other variables listed as possible causes. If a person presented with a new 31 onset seizure after head injury, both the new onset seizure and head injury were listed to increase the information gleaned from the chart review. In addition, if a person with known epilepsy had no identified cause, the variable of “breakthrough seizure” was used. The various causes of seizure included in our chart reviews are shown in Table 13. Table 13. Complete list of cause of seizure • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • AED above therapeutic level Accidental near drowning Accidental poisoning Anorexia Anxiety Breakthrough seizure CVA Dehydration Drug overdose Electrolyte imbalance Fever Head injury Herbal ingestion Illness Injury Job injury Medication changes Medication reaction Menses MVA N/A (no seizure related to visit) New onset seizure Noncompliance Postpartum < 1 month Pregnancy Respiratory distress Seizure versus pseudoseizure Shunt failure Sleep deprivation Stress Structural brain issue Substance abuse versus using alcohol and/or drugs Subtherapeutic AED level Suicide attempt TIA Unknown VEEG admit VNS (vagal nerve stimulator) placement VNS replacement 32 In analyzing the immediate seizure causes, cases that were abstracted in both 2001 and 2002 were retained, since seizures may be triggered by different factors at different times. In this subset, there were 44 cases with data from both years. 542 of the visits did not involve a seizure, so these were removed, leaving analysis on 952 seizures that occurred in 928 individuals. More than one cause was allowed making the total more than 100 percent. The top causes (those >1% in frequency) are listed below in Figure 4. The most common cause is a “breakthrough seizure”, which indicates the cause is unknown. That is followed by illness, which indicates a current or recent illness. The next three causes could be considered preventable: noncompliance, subtherapeutic antiepileptic medication level, and substance abuse. They could be prevented with intervention. There are other causes on the list that could be affected by intervention, and include antiepileptic levels above a therapeutic level, sleep deprivation, and injury. Counting just the six causes mentioned, over 50% of these epileptic seizures could possibly be prevented with appropriate treatment, monitoring, and counseling. Medication changes included only changes to antiepileptic medications, and a video electroencephalogram admission (VEEG admit) indicated someone admitted for a VEEG who may have had seizures during the course of the procedure. 28.2% BREAKTHROUGH SEIZURE 20.6% ILLNESS 19.7% NONCOMPLIANCE 15.8% SUBTHERAPEUTIC AED 13.7% SUBSTANCE ABUSE 6.2% MEDICATION CHANGES 4.0% FEVER 2.8% AED ABOVE THERAPEUTIC ELECTROLYTE IMBALANCE 1.8% VEEG ADMIT 1.6% SLEEP DEPRIVATION 1.3% INJURY 1.3% PREGNANCY 1.2% 0.0% 5.0% 10.0% Figure 4. Causes of seizure 33 15.0% 20.0% 25.0% 30.0% We performed similar analysis on the physician office data. This work is still in progress and results presented in this report reflects only those with a final clinicianreviewed diagnosis of epilepsy (n=276). In looking at etiology, only items listed under the variables ‘etiology’ and ‘family history’ were considered, with the project epidemiologist categorizing the etiologies according to the same categories as used for the HD/ED data. There should be additional information after clinician review. More than one etiology is allowed making the total is greater than 100%. As with the inpatient/ED data, the most common category is unknown/not stated. There are differences between the etiologies of the HD/ED and the POV patients, which could be due to the differences in both patients and physicians (see Figure 5). For instance, the increase in known relatives with epilepsy may reflect the more in-depth history taken in a physician’s office, and the decrease in substance abuse as an etiology may reflect a difference in where health care is accessed. UNKNOWN/NOT STATED 62.7% RELATIVE W/EPILPESY 17.0% HEAD TRAUMA CVA 6.9% 5.4% CNS INFECTIONS 3.6% NATAL FACTORS 2.9% CONGENITAL OTHER DISEASE PROCESS 2.5% 2.2% SUBSTANCE ABUSE 1.1% FEBRILE SEIZURES 1.1% TUMORS(BRAIN) 0.7% DEVELOPMENTAL 0.7% GENETIC 0.4% 0.0% 10.0% 20.0% 30.0% 40.0% 50.0% 60.0% 70.0% Figure 5. Etiology categories of physician office epilepsy cases The categorized causes are shown below in Figure 6. Only information already categorized in the data by the abstractors is included. This work is in progress and there 34 will be an update after clinician review of all information. More than one cause is allowed and the total percent exceeds 100. More than half of the records had no cause listed for the most recent seizure. The second most common cause was a change in antiepileptic medication, followed by injury (other than head) and illness. Noncompliance and substance abuse are not as common among the POV patients as the HD/ED patients. AED level in these patients was the most recent level done, which may or may not have been connected to the most recent seizure, so it was examined separately. In 20% of patients, the AED levels were subtherapeutic, and 17% were above the therapeutic range. 58.7% NONE OF THESE LISTED MEDICATION CHANGES 14.5% OTHER INJURY 11.6% 10.1% FEVER ILLNESS 9.4% 6.2% NONCOMPLIANCE HEAD INJURY 5.1% SUBSTANCE USE/ABUSE 1.8% SLEEP DEPRIVATION 1.8% WEIGHT GAIN (PED) 0.7% PREGNANCY 0.4% 0.0% 10.0% 20.0% 30.0% 40.0% 50.0% 60.0% 70.0% Figure 6. Causes of seizures among POV patients with epilepsy IV. Target populations for intervention A. Medication use - HD/ED data In abstracting patient charts, we asked the abstractors to note any antiepileptic medication that the patient was taking at the onset of the visit. To examine medication use, we looked only at those with epilepsy that is not new onset. If they were abstracted in both 2001 and 2002 we used the data from 2002. Of 2,950 such cases, 2226 (75.5%) had their current AEDs listed on the chart. We classified drugs approved by the Federal 35 Drug Administration since 1990 for use in epilepsy as ‘new’, based on their recognition by the American Academy of Neurology and the American Epilepsy Society as new generation AEDs. Drugs released prior to this we classified as ‘old’. Table 14 below shows the percent of patients taking each drug, and its classification. Patients often took more than one drug making total percent greater than 100. More than half of the patients were taking phenytoin, an older drug with a fairly significant profile of side effects and difficult pharmacokinetics. Of the 263 women between the ages of 15 and 44 years, 263 (60%) were taking either phenytoin, carbamazepine, lamotrigine, or valproate, all of which have recently been shown to be connected to an increased rate of miscarriage and birth defects. Table 14. AEDs being taken by inpatient and ED patients (N=2,226). Trade name Dilantin, Phenytek Generic name Phenytoin % taking 55% Type Old Depakote/Depakene Divalproex sodium/Valproic acid 19% Old Carbatrol, Tegretol Carbamazepine 18% New Luminal Phenobarbital 13% Old Neurontin Gabapentin 6% New Keppra Levetiracetam 5% New Ativan Lorazepam 4% Old Lamictal Lamotrigine 4% New Topamax Topiramate 3% New Klonopin Clonazepam 3% Old Trileptal Oxcarbazepine 2% New Valium Diazepam 2% Old Zonegran Zonisamide 1% New Gabitril Tiagabine 1% New Mysoline Primidone 1% Old Felbatrol Felbamate <1% New Zarontin Ethosuximide <1% Old Tranxene Clorazepate <1% Old Celontin Methsuximide <1% Old Depacon Valproate sodium <1% New 36 Patients were grouped according to the number of AEDs they had been prescribed prior to admission. 70.5% took only one AED, 24.2% took two AEDs, and 5.3% took three or more AEDs. A logistic regression model was built utilizing the independent variables listed in Table 15. Among individuals who had no insurance were included individuals who were listed as indigent, or who whose visit was covered under charity. Substance use/abuse could indicate either past or present, and included alcohol or illicit drugs. Abstractors could mark either ‘yes’ to questions concerning past and present use or abuse if such information was charted, could mark ‘no’ if past and present use or abuse was specifically denied in the chart, or ‘unknown’ if use or abuse was not mentioned. The outcome was monotherapy versus polytherapy. Table 15. Odds of taking more than one antiepileptic drug. Independent variable Adjusted odds ratio (95% CI) Age group 0-21 1.74 (1.14, 2.65) 22-44 2.78 (2.01, 3.84) 45-64 2.15 (1.58, 2.93) 65+ Reference Gender Female 0.99 (0.82, 1.20) Male Reference Race Nonwhite 0.73 (0.60, 0.89) White Reference Insurance None (plus indigent/charity) 0.50 (0.35, 0.73) Medicaid 1.30 (0.98, 1.72) Medicare 1.72 (1.29, 2.30) Private Reference Substance use/abuse Yes 0.24 (0.14, 0.42) Unknown 0.42 (0.26, 0.68) No Reference Seizure-related visit Yes 1.36 (1.12, 1.67) No Reference Individuals under age 65, whites, individuals on Medicare, individuals without substance abuse or use, and individuals whose visit was seizure-related were more likely to be taking more than one antiepileptic drug. Gender did not appear to affect whether an individual was on 37 monotherapy or polytherapy. It was also noted that individuals without insurance were more likely to be on monotherapy than individuals with commercial insurance. Patients were also grouped according to whether they took only newer AEDS, only older AEDs, or took a mix of older and newer AEDs. 79.6% took only older medications, 9.5% took only newer medications, and 10.9% took a mix of the two. Another logistic regression model was built using the same variables as above, but with an outcome of taking either only a new AED(s) or a mix of old & new AEDs versus taking only an older AED(s). Individuals were more likely to be recorded as taking only an older AED(s) if they were over age 65 years, male, nonwhite, had no insurance, were noted to have substance use/abuse or if that knowledge was unknown, and if their visit was not seizure-related (Table 16). Table 16. Odds of taking only ‘older’ antiepileptic drugs. Independent variable Adjusted odds ratio (95% CI) Age group 0-21 0.24 (0.15, 0.39) 22-44 0.27 (0.18, 0.40) 45-64 0.65 (0.43, 0.97) 65+ Reference Gender Female 0.67 (0.53, 0.84) Male Reference Race Nonwhite 1.94 (1.54, 2.45) White Reference Insurance None/indigent 4.82 (2.94, 7.90) Medicaid 0.99 (0.74, 1.34) Medicare 1.06 (0.77, 1.48) Private Reference Substance use/abuse Yes 3.29 (1.79, 6.04) Unknown 2.34 (1.42, 3.87) No Reference Seizure-related visit Yes 0.67 (0.53, 0.86) No Reference 38 B. Medication use - POV data As stated earlier, we asked the abstractors to note what AEDs the patient was listed as taking at the onset of the most recent visit. As a reminder, these are individuals receiving Medicare, Medicaid, or State Health Plan, and who did not have an inpatient or ED visit during 2001 and 2002. To examine medication use, we looked only at those with epilepsy that is not new onset. Patients may be taking more than one AED. Table 17 below shows the percent taking each type. Of 276 such patients, 270 (97.8%) had Table 17. AEDs being taken by patients attending POV (n=270). Trade name Generic name Carbatrol, Tegretol Carbamazepine 32% New Dilantin, Phenytek Phenytoin 30% Old Depakote/Depakene Divalproex sodium/Valproic acid 18% Old Keppra Levetiracetam 14% New Lamictal Lamotrigine 11% New Topamax Topiramate 11% New Neurontin Gabapentin 9% New Luminal Phenobarbital 8% Old Klonopin Clonazepam 4% Old Mysoline Primidone 3% Old Zonegran Zonisamide 3% New Trileptal Oxcarbazepine 3% New Ativan Lorazepam 2% Old Gabitril Tiagabine 1% New Felbatrol Felbamate 1% New Zarontin Ethosuximide 1% Old Tranxene Clorazepate 1% Old Valium Diazepam <1% Old Celontin Methsuximide <1% Old Depacon Valproate sodium <1% New 39 % taking Type medications listed. 93.7% were seen by a neurologist or epileptologist. While the three most common AEDs are the same, only 30% are taking phenytoin, compared to 55% of the HD/ED patients. Also, 8% are taking Phenobarbital, compared to 13% of the HD/ED patients. Newer AEDs are being used among these patients than among the HD/ED patients. 60.0% took only one AED, 27.8% took two AEDs, and 12.2% took three or more AEDs. 52.2% took only older medications, 23.7% took only newer medications, and 24.1% took a mix of the two. C. Posttraumatic epilepsy A possible target for intervention in preventing epilepsy could be individuals who sustain TBI. It was noted in an earlier section of the report that individuals who developed posttraumatic epilepsy within the first three years of discharge from hospitalization were more likely to have certain characteristics. Individuals who experienced early posttraumatic seizures, with more severe head injury, with a history of depression, with three or more comorbid conditions, and under Medicaid, were more likely to develop posttraumatic epilepsy. While some of these characteristics are markers of risk, some of them are amenable to early intervention to allay posttraumatic epilepsy. Depression as a predisposing factor for epilepsy is fairly new in the literature, but offers an intriguing avenue for further exploration to determine whether treatment of depression could affect the incidence of posttraumatic epilepsy. D. Behavioral Risk Factor Surveillance System data The BRFSS questions on epilepsy include a question on whether an individual reporting having been told they have epilepsy have experienced a seizure within the past three months. Of those individuals reporting one or more seizures within the previous three months, only 74.0% (95% CI 58.6-89.4) are currently taking a medication to control epilepsy. While these cases could involve issues of personal choice, nonepileptic seizures, etc., an important area to explore further would be why over a quarter of individuals reporting seizures are not taking medication. The BRFSS also included a question on multivitamin use. Whereas 85% of the general population reported taking a multivitamin, only 69% of individuals who had been told they had epilepsy reported taking a multivitamin. This may be an important area of intervention due to the potential 40 effects of antiepileptic drugs. Levels of heavy drinking and obesity were not significantly different between those reporting epilepsy and the general population. However, significantly more individuals with active epilepsy (i.e., those taking medication and/or reporting a recent seizure) were current smokers (36.8%, 95% CI 32.5-45.7) than those in the general population (22.5%, 95% CI 21.4-23.7). This is another important area for intervention, considering the cardiovascular effects of smoking on the brain, plus the interplay of untreated mood disorders and smoking. E. Focus group data Six focus groups were held in four different geographic locations throughout SC. The topic to be explored was to learn what problems and barriers people with epilepsy in SC face. The specific questions are in Table 18 below. There were 41 individuals who Table 18. Focus group questions. 1) Please tell the group: a) Your first name, when you were diagnosed with epilepsy and how controlled your seizures are now. b) If you are a parent of someone with epilepsy, please tell us the age of your child at diagnosis and how controlled their seizures are now. 2) What services, help, information or care exists to help people with their epilepsy needs or issues? Probes: a) Have you used any of these services you listed over time? If yes, what was your experience in using these services? b) Are there any services you wanted but could not get? If yes, what was the problem? c) What services or help are missing that you need/needed? 3) How does a person find out about services or help for epilepsy? How did you find out? a) How might finding out about services be improved? 4) What challenges or problems has epilepsy created in your life over time? Probes: a) How have these challenges changed throughout your life? b) What types of things have you found helpful in dealing with these challenges and managing your condition? c) What are your thoughts about your future, in relation to epilepsy? 5) In closing, I have one last question that allows you to use your imagination: a) If you could tell health care providers and policy makers about what it’s like to live with epilepsy, and what you think they should know, what are some things you would say? 41 participated, who together represented 31 persons with epilepsy, either themselves or a family member. While gender and race reflected the population quite well, there were no individuals over the age of 65 years. All categories of education, employment, insurance, and marital status were represented. A manuscript detailing the results of the focus groups has been published (Sample PL. Ferguson PF. Wagner JL. Pickelsimer EE. Selassie AW. Experiences of persons with epilepsy and their families as they look for medical and community care: A focus group study from South Carolina. Epilepsy & Behavior. 9:649-662, 2006.). The groups identified numerous areas where intervention is warranted among people with epilepsy. Clinical issues will be addressed later in the report. Other areas they identified were social isolation, difficulty finding information, difficulty finding and keeping employment, lack of transportation, the perception and/or reality of epilepsy not being considered a disability despite the limitations it can carry, and the stigma that epilepsy continues to carry. They painted a picture of individuals living on the edge of life, apart from the mainstream, stymied in following their dreams, considered by much of the population to be disabled yet unable to obtain disability assistance. These are areas in which intervention is not held back by the need for new clinical developments. Public education and government action could greatly impact on these outcomes. V. Severity and subtypes of epilepsy Aim—Describe the specific categories, subtypes, and severity of epilepsy and seizure disorder A. HD/ED data—Determining severity of seizures is a difficult task under the best of circumstances. There is no universally agreed upon scale to measure seizure severity. While there are some seizure characteristics that can be used, such as frequency of seizures or type of seizures, it can be argued that severity should be decided by the individual him- or herself. In addition, severity could include medication side effects, as was noted by the focus groups. While this information was not collected from the chart abstractions, information was collected on the number of people with epilepsy on polytherapy, which increases the likelihood of side effects occurring (see earlier section on target populations for intervention). We included several variables in the data abstraction to attempt to get a picture of severity. Unfortunately, there was limited 42 information in the patient charts concerning seizure lengths, types of seizures, etc. The following is the information that was obtained from the HD/ED charts. The subset analyzed consists of those cases with a diagnosis of epilepsy after clinician review. Since information was scarce in many charts, individuals whose charts were abstracted in 2001 and 2002 had information retained in both years. If the hospital visit involved a seizure (ie, person went to ED because of a seizure, was admitted because of a seizure, the person had a seizure while at the hospital, etc.) then the abstractor had to attempt to record the number of seizure episodes, the time of onset of the seizure(s), and the length of the seizures(s). There were 1701 cases of epilepsy with a seizure-related hospital visit, which included 48 duplicates. • Of the 1,701, 1,281 (75%) had information on number of seizure episodes. Of those with information, 64.4% had one seizure, 14.1% had two seizures, and 21.5% had more than two seizures. • 714 cases (42.0%) had information on time of onset of the seizure(s). Of those with information, 33.8% had a morning seizure event, 28.0% had an afternoon seizure event, 15.4% had an evening seizure event, and 22.8% had a seizure event during the night. • 510 cases (30.0%) had information on the length of the seizure event. Of those with information, 9.0% had a seizure event lasting less than 30 seconds, 23.5% had an event lasting up to 2 minutes, 31.2% had an event lasting up to 5 minutes, and 36.3% had a seizure event lasting greater than 5 minutes. If there was information in the chart concerning a history of seizures, the abstractor was to collect information on the number of types of seizures and the frequency of seizures. There were 2,751 cases of epilepsy with a noted history of seizures, which included 78 duplicates. • Of the 2,751, 300 (10.9%) had information on number of types of seizures. Of those with information, 68.0% had one type of seizure, 17.3% had two types of seizures, and 14.7% had more than two types of seizures. • Only 172 cases (6.3%) had information on the frequency of their seizure events. Of those with information, 18.6% had seizures less than once a year, 30.8% had seizures more than once a year, 11.0% had seizures more than once a month, 43 21.5% had seizures more than once a week, and 18.0% had seizures more than once a day. There were 959 charts (including 44 patients abstracted both in 2001 and 2002) with an epilepsy diagnosis originally in our sample. Four of these charts had two epilepsy diagnoses recorded. Status epilepticus codes were excluded since they do not necessarily indicate epilepsy. Figure 7 shows the frequency of each diagnosis. The following are the definitions of the ICD-9-CM epilepsy codes: 345.0 = generalized nonconvulsive epilepsy, 345.1 = generalized convulsive epilepsy, 345.2 = petit mal status (epileptic absence status), 345.3 = grand mal status (status epilepticus, not otherwise specified), 345.4 = partial epilepsy, with impairment of consciousness, 345.5 = partial epilepsy, without mention of impairment of consciousness, 345.6 = infantile spasms, 345.7 = epilepsia partialis continua (Kojevnikov’s epilepsy), 345.8 = other forms of epilepsy, 345.9 = epilepsy, unspecified. 70.0% 65.4% 60.0% 50.0% 40.0% 30.0% 20.0% 10.0% 14.0% 3.6% 2.7% 7.4% 1.0% 2.3% 4.0% 345.6 345.7 345.8 0.0% 345.0 345.1 345.4 345.5 345.9 Figure 7. Original epilepsy diagnoses recorded on the HD/ED charts (n=959). After clinician review of the abstracted information and the previous diagnoses of seizures and/or epilepsy, there were 3,056 patients with epilepsy diagnoses, with 88 of the records of these patients abstracted in both 2001 and 2002. Fourteen percent had more than one epilepsy diagnosis listed. Figure 8 shows the frequencies of epilepsy diagnoses after clinical review of the abstracted information. These cases include status codes, as an indication of severity. Furthermore, post-chart review evaluation indicated 44 that 13.1% of those with epilepsy diagnoses were being seen for a status seizure (there was only one case with a 345.2 diagnosis, representing less than 0.1%). The large percent of cases with 345.9 is an indication of the scarcity of specific information on the charts with regard to seizure characteristics. Perhaps, the assignment of this code may also suggest the lack of continuity of care by a neurologist/epileptologist. It should be noted that large proportions of epilepsy patients who receive their cares through the ED are uninsured or underinsured to get the appropriate evaluation. 90.0% 83.7% 80.0% 70.0% 60.0% 50.0% 40.0% 30.0% 20.0% 10.0% 0.0% 7.9% 0.5% 13.1% 4.5% 2.9% 0.2% 0.9% 0.6% 345.0 345.1 345.4 345.5 345.6 345.7 345.8 345.9 0.0% 345.2 345.3 Figure 8. Epilepsy diagnoses after clinician review of the HD/ED charts (n=3,056). B. Physician Office Visit (POV) data Data from the 302 patients from POV reflect visits in either 2001 or 2002 and not seen in an ED or HD during the period of surveillance. Further, these are patients whose insurance provider is Medicare, Medicaid, or State Health Plan. After clinician review, 276 of the cases were determined to be epilepsy. Salient findings of this evaluation indicate the following key points. • Of the 276 epilepsy cases, 182 (65.9%) had information concerning number of seizures experienced during their most recent episode of seizure(s). Of those with information, 75.8% had had one seizure, 15.4% had had two seizures, and 8.8% had had more than two seizures. 45 • Only 21 (7.6%) had information recorded concerning the time of the most recent seizure event. Of those with information, 28.6% had a morning seizure, 14.3% had an afternoon seizure, 4.8% had an evening seizure, and 52.4% had had a seizure during the night. • 102 cases (37.0%) had information about the length of the most recent seizure event. Of those with information, 30.4% had a seizure lasting 30 seconds or less, 36.3% had a seizure lasting 2 minutes or less, 20.6% had a seizure lasting 5 minutes or less, and 12.7% had a seizure lasting more than 5 minutes. • 229 cases (83.0%) had information about how many types of seizures the patient experiences. Of those with information, 83.0% had one type of seizure, 10.9% had two types of seizures, and 6.1% had more than two types of seizures. • In 275 of the charts, there was no indication that this was a visit for a first seizure. In these cases, the abstractor was asked to collect information on the frequency of the seizures in the past year. 181 (65.8%) had information on seizure frequency. Of those with information, 24.3% had seizures less than once a year, 29.3% had seizure more than once a year, 26.0% had seizures more than once a month, 12.7% had seizures more than once a week, and 7.7% had seizures more than once a day. Of the entire sample of 302 physician office visit cases, there were 221 cases with an original diagnosis of epilepsy. Only one diagnosis for each case was sent to us from the Office of Research and Statistics (ORS). Figure 9 shows the frequency of the epilepsy diagnoses. After clinician review, there were 276 cases of epilepsy (see Figure 10). Ten percent were given two epilepsy diagnoses. There were no cases of status, as would be expected in an office setting. There are far fewer cases of 345.9 (epilepsy, unspecified) in the office charts, both among the original diagnoses and after review, due to the increased information available on seizure specifics in these charts compared to the HD and ED charts. The most common diagnoses originally on these charts were 345.0, 345.4, and 345.1. After review, the three most common diagnoses were 345.4, 345.1, and 345.9. We questioned if perhaps the original 345.0 diagnoses were actually recorded as 345, without the fourth digit, but ORS confirmed that they were actually listed as 345.0. 46 Taken together, information gleaned from HD/ED is less informative than information gleaned from POV with regard to determination of seizure type and severity. The limited information on HD/ED data is yet additional evidence about the type and continuity of care provided to epilepsy patients. Furthermore, the distinct difference is the large number of patients with epilepsy attending physician offices have insurance while the large proportion of patients (14.3%) evaluated in HD/ED have no insurance. 45.0% 40.3% 40.0% 35.0% 30.3% 30.0% 25.0% 20.4% 20.0% 15.0% 10.0% 5.9% 5.0% 0.0% 345.0 345.1 345.4 345.5 3.2% 0.0% 0.0% 0.0% 345.6 345.7 345.8 345.9 Figure 9. Original epilepsy diagnoses recorded on the POV charts (n=221). 70.0% 59.1% 60.0% 50.0% 40.0% 30.0% 25.0% 19.2% 20.0% 10.0% 6.2% 0.7% 0.0% 345.0 345.1 345.4 345.5 0.0% 0.0% 0.0% 345.6 345.7 345.8 345.9 Figure 10. Epilepsy diagnoses after clinician review of POV charts (n=276). 47 C. Behavioral Risk Factor Surveillance System (BRFSS) data The BRFSS collects information on the number of mentally unhealthy, physically unhealthy, overall unhealthy, and activity-limited days individuals report for the preceding 30 days. In addition, they ask about disability with the question, ‘Are you limited in any way in any activities because of physical, mental, or emotional problems?’. After receipt of the first two years of BRFSS data (2003 and 2004), we analyzed the epilepsy responses in conjunction with these five variables, which may be considered a reflection of the severity of epilepsy when compared to the general population results. The results were published in an MMWR article entitled, Prevalence of epilepsy and health-related quality of life and disability among adults with epilepsy - South Carolina, 2003 and 2004. (MMWR, October 28, 2005 / 54(42);1080-1082; Centers for Disease Control and Prevention). We found that individuals with active epilepsy (i.e., taking medication for epilepsy and/or reported seizure activity in the previous three months) had more than twice as many mentally unhealthy days, physically unhealthy days, overall unhealthy days, and activity-limited days than the general population. 17.9% (95% CI 17.1-18.8) reported having disability, compared to 63.5% (95% CI 52.8-73.1) of individuals with active epilepsy. Even individuals with inactive epilepsy (not currently taking antiepileptic medications, and no seizure activity in the preceding three months) had a significantly higher percentage reporting disability (28.2%, 95% CI 19.6-38.7) than the general population. VI. Venues and Levels of Care Aim—Document the venues and levels of care provided to persons with epilepsy A. Behavioral Risk Factor Surveillance System data—In the final two years of the epilepsy module for SC BRFSS (2004 and 2005) we asked a fifth question, ‘In the past year have you seen a neurologist or epilepsy specialist for you epilepsy or seizure disorder?’. Of individuals who responded that they had ever been told by a doctor that they had epilepsy, 37.8% (95% CI: 29.4-46.3) responded affirmatively. Of those with active epilepsy (i.e., currently taking medicine for epilepsy and/or had seizure activity within the previous three months), 65.9% (95% CI: 56.4-75.5) reported they had seen a 48 neurologist within the past year. Unfortunately, numbers are too small to report by specific characteristics. 82.6% (95% CI 73.6-91.6) of individuals with active epilepsy reported they had a personal doctor or health care provider, and 72.9% (95% CI 61.9-84.0) had had a routine check-up in the past twelve months. Far fewer, however, reported seeing a dentist within the past twelve months—only 53.5% (95% CI 41.4-65.5) of individuals with active epilepsy. Since some AEDs can affect oral health, this can be considered an important part of routine health care for individuals with epilepsy. A large percentage of individuals with active epilepsy reported there was a time in the past 12 months when they needed to see a doctor but could not because of cost 38.2% (95% CI 29.0-47.3). In the general population, only 15.6% (95% CI 15.0-16.3) reported not being able to see a doctor because of the cost. 77.5% (95% CI 68.3-86.7) of individuals with active epilepsy reported having health care coverage. Although the confidence intervals do not indicate a statistically significant difference, 86.5% of individuals with active epilepsy who did not have a seizure within the past three months had health care coverage, whereas only 68.5% of individuals with a recent seizure had health care coverage. B. Physician office visit data—Information gleaned from the SC ORS, for 2001 and 2002 Medicaid and State Health Plan participants suggests gaps in the receipts of care among those individuals with seizure unspecified diagnosis. Comparison between persons coded with 345.x and 780.3 indicate differences in the type of care received. When analyzing only individuals with a 345.x or 780.3 diagnoses, and who had not gone to an ED or been an inpatient during that time, 75% of those with a 345.x diagnosis had seen a neurologist, versus 35% of those with a 780.3 diagnosis. This suggests that diagnosis of 345.x is more likely to be assigned among patients evaluated by neurologists compared to other physicians. This may have numerous implications on the type of care received by these patients. First, without specific diagnosis, appropriate medical care, particularly medication, is less likely to be effective. Second, this may also suggest lack of neurologists/epileptologists in the state, limiting access to specialty care particularly for those without good insurance. Third, although blacks account for 30% of the state 49 residents, they represent 38% of the patients who received care through HD/ED in 2001 and 2002, suggesting limited access to specialty care in private practices. Further, among blacks16.3% and 25.2% are uninsured and Medicaid insured respectively while among whites 12.5% and 11.6% are respectively uninsured and Medicaid insured. This might also have implications in access to specialty care since the case mix of uninsured and Medicaid is very low in POV settings. C. Focus group data—Barriers to care and accessing epilepsy-related services, and what life with epilepsy is like were discussed among the focus groups held throughout South Carolina. Participants reported problems in obtaining an epilepsy diagnosis, with delays in diagnosis consisting of years. Other findings include the difficulty of life with epilepsy and the permanency of these difficulties, life-long searches for help with epilepsy, and how those searches are seldom easy or successful. Also highlighted are several services that have provided helpful support, and recommendations for potential improvements in public education, professional training, and helpful interventions. Participants mentioned difficulty in obtaining clinical information from physicians, and turning to other sources for information, such as the internet. In addition, there was a need for information on assistance with low-cost health care providers, and financial assistance with medication. Additional details on barriers to care are described in detail in the published manuscript (Sample PL. Ferguson PF. Wagner JL. Pickelsimer EE. Selassie AW. Experiences of persons with epilepsy and their families as they look for medical and community care: A focus group study from South Carolina. Epilepsy & Behavior. 9:649-662, 2006.). VII. TBI Among Persons with Seizure Disorder Aim—Determine the burden of TBI among persons with seizure disorders A. Pre-existing seizure disorders and TBI—This evaluation utilized surveillance data collected from statewide HD and ED from1996 through 2005. Pre-existing epilepsy and/or seizure disorders (ESD) were identified using a comorbid diagnosis of ICD-9-CM codes 345.xx and 780.39 from either primary or secondary diagnosis fields among persons with TBI. The underlying assumption for this analysis is that these conditions, 50 by virtue of being chronic in nature, are temporally antecedent to the TBI. Further, the coding conventions of ICD-9-CM clearly indicate neither 345.x nor 780.39 should be used for convulsions precipitated by head trauma either immediately or within the first week of head injury. The corollary of that convention assumes that the higher rate of TBI among persons with ESD, compared to those without, may suggest the excess burden of TBI is attributable to pre-existing ESD. In this evaluation, we systematically analyzed surveillance data on a total of 128,882 unduplicated persons with a diagnosis of TBI. Of these, 4,043 (2.7%) had a comorbid diagnosis of 345.xx (epilepsy) and/or 780.39 (seizures not otherwise specified). Among the 4,043 individuals with a comorbid diagnosis of ESD, a subset of records identified through 2001 (N=2,170), were stratified by diagnosis and a random sample of 100% of the persons with 345.xx and 50% of the 780.3x diagnoses were selected resulting in 1,145 cases. Of these, the SC Department of Health and Environmental Control abstractors were able to find and abstract the records of 872 (76.2%) cases. A total of 528 (60.6%) satisfied the case definition of pre-existing ESD, which is defined as evidence of epilepsy or seizures prior to the onset of TBI validated by: 1) eyewitness history indicating that the seizure triggered the fall that lead to the TBI, 2) a seizure related visit at least one month prior to the TBI, or 3) use of antiepileptic drugs prior to the TBI. There was no significant difference between persons whose records were found and those whose records were missing regarding demographic and clinical characteristics (p>0.10). Thus, the data analyzed are at least 61% true ESD patients. Data analysis included descriptive and analytical comparisons of demographic and clinical characteristics between those with and without pre-existing ESD. Bivariate and multivariable logistic regression analyses indicated that all variables retained in the final model showed association with pre-existing ESD (Table 19). Comparison of crude proportions and odds ratios showed significant differences between persons with ESD and those without regarding age, race, severity, payer type, type of injury, other comorbidities, and repetitiveness of the trauma. Persons with ESD tend to be older, black, with more than 2 TBI during the time interval, more likely to be injured due to fall or adverse effects of drugs, and to be insured by Medicaid and/or Medicare then those without pre-existing TBI. Table 20 shows multivariable adjusted odds ratios. All the 51 variables noted earlier retained statistical significance with the exception of the 65 and older age group. This effect is due to collinearity with Medicare payer status, which retained the comparable degree of association with presence of ESD. The observed associations suggest the implications of ESD in contributing to the cycle of seizure and the repetitiveness of these injuries. The risk of repetitive TBI is 3-fold among persons with ESD compared to those without. To further elucidate the burden of other comorbid conditions associated with persons with ESD and TBI, we used the comorbidity index developed by Elixhasuer et al. (1) and evaluated the odds of having the conditions listed in Table 21 as a function of preexisting ESD. Accordingly, age-sex-race adjusted odds ratios indicate significant association (P<0.05) for six conditions (AIDS, alcohol abuse, paralysis, neurological disorders, fluid & electrolyte disorders, and coagulopathies). In conclusion, our findings suggest that persons with ESD carry a disproportionate burden of TBI. The standardized morbidity ratio of TBI is at least 1.6 to 2.0 times greater among persons with ESD than what is noted in the general population of the state even when we assumed the lifetime prevalence rate of epilepsy (2.0%). When the prevalence is defined as 1.0%, which is the case based on our prevalence report, the morbidity ratio is 3 to 4 times greater that the rate in the general population of the state. Furthermore, the increased risk of repetitiveness of TBI among persons with ESD is a profound concern since more head trauma has the potential to aggravate the intensity and severity of seizure. 52 Table 19. Characteristics of TBI patients by ESD, SC, 1996-2006 (N=128,882). Characteristics Age 65+ 45-64 35-44 20-34 10-19 0 -9 Mean (SD) Race/Sex Black Female Black Male White Female White Male TBI Severity Severe Moderate Mild Cause of Injury Drug adverse/Poison All Others Fall Transport Struck by/against Violence Payer Medicare Medicaid Other government Uninsured Commercial Care Type Inpatients Outpatients TBI Injury Type Intracranial Skull fracture Concussions Unspecified injury Comorbidity 3+ 2 1 0 Concomitant Injury No Yes Repetitive TBI 2+ 1 0 Pre-existing ESD Yes % No % (N=4,043) (N=124,839) Crude Odds Ratio 95% Confidence Interval 20.2 23.4 16.2 16.8 11.9 11.5 41.2 (24.5) 12.9 13.0 12.1 23.5 19.6 19.0 31.5 (24.0) 2.59 2.97 2.21 1.18 1.00 1.00 2.31-2.91 2.66-3.33 1.96-2.50 1.15-1.33 0.88-1.14 Reference 10.9 26.5 25.8 36.8 12.0 20.5 28.5 39.1 0.97 1.37 0.96 1.00 0.87-1.08 1.27-1.49 0.89-1.04 Reference 39.6 8.3 52.1 11.9 5.8 82.3 5.26 2.26 1.00 4.91-5.62 2.01-2.54 Reference 1.3 28.8 36.8 20.0 5.2 8.0 0.6 14.7 28.4 31.9 13.7 10.8 2.90 2.64 1.75 0.85 0.52 1.00 2.14-3.92 2.33-3.00 1.55-1.98 0.74-0.96 0.43-0.61 Reference 28.5 18.4 2.5 23.2 27.5 13.1 14.8 3.1 30.9 38.1 3.02 1.72 1.12 1.04 1.00 2.77-3.28 1.57-1.89 0.91-1.38 0.95-1.14 Reference 63.6 36.4 23.5 76.5 5.69 1.00 5.33-6.08 Reference 44.4 4.5 23.2 28.0 13.7 4.8 38.2 43.3 5.01 1.45 0.94 1.00 4.64-5.41 1.24-1.70 0.86-1.03 Reference 10.9 14.4 23.6 51.2 2.2 4.3 11.1 82.4 7.99 5.39 3.42 1.00 7.16-8.91 4.90-5.94 3.16-3.70 Reference 57.4 42.6 44.3 55.7 1.70 1.00 1.59-1.81 Reference 6.9 20.3 72.8 2.0 11.9 86.1 4.00 2.02 1.00 3.51-4.55 1.86-2.18 Reference 53 Table 20. Adjusted odds ratios comparing TBI patients with ESD and without (N=128,882). Characteristic Age 65+ 45-64 35-44 20-34 10-19 0 -9 Race/Sex Black Female Black Male White Female White Male TBI Severity Severe Moderate Mild Cause of Injury Drug adverse/Poison All Others Fall Transport Struck by/against Violence Payer Medicare Medicaid Other government Uninsured Commercial Care Type Inpatient Outpatient Comorbidity 3+ 2 1 0 Concomitant Injury No Yes Repetitive TBI 2+ 1 0 Adjusted Odds Ratio 95% Confidence Interval 0.51 1.81 2.15 1.42 1.32 1.00 0.43-0.62 1.58-2.08 1.87-2.47 1.24-1.63 1.15-1.52 Reference 1.20 1.48 0.98 1.00 1.07-1.35 1.36-1.61 0.90-1.06 Reference 1.58 1.09 1.00 1.45-1.73 0.96-1.24 Reference 2.53 2.68 2.02 0.96 0.99 1.00 1.84-3.48 2.35-3.06 1.77-2.31 0.83-1.10 0.82-1.19 Reference 2.50 1.93 1.19 1.17 1.00 2.19-2.85 1.74-2.14 0.96-1.47 1.06-1.28 Reference 3.40 1.00 3.11-3.72 Reference 2.19 2.14 1.83 1.00 1.91-2.51 1.90-2.40 1.67-2.00 Reference 1.65 1.00 1.54-1.77 Reference 2.33 1.33 1.00 2.03-2.68 1.22-1.44 Reference 54 Table 21. Odds of comorbid conditions among persons with TBI by pre-existing ESD Congestive heart failure Pre-existing ESD Yes No Freq. (%) Freq. (%) 117 (2.89) 1235 (0.99) Cardiac arrhythmias 165 (4.08) Comorbid Condition Unadjusted OR (95%CI) Adjusted† OR (95%CI) 2.98 (2.46-3.62) 0.78 (0.63-0.96) 1932 (1.55) 2.71 (2.30-3.18) 0.83 (0.70-0.995) 49 (1.21) 517 (0.41) 2.95 (2.20-3.96) 0.79 (0.58-1.08) 3 (0.07) 63 (0.05) 1.46 (0.46-4.69) 0.37 (0.11-1.21) 32 (0.79) 329 (0.26) 3.02 (2.10-4.35) 0.75 (0.51-1.10) 662 (16.37) 9050 (7.25) 2.51(2.30-2.73) 0.77 (0.69-0.86) Paralysis 126 (3.12) 518 (0.41) 7.72 (6.34-9.40) 1.77 (1.43-2.19) Neurological disorders 142 (3.51) 921 (0.74) 4.90 (4.09-5.86) 1.39 (1.14-1.69) COPD/Asthma 243 (6.01) 2748 (2.20) 2.84 (2.48-3.25) 1.02 (0.88-1.19) Diabetes 245 (6.06) 3817 (3.06) 2.05 (1.79-2.34) 0.64 (0.55-0.74) Renal failure 74 (1.83) 454 (0.36) 5.11 (3.99-6.55) 0.99 (0.75-1.30) Liver disease 50 (1.24) 221 (0.18) 7.07 (5.19-9.62) 1.13 (0.81-1.60) 9 (0.22) 155 (0.12) 1.80 (0.92-3.52) 0.48 (0.24-0.96) 21 (0.52) 45 (0.04) 14.48 (8.62-24.3) 2.18 (1.22-3.92) 8 (0.20) 72 (0.06) 3.44 (1.65-7.14) 0.79 (0.37-1.69) 16 (0.40) 143 (0.11) 3.47(2.07-5.81) 0.81 (0.48-1.39) 76 (1.88) 1000 (0.80) 2.37 (1.88-3.00) 0.70 (0.55-0.90) 22 (0.54) 222 (0.18) 3.07 (1.98-4.77) 0.93 (0.58-1.47) 214 (5.29) 1176 (0.94) 5.88 (5.06-6.82) 1.36 (1.15-1.61) 76 (1.88) 487 (0.39) 4.90 (3.84-6.25) 1.13 (0.87-1.46) 499 (12.34) 3210 (2.57) 5.34 (4.83-5.90) 1.33 (1.18-1.49) 23 (0.57) 227 (0.18) 3.14 (2.04-4.83) 0.91 (0.58-1.43) Alcohol abuse 253 (6.26) 897 (0.72) 9.23 (8.00-10.65) 2.20 (1.86-2.59) Drug abuse 145 (3.59) 1637 (1.31) 2.80 (2.36-3.33) 0.94 (0.77-1.13) Psychoses 92 (2.28) 844 (0.68) 3.42 (2.75-4.25) 0.95 (0.75-1.20) Depression 105 (2.60) 1144 (0.92) 2.88 (2.36-3.53) 0.88 (0.71-1.09) Stroke 97(2.40) 590 (0.47) 5.18 (4.17-6.43) 1.13 (0.89-1.42) Hypotension 69 (1.71) 951 (0.76) 2.27 (1.77-2.90) 0.67 (0.52-0.88) Valvular disease Pulmonary circulation disorders Peripheral vascular disorders Hypertension Peptic ulcer disease AIDS Lymphoma Metastatic cancer Solid tumor without metastasis Rheumatoid arthritis Coagulopathy Weight loss Fluid and electrolyte disorders Blood loss anemia † Adjusted for age, sex, and race 55 VIII. Data quality Aim—Evaluate quality of the data and accuracy of the estimates A. Population and sample information—the sources of the data and the methods of sampling play a significant role in the determination of data quality. These concepts are tied to coverage of the data sources and the representativeness of the sample, each of which determine the external validity of the findings. Coverage refers to the extent to which all persons in the geopolitical jurisdiction are included in the numerator and such numerator is reflective of the pool of individuals in the referent population. Sample representativeness indicates the extent to which every single unit in the frame has an equal probability of inclusion in the sample such that the aggregate nature of the sample is truly reflective of the population in the frame. In our surveillance of epilepsy, we made every effort to insure that the information collected is inclusive and selection is unbiased. To further elucidate these points, we present key information about the data sources. 1. Hospital Discharge and Emergency Department—the South Carolina Office of Research and Statistics (SC ORS) houses an administrative billing dataset with information from all non-Federal acute care hospital discharges and ED visits in SC. Among the various variables included billing abstracted data are personal identifiers and up to ten diagnosis codes. As a major partner to the project and a stakeholder in statewide health and demography data collection, ORS identified a sample of individuals from 2001 and 2002 with epilepsy (345.x), convulsions (780.3), syncope and collapse (780.2), and acute delirium (293.0). The latter two codes were included on the chance that some seizures might be miscoded as syncope or delirium. The initial plan was to choose 35% of 345.x codes, 5% of 780.3 codes, 1% of 780.2 codes, and 5% of 293.0 codes for 2001 for a total of approximately 3,000 charts. From the 2002 population we wanted to have a total abstraction of approximately 1,500 charts, and based on our experience with the 2001 data, inflated that number upward to adjust for expected rate of 15% unlocated charts. See Appendix G for the HD/ED sampling plan. Accordingly, ORS identified 3,000 and 1,742 charts from 2001 and 2002 encounters respectively for abstraction. When drawing the 2002 sample, we decided to allow cases from 2001 to be included in the population. Of the final 2002 56 chart abstractions, 102 (7.0%) of the cases had also been abstracted in the 2001 sample. ORS sent information from these records to SC Department of Health and Environmental Control (SC DHEC). DHEC abstractors went to the hospitals and abstracted information from those records based on the abstraction tool we had developed. DHEC then sent the abstracted information to ORS, where all identifying information was removed. The de-identified data was then sent to MUSC. To assist MUSC in determining history of seizures, ORS also sent a database consisting of the unique IDs they had assigned each case, along with the diagnosis codes from any earlier hospital or ED visits with a seizure-related code. MUSC received 3,998 abstractions from ORS (2001 N=2538, 2002 N=1460). This included 5 duplicates (same chart abstracted twice) and 5 blank abstractions, plus 2 abstractions done that appeared not to have been part of original sample. Thus, there was a total of 3988 unique abstractions completed (2001 N=2535, 2002 N=1453). We further found that 3 of the charts appeared to have been erroneously coded with 780.39 or 345.x, so we removed them from analysis, reducing the total to 3985 charts (2001 N=2532, 2002 N=1453). After matching the charts back to the original population, there were two charts abstracted that were never in original sample. These were excluded from analysis yielding a total of 3,983 charts with seizure-related (ie, epilepsy, convulsion/seizure, syncope & collapse, and/or acute delirium) diagnoses (2001 N=2530, 2002 N=1453). In order to evaluate the representativeness of our sample to the population, we requested from ORS a dataset with the entire population of unduplicated cases with a seizure-related diagnosis code in 2001 and 2002, with age group, sex, race, payer, and diagnosis code grouping (if more than one seizure-related code, grouping assigned with the following hierarchy: 345.x > 780.3 > 780.2 > 293.0), with abstracted cases marked. Below is a summary of how well the data for those marked cases matched with our abstracted data. See Appendix C for a comparison of the proportions of sex, race, age group, and payer for 780.3 and 345.x diagnoses and year in cases marked as abstracted versus the population. The following is a brief summary of the finding regarding comparability of the records abstracted to the information in the frame. 57 1. Age and gender have accuracy of 99.8% 2. Race has a match rate of 99.4%. 3. Payer is 98.1% accurate. 4. Diagnosis from chart review is accurate in 95.2 % of the cases with the diagnosis listed in UB-92 when arranged hierarchically as 345.x > 780.3 > 780.2 > 293.0. 2. Physician Office Visits—captures individuals not seen in HD/ED setting. ORS is the repository of the data from physician offices when primary payer is Medicare, Medicaid, or State Health Plan—an insurance plan for state employees, their dependents, and retirees. The data format for reporting is based on CMS-1500, a billing form that includes all variables but inpatient care and related procedures. The POV report does not include “out-of pocket” payers (self-pay) and these are expected to be <5% of the patient pool in physician offices. The proportion of self-pay in POV setting is even expected to be much lower when it comes to persons with chronic conditions like seizure disorders. We sampled from 2001 and 2002, except for Medicare, for which we only had 2001 data available at the time of the sampling in 2003. Data abstraction from physician offices was entirely voluntary on their part. Due to this, the practice from which the sample of charts abstracted from participating practices was neither random nor representative of all such practices in the state. The practices that volunteered for chart abstraction were identified and contacted by the project epileptologist and investigator (BW), asking them whether they would be willing to have their records abstracted by SC DHEC personnel. He strove to include a variety of specialties. Overall 90 practitioners agreed to participate, representing 32 family practice, 8 internal medicine, 29 neurology, 8 obstetrics/gynecology, and 13 pediatric specialties. However, some practitioners had no cases represented in the sample, and some offices could not find the requested records. The final abstracted sample included data from 46 practices, representing 8 family practice, 6 internal medicine, 26 neurology, 2 obstetrics/gynecology, and 4 pediatric specialties, however most of the population was from neurology offices. We asked the abstractors to note the type of specialty visit for each case. They recorded the following distribution: 9 from primary care, 199 from neurology, 83 from epileptology, and 11 unknown. 58 Cases were restricted to those individuals who had no encounter in either HD or ED in SC during 2001 or 2002. The rationale for this was to have two exclusive groups—those seen in HD/ED setting, and those seen in a POV setting. ORS estimates that during the surveillance period, Medicare, Medicaid, and SHP covered approximately 39% of the population of SC. We initially attempted to abstract 400 charts. Due to difficulty finding the charts, a series of four samples were eventually drawn to reach a revised target of 300. Out of a total drawn sample of 581 charts, 302 (52.0%) were found and abstracted. We initially planned on pulling a ratio of 8:2, epilepsy to seizure codes. However, as the sampling progressed, more seizure codes were included as necessary to reach our target goal. Due to limited information at ORS, Medicare charts were the most difficult to find, and could only be sampled from 2001. (Please see Appendix H for the physician office visit sampling plan.) The HD/ED sample showed that all cases pulled with a 293.0 or 780.2 code also had either a 345.x or 780.3 codes. The sampling strategy for the office charts was restricted to only 345.x and 780.3 diagnoses. Rather than allowing cases with more than one seizure-related diagnosis to be pulled under either diagnosis, this sample was pulled using only one diagnosis for each case, and it is unknown whether the cases had a second seizure-related diagnosis. However, we feel second diagnoses were likely rare, since in examining the HD/ED data, and looking only at 780.3 and 345.x codes on the charts, there were only 16 (0.8%) cases in 2001 that had more than one of those diagnoses, and there were only 12 (0.6%) cases in 2002 that had more than one of those diagnoses. Assuming that more than one epilepsy or seizure code per chart is equally as rare in the POV charts, there should be minimal effect from limiting to one seizure-related code. As mentioned above, to compare our sample with the population, we requested from ORS a dataset with the entire population of unduplicated cases with a seizure-related diagnosis code in 2001 and 2002 (only 2001 available for Medicare), with age group, sex, race, payer, and diagnosis, with abstracted cases marked, for each of the three payer groups. Below is a summary of how well information acquired through chart review correctly matched with the POV data from ORS. Due to initial matching error of the unique identifiers, this analysis is restricted to 246 (81.5%) of the records. Please see Appendix D for a comparison of 59 the proportions of the characteristics seen below in cases marked as abstracted versus the population. The following is a brief summary of the finding regarding comparability of the records abstracted to the information in the frame. 1. Age and gender have accuracy of 98.8% 2. Race has a match rate of 99.6%. 3. Payer is 75.6% accurate. 4. Diagnosis from chart review is accurate in 92.3 % of the cases with the diagnosis listed in CMS-1500 when arranged hierarchically as 345.x > 780.3. B. Predictive Value Positive (PVP) and Sensitivity of seizure and epilepsy codes—PVP reflects the ability of the ICD-9-CM codes assigned in administrative datasets to correctly identify epilepsy cases. Conversely, sensitivity implies the extent to which true cases and true non-cases of epilepsy from medical records are correctly differentiated by the codes in the administrative datasets. While sampling is often driven by the codes assigned in the administrative datasets, the prospect of correctly conducting sensitivity is likely to be biased since a priori case selection from medical charts is difficult at best and impossible at worst. Thus, emphasis of this evaluation is on PVP while Sensitivity will be interpreted with caution. 1. HD and ED Data—to determine PVP and Sensitivity, codes were arranged in hierarchical manner: epilepsy (345, 345.0, 345.1, 345.4, 345.5, 345.6, 345.7, 345.8, 345.9) > status (345.2, 345.3) > seizure (780.3, 780.31, or 780.39.) > syncope (780.2) > delirium (293.0). In our sample it happened that any original 780.2 and 293.0 codes were always accompanied by either a 780.3 or a 345.x. In some cases, some of the charts have more than one seizure-related codes and any one of these codes were used to validate PVP and sensitivity. Table 22 summarizes the diagnosis codes. 60 Table 22. Inpatient/ED diagnoses before and after abstraction review. Diagnosis after Clinician Review Original Diagnosis Epilepsy 2001 2002 Seizure 2001 2002 Status 2001 2002 All • Total Epilepsy Seizure Status Syncope Unknown 545 (86.8%) 296 (89.4%) 24 (3.8%) 15 (4.5%) 3 (0.5%) 3 (0.9%) 0 0 56 (8.9%) 17 (5.1%) 628 (15.8%) 331 (8.3%) 1231 (69.8%) 809 (78.9%) 303 (17.2%) 127 (12.4%) 74 (4.2%) 20 (2.0%) 2 (0.1%) 1 (0.1%) 155 (8.8%) 69 (6.7%) 1765 (44.3%) 1026 (25.8%) 99 (72.3%) 76 (79.2%) 1 (0.7%) 1 (1.0%) 36 (26.3%) 19 (19.8%) 0 0 1 (0.7%) 0 137 (3.4%) 96 (2.4%) 3056 (76.7%) 471 (11.8%) 155 (3.9%) 3 (0.1%) 298 (7.5%) 3983 (100%) Epilepsy—there were 959 cases coded as epilepsy in the HD/ED coding scheme. After clinician review, 841 (87.7%) of these were considered correct as epilepsy codes. 39 (4.1%) of them were considered to be a 780.3 code, 6 (1%) were considered to be status epilepticus (345.2, 345.3), and 73 (7.6%) were considered to not have enough information to assign a diagnosis. • Seizures—there were 2,791 cases with seizure codes. After clinician review, 430 (15.4%) were considered correct as a seizure code. 2,040 (73.1%) were considered to have a history of epilepsy, 94 (3.4%) were considered to be status epilepticus, 3 (<1%) were considered to be syncope (780.2), and 224 (8.0%) did not have enough information to assign a diagnosis. • Status—there were 233 cases with status epilepticus codes (345.2, 345.3). After clinician review, 55 (23.6%) were considered to be correct as status epilepticus. 175 (75.1%) were considered to have a history of epilepsy, 2 (<1%) were considered to be 780.3, and 1 (<1%) was considered to not have enough information to assign a diagnosis. • Febrile Seizure—there were 110 patients coded as febrile seizure, of which 96 (87.3%) were correct. 8 (7.3%) were determined to have epilepsy after clinician review. 2. POV Data—determination of PVP and Sensitivity from POV generally followed the same principle with codes were arranged in hierarchical manner but only for 61 epilepsy and seizure codes: epilepsy (345, 345.0, 345.1, 345.4, 345.5, 345.6, 345.7, 345.8, 345.9) > status (345.2, 345.3) > seizure (780.3, 780.31, or 780.39.). There were no other codes sampled from the POV dataset. As mentioned earlier, while the sampled observations were randomly selected, the offices from which the sample of charts was pulled were a convenience sample. However, there appears to be comparability between the samples drawn and the frame regarding some of the characteristics. After removing the 11 cases in which there was not enough information to determine a diagnosis after review, there were 165 and 126 records sampled for 2001 and 2002 respectively. Seventy-three percent (213) of the records had an original diagnosis code of epilepsy while 27% (78) records had seizure as the original diagnosis. 3. Estimates of PVP and Sensitivity for HD/ED data—using the clinical reviewers’ evaluation as the “correct” classification, we generated the following estimate for the sources. While the PVP of epilepsy codes range between 95% and 96%, the corresponding value for seizure codes remain very low (20% and 14% for 2001 and 2002 respectively). This suggests that 80-86% of the cases coded as seizure are epilepsy codes, while 95% of the cases coded as epilepsy are true epilepsy cases. Unfortunately, 780.3x is the most prolific code in the seizure category and the ratio of 345.x to 780.3x is 1:11. It should be noted that until proven otherwise, 780.3 codes should be presumed as epilepsy cases for surveillance purposes. With the constraint of the sampling scheme in mind, 780.3x codes have high sensitivity due to the large proportion false positive cases they capture. Conversely, the sensitivity of 345.x codes is very low due to the small number of false positive cases they capture (please see the calculation in the prototype tables shown by year). Code assigned after clinician review Epilepsy Seizure Total HD/ED Epilepsy 545 24 569 UB-92 Seizure 1236 304 1540 code Total 1781 328 2109 PVP of 345.x = 545/569 = 96% PVP of 780.3 = 304/1540 = 20% Sensitivity of 345.x = 545/1781 = 31% Sensitivity of 780.3 = 304/328 = 93% Code assigned after clinician review Epilepsy Seizure Total HD/ED Epilepsy 296 15 311 UB-92 Seizure 811 127 938 code Total 1107 142 1249 PVP of 345.x = 296/311 = 95% PVP of 780.3 = 127/938 = 14% Sensitivity of 345.x = 296/1107 = 27% Sensitivity of 780.3 = 127/142 = 89% 2001 2002 62 4. Estimates of PVP and Sensitivity for POV data—as indicated in the aforementioned description, all procedures for estimating PVP and sensitivity remained the same. The estimated PVP for POV is even higher than what was noted in the HD/ED dataset. Estimate range between 94% and 97%, the corresponding value for seizure codes remain much lower at 4% and 6% for 2001 and 2002 respectively. This lower rate for seizure codes is mainly accounted by the very small number of patients coded with 780.3x in POVs. Perhaps the large part of this might be the client pool of patients in physician offices who already have well-established diagnoses. The tables below show the calculated values for the two measures for 2001 and 2002 Code assigned after clinician review Epilepsy Seizure Total POV Epilepsy 108 7 115 code Seizure 47 3 50 Total 155 10 165 PVP of 345.x = 108/115 = 94% PVP of 780.3 = 3/50 = 6% Sensitivity of 345.x = 108/155 = 70% Sensitivity of 780.3 = 3/10 = 30% 2001 2002 POV code Code assigned after clinician review Epilepsy Seizure Total Epilepsy 95 3 98 Seizure 27 1 28 Total 122 4 126 PVP of 345.x = 95/98 = 97% PVP of 780.3 = 1/28 = 4% Sensitivity of 345.x = 95/122 = 78% Sensitivity of 780.3 = 1/4 = 25% As a summary, the PVP of 345.x codes is very high across the two data sources. Persons coded with 345.x are about 95% of the time true epilepsy cases. However, 70-80% the persons coded with780.3x are misclassified as seizure. Sensitivity as a measure of data validity is not reliable due to the difficulty of the sampling scheme. C. Algorithm to identify epilepsy patients from administrative datasets—among the major achievements of our project is the lessons learned and the experience amassed on identifying critical components of the clinical decision making process. The two epileptologists (BW and RT) and the neurology nurse practitioner (GS) have a cumulative total of 65 years of clinical experience in neurology and epileptology. The extensive and elaborate data abstraction tool along with the 3,983 charts that have been reviewed provided very useful information to develop an algorithm among commonly miscoded differential diagnoses of seizure disorders. Hence, we developed four sets of algorithms regarding persons coded as Syncope (780.2), Febrile Seizure (780.31), Seizure NOS (780.39), and Epilepsy (345.xx), each of which are described below. 63 1. For cases coded as 780.2: Syncope--If a visit is coded as 780.2 but previous visits are noted as coded for 780.39 or 345, then the diagnosis is less likely to be 780.2 and it is important to consider 780.39 or 345. If antiepileptic medications (AEDs) are listed or an AED drug level was done, then it is less likely to be 780.2 and 345 should be more strongly considered. If there is presence of a vagal nerve stimulator (VNS), VNS interrogation/parameters adjusted (CPT codes 95970 or 95974) or VNS implanted, the code 345 must be considered. If there is a diagnosis for a genetic, neurologic, developmental or metabolic disorder, it is less likely to be 780.2 and more likely to be 345. If there is a code for epilepsy surgery, then the code is a 345 and 780.2 should not be coded. Figure 11 depicts the decision path for 780.2. 2. For visit coded as 780.31: Febrile seizure disorder—if a person is less than 6 months or greater than 6 years and has been coded as a febrile seizure disorder, then that code is incorrect due to the diagnostic criteria and should be changed to a 780.39 or 345 code. If there is a concurrent illness that could cause fever and/or fever is coded, then 780.31 is the likely code. If a visit is coded as 780.31 but previous visits are noted as coded for 780.39 or 345, then the diagnosis is less likely to be 780.31 and it is important to consider 780.39 or 345. If antiepileptic medications (AEDs) are listed or an AED drug level was done, then it is less likely to be 780.31 and 345 should be more strongly considered. If there is presence of a vagal nerve stimulator (VNS), VNS interrogation/parameters adjusted (CPT codes 95970 or 95974) or VNS implanted, the code 345 must be considered. If there is a diagnosis for a genetic, neurologic, developmental or metabolic disorder, it is less likely to be 780.31 and more likely to be 345. If there is a code for epilepsy surgery, then the code is a 345 and 780.31 should not be coded. Figure 12 is a graphic depiction of the decision rules for 780.31. 64 Figure 11 Decision algorithm for 780.2 code 3. For visits coded as 780.39: Seizure, NOS—if a visit is coded as 780.39 but previous visits are noted as coded for 780.39 or 345, then the diagnosis is less likely to be 780.39 and it is important to consider 345. If antiepileptic medications (AEDs) are listed or an AED drug level was done, then it is less likely to be 780.39 and 345 should be more strongly considered. If there is presence of a vagal nerve stimulator (VNS), VNS interrogation/parameters adjusted (CPT codes 95970 or 95974) or VNS implanted, the code 345 must be considered. If there is a diagnosis for a genetic, neurologic, developmental or metabolic disorder, it is less likely to be 780.39 and more likely to be 345. If there is a code for epilepsy surgery, then the code is a 345 65 and 780.39 should not be coded. Figure 13 is a graphic depiction of the decision rules for 780.39. Figure 12. Decision algorithm for 780.31 code 4. For visits coded as 345.xx: Epilepsy disorders—if a visit is coded as 345 but previous visits are noted as coded for 780.39 or 345, then the diagnosis is less likely to be 780.39 and it is important to consider 345. If antiepileptic medications (AEDs) are listed or an AED drug level was done, then 345 should be more strongly considered. If there is presence of a vagal nerve stimulator (VNS), VNS interrogation/parameters adjusted (codes 95970 or 95974) or VNS implanted, the code 345 must be considered. If there is a diagnosis for a genetic, neurologic, 66 developmental or metabolic disorder, it is more likely to be 345. If there is a code for epilepsy surgery, then the code is a 345. Figure 14 is illustrative chart of the decision procedures to establish or rule out the diagnosis of epilepsy. Figure 13. Decision algorithm for 780.39 code 67 Figure 14. Decision algorithm for 345.xx code D. Data quality measures—we implemented various measures to ensure the quality of the data collected and to improve the process by which medical charts are abstracted. The following are the processes and activities involved to asses the quality of the data collected. 1. Abstraction tool—the tool for collecting standardized data from HD/ED medical charts was developed by the epidemiologists and clinicians. It was piloted by the abstractors, and changes were made accordingly. After the 2001 charts had been abstracted, improvements were made in the abstraction tool based on the data collection experience, including the removal of some variables with low yield, prior 68 to collection of data from the 2002 charts. In addition, as needed, the tool was modified to collect information from physician office charts, and to collect information from charts of individuals with TBI with a diagnosis or self-report of seizures or epilepsy. 2. Training workshops for data abstractors. An initial 2 ½ day training workshop was held for the abstractors. It presented information on epilepsy and data quality, and introduced the data abstraction tool. It included group practice with practice charts and the abstraction tool, and then individual chart abstractions in a group setting, allowing for questions and discussion on the abstraction tool. Feedback collected was then incorporated into the abstraction tool. Subsequent annual workshops were held to review key information, provide feedback, introduce changes in the abstraction tool, and allow for discussion and practice abstraction. 3. Training for abstraction reviewers assigning diagnosis codes. Initially, one clinician—a pediatric nurse practitioner specializing in epilepsy (GS)—reviewed each abstraction to determine, based on the available information, whether the originally assigned code was correct, and if not, to assign the most appropriate code. An epileptologist agreed to act as our ‘gold standard’ for the review. Due to the volume of cases, two additional clinicians were hired to review cases (one adult epilepsy nurse, and one pediatric epilepsy nurse). They were given information on the project and the protocol, as well as 40 practice cases to review. After review, all four clinicians met to discuss the practice cases prior to beginning reviews. After official reviews began, the three reviewers plus the epileptologist were twice given the same set of cases to review individually, and then the group would meet to discuss any discrepancies or other issues. 20 of the 2001 cases were reviewed like this, with 80% agreement. 26 of the 2002 cases were also reviewed. In this group, there was low agreement when looking for agreement to fourth place (345.x). However, if looking at whether seizure/status epilepticus (780.3 or 345.3) versus epilepsy (all other 345), then the proportion of agreement improved. There were a number of cases in which 1 or 2 reviewers noted ‘not enough information’ rather than putting a diagnosis but the rest of the reviewers agreed on the diagnosis. In addition, there were 100 cases 69 reviewed by two of the reviewers, with agreement between these reviewers ranging between 86-89%. 4. Protocol for assigning diagnosis codes. A protocol, based on International League Against Epilepsy guidelines, was developed (see Table 23) for use in assigning ICD-9 epilepsy and/or seizure diagnoses to cases after abstraction review. This was created with input from our CDC technical advisor and consultant epileptologists. As work progressed the protocol was added to or changed as needed. If changes were instituted that might affect previous abstraction reviews, those cases were pulled and reviewed again to maintain consistency in assignment of diagnoses. 5. Additional information on previous epilepsy/seizure diagnoses. To assist the clinicians in assigning diagnosis codes and to collect incidence information, ORS provided a database containing the assigned unique IDs of the cases, and any instances of previous inpatient or emergency department diagnoses of seizure or epilepsy. 6. Re-abstractions of medical charts. Approximately a 5% sample of the 2001 charts was drawn (n=126), and were re-abstracted by one of the abstractors. These reabstractions were then compared to the original abstractions of the same charts (n=122; 4 charts did not match back), using McNemar's test for 2 ×2 tables and Bowker's test of symmetry for tables with more than two response categories. Kappa statistics were calculated on categorical variables to determine consistency in abstraction. Kappa values (simple, weighted, and exact, as appropriate) ranged from negative numbers through 1.0. Weighted kappas could not be done on some ordered values, however, due to inclusion of a missing category. Most 100% agreements were in variables in which no data was found in the chart. Excluding the variables discussed below, of the 42 categorical variables compared, the test of H0:kappa=0 resulted in 30 comparisons (71%) with p-values <0.05, and the remaining 12 (29%) with p-values >= 0.05. In two areas of abstraction, there was the chance to record information in text form 70 Table 23. Protocol for reviewing ICD-9-CM codes of abstracted cases. A. Definition of correct coding entries: 1. Yes – correct code noted 2. No – incorrect code noted 3. No* - correctly identified seizure diagnosis but this is an unknown type of epilepsy disorder due to history of seizures 4. ? – not enough information to know correct diagnosis code to assign If there are two seizure-related codes assigned, and both are correct, then Correct Code = ‘yes’. If there are two seizure-related codes assigned, and both are incorrect, then Correct Code=’no’, and the correct codes are filled into the Primary Diagnosis and Secondary Diagnosis boxes. If there are two seizure-related codes assigned, and one is correct and one is incorrect, then Correct Code=’no’, and the correct codes are filled into the Primary Diagnosis and Secondary Diagnosis boxes, meaning one of the codes (the correct one) will remain the same. B. Reviewers will use an Epi Info program to enter information to ease both entry and retrieval. C. Coding review protocol 1. Clinical definition: A case of epilepsy is a person who has ever experienced episodic, sudden, and usually unprovoked attacks of subjective experiential phenomena, altered awareness, involuntary movements, or convulsions due to a chronic, underlying CNS process (Pedley, 1996). 2. Operational definition: A case of epilepsy is a person who experienced ≥ 2 usually unprovoked seizures. D. Rules agreed to: 1. An alcohol-provoked seizure with a history of seizures (even if also most likely all alcohol provoked) should be coded as epilepsy (345.9). 2. A first seizure with an abnormal EEG is still coded as seizure (780.39) unless there is information noting a defined epilepsy syndrome on EEG. 3. If there is a history of epilepsy, even if remote, and a seizure (provoked or unprovoked) occurs, it should be coded as epilepsy. 4. If the child has a new onset seizure with fever and is less than or equal to 6 years old, it is coded as febrile seizure (780.31). If verified as 780.31, but h/o seizures, include 345.x code only if other seizures not febrile seizures (see #3 above). 5. If there is incomplete information about the seizure, and there are no contradictions to the assigned diagnosis, defer to the MD’s diagnosis. 6. If a case is coded as 780.39, and there is a history of seizures, but not enough information provided to determine type of epilepsy, it should be coded as 345.9 and noted as No* to capture the under diagnosis of epilepsy cases. 7. If a child has an extended febrile seizure (>5 minutes), code for both 780.31 and status (best to use both for all status, for purposes of billing, would probably use status code which explains higher costs – unsure what coding rule is). 8. Except in the case of febrile seizure, a first seizure that is prolonged is coded as status rather than as 780.39. 9. If there is a history of non epileptic seizures, but there is other evidence that there is also a history of seizures (ie, taking multiple AEDs, h/o etiologic event such as CHI, etc.) then must code as epilepsy, and include 780.39 if pseudoseizure involved. 10. What about an individual with h/o seizures, who comes in with a seizure, and is coded with 780.39 as a primary diagnosis, and a 345.x as a secondary diagnosis? Should both codes be used? No, use 345.x. 11. If individual has 345.7, but does not have epilepsia partialis continua, then change to 345.3, and 345.4 or 345.5. Nonstop partial seizure activity required for a diagnosis of 345.7. 12. If individual has status epilepticus diagnosis is given, and rescue medications are used, then status diagnosis is kept regardless of time noted for seizure. If diagnosis of status epilepticus is given, but no seizure time or medication intervention is noted, keep status code as all information may not be available in chart, and defer to physician coding. 71 (narrative of seizure, narrative of cause of seizure) and categorical form (list of seizure characteristics, list of possible seizure causes). The categorical variables usually formed 2x2 tables, with most values clustering in one cell, causing difficulty in interpreting the Kappa value. In addition, the abstractors might record the information in only text or category form, rather than recording in both places. Text variables were compared on percent agreement in meaning. Of the 13 analyzed, 1 was in the 60-69% range, 4 in the 70-79% range, 3 in the 80-89% range, and 5 in the 90-99% range. Most often, discrepancies were not disagreement, but rather missing information in one abstraction versus the other. One key variable, whether the seizure, if one occurred during the visit, was considered new onset, had low agreement. However, clinician reviewers re-assessed this variable for all cases, and had additional information from ORS for some of the cases (see #5, above). 7. Individual reviews of abstractions and diagnosis code assignment. Data was cleaned by looking for outlying or illogical values and addressing the issue with ORS or the clinical reviewers, addressing those cases noted by the clinician reviewers to have a clinical or administrative issue, and updating any previously reviewed cases if there was a protocol change. IX. Dissemination of Results Aim—Maximize the scientific impact of the study through dissemination of the findings and involvement of field experts in data analysis and manuscript development. A. Purpose and plan of dissemination—information dissemination has remained an important component of SCESESD. It includes various methods that allow diffusion of knowledge and experience through various medium of communication. The project strove continuously to keep stakeholders, i.e., policy makers, families, healthcare providers, state and federal agencies, advocacy groups, and persons with epilepsy, informed about the progress on all activities including data findings. Dissemination of the study findings in scientific publications and presentation is a critical priority. The project encourages all investigators, staff, and the CDC Technical Advisor to participate 72 in manuscript development in accordance with the suggested set of rules by the International Council of Medical Journal Editors and the SC Data Oversight Council. The following are venues that allow dissemination of the study methods, approaches, activities, and findings. 1. Reports—prepared quarterly, semiannually, and annually were the major venues utilized to engage the CDC and AAMC advisors and other stakeholders. We attempted to provide interim results of the data generated and analyzed. The Steering Committee and the Epilepsy Foundation of South Carolina (EFSC) have received customized data for seeking legislative assistance. 2. Data Codebook and Dictionary—plan is underway to compile the extensive documentation developed during the course of the study. This included the data abstraction manual and data dictionary, the various coding protocols, the algorithm, the IRB protocol, and the procedure manual for quality assurance and analysis. 3. Website—we developed a link to our sister project on epilepsy, the SC Health Outcome Project on Epilepsy (SCHOPE), to provide useful information to local support groups and families of persons with epilepsy. Once our data analysis is completed, preliminary results will be posted at the website. 4. Presentations—we have actively pursued dissemination of interim data through scientific presentation both locally, nationally, and internationally. The following 12 presentations were made during the past three years. 1. Selassie AW. A multi-faceted epidemiological investigation of epilepsy and seizure disorders in South Carolina: how big is the problem? South Carolina Department of Health and Environmental Control epidemiology seminar, Columbia, SC, November 5, 2004. 2. Ferguson PL. Smith GS. Selassie AW. Turner RP. Wannamaker B. Use of a statewide administrative dataset to determine number of seizure and epilepsy cases. American Epilepsy Society 58th Annual Meeting. New Orleans, LA, December 7, 2004. 3. Selassie AW. Ferguson PL. Smith G. Pickelsimer EE. Wannamaker BW. Turner RP. The frequency of epilepsy and seizure disorders among persons with traumatic brain injury: a population-based evaluation of hospital discharges and 73 emergency department visits in South Carolina, 2996-2001. American Epilepsy Society 58th Annual Meeting. New Orleans, LA, December 7, 2004. 4. Ferguson PL. Highlights of the first year findings (TBI & epilepsy among the topics covered). CDC Ground Rounds. Atlanta, GA, April 14, 2005. 5. Ferguson PF. Selassie AW. South Carolina Epidemiological Studies of Epilepsy and Seizure Disorder (SCESESD): 2002 - 2005. AAMC Grantee’s Meeting. Atlanta, GA, March 16, 2005. 6. Wagner JL. Sample PL. Phillips L. Selassie AW. Psychosocial concerns in pediatric epilepsy: a qualitative analysis. American Psychological Association National Convention. Washington, DC, August 18, 2005. 7. Ferguson PL. Smith G. Selassie AW. Turner RP. Wannamaker BB. Tyrell M. Pope A. Cavins KA. Topping KB. Use of seizure and epilepsy codes in emergency department and inpatient discharges – South Carolina, USA. 26th International Epilepsy Conference. Paris, France, August 28-September 1, 2005. 8. Pickelsimer EE. Sample PL. Ferguson PL. Selassie AW. Personal perspectives of living with epilepsy or a seizure disorder. 26th International Epilepsy Conference. Paris, France, August 28-September 1, 2005. 9. Ferguson PL. Smith GS. Selassie AW. Use of a statewide administrative dataset to determine use of monotherapy vs polytherapy in epilepsy. American Epilepsy Society 59th Annual Meeting. Washington, DC, December 2-6, 2005. 10. Ferguson PL. Special Interest Group presentation on Epidemiology: CDC Epidemiologic Research Initiative. American Epilepsy Society 59th Annual Meeting. Washington, DC, December 2-6, 2005. 11. Pickelsimer EE. Ferguson PL. Smith G. Selassie AW. Occurrence of Epilepsy and Seizure Disorder in a Population Ages >15 Years Before and After Hospitalization for Traumatic Brain Injury – a 3-year Follow-up Study. 7th European Congress on Epileptology. Helsinki, Finland, July 2-6, 2006. 12. Ferguson PL. Selassie AW. South Carolina Epidemiological Studies of Epilepsy and Seizure Disorder (SCESESD): 2002 - 2006. CDC Skill Building Institute. Charleston, SC, May 17, 2006. 5. Manuscripts completed and in preparation—the project pursued manuscript development since data became available in 2005. This activity will remain engaging for the next couple of years to develop as many as 10-12 manuscripts. The following is the list of manuscripts published and in preparation for submission. 74 1. Centers for Disease Control and Prevention. Prevalence of epilepsy and health-related quality of life and disability among adults with epilepsy South Carolina, 2003 and 2004. MMWR 2005;54:1080-2. 2. Sample PL. Ferguson PF. Wagner JL. Pickelsimer EE. Selassie AW. Experiences of persons with epilepsy and their families as they look for medical and community care: A focus group study from South Carolina. Epilepsy & Behavior. 2006;9:649-662. 3. Ferguson PL. Smith GM. Wannamaker BW. Thurman DJ. Pickelsimer EE. Selassie AW. A population-based study of risk of epilepsy after hospitalization for traumatic brain injury. Submitted to CDC for clearance. 4. Wagner JL. Sample PL. Smith G. Pickelsimer EE. Ferguson PL. Selassie AW. Impact of Pediatric Epilepsy: Voices from a Focus Group in South Carolina. Written, final edits being done. 5. Ferguson PL. Chiprich J. Dong B. Thurman DJ. Wannamaker BW. Smith GM. Prevalence of epilepsy and health care access among adults with epilepsy in South Carolina: 2003-2005. Methods & analyses complete, writing not yet complete. 6. Selassie AW. Ferguson PL. Smith GM. Lineberry L. Wannamaker BW. Thurman DJ. Epilepsy and Seizure Disorders among persons with TBI: A population-based Evaluation of Hospital Discharges and ED Visits in SC, 1997-2001. Analyses being revised, writing not complete at this time. 7. Ferguson PL. Smith GM. Wannamaker BW. Thurman DJ. Selassie AW. Antiepileptic drug use in a population-based sample of individuals with epilepsy. Most analyses complete, writing not started. 8. Selassie AW. Ferguson PL. Smith GM. Gu J. Wannamaker BW. Thurman DJ. Predictive value positive and sensitivity of administrative data for epilepsy and seizures. Most analyses complete, writing not started. 75 Appendices 76 Appendix A - Abstraction manual for inpatient and emergency department charts SC Epidemiological Studies of Epilepsy & Seizure Disorders Data Abstraction Manual SECTION 1: RANDOM SAMPLE INFORMATION (Verbatim from Random Sample) 1a. Date of Abstraction (DATEABS) Description: Date on which record was abstracted. Field Length: 8 Values: Date record actually abstracted -- mm/dd/ccyy 1b. Abstracter Initials (INITIALS) Description: Initials of data abstractor. Field Length: 2 Values: First and last name initials 1c. Hospital ID Number (HID) Random Sample Information Description: Number identifying the hospital in which the individual was treated. Field Length: 3 Values: A combination of numbers, which make up the hospital ID number Note: may not be applicable if individual not seen in a hospital. Enter ‘999’ if NA. 1d. Medical Record ID Number (MEDRECID) Random Sample Information Description: A required, unique identification code. Field length: 20 Values: The actual number found on the abstraction list Note: Please enter medical record number from the random sample list we send you. Include dashes, letters, etc., and type exactly as the number appears on the abstraction list. This is only the number to link the data you send to the original file. 1e. Type of medical visit (TYPEVIS) Random Sample Information Description: Type of medical visit. Field length: 1 Values: 1 = hospital inpatient 2 = emergency department 3 = outpatient, primary care provider (GP, family practice, internist, pediatrician) 4 = outpatient, neurologist 5 = outpatient, epileptologist 6 = outpatient, other 9 = not STATED on abstraction list 77 1f. Admittance or Outpatient Visit Date (ADMIT) Random Sample Information Description: Date when individual was admitted to the hospital, ED, or seen in clinic/office. Field length: 8 Values: Date -- mm/dd/ccyy Note: Enter 11/11/1111 if not found on the abstraction list. If admittance date differs on the medical record, make a note of the different date in the comments section. 1g. Discharge Date (DISD) Random Sample Information Description: Date when individual was discharged from the hospital, if applicable. Field length: 8 Values: Date in -- mm/dd/ccyy Note: Enter 11/11/1111 if not found on the abstraction list. If discharge date differs on the medical record, make a note of the different date in the comments section. 1h. Case ID Number (CASEID) Random Sample Information Description: A required, unique identification code. Field length: 8 Values: The actual number found on the abstraction list. Note: Please enter the Case ID Number from the random sample list we send you. Please type exactly as provided. 1i. Is Medical Record Available (RECAVAIL) Description: Determines whether the chart was available to be abstracted. Field length: 1 Values: 1 = Yes 2 = No, clerk states (verbal or written) that record is being microfilmed/ microfiched or off-site storage 3 = No, clerk states (verbal or written) unable to locate but no reason given 4 = No, clerk states (verbal or written) other reason 5 = No identifiable reason Note: If answers 2 through 5, program skips to Missing Record Comments at end of program. If Yes, program continues to next question. SECTION 2: PERSONAL INFORMATION (From Here Forward from Medical Chart) 2a. Individual’s Last Name (LNAME) Description: The last name of the individual. Field length: 40 Values: Any valid name Note: Enter 999 in first column if not STATED in medical record. Note: If the name on the medical record is different than the name on the random sample, enter the name from the medical record, including incorrect spelling. Document in the comment section and flag the record. 78 2b. Individual’s First Name (FNAME) Description: The last name of the individual. Field length: 20 Values: Any valid name Note: Enter 999 in first column if not STATED in medical record. 2c. Date of Birth (DOB) Description: Individual's date of birth. Field Length: 8 Values: mm/dd/ccyy Note: Enter 11/11/1111 if not found in medical record. 2d. Age (AGE) Description: Individual’s age Field Length: 3 Values: Age in years (acceptable range 0 to 120) Note: Enter 999 if no age or DOB in medical record. 2e. Sex (SEX) Description: Field Length: Values: The gender of the individual. 1 M= Male F= Female U= Unknown/Not STATED in medical record 2f. Race (RACE) Description: The race or ethnicity of the individual. Field Length: 1 Values: 1 = White 2 = Black 3 = Oriental / Asian 4 = American Indian 5 = Hispanic 6 = Native Hawaiian or other Pacific Islander 8 = Other 9 = Unknown/Not STATED in medical record 2g. Marital Status (MARISTAT) Description: Current marital status of the individual. Field length: 1 Values: 1 = Married (includes separated, stated common law marriages) 2 = Widowed 3 = Single (includes divorced) 5 = Minor child (<18, unless stated as married) 79 9 = Unknown/Not STATED in medical record 2h. Individual’s Street Address (INDADDR) Description: The street address of residence of the individual. Field length: 40 Values: Any street address Note: Be sure to get apartment and route numbers if applicable. Note: Enter 999 in first column if not STATED in medical record. 2i. Individual’s City (INDCITY) Description: The city of residence of the individual. Field length: 20 Values: Any city name Note: Enter 999 in first column if not STATED in medical record. 2j. Individual’s State (INDSTATE) Description: The state of residence of the individual. Field length: 2 Values: SC = South Carolina NC = North Carolina GA = Georgia OT = Other UK = Unknown/Not STATED in medical record 2k. Individual’s Zip Code (INDZIP) Description: The zip code of residence of the individual. Field length: 5 Values: Five-digit number Note: Enter 999 in first column if not STATED in medical record. 2l. Telephone Number (PHONE) Description: Telephone number of the individual. Field length: 13 Values: A telephone number with area code as stated in the medical record. Note: Enter (000) 000-0000 if the number is not STATED in medical record. If the area is missing, enter (000) for area code and then the 7-digit number. 2m. Employment Status (EMPSTAT) Description: Employment status of the individual. Field length: 1 Values: 1 = Student 2 = Employed (full time or part time) 3 = Employed & attending school 4 = Retired 5 = Disabled 80 6 = Unemployed 7 = Minor Child under age 5 8 = None of the above 9 = Not STATED in medical record 2n. Family Structure (FAMSTRU) Description: Family structure of individual Field length: 2 Values: 1 = Lives alone 2 = Lives with spouse/significant other 3 = Lives with parent(s) 4 = Lives with relative other than parent 5 = Lives with foster parent 6 = Lives with hired caregiver 7 = Lives in rehabilitation facility 8 = Lives in nursing home 10 = Lives in state residential facility 11 = Lives in group home 12 = Lives in detention or corrections facility 13 = Other 9 = Not STATED in medical record 2o. Insurance Status (INSSTA) Description: Insurance of individual Field length: 1 Values: 1 = no medical insurance/self-pay 2 = Medicaid (includes Select Health, a SC Medicaid HMO) 3 = Medicare 4 = HMO 5 = Other private insurance 6 = Champus, TRI Care, or VA 9 = not STATED in medical record SECTION 3: DIAGNOSIS & SEIZURE SPECIFICS Enter ‘999’ in text fields if no information in medical record. 3a. Chief Complaint (PREDIAG) Description: Chief complaint Field length: 40 Values: Medical reason for the visit or diagnosis 3b. Discharge Diagnosis (POSTDIAG) Description: Discharge diagnosis Field length: 40 Values: MD’s discharge diagnosis 3c. Primary Diagnosis Code/ICD-9 Code (DIAG1) 81 Description: An ICD-9 CM code and rubric assigned to the diagnosis. Field Length: 5 Values: Code that the hospital, clinic, or office listed first. Note: Decimal points are not entered. Enter 5 nine's (99999) if no code is found. Note: Do not Add Zeros to any three digit N-Code found in the record. 3d-l. Secondary ICD-9 Codes (DIAG2-DIAG10) Description: An ICD-9 CM code and rubric assigned to the diagnoses. Field Length: 5 for each, for up to 9 codes Values: Any code listed in the medical record. Note: Decimal points are not entered. Enter 5 nine's (99999) if no code is found. Note: Do not Add Zeros to any three digit N-Code found in the record. 3m. Discharge medication prescribed (MEDRX1) Description: Maintenance anti-epileptic medications prescribed this visit at discharge. Field Length: Checklist of anti-epileptic medications Values: Check off medication if listed in medical record as discharge medication. 3n. Discharge medication prescribed (MEDRX2) Description: Maintenance anti-epileptic medication prescribed this visit at discharge. Field Length: Checklist of anti-epileptic medications Values: Check off medication if listed in medical record as discharge medication. 3o. Discharge medication prescribed (MEDRX3) Description: Maintenance anti-epileptic medication prescribed this visit at discharge. Field Length: Checklist of anti-epileptic medications Values: Check off medication if listed in medical record as discharge medication. 3p. Other discharge medication(s) prescribed (MEDRX4) Description: Maintenance anti-epileptic medication prescribed this visit at discharge. Field Length: Checklist of anti-epileptic medications Values: Check off medication if listed in medical record as discharge medication. 3q. Seizure treatments prescribed other than medication (OTHTRT) Description: Other non-drug treatments prescribed to individual. Field Length: 1 Values: 1 = Vagal nerve stimulator 2 = Neurosurgery 3 = Ketogenic diet 9 = not STATED in medical record 82 3r. Seizure visit (SZVISIT) Description: Is this visit in reference to a current (day of visit or occurring since last visit) seizure, possible seizure, seizure-like episode, or epilepsy? Field Length: 1 Values: 1 = Purpose of visit related to a current seizure 2 = Initial visit not related to seizure, but seizure occurred during visit 3 = No current seizure, but visit related to seizure treatment, assessment, etc. (If 1, 2, or 3, rest of sections 3, 4, & 5 refer to this seizure or seizurelike episode) 4 = None of the above, visit not related to seizures (skip to 6a) Following questions in this section and next 2 sections refer to the CURRENT seizure or seizure-like episode. 3s. Date of current seizure or seizure-like episode (SZDATE) Description: Date of recent seizure or seizure-like episode (may be date of visit or date of most recent episode since last visit) Field Length: 8 Values: Any month, day, and year – mm/dd/ccyy Note: may be estimated if not explicitly stated in record. If not STATED at all in record, put 11/11/1111 3t-v. Narrative of current seizure or seizure-like episode (NTSEIZ) Description: Description of the current seizure and/or seizure-like episode (what happened to patient before, during, and after this episode, or during typical episode) Field Length: 40 for each of three lines Values: Any description of the seizure episode(s). Please be as descriptive as possible. 3w. Seizure Type (SZTYPE1) Description: Clinical determination by MD of current seizure type(s), seizure disorder, seizure-like episode, &/or epilepsy syndrome Field Length: Checklist of seizure types Values: Check off seizure type if listed in medical record describing this episode, or typical episode. 3x. Seizure Type (SZTYPE2) Description: Clinical determination by MD of current seizure type(s), seizure disorder, seizure-like episode, &/or epilepsy syndrome Field Length: Checklist of seizure types Values: Check off seizure type if listed in medical record describing this episode, or typical episode. 3y. Seizure Type (SZTYPE3) 83 Description: Field Length: Values: Clinical determination by MD of current seizure type(s), seizure disorder, seizure-like episode, &/or epilepsy syndrome Checklist of seizure types Check off seizure type if listed in medical record describing this episode, or typical episode. 3z. Other seizure type(s) (SZTYPE4) Description: Other clinical determination by MD of current seizure type or seizurelike episode. Field Length: 40 Values: Any additional seizure type(s) or epilepsy not included on above lists (refer to list provided in class) 3aa. Number of seizures or seizure-like episodes (NUMSZ) Description: How many seizures or seizure-like episodes did the individual experience during this episode? Field Length: 1 Values: 1 = One seizure 2 = Two seizures 3 = More than two seizures 9 = Not STATED in medical record 3bb. Time of seizure or seizure-like episode onset (SZTIME) Description: Time of onset of seizure or seizure-like episode (if cluster of seizures or seizure-like episodes, use time of onset of cluster). Field Length: 1 Values: 1 = Morning (7 am to 11:59 am) 2 = Afternoon (noon to 5:59 pm) 3 = Evening (6 pm to 9:59 pm) 4 = Nighttime (10 pm to 6:59 am) 9 = Not STATED in medical record Note: May be estimated if approximate time mentioned in medical record. 3cc. Length of seizure (LENGTH) Description: How long did the seizure or seizure-like episode last? Field Length: 1 Values: 1 = 30 seconds or less 2 = >30 secs & up to 2 minutes 3 = > 2 minutes & up to 5 minutes 4 = More than 5 minutes 9 = Not STATED in medical record 3dd. Seizure or seizure-like episode witness (SZOBS) Description: Who gave the details of the seizure or seizure-like episode? Field Length: 1 84 Values: 1 = A health-care professional who directly witnessed the seizure or episode 2 = A non-health care professional who directly witnessed the seizure or episode 3 = Someone who spoke with an individual who directly witnessed the seizure or episode 9 = Not STATED in medical record 3ee-gg. Emergent medication(s) given (MEDTRT1-3) Description: Medications given pre-hospital, or by ED, if applicable. Field Length: Checklist of anti-epileptic medications Values: Check off medication if listed in medical record as given for seizure or seizure-like episode. SECTION 4: DETERMINATION OF SEIZURE TYPE All questions in Section 4 refer to the current seizure or seizure-like episode. 4a. Change in Mood before Seizure (ASMOOD) Description: Did the individual have a change in mood (i.e. quiet, agitated, scared) PRIOR to this seizure? Field Length: 1 Values: 1 = Yes 2 = No 9 = not STATED in medical record 4b. Restlessness before Seizure (ASREST) Description: Did the individual have restlessness PRIOR to this seizure? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 4c. Bad taste in mouth (BADTST) Description: Did the individual experience a bad taste in mouth PRIOR to seizure? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 4d. Smelling foul odor (BADSML) Description: Did the individual smell a foul odor PRIOR to seizure? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 85 4e. Seeing spots or flashing lights or colors in front of eyes (BADSEE) Description: Did the individual see spots or flashing lights or colors PRIOR to seizure? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 4f. Strange feeling in stomach (STOMACH) Description: Did the individual experience a strange feeling in stomach PRIOR to seizure? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 4g. Nausea (NAUSEA) Description: Did the individual experience nausea PRIOR to seizure? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 4h. Headache (HEADACHE) Description: Did the individual experience a headache PRIOR to seizure? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 4i. Weakness (WEAK) Description: Did the individual experience weakness PRIOR to seizure? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 4j. Numbness (NUMB) Description: Did the individual experience numbness PRIOR to seizure? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 4k. Tingling (TINGLE) 86 Description: Did the individual experience tingling in hands, feet, arms, legs, or face PRIOR to seizure? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 4l. Crying out (CRYOUT) Description: Did the individual cry out PRIOR to seizure? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 4m. Mumbling (MUMBLE) Description: Did the individual mumble PRIOR to seizure? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 4n. Making sounds (MAKSNDS) Description: Did the individual make sounds but not real words PRIOR to seizure? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 4o. Other aura (OTHAURA) Description: Did the individual have a feeling or sense that a seizure was about to occur that cannot be described? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 4p. Start of seizure (START) Description: Did the seizure start suddenly, without any warning? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 4q. Fall (FALL) Description: Did the individual fall DURING the seizure? 87 Field Length: Values: 4r. Stiff (STIFF) Description: Field Length: Values: 1 1 = Yes, fell forward 2 = Yes, fell backward 3 = No, did not fall (skip to 4t) 8 = Not applicable, since individual was sitting, lying down, etc. (skip to 4t) 9 = Not STATED in medical record (skip to 4t) Did the individual go stiff and fall like a tree DURING the seizure? 1 1 = Yes 2 = No 9 = Not STATED in medical record 4s. Limp (LIMP) Description: Did the individual go limp and fall over DURING the seizure? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 4t. Blacked out (BLACKOUT) Description: Did the individual black out DURING the seizure? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 4u. Pale Face (PALFACE) Description: Did the individual’s face color become pale DURING the seizure? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 4v. Red face (REDFACE) Description: Did the individual’s face color flush, or become red, DURING the seizure? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 4w. Blue face (BLUFACE) 88 Description: Field Length: Values: Did the individual’s face color become blue DURING the seizure? 1 1 = Yes 2 = No 9 = Not STATED in medical record 4x. Tongue/lip/cheek biting (TONGUE) Description: Did the individual bite their tongue, lip or cheek DURING the seizure? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 4y. Remembers seizure (MEMORY) Description: Does the individual remember what happened DURING the seizure? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 4z. Hear during seizure (HEAR) Description: Was the individual able to hear DURING the seizure? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 4aa. Incontinence (INCONT) Description: Did the individual lost control of bladder &/or bowels DURING seizure? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 4bb. Ability to talk (TALK) Description: Was the individual able to speak clearly DURING the seizure? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 4cc. Affected side (SIDE) Description: What side of the body was affected DURING the seizure? Field Length: 1 89 Values: 1 = Right side only 2 = Left side only 3 = Both sides 4 = No activity on either side 9 = Not STATED in medical record 4dd. Chewing motions (CHEW) Description: Did the individual exhibit chewing motions DURING the seizure? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 4ee. Lip motions (LIP) Description: Did the individual exhibit lip-smacking DURING the seizure? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 4ff. Face pulling (PULLING) Description: Did the individual’s face pull to one side DURING the seizure? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 4gg. Facial trembling/shaking/twitching (FACE) Description: Did the individual’s face tremble, shake or twitch DURING the seizure? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 4hh. Head turning (HEAD) Description: How did the individual’s head turn DURING the seizure? Field Length: 1 Values: 1 = Head turned to the right 2 = Head turned to the left 3 = Head turned back and forth 4 = Head did not turn 9 = Not STATED in medical record 4ii. Eye movement (EYEMOVE) Description: How did the individual’s eyes move DURING the seizure? 90 Field Length: Values: 1 1 = Eyes pointed to the right 2 = Eyes pointed to the left 3 = Eyes rolled upward 4 = Eyes moved around randomly 5 = Eyes were straight ahead 6 = Eyes were closed 9 = Not STATED in medical record 4jj. Hand movement (HANDS) Description: Did the individual’s hands pick at or fumble with clothes or objects DURING the seizure? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 4kk. Staring (STARE) Description: Did the individual exhibit a blank stare DURING the seizure? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 4ll. Eyelid fluttering/blinking (EYELID) Description: Did the individual exhibit eyelid fluttering or blinking DURING the seizure? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 4mm. Extremity stiffening (STIFFEN) Description: Did the individual exhibit stiffening (or spasms) of the arms &/or legs DURING the seizure? Field Length: 1 Values: 1 = Arms stiffened 2 = Legs stiffened 3 = Both arms & legs stiffened 4 = Neither arms nor legs stiffened 9 = Not STATED in medical record 4nn. Extremity jerking (JERKING) Description: Did the individual exhibit jerking (or twitching) movements of the arms &/or legs DURING the seizure? 91 Field Length: Values: 1 1 = Arms had jerking movements 2 = Legs had jerking movements 3 = Both arms & legs had jerking movements 4 = Neither arms nor legs had jerking movements 9 = Not STATED in medical record 4oo. Body trembling (TREMBLE) Description: Did the individual exhibit trembling (other than in face) DURING the seizure? Field Length 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 4pp. Grunting (GRUNT) Description: Did the individual make grunting noises DURING the seizure? Field Length 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 4qq. Stopped breathing (NOBREATH) Description: Did the individual stop breathing DURING the seizure? Field Length 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 4rr. Noisy labored breathing (STERTOR) Description: Did the individual have noisy labored (stertorous) breathing DURING the seizure? Field Length 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 4ss. Salivation, drooling, foaming (DROOL) Description: Did the individual have increased salivation, drooling, or foaming at the mouth DURING the seizure? Field Length 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 92 4tt. Heart racing (TACHY) Description: Did the individual have experience tachycardia, or a feeling of their heart racing, DURING the seizure? Field Length 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 4uu. Panic feelings (PANIC) Description: Did the individual experience feelings of panic DURING the seizure? Field Length 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 4vv. Recovery after seizure (RECOVER) Description: Did the individual immediately recover after the seizure, and was able to return to previous activities AFTER the seizure? Field Length 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 4ww. Sleepy after seizure (SLEEPY) Description: Was the individual sleepy AFTER the seizure? Field Length 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 4xx. Restless after seizure (RESTLESS) Description: Was the individual agitated or restless AFTER the seizure? Field Length 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 4yy. Combative after seizure (COMBAT) Description: Was the individual combative AFTER the seizure? Field Length 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 4zz. Confused after seizure (CONFUSED) 93 Description: Field Length Values: Was the individual confused AFTER the seizure? 1 1 = Yes 2 = No 9 = Not STATED in medical record 4aaa. Trouble speaking (TRBSPEAK) Description: Did the individual have trouble saying words (mumbling, nonsense speech) AFTER the seizure? Field Length 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 4bbb. Trouble finding word (TRBWORD) Description: Did the individual have trouble finding the right word to say AFTER the seizure? Field Length 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record SECTION 5: CAUSE OF SEIZURE Enter ‘999’ in text fields if no information in medical record. 5a-c. Narrative of Cause of Seizure or seizure-like episode (CAUSE) Description: Brief description of current circumstances that might have contributed to this seizure or seizure-like episode (what provoked this seizure?). Field Length: 40 characters for each of three lines Values: Any narrative that describes a possible current cause of seizure or seizure-like episode. Note: Be as specific as possible. 5d. Injury (INJURY) Description: Is there report that the individual’s seizure or seizure-like episode occurred due to current injury? Field Length: 1 Values: 1 = Yes, head injury 2 = Yes, injury other than to head 3 = Yes, injury to both head and elsewhere 4 = No (skip to 5f) 9 = Not STATED in medical record (skip to 5f) 5e. Injury Incident Location Type (INJLOC) Description: Describes the type of place of occurrence of the current injury. 94 Field Length: Values: 2 1 = Home 2 = Residential Institution 3 = School, other Institution and Public Administrative Area 4 = Sports or Recreation area 5 = Street or Highway 6 = Trade or Service Area 7 = Industrial or Construction Area 8 = Other Specified Place 9 = Unspecified Place or not STATED in medical record 10 = Farm 5f. Illness prior to current seizure or seizure-like episode (ILLNESS) Description: Is there report that the individual is or has recently been ill? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 5g. Fever prior to seizure or seizure-like episode (FEVER) Description: Is there report that the individual has had an elevated temperature? Field Length: 1 Values: 1 = Yes, stated in record but no temperature recorded 2 = Yes, 99.6 to 101.5 F degrees (37.6 to 38.6 C) 3 = Yes, greater than 101.5 F degrees (> 38.6 C) 4 = No 9 = Not STATED in medical record 5h. Sleep deprivation (SLEEP) Description: Is there report that the individual has experienced sleep deprivation recently? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 5i. Pregnancy (PREG) Description: Is the individual pregnant? Field Length: 1 Values: 1 = Yes, record states female is pregnant 2 = Teen or adult female is not pregnant (skip to 5m) 9 = Not STATED in medical record (skip to 5m) 5j. Eclampsia (ECLAMP) 95 Description: Field Length: Values: Does the individual have eclampsia (or pre-eclampsia, pregnancyinduced hypertension, PIH, toxemia)? 1 1 = Yes 2 = No 9 = Not STATED in medical record 5k. Alcohol use (ETOHUSE) Description: Is there report that the individual is presently using alcohol? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 5l. Drug use (DRUGUSE) Description: Is there report that the individual is presently using illicit drugs? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 5m. New Onset (NEWONSET) Description: Is this visit in regard to the patient’s first seizure or seizure-like episode? Field Length: 1 Values: 1 = Yes (skip to 6aa) 2 = No (previous seizures noted, or history of seizures in medical record) 9 = Not STATED in medical record 5n. Medication change (MEDCHNG) Description: Is there report that the individual has recently changed anti-epileptic medications? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 5o. Weight gain in children (WTGAIN) Description: Is there report that the child has gained weight recently? Field Length: 1 Values: 1 = Yes, pt aged 1-18 years has gained weight recently 2 = No, pt aged 1-18 years had not gained weight recently 8 = Not applicable (individual 18 years or older) 9 = Aged 1-18 years, but weight change not STATED in medical record 96 5p. Use of personal protective equipment (EQUIP) Description: Is there report that the individual was using a prescribed helmet when seizure occurred (this would be different from the usual bicycle helmet, etc.)? Field Length: 1 Values: 1 = Yes, individual was using prescribed helmet 2 = No, individual was not using prescribed helmet 9 = Not STATED in medical record 5q. Noncompliance with medication (NONCOMP) Description: Is there report that the individual has recently not been compliant with his/her seizure medication regimen? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record SECTION 6: MEDICAL HISTORY Note: Enter “YES” if the record states previous history for the following. Enter “NO” if the record states NO previous history for the following. Enter “999” if no previous history information is documented in the medical record. 6a. Previous seizure disorder or epilepsy diagnosis (PREVSEIZ) Description: Is there an earlier diagnosis, or some evidence of, seizure disorder or epilepsy in the past? Field Length: 1 Values: 1 = Yes (some evidence of previous seizures in chart) 2 = No (no mention of any previous seizures) (skip to 6aa) 9 = Not STATED in medical record (skip to 6aa) 6b. Previous seizure type (PREVTYP1) Description: Previous clinical determination of type of seizure(s), seizure disorder(s), or epilepsy? Field Length: Checklist of seizure types Values: Check off seizure type if previous determination of seizure, seizure disorder, or epilepsy 6c. Previous seizure type (PREVTYP2) Description: Previous clinical determination of type of seizure(s), seizure disorder(s), or epilepsy? Field Length: Checklist of seizure types Values: Check off seizure type if previous determination of seizure, seizure disorder, or epilepsy 6d. Previous seizure type (PREVTYP3) 97 Description: Field Length: Values: Previous clinical determination of type of seizure(s), seizure disorder(s), or epilepsy? Checklist of seizure types Check off seizure type if previous determination of seizure, seizure disorder, or epilepsy 6e. Other previous seizure type(s) (PREVTYP4) Description: Other previous clinical determination of type of seizure(s), seizure disorder(s), or epilepsy? Field Length: 40 Values: Any previous seizure type(s) not included on above lists (refer to list provided in class) 6f-h. Narrative of past seizure(s) (PREVNAR) Description: Description of past seizure(s) – what happens to patient before, during, & after. Field Length: 40 for each of 3 lines Values: Any description of previous seizure(s). Please be as descriptive as possible. 6i. Onset of seizure disorder/epilepsy (SZONSET) Description: Date of original onset of seizure disorder or epilepsy. Field Length: 8 Values: mm/dd/ccyy Note: enter 11/11/1111 if not found in medical record. Can estimate date if an approximate date is mentioned in medical record. 6j. Date of Seizure (PREVDATE) Description: Date when individual had most recent seizure other than current seizure or seizure-like episode. Field Length: 8 Values: Any month, day, and year -- mm/dd/ccyy Note: Put 11/11/1111, if not STATED in medical record. May be estimated if approximate date mentioned in medical record. 6k. Multiple seizure types (MULTTYPE) Description: How many types of seizures has the individual experienced? Field Length: 1 Values: 1 = One 2 = Two 3 = More than two 9 = Not STATED in medical record. 6l. Frequency of seizures (FREQSEIZ) Description: How frequently does individual experience any seizure activity? 98 Field Length: Values: 1 1 = Less than once a year 2 = More than once a year 3 = More than once a month 4 = More than once a week 5 = More than once a day 9 = Not STATED in medical record 6m. Previous EEG (PREVEEG) Description: Has individual previously had an EEG? Field Length: 1 Values: 1 = Yes, but no comment as to results in chart 2 = Yes, normal 3 = Yes, abnormal 9 = Not STATED in chart 6n. Previous video EEG (PREVVEEG) Description: Has individual previously had a video EEG? Field Length: 1 Values: 1 = Yes, but no comment as to results in chart 2 = Yes, normal 3 = Yes, abnormal 9 = Not STATED in chart 6o. Previous CT scan of the head (PREVCT) Description: Has individual previously had a CT scan of the head? Field Length: 1 Values: 1 = Yes, but no comment as to results in chart 2 = Yes, normal 3 = Yes, abnormal 9 = Not STATED in chart 6p. Previous MRI of the head (PREVMRI) Description: Has individual previously had a MRI of the head? Field Length: 1 Values: 1 = Yes, but no comment as to results in chart 2 = Yes, normal 3 = Yes, abnormal 9 = Not STATED in chart 6q. Previous seizure treatments - medication (PREVMED) Description: Has individual been on anti-epileptic medication in past? Field Length: 1 Values: 1 = Yes 2 = No 99 9 = Not STATED in medical record 6r. Previous seizure treatments – neurosurgery (PREVSURG) Description: Has individual had neurosurgery for epilepsy (lobectomy, resection, hemispherectomy, corpus callosotomy, subpial transection) in the past? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 6s. Previous seizure treatments – vagal nerve stimulator (PREVVNS) Description: Has individual had a vagal nerve stimulator implanted for epilepsy in the past? Field Length: Values: 1 1 = Yes 2 = No 9 = Not STATED in medical record 6t. Previous seizure treatments – ketogenic diet (PREVKETO) Description: Has individual been on a ketogenic diet in the past for epilepsy? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 6u. Previous seizure treatments – other (PREVOTH) Description: Has individual received some other treatment for epilepsy other than medication, ketogenic diet, vagal nerve stimulator, or neurosurgery? Field Length: Values: 40 Description of other treatments, including alternative therapies. 6v. Present medication (PRESMED1) Description: All present anti-epileptic medication. Field Length: Checklist of anti-epileptic medications. Values: Check off medication patient taking at onset of visit. 6w. Present medication (PRESMED2) Description: All present anti-epileptic medication. Field Length: Checklist of anti-epileptic medications. Values: Check off medication patient taking at onset of visit 6x. Present medication (PRESMED3) Description: All present anti-epileptic medication. Field Length: Checklist of anti-epileptic medications. 100 Values: 6y. Check off medication patient taking at onset of visit Other present medication(s) (PRESMED4) Description: All present anti-epileptic medication. Field Length: Checklist of anti-epileptic medications. Values: Check off medication patient taking at onset of visit 6z. Previous medication noncompliance (PREVNONC) Description: Prior to this episode, does the individual have a history of epilepsy medication noncompliance? Field Length: 1 Values: 1 = Yes, skipping doses 2 = Yes, stopping medication 3 = Yes, but not specified 4 = No 9 = Not STATED in medical record Note: If ‘1’ and ‘2’, choose ‘2’. 6aa. Previous TBI (PREVTBI) Description: Does the patient’s medical history include a previous TBI? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 6bb. Previous Alcohol Misuse (PREVALC) Description: Prior to this episode, does the patient’s medical history include alcohol misuse? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 6cc. Previous Illicit Drug Use (PREVDRUG) Description: Prior to this episode, does the patient’s medical history include illicit drug use? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 6dd. Previous illness or injury (ETIOLOGY) Description: Any past illness, condition, or injury in the patient’s history that initially caused the seizure disorder/epilepsy. Field Length: 40 101 Values: Previous illness, seizures/epilepsy. condition, or injury that initially caused 6ee-hh.Co-morbidities (COMORBID) Description: Any other current medical &/or psychological conditions. Field Length: 40 for each of 4 lines Values: List all other current medical conditions. 6ii. Family history of seizures (FAMHX) Description: Family history of seizures, seizure disorders, or epilepsy. Field Length: 1 Values: 1 = Yes 2 = No (skip to section 7) 9 = Not STATED in medical record (skip to section 7) 6jj. Specifics of seizure family history (SPECFMHX) Description: What type of seizure, etc., is mentioned in family history? Field Length: 40 Values: Write whether family history of seizures, seizure disorder, or epilepsy, and what type, as well as family member(s) with disorder. SECTION 7: DIAGNOSTIC TESTS A physician’s interpretation, such as in the H&P or notes, is preferable to determining results from departmental print-out. 7a. Anti-epileptic drug level 1 (AED1) Description: Blood drug level of an anti-epileptic drug and results Field Length: 1 Values: 1 = Test ordered, no results/interpretation in chart 2 = Test done, and within therapeutic range 3 = Test done, and below therapeutic range 4 = Test done, and above therapeutic range 9 = Not STATED in medical record 7b. Anti-epileptic drug level 2 (AED2) Description: Blood drug level of an anti-epileptic drug and results Field Length: 1 Values: 1 = Test ordered, no results/interpretation in chart 2 = Test done, and within therapeutic range 3 = Test done, and below therapeutic range 4 = Test done, and above therapeutic range 9 = Not STATED in medical record 7c. Blood Alcohol Level (BAL) Description: BAL and Results? Field Length: 1 102 Values: 1 = Test ordered, no results/interpretation in chart 2 = BAL was done and value is 0-10 BAC in mg/dl (Negative) 3 = BAL was done and value is > 10 BAC in mg/dl 9 = Not STATED in medical record 7d. Toxicology Screen (TOX) Description: Toxicology screen and Results? Field Length: 1 Values: 1 = Test ordered, no results/interpretation in chart 2 = Toxicology screen was done and Negative 3 = Toxicology was done and At least one positive result 9 = Not STATED in medical record 7e. CT Information (CT) Description: CT of the head and Results? Field length: 1 Values: 1 = Test ordered, no results/interpretation in chart 2 = CT was taken and Normal 3 = CT was taken and Abnormal 9 = Not STATED in medical record 7f. MRI Information (MRI) Description: MRI of the head and Results? Field length: 1 Values: 1 = Test ordered, no results/interpretation in chart 2 = MRI was taken and Normal 3 = MRI was taken and Abnormal 9 = Not STATED in medical record 7g. EEG Information (EEG) Description: EEG and Results? Field length: 1 Values: 1 = Test ordered, no results/interpretation in chart 2 = EEG was taken and Normal 3 = EEG was taken and Abnormal 9 = Not STATED in medical record 7h. Video EEG Information (VEEG) Description: Video EEG and Results? Field length: 1 Values: 1 = Test ordered, no results/interpretation in chart 2 = EEG was taken and Normal 3 = EEG was taken and Abnormal 9 = Not STATED in medical record 103 7i. EKG Information (EKG) Description: EKG and results? Field length: 1 Values: 1 = Test ordered, no results/interpretation in chart 2 = EKG was taken and Normal 3 = EKG was taken and Abnormal 9 = Not STATED in medical record 7j. Neurology consultation (NEUROCON) Description: Was the patient seen by a neurologist? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record. 7k. Other consultation (OTHERCON) Description: Was the patient seen by some other specialist regarding their seizure or seizure-like episode? Field Length: 1 Values: 1 = Yes 2 = No (skip to section 8) 9 = Not STATED in medical record (skip to section 8) 7l. Other consultation specifics (CONSPEC) Description: By what other type of specialist was the patient seen? Field Length: 40 Values: Specify what other specialist (list only specialties - if only name available, leave empty). SECTION 8: DISCHARGE DISPOSITION 8a. Patient discharged to (DISCHTO) Description: Determines the type of facility the patient was discharged to. Field length: 1 Values: 0 = Transferred to another acute care hospital 1 = Returned home, self-care 2 = Returned home, requiring non-skilled assistance (family member, etc.) 3 = Returned home, requiring home health services and/or outpatient rehabilitation 4 = Transferred to a residential facility without skilled nursing services or with an unknown level of nursing care 5 = Transferred to a residential facility with skilled nursing services 6 = Transferred to an inpatient rehabilitation facility 7 = Died 10 = Left against medical advice (AMA) 104 11 = Correctional facility - includes prison, jail and detention centers 8 = Other 9 = Unknown/Not STATED in medical record SECTION 9: REFERRALS 9a. Primary Care Physician or pediatrician (PCMD) Description: Is a primary care physician listed (may include PCP, pediatrician, internist, family health clinic, family practice physician, etc)? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 9b. Primary care physician or pediatrician referral (REFPCMD) Description: Did the individual receive a referral to a primary care physician (see above)? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 9c. Neurologist (NEURO) Description: Is a neurologist listed? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 9d. Neurologist referral (REFNEURO) Description: Did the individual receive a referral to a neurologist? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 9e. Epileptologist (EPIL) Description: Is the individual currently under the care of an epileptologist/epilepsy specialist? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 9f. Epileptologist referral (REFEPIL) 105 Description: Field Length: Values: Did the individual receive a referral to an epileptologist/epilepsy specialist? 1 1 = Yes 2 = No 9 = Not STATED in medical record 9g. Neurosurgical referral (REFNS) Description: Did the individual receive a referral to a neurosurgeon? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 9h. Referral for EEG or VEEG tests (REFEEG) Description: Did the individual receive a referral for an EEG or VEEG? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 9i. Referral for an MRI (REFMRI) Description: Did the individual receive a referral for an MRI? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 9j. Referral to an Epilepsy Center (REFEC) Description: Did the individual receive a referral to an epilepsy center? Field length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record SECTION 10: CLOSING 10a. Abstractor Comments (COMMENTS) Description: Does the abstractor have any comments about this record? Field length: 1 Values: 1 = Yes 2 = No (skip to 10g) 10b-c. Abstracter Administrative Comments (COMMENT1-2) Description: Informative remarks made by the data abstracter to add additional insight to this particular patient record. 106 Field Length: Values: 40 for each of 2 lines Any narrative (administrative problems – random sample info different from chart, chart difficult to read, etc.) 10d-f. Abstractor Clinical Comments (COMMENT3-5) Description: Informative remarks made by the data abstracter to add additional insight to this particular patient record. Field Length: 40 for each of 3 lines Values: Any narrative (clinically-related problems – need to add something extra for which there was no entry place, note that question mark written in text was copied from chart, note that question mark written in text due to illegible chart, etc.) 10g. Flagged for Special Review (FLAG) Description: Does this record need to be flagged for special review? Field length: 1 Values: 1 = Yes 2 = No Note: A record would need to be reviewed by another abstractor if there is contradictory information in the chart (ex. a pregnant male), information in medical record not matching random sample or other problems. Note: Reason(s) for flag should be recorded in Abstractor Comments. 107 Appendix B - Abstraction manual for physician office visit charts SC Epidemiological Studies of Epilepsy & Seizure Disorders Data Abstraction Manual for Physician Office Visits SECTION 1: RANDOM SAMPLE INFORMATION (Verbatim from Random Sample) 1a. Date of Abstraction (DATEABS) Description: Date on which record was abstracted. Field Length: 8 Values: Date record actually abstracted -- mm/dd/ccyy 1b. Abstracter Initials (INITIALS) Description: Initials of data abstractor. Field Length: 2 Values: First and last name initials 1c. Physician ID Number (PID) Random Sample Information Description: Number identifying the individual’s physician. Field Length: 3 Values: A combination of numbers, which make up the physician ID number Note: may not be applicable if individual not seen in a hospital. Enter ‘999’ if NA. 1d. Medical Record ID Number (MEDRECID) Random Sample Information Description: A required, unique identification code. Field length: 20 Values: The actual number found on the abstraction list Note: Please enter medical record number from the random sample list we send you. Include dashes, letters, etc., and type exactly as the number appears on the abstraction list. This is only the number to link the data you send to the original file. 1e. Type of physician (TYPEMD) Random Sample Information Description: Specialty of physician or other provider (ie, nurse practitioner). Field length: 1 Values: 1 = primary care (ie, GP, family practice, internal medicine, pediatrics) 2 = neurology 3 = epileptology 4 = obstetrics and/or gynecology 5 = other 9 = not STATED on abstraction list 1f. Outpatient Visit Date (ADMIT) Random Sample Information Description: Date when individual was seen by physician. 108 Field length: 8 Values: Date -- mm/dd/ccyy Note: Enter 11/11/1111 if not found on the abstraction list. If date differs on the medical record, make a note of the different date in the comments section. 1g. Case ID Number (CASEID) Random Sample Information Description: A required, unique identification code. Field length: 8 Values: The actual number found on the abstraction list. Note: Please enter the Case ID Number from the random sample list we send you. Please type exactly as provided. 1h. ICD-9-CM code, most recent (ICD9ORS) Random Sample Information Description: Seizure-related code provided in random sample. Field length: 1 Values: 0 = 345.0 1 = 345.1 2 = 345.2 3 = 345.3 4 = 345.4 5 = 345.5 6 = 345.6 7 = 345.7 8 = 345.8 9 = 345.9 10 = 780.31 11 = 780.39 12 = 780.3 1i. Is Medical Record Available (RECAVAIL) Description: Determines whether the chart was available to be abstracted. Field length: 1 Values: 1 = Yes 2 = No, clerk states (verbal or written) that record is being microfilmed/ microfiched or off-site storage 3 = No, clerk states (verbal or written) unable to locate but no reason given 4 = No, clerk states (verbal or written) other reason 5 = No identifiable reason Note: If answers 2 through 5, program skips to Missing Record Comments at end of program. If Yes, program continues to next question. SECTION 2: PERSONAL INFORMATION (From Here Forward from Medical Chart) 2a. Individual’s Last Name (LNAME) 109 Description: The last name of the individual. Field length: 40 Values: Any valid name Note: Enter 999 in first column if not STATED in medical record. Note: If the name on the medical record is different than the name on the random sample, enter the name from the medical record, including incorrect spelling. Document in the comment section and flag the record. 2b. Individual’s First Name (FNAME) Description: The last name of the individual. Field length: 20 Values: Any valid name Note: Enter 999 in first column if not STATED in medical record. 2c. Date of Birth (DOB) Description: Individual's date of birth. Field Length: 8 Values: mm/dd/ccyy Note: Enter 11/11/1111 if not found in medical record. 2d. Age (AGE) Description: Individual’s age at time of most recent visit Field Length: 3 Values: Age in years (acceptable range 0 to 120) Note: Enter 999 if no age or DOB in medical record. 2e. Sex (SEX) Description: Field Length: Values: 2f. The gender of the individual. 1 M= Male F= Female U= Unknown/Not STATED in medical record Race (RACE) Description: The race or ethnicity of the individual. Field Length: 1 Values: 1 = White 2 = Black 3 = Oriental / Asian 4 = American Indian 5 = Hispanic 6 = Native Hawaiian or other Pacific Islander 8 = Other 9 = Unknown/Not STATED in medical record 110 2g. Marital Status (MARISTAT) Description: Current marital status of the individual. Field length: 1 Values: 1 = Married (includes separated, stated common law marriages) 2 = Widowed 3 = Single (includes divorced) 5 = Minor child (<18, unless stated as married) 9 = Unknown/Not STATED in medical record 2h. Individual’s Street Address (INDADDR) Description: The street address of residence of the individual. Field length: 40 Values: Any street address Note: Be sure to get apartment and route numbers if applicable. Note: Enter 999 in first column if not STATED in medical record. 2i. Individual’s City (INDCITY) Description: The city of residence of the individual. Field length: 20 Values: Any city name Note: Enter 999 in first column if not STATED in medical record. 2j. Individual’s State (INDSTATE) Description: The state of residence of the individual. Field length: 2 Values: SC = South Carolina NC = North Carolina GA = Georgia OT = Other UK = Unknown/Not STATED in medical record 2k. Individual’s Zip Code (INDZIP) Description: The zip code of residence of the individual. Field length: 5 Values: Five-digit number Note: Enter 999 in first column if not STATED in medical record. 2l. Employment Status (EMPSTAT) Description: Employment status of the individual. Field length: 1 Values: 1 = Student 2 = Employed (full time or part time) 3 = Employed & attending school 4 = Retired 5 = Disabled 111 6 = Unemployed 7 = Minor Child under age 5 8 = None of the above 9 = Not STATED in medical record 2m. Family Structure (FAMSTRU) Description: Family structure of individual Field length: 2 Values: 1 = Lives alone 2 = Lives with spouse/significant other 3 = Lives with parent(s) 4 = Lives with relative other than parent 5 = Lives with foster parent 6 = Lives with hired caregiver 7 = Lives in rehabilitation facility 8 = Lives in nursing home 10 = Lives in state residential facility 11 = Lives in group home 12 = Lives in detention or corrections facility 13 = Other 9 = Not STATED in medical record 2n. Insurance Status (INSSTA) Description: Insurance of individual Field length: 1 Values: 1 = no medical insurance/self-pay 2 = Medicaid (includes Select Health, a SC Medicaid HMO) 3 = Medicare 4 = HMO 5 = Other private insurance 6 = Champus, TRI Care, or VA 9 = not STATED in medical record SECTION 3: DIAGNOSIS & SEIZURE SPECIFICS Enter ‘999’ in text fields if no information in medical record. 3a. Date of most current seizure (SZDATE) Description: Date of most recent seizure (in relation to Outpatient Visit Date from random sample information) Field Length: 1 Values: 0 = prior to 2001 1 = 2001 2 = 2002 9 = not STATED in medical record 3b-g. Narrative of most current seizure (NTSEIZ) 112 Description: Field Length: Values: possible. Description of the most current seizure(s) (what happened to patient before, during, and after this episode, or during typical episode) 40 for each of three lines Any description of the seizure episode(s). Please be as descriptive as 3h. Multiple seizure types (MULTTYPE) Description: How many types of seizures has the individual experienced? Field Length: 1 Values: 1 = One 2 = Two 3 = More than two 9 = Not STATED in medical record. 3i. Seizure Type (SZTYPE1) Description: Clinical determination by MD of most current seizure type(s), seizure disorder, &/or epilepsy syndrome Field Length: Checklist of seizure types Values: Check off seizure type if listed in medical record describing this episode, or typical episode. 3j. Seizure Type (SZTYPE2) Description: Clinical determination by MD of most current seizure type(s), seizure disorder, &/or epilepsy syndrome Field Length: Checklist of seizure types Values: Check off seizure type if listed in medical record describing this episode, or typical episode. 3k. Seizure Type (SZTYPE3) Description: Clinical determination by MD of most current seizure type(s), seizure disorder, &/or epilepsy syndrome Field Length: Checklist of seizure types Values: Check off seizure type if listed in medical record describing this episode, or typical episode. 3l. Other seizure type(s) (SZTYPE4) Description: Other clinical determination by MD of current seizure type. Field Length: 40 Values: Any additional seizure type(s) or epilepsy not included on above lists (refer to list provided in class) 3m. Number of seizures (NUMSZ) Description: How many seizures did the individual experience during most recent episode of seizures? Field Length: 1 113 Values: 1 = One seizure 2 = Two seizures 3 = More than two seizures 9 = Not STATED in medical record 3n. Time of seizure onset (SZTIME) Description: Time of onset of most recent seizure (if cluster of seizures use time of onset of cluster). Field Length: 1 Values: 1 = Morning (7 am to 11:59 am) 2 = Afternoon (noon to 5:59 pm) 3 = Evening (6 pm to 9:59 pm) 4 = Nighttime (10 pm to 6:59 am) 9 = Not STATED in medical record Note: May be estimated if approximate time mentioned in medical record. 3o. Length of seizure (LENGTH) Description: How long did the most recent seizure last? Field Length: 1 Values: 1 = 30 seconds or less 2 = >30 secs & up to 2 minutes 3 = > 2 minutes & up to 5 minutes 4 = More than 5 minutes 9 = Not STATED in medical record SECTION 4: CAUSE OF SEIZURE Enter ‘999’ in text fields if no information in medical record. 4a-c. Narrative of Cause of most recent Seizure (CAUSE) Description: Brief description of recent circumstances that might have contributed to the most recent seizure (ie, what provoked this seizure?). Field Length: 40 characters for each of three lines Values: Any narrative that describes a possible current cause of seizure Note: Be as specific as possible. 4d. Injury (INJURY) Description: Is there report that the individual’s most recent seizure occurred due to a recent injury? Field Length: 1 Values: 1 = Yes, head injury 2 = Yes, injury other than to head 3 = Yes, injury to both head and elsewhere 4 = No 9 = Not STATED in medical record 4e. Illness prior to most recent seizure (ILLNESS) 114 Description: Field Length: Values: Is there report that the individual was ill prior to their most recent seizure? 1 1 = Yes 2 = No 9 = Not STATED in medical record 4f. Fever prior to most recent seizure (FEVER) Description: Is there report that the individual had an elevated temperature prior to their most recent seizure? Field Length: 1 Values: 1 = Yes, stated in record but no temperature recorded 2 = Yes, 99.6 to 101.5 F degrees (37.6 to 38.6 C) 3 = Yes, greater than 101.5 F degrees (> 38.6 C) 4 = No 9 = Not STATED in medical record 4g. Sleep deprivation (SLEEP) Description: Is there report that the individual experienced sleep deprivation prior to their more recent seizure? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 4h. Pregnancy (PREG) Description: Was the individual pregnant during their most recent seizure? Field Length: 1 Values: 1 = Yes, record states female was pregnant 2 = Teen or adult female was not pregnant (skip to 4j) 9 = Not STATED in medical record (skip to 4j) 4i. Eclampsia (ECLAMP) Description: Did the individual have eclampsia (or pre-eclampsia, pregnancy-induced hypertension, PIH, toxemia)? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 4j. Alcohol use (ETOHUSE) Description: Is there report that the individual was using alcohol prior to the most recent seizure? Field Length: 1 Values: 1 = Yes 115 2 = No 9 = Not STATED in medical record 4k. Drug use (DRUGUSE) Description: Is there report that the individual was using illicit drugs prior to the most recent seizure? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 4l. First seizure (SZFIRST) Description: Is the most recent seizure the individual’s first seizure? Field Length: 1 Values: 1 = Yes (skip to 5l) 2 = No 9 = Not STATED in medical record 4m. Medication change (MEDCHNG) Description: Is there report that the individual had recently changed anti-epileptic medications before their most recent seizure? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 4n. Weight gain in children (WTGAIN) Description: Is there report that the child had gained weight recently prior to their most recent seizure? Field Length: 1 Values: 1 = Yes, pt aged 1-18 years had gained weight recently 2 = No, pt aged 1-18 years had not gained weight recently 8 = Not applicable (individual 18 years or older) 9 = Aged 1-18 years, but weight change not STATED in medical record 4o. Noncompliance with medication (NONCOMP) Description: Is there report that the individual had recently not been compliant with his/her seizure medication regimen prior to the most recent seizure? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record SECTION 5: MEDICAL HISTORY Note: Enter “YES” if the record states previous history for the following. 116 5a. Enter “NO” if the record states NO previous history for the following. Enter “999” if no previous history information is documented in the medical record. Onset of seizure disorder/epilepsy (SZONSET) Description: Year of original onset of seizure disorder or epilepsy. Field Length: 4 Values: ccyy Note: enter 1111 if not found in medical record. Can estimate date if an approximate date is mentioned in medical record. 5b. Frequency of seizures (FREQSEIZ) Description: In the year prior to the most recent visit (Outpatient Visit Date from random sample information) how many times did the individual experience any seizure activity? Field Length: 1 Values 1 = None in previous year 2 = 1-3 times 3 = 4-11 times 4 = 12-24 times, or once a month 5 = 25-51 times 6 = 52-104 times, or once a week 7 = 104-364 times 8 = 365 times or more, or once a day or more 9 = Not STATED in medical record 5c. Seizure treatments – neurosurgery (PREVSURG) Description: Has individual had neurosurgery for epilepsy (lobectomy, resection, hemispherectomy, corpus callosotomy, subpial transection? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 5d. Seizure treatments – vagal nerve stimulator (PREVVNS) Description: Has individual had a vagal nerve stimulator implanted for epilepsy? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 5e. Seizure treatments – ketogenic diet (PREVKETO) Description: Has individual been on a ketogenic diet for epilepsy? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 117 5f. Medication (PRESMED1) Description: Any anti-epileptic medication at most recent visit (Outpatient Visit Date from random sample information). Field Length: Checklist of anti-epileptic medications. Values: Check off medication patient taking at onset of visit. 5g. Medication (PRESMED2) Description: Any anti-epileptic medication at most recent visit (Outpatient Visit Date from random sample information). Field Length: Checklist of anti-epileptic medications. Values: Check off medication patient taking at onset of visit 5h. Medication (PRESMED3) Description: Any anti-epileptic medication at most recent visit (Outpatient Visit Date from random sample information). Field Length: Checklist of anti-epileptic medications. Values: Check off medication patient taking at onset of visit 5i. Medication (PRESMED4) Description: Any anti-epileptic medication at most recent visit (Outpatient Visit Date from random sample information). Field Length: Checklist of anti-epileptic medications. Values: Check off medication patient taking at onset of visit 5j. Previous medication (PREVMED) Description: If individual not on anti-epileptic medication at most recent visit, has individual ever been on anti-epileptic medication? Field Length: 1 Values: 1 = Yes, took anti-epileptic medication in the past 2 = No evidence in chart that individual has ever taken anti-epileptic medication (skip to 5l) 3 = N/A, individual taking anti-epileptic medication at most recent visit 9 = Not STATED in medical record 5k. History of medication noncompliance (PREVNONC) Description: Does the individual have a history noncompliance? Field Length: 1 Values: 1 = Yes, skipping doses 2 = Yes, stopping medication 3 = Yes, but not specified 4 = No 9 = Not STATED in medical record Note: If ‘1’ and ‘2’, choose ‘2’. 118 of epilepsy medication 5l. History of TBI (PREVTBI) Description: Does the patient’s medical history include a TBI? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 5m. History of Alcohol Misuse (PREVALC) Description: Does the patient’s medical history include alcohol misuse? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 5n. History of Illicit Drug Use (PREVDRUG) Description: Does the patient’s medical history include illicit drug use? Field Length: 1 Values: 1 = Yes 2 = No 9 = Not STATED in medical record 5o. Previous illness or injury (ETIOLOGY) Description: Any past illness, condition, or injury in the patient’s history that initially caused the seizure disorder/epilepsy. If there is mention of any of the following terms, please make note under ‘etiology’: CRYPTOGENIC, IDIOPATHIC, FAMILIAL, SYMPTOMATIC, REFLEX, PRIMARY, or SECONDARY. Field Length: 40 Values: Previous illness, condition, or injury that initially caused seizures/epilepsy. 5p-s. Co-morbidities (COMORBID) Description: Any other medical &/or psychological conditions at most recent visit (Outpatient Visit Date from random sample information). Field Length: 40 for each of 4 lines Values: List all other current medical conditions. 5t. Family history of seizures (FAMHX) Description: Family history of seizures, seizure disorders, or epilepsy. Field Length: 1 Values: 1 = Yes 2 = No (skip to section 6) 9 = Not STATED in medical record (skip to section 6) 119 5u. Specifics of seizure family history (SPECFMHX) Description: What type of seizure, etc., is mentioned in family history? Field Length: 40 Values: Write whether family history of seizures, seizure disorder, or epilepsy, and what type, as well as family member(s) with disorder. SECTION 6: DIAGNOSTIC TESTS A physician’s interpretation, such as in the H&P or notes, is preferable to determining results from departmental print-out. 6a. EEG (PREVEEG) Description: Has individual had an EEG? Field Length: 1 Values: 1 = Yes, but no comment as to results in chart 2 = Yes, normal 3 = Yes, abnormal 9 = Not in medical record 6b. Video EEG (PREVVEEG) Description: Has individual had a video EEG? Field Length: 1 Values: 1 = Yes, but no comment as to results in chart 2 = Yes, normal 3 = Yes, abnormal 9 = Not STATED in medical record 6c. CT scan of the head (PREVCT) Description: Has individual had a CT scan of the head? Field Length: 1 Values: 1 = Yes, but no comment as to results in chart 2 = Yes, normal 3 = Yes, abnormal 9 = Not STATED in medical record 6d. MRI of the head (PREVMRI) Description: Has individual had a MRI of the head? Field Length: 1 Values: 1 = Yes, but no comment as to results in chart 2 = Yes, normal 3 = Yes, abnormal 9 = Not STATED in medical record 6e. Anti-epileptic drug level 1 (AED1) Description: Most recent blood drug level of an anti-epileptic drug and results Field Length: 1 Values: 1 = Test ordered, no results/interpretation in chart 120 2 = Test done, and within therapeutic range 3 = Test done, and below therapeutic range 4 = Test done, and above therapeutic range 9 = Not STATED in medical record 6f. Anti-epileptic drug level 2 (AED2) Description: Most recent blood drug level of a second anti-epileptic drug and results Field Length: 1 Values: 1 = Test ordered, no results/interpretation in chart 2 = Test done, and within therapeutic range 3 = Test done, and below therapeutic range 4 = Test done, and above therapeutic range 9 = Not STATED in medical record SECTION 7 CLOSING 7a. Abstractor Comments (COMMENTS) Description: Does the abstractor have any comments about this record? Field length: 1 Values: 1 = Yes 2 = No (skip to7g) 7b-c. Abstracter Administrative Comments (COMMENT1-2) Description: Informative remarks made by the data abstracter to add additional insight to this particular patient record. Field Length: 40 for each of 2 lines Values: Any narrative (administrative problems – random sample info different from chart, chart difficult to read, etc.) 7d-f. Abstractor Clinical Comments (COMMENT3-5) Description: Informative remarks made by the data abstracter to add additional insight to this particular patient record. Field Length: 40 for each of 3 lines Values: Any narrative (clinically-related problems – need to add something extra for which there was no entry place, note that question mark written in text was copied from chart, note that question mark written in text due to illegible chart, etc.) 7g. Flagged for Special Review (FLAG) Description: Does this record need to be flagged for special review? Field length: 1 Values: 1 = Yes 2 = No Note: A record would need to be reviewed by another abstractor if there is contradictory information in the chart (ex. a pregnant male), information in medical record not matching random sample or other problems. Note: Reason(s) for flag should be recorded in Abstractor Comments. 121 Appendix C - Comparison of inpatient/ED sample to population Using population database from ORS, with sampled cases marked, and using abstracted values of sex, age group, payer, and diagnosis for sample group (using race assigned by ORS for sample since more complete). ‘ABS’ = ABSTRACTED CASES (IE, SAMPLE); ‘POP’ = WHOLE POPULATION OF CASES WITH SEIZURE-RELATED DIAGNOSES ‘CNT’ = COUNT; ‘PCNT’ = PERCENT; ‘LL’ = LOWER 95% CI; ‘UL’ = UPPER 95% CI IF >1 DIAGNOSIS, POP. DIAGNOSIS ASSIGNED ACCORDING TO FOLLOWING HIERARCHY: 345.X > 780.3 > 780.2> 293.0. 2001, 345.X (n=765, n=1731) SEX_ABS_ SEX_ABS_ SEX CNT PCNT F 388 50.72 M 377 49.28 RACE_ ABS_CNT 315 12 438 RACE BLACK OTH/UNK WHITE PAYER INSURANCE MEDICAID MEDICARE NOT STATED UNINSURED AGEGRP 0-9 10-19 20-29 30-39 40-49 50-59 60-69 70-79 80-89 90+ RACE_ ABS_PCNT 41.18 1.57 57.25 PAYER_ ABS_CNT 196 210 260 3 96 AGEGRP_ ABS_CNT 88 49 87 125 125 109 93 56 30 3 AGEGRP_ ABS_PCNT 11.50 6.41 11.37 16.34 16.34 14.25 12.16 7.32 3.92 0.39 RACE_ ABS_CNT 819 26 920 RACE_ ABS_PCNT 46.40 1.47 52.12 SEX_ ABS_UL 57.26 55.91 RACE_ ABS_LL 34.03 0.00 51.17 PAYER_ ABS_PCNT 25.62 27.45 33.99 0.39 12.55 2001, 7803 (n=1765, N=18,685) SEX_ABS_ SEX_ABS_ SEX CNT PCNT F 839 47.54 M 926 52.46 U . . RACE BLACK OTH/UNK WHITE SEX_ ABS_LL 44.18 42.65 RACE_ ABS_UL 48.32 10.81 63.34 PAYER_ ABS_LL 17.59 19.52 26.42 0.00 3.84 AGEGRP_ ABS_LL 2.74 0.00 2.61 7.82 7.82 5.62 3.43 0.00 0.00 0.00 SEX_ ABS_LL 43.09 48.24 . RACE_ ABS_LL 41.91 0.00 47.88 SEX_POP_ CNT 862 869 RACE_ POP_CNT 734 33 964 PAYER_ ABS_UL 33.65 35.38 41.55 9.69 21.26 AGEGRP_ ABS_UL 20.26 15.41 20.14 24.86 24.86 22.87 20.89 16.29 13.05 9.69 SEX_ ABS_UL 51.98 56.69 . RACE_ ABS_UL 50.89 7.56 56.37 SEX_POP_ PCNT 49.80 50.20 SEX_POP_ CNT 8935 9749 1 RACE_ POP_CNT 8004 356 10325 122 RACE_ POP_PCNT 42.40 1.91 55.69 PAYER_ POP_CNT 451 467 571 5 237 AGEGRP_ POP_CNT 209 142 192 266 306 236 175 134 61 10 SEX_ POP_LL 45.41 45.83 RACE_ POP_LL 37.70 0.00 51.57 PAYER_ POP_PCNT 26.05 26.98 32.99 0.29 13.69 AGEGRP_ POP_PCNT 12.07 8.20 11.09 15.37 17.68 13.63 10.11 7.74 3.52 0.58 SEX_POP_ PCNT 47.82 52.18 0.01 RACE_ POP_PCNT 42.84 1.91 55.26 SEX_ POP_UL 54.18 54.57 RACE_ POP_UL 47.10 8.04 59.81 PAYER_ POP_LL 20.73 21.69 27.92 0.00 7.94 AGEGRP_ POP_LL 6.27 2.27 5.25 9.67 12.06 7.88 4.24 1.79 0.00 0.00 SEX_ POP_LL 46.46 50.87 0.00 RACE_ POP_LL 41.41 0.04 54.00 PAYER_ POP_UL 31.38 32.27 38.05 6.47 19.44 AGEGRP_ POP_UL 17.88 14.14 16.93 21.06 23.29 19.39 15.98 13.69 9.61 6.75 SEX_ POP_UL 49.18 53.48 1.89 RACE_ POP_UL 44.26 3.77 56.52 PAYER INSURANCE MEDICAID MEDICARE NOT STATED UNINSURED AGEGRP 0-9 10-19 20-29 30-39 40-49 50-59 60-69 70-79 80-89 90+ PAYER_ ABS_CNT 404 397 704 7 253 AGEGRP_ ABS_CNT 207 113 161 215 275 253 194 210 115 22 PAYER_ ABS_PCNT 22.89 22.49 39.89 0.40 14.33 AGEGRP_ ABS_PCNT 11.73 6.40 9.12 12.18 15.58 14.33 10.99 11.90 6.52 1.25 2002, 345X (n=427, N=1656) SEX_ABS_ SEX_ABS_ SEX CNT PCNT F 220 51.52 M 207 48.48 RACE_ ABS_CNT 197 10 220 RACE BLACK OTH/UNK WHITE PAYER INSURANCE MEDICAID MEDICARE NOT STATED UNINSURED AGEGRP 0-9 10-19 20-29 30-39 40-49 50-59 60-69 70-79 80-89 90+ RACE_ ABS_PCNT 46.14 2.34 51.52 PAYER_ ABS_CNT 117 147 117 . 46 AGEGRP_ ABS_CNT 62 46 49 69 68 63 30 29 10 1 AGEGRP_ ABS_LL 5.97 0.47 3.28 6.44 9.95 8.66 5.21 6.14 0.59 0.00 SEX_ ABS_LL 42.84 39.53 RACE_ ABS_LL 36.99 0.00 42.84 PAYER_ ABS_PCNT 27.40 34.43 27.40 . 10.77 AGEGRP_ ABS_PCNT 14.52 10.77 11.48 16.16 15.93 14.75 7.03 6.79 2.34 0.23 PAYER_ ABS_LL 17.51 17.10 35.13 0.00 8.66 AGEGRP_ ABS_UL 17.49 12.33 14.97 17.93 21.21 20.01 16.78 17.65 12.44 7.34 SEX_ ABS_UL 60.20 57.43 PAYER_ POP_CNT 4548 4284 6936 91 2826 AGEGRP_ POP_CNT 2303 1259 1764 2303 3086 2683 1960 1903 1187 237 SEX_POP_ CNT 834 822 RACE_ ABS_UL 55.28 14.66 60.20 PAYER_ ABS_LL 16.78 24.33 16.78 . 0.00 AGEGRP_ ABS_LL 3.00 0.00 0.00 4.75 4.50 3.25 0.00 0.00 0.00 0.00 PAYER_ ABS_UL 28.27 27.89 44.64 6.52 20.01 AGEGRP_ POP_PCNT 12.33 6.74 9.44 12.33 16.52 14.36 10.49 10.18 6.35 1.27 SEX_POP_ PCNT 50.36 49.64 RACE_ POP_CNT 679 35 942 PAYER_ ABS_UL 38.02 44.52 38.02 . 22.55 AGEGRP_ ABS_UL 26.04 22.55 23.20 27.57 27.35 26.26 19.05 18.83 14.66 12.68 123 PAYER_ POP_CNT 460 489 489 1 217 AGEGRP_ POP_CNT 226 157 216 243 265 233 138 125 42 11 PAYER_ POP_PCNT 24.34 22.93 37.12 0.49 15.12 AGEGRP_ POP_LL 10.56 4.92 7.65 10.56 14.79 12.62 8.71 8.40 4.53 0.00 SEX_ POP_LL 45.90 45.15 RACE_ POP_PCNT 41.00 2.11 56.88 PAYER_ POP_PCNT 27.78 29.53 29.53 0.06 13.10 AGEGRP_ POP_PCNT 13.65 9.48 13.04 14.67 16.00 14.07 8.33 7.55 2.54 0.66 PAYER_ POP_LL 22.70 21.27 35.63 0.00 13.39 RACE_ POP_LL 36.14 0.00 52.73 AGEGRP_ POP_UL 14.09 8.56 11.23 14.09 18.24 16.10 12.27 11.97 8.18 3.14 SEX_ POP_UL 54.82 54.13 RACE_ POP_UL 45.86 8.38 61.04 PAYER_ POP_LL 22.40 24.22 24.22 0.00 7.20 AGEGRP_ POP_LL 7.77 3.46 7.14 8.83 10.20 8.20 2.27 1.46 0.00 0.00 PAYER_ POP_UL 25.98 24.58 38.61 2.37 16.86 PAYER_ POP_UL 33.16 34.84 34.84 6.39 19.00 AGEGRP_ POP_UL 19.53 15.50 18.95 20.52 21.80 19.94 14.39 13.63 8.79 6.97 2002, 7803 (n=1026, N=19,511) SEX_ABS_ SEX_ABS_ SEX CNT PCNT F 463 45.13 M 563 54.87 RACE BLACK OTH/UNK WHITE RACE_ ABS_CNT 476 18 532 PAYER INSURANCE MEDICAID MEDICARE NOT STATED UNINSURED 174 AGEGRP 0-9 10-19 20-29 30-39 40-49 50-59 60-69 70-79 80-89 90+ RACE_ ABS_PCNT 46.39 1.75 51.85 SEX_ ABS_LL 39.17 49.47 RACE_ ABS_LL 40.51 0.00 46.27 PAYER_ PAYER_ ABS_CNT ABS_PCNT 186 18.13 248 24.17 411 40.06 7 0.68 16.96 9.63 AGEGRP_ ABS_CNT 90 57 113 131 190 160 111 92 70 12 AGEGRP_ ABS_PCNT 8.77 5.56 11.01 12.77 18.52 15.59 10.82 8.97 6.82 1.17 SEX_ ABS_UL 51.08 60.28 RACE_ ABS_UL 52.28 9.73 57.43 PAYER_ ABS_LL 10.85 17.17 33.83 0.00 24.29 AGEGRP_ ABS_LL 1.09 0.00 3.43 5.26 11.26 8.21 3.22 1.29 0.00 0.00 SEX_POP_ CNT 9261 10250 RACE_ POP_CNT 8198 381 10932 SEX_POP_ PCNT 47.47 52.53 RACE_ POP_PCNT 42.02 1.95 56.03 SEX_ POP_LL 46.13 51.26 RACE_ POP_LL 40.61 0.13 54.81 PAYER_ PAYER_ PAYER_ ABS_UL POP_CNT POP_PCNT 25.40 4505 23.09 31.17 4772 24.46 46.28 7208 36.94 8.70 123 0.63 2903 14.88 13.18 AGEGRP_ ABS_UL 16.45 13.37 18.60 20.28 25.78 22.98 18.41 16.64 14.59 9.16 124 AGEGRP_ POP_CNT 2464 1287 1797 2439 3295 2803 2048 1903 1220 255 AGEGRP_ POP_PCNT 12.63 6.60 9.21 12.50 16.89 14.37 10.50 9.75 6.25 1.31 SEX_ POP_UL 48.80 53.80 RACE_ POP_UL 43.42 3.78 57.25 PAYER_ POP_LL 21.47 22.86 35.48 0.00 16.58 AGEGRP_ POP_LL 10.91 4.81 7.45 10.78 15.21 12.66 8.75 8.00 4.47 0.00 PAYER_ POP_UL 24.71 26.06 38.41 2.47 AGEGRP_ POP_UL 14.35 8.38 10.97 14.23 18.57 16.07 12.24 11.51 8.04 3.14 Appendix D - Comparison of physician office visit sample to population Using population database from ORS, combining years and payer groups, with sampled cases marked, and using population values of sex, age group, payer, and diagnosis for both sample and population group. 345X (n=174, N=1157) SEX F M SEX_ABS_ CNT 89 85 RACE BLACK OTH/UNK WHITE PAYER COMM-SHP M/CAID M/CARE SEX_ABS_ PCNT 51.15 48.85 RACE_ ABS_CNT 44 55 75 RACE_ ABS_PCNT 25.29 31.61 43.10 PAYER_ ABS_CNT 48 67 59 AGEGRP_ ABS_CNT 37 6 13 31 32 27 12 13 3 AGEGRP 10-19 2-9 20-29 30-39 40-49 50-59 60-69 70-79 80+ SEX_ABS_ CNT 34 38 RACE BLACK OTH/UNK WHITE PAYER COMM-SHP M/CAID M/CARE RACE_ ABS_LL 8.41 15.46 28.37 PAYER_ ABS_PCNT 27.59 38.51 33.91 AGEGRP_ ABS_PCNT 21.26 3.45 7.47 17.82 18.39 15.52 6.90 7.47 1.72 7803 (n=72, N=2096) SEX F M SEX_ ABS_LL 37.50 34.88 SEX_ABS_ PCNT 47.22 52.78 RACE_ ABS_CNT 7 31 34 PAYER_ ABS_CNT 31 23 18 PAYER_ ABS_LL 10.97 23.19 18.03 AGEGRP_ ABS_LL 3.94 0.00 0.00 0.11 0.75 0.00 0.00 0.00 0.00 SEX_ ABS_LL 25.17 31.92 RACE_ ABS_PCNT 9.72 43.06 47.22 PAYER_ ABS_PCNT 43.06 31.94 25.00 RACE_ ABS_LL 0.00 20.15 25.17 PAYER_ ABS_LL 20.15 6.90 0.00 SEX_ ABS_UL 64.80 62.82 SEX_POP_ CNT 587 570 RACE_ ABS_UL 42.17 47.76 57.83 RACE_ POP_CNT 238 476 443 SEX_POP_ PCNT 50.73 49.27 RACE_ POP_PCNT 20.57 41.14 38.29 SEX_ POP_LL 45.42 43.87 SEX_ POP_UL 56.05 54.66 RACE_ POP_LL 13.82 35.33 32.34 RACE_ POP_UL 27.32 46.95 44.24 PAYER_ ABS_UL 44.20 53.82 49.78 PAYER_ POP_CNT 409 528 220 PAYER_ POP_PCNT 35.35 45.64 19.01 PAYER_ POP_LL 29.26 40.05 12.20 AGEGRP_ ABS_UL 38.59 22.64 26.25 35.52 36.03 33.47 25.74 26.25 21.08 AGEGRP_ POP_CNT 232 86 159 174 185 163 87 47 24 AGEGRP_ POP_PCNT 20.05 7.43 13.74 15.04 15.99 14.09 7.52 4.06 2.07 AGEGRP_ POP_LL 13.28 0.15 6.71 8.06 9.05 7.07 0.24 0.00 0.00 SEX_ ABS_UL 69.28 73.64 SEX_POP_ CNT 1060 1036 RACE_ ABS_UL 38.57 65.96 69.28 RACE_ POP_CNT 606 715 775 PAYER_ ABS_UL 65.96 56.99 51.29 125 PAYER_ POP_CNT 611 1151 334 SEX_POP_ PCNT 50.57 49.43 RACE_ POP_PCNT 28.91 34.11 36.98 PAYER_ POP_PCNT 29.15 54.91 15.94 SEX_ POP_LL 46.62 45.43 RACE_ POP_LL 24.17 29.55 32.51 PAYER_ POP_LL 24.42 51.14 10.78 PAYER_ POP_UL 41.44 51.22 25.83 AGEGRP_ POP_UL 26.82 14.72 20.78 22.02 22.93 21.11 14.80 11.48 9.57 SEX_ POP_UL 54.53 53.43 RACE_ POP_UL 33.66 38.68 41.44 PAYER_ POP_UL 33.89 58.69 21.09 AGEGRP 2-9 10-19 20-29 30-39 40-49 50-59 60-69 70-79 80+ AGEGRP_ ABS_CNT . 5 6 15 14 11 12 8 1 AGEGRP_ ABS_PCNT . 6.94 8.33 20.83 19.44 15.28 16.67 11.11 1.39 AGEGRP_ ABS_LL . 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 AGEGRP_ ABS_UL . 36.23 37.40 47.84 46.69 43.22 44.38 39.73 31.53 126 AGEGRP_ POP_CNT 155 244 258 314 351 339 233 123 79 AGEGRP_ POP_PCNT 7.40 11.64 12.31 14.98 16.75 16.17 11.12 5.87 3.77 AGEGRP_ POP_LL 1.98 6.35 7.04 9.79 11.61 11.02 5.81 0.41 0.00 AGEGRP_ POP_UL 12.81 16.93 17.58 20.17 21.88 21.32 16.42 11.33 9.29 Appendix E. Likelihood of an epilepsy diagnosis conditional on satisfying the model parameters (n=Random Sample 87 records) Obs id1 agegp racegp comorbgp VISIT dxgp payer beta0 beta11 beta12 beta13 beta14 beta15 beta16 38 955337687 10-19 BLACK 1)hi-risk 2-5 345X MEDICAID -1.0658 1.9364 0.0000 0.0000 0.0000 0.0000 0 207 959037307 10-19 WHITE 1)hi-risk 2-5 7803 MEDICAID -1.0658 1.9364 0.0000 0.0000 0.0000 0.0000 0 678 955780301 20-39 WHITE 1)hi-risk 2-5 345X MEDICARE -1.0658 0.0000 2.3996 0.0000 0.0000 0.0000 0 1478 950281255 0-9 WHITE 1)hi-risk 1 345X MEDICAID -1.0658 0.0000 0.0000 0.0000 0.0000 0.0000 0 2323 959696611 10-19 WHITE 2)lo-risk 1 7802+ MEDICAID -1.0658 1.9364 0.0000 0.0000 0.0000 0.0000 0 2877 955165719 10-19 BLACK 2)lo-risk 1 7803 MEDICAID -1.0658 1.9364 0.0000 0.0000 0.0000 0.0000 0 3462 959343277 60-79 BLACK 1)hi-risk 2-5 7802+ MEDICARE -1.0658 0.0000 0.0000 0.0000 1.8278 0.0000 0 3944 955331223 10-19 BLACK 2)lo-risk 1 7802+ MEDICARE -1.0658 1.9364 0.0000 0.0000 0.0000 0.0000 0 5495 955387868 60-79 BLACK 2)lo-risk 2-5 7803 MEDICAID -1.0658 0.0000 0.0000 0.0000 1.8278 0.0000 0 6085 959979595 20-39 BLACK 2)lo-risk 2-5 7803 MEDICAID -1.0658 0.0000 2.3996 0.0000 0.0000 0.0000 0 7885 959298518 10-19 BLACK 2)lo-risk 1 7803 INSURANCE -1.0658 1.9364 0.0000 0.0000 0.0000 0.0000 0 9761 955829918 20-39 WHITE 2)lo-risk 1 345X INSURANCE -1.0658 0.0000 2.3996 0.0000 0.0000 0.0000 0 11066 950358501 20-39 BLACK 2)lo-risk 1 7802+ MEDICAID -1.0658 0.0000 2.3996 0.0000 0.0000 0.0000 0 11527 955739597 40-59 BLACK 2)lo-risk 1 7802+ MEDICAID -1.0658 0.0000 0.0000 2.1972 0.0000 0.0000 0 11587 955861570 20-39 BLACK 2)lo-risk 1 7802+ MEDICAID -1.0658 0.0000 2.3996 0.0000 0.0000 0.0000 0 12246 955868655 10-19 WHITE 2)lo-risk 1 7803 INSURANCE -1.0658 1.9364 0.0000 0.0000 0.0000 0.0000 0 14522 959548081 20-39 WHITE 1)hi-risk 1 7803 MEDICAID -1.0658 0.0000 2.3996 0.0000 0.0000 0.0000 0 14964 954090312 60-79 BLACK 2)lo-risk 2-5 7803 MEDICARE -1.0658 0.0000 0.0000 0.0000 1.8278 0.0000 0 15886 959043134 40-59 BLACK 2)lo-risk 2-5 7803 MEDICARE -1.0658 0.0000 0.0000 2.1972 0.0000 0.0000 0 16890 950895712 20-39 WHITE 2)lo-risk 1 7802+ MEDICAID -1.0658 0.0000 2.3996 0.0000 0.0000 0.0000 0 17260 955901033 20-39 WHITE 2)lo-risk 1 7802+ MEDICAID -1.0658 0.0000 2.3996 0.0000 0.0000 0.0000 0 17531 959560662 40-59 WHITE 2)lo-risk 1 7802+ MEDICAID -1.0658 0.0000 0.0000 2.1972 0.0000 0.0000 0 17647 950781243 40-59 WHITE 1)hi-risk 1 7802+ INSURANCE -1.0658 0.0000 0.0000 2.1972 0.0000 0.0000 0 18596 959678822 60-79 WHITE 1)hi-risk 1 7802+ MEDICARE -1.0658 0.0000 0.0000 0.0000 1.8278 0.0000 0 18823 955235671 10-19 WHITE 2)lo-risk 1 7803 UNINSURED -1.0658 1.9364 0.0000 0.0000 0.0000 0.0000 0 19011 955879589 40-59 BLACK 2)lo-risk 1 7803 MEDICAID -1.0658 0.0000 0.0000 2.1972 0.0000 0.0000 0 127 Appendix E. Likelihood of an epilepsy diagnosis conditional on satisfying the model parameters (n=Random Sample 87 records) Obs id1 agegp racegp comorbgp VISIT dxgp payer beta0 beta11 beta12 beta13 beta14 beta15 beta16 20199 955668458 40-59 BLACK 2)lo-risk 1 7803 MEDICAID -1.0658 0.0000 0.0000 2.1972 0.0000 0.0000 0 20669 959190026 40-59 BLACK 2)lo-risk 1 7803 MEDICAID -1.0658 0.0000 0.0000 2.1972 0.0000 0.0000 0 21256 952417855 60-79 WHITE 2)lo-risk 2-5 7803 MEDICARE -1.0658 0.0000 0.0000 0.0000 1.8278 0.0000 0 23700 959634019 40-59 WHITE 2)lo-risk 2-5 7803 MEDICARE -1.0658 0.0000 0.0000 2.1972 0.0000 0.0000 0 24169 950222829 40-59 BLACK 1)hi-risk 1 7803 INSURANCE -1.0658 0.0000 0.0000 2.1972 0.0000 0.0000 0 25195 955715653 40-59 BLACK 2)lo-risk 2-5 7803 UNINSURED -1.0658 0.0000 0.0000 2.1972 0.0000 0.0000 0 26980 954290003 40-59 BLACK 2)lo-risk 1 7802+ INSURANCE -1.0658 0.0000 0.0000 2.1972 0.0000 0.0000 0 27630 954863693 40-59 BLACK 2)lo-risk 1 7802+ INSURANCE -1.0658 0.0000 0.0000 2.1972 0.0000 0.0000 0 28218 955302959 40-59 BLACK 2)lo-risk 1 7802+ INSURANCE -1.0658 0.0000 0.0000 2.1972 0.0000 0.0000 0 29065 955897618 80-89 BLACK 2)lo-risk 1 7802+ MEDICARE -1.0658 0.0000 0.0000 0.0000 0.0000 1.9839 0 29884 959378017 40-59 BLACK 2)lo-risk 1 7802+ INSURANCE -1.0658 0.0000 0.0000 2.1972 0.0000 0.0000 0 32917 954426231 60-79 WHITE 1)hi-risk 1 7803 MEDICARE -1.0658 0.0000 0.0000 0.0000 1.8278 0.0000 0 33715 959825560 60-79 WHITE 1)hi-risk 1 7803 MEDICARE -1.0658 0.0000 0.0000 0.0000 1.8278 0.0000 0 34359 959975360 20-39 WHITE 2)lo-risk 2-5 7803 UNINSURED -1.0658 0.0000 2.3996 0.0000 0.0000 0.0000 0 35955 952252831 60-79 WHITE 2)lo-risk 1 7802+ MEDICARE -1.0658 0.0000 0.0000 0.0000 1.8278 0.0000 0 36432 952468497 80-89 WHITE 2)lo-risk 1 7802+ MEDICARE -1.0658 0.0000 0.0000 0.0000 0.0000 1.9839 0 36745 952638795 80-89 WHITE 2)lo-risk 1 7802+ MEDICARE -1.0658 0.0000 0.0000 0.0000 0.0000 1.9839 0 37754 954012591 60-79 WHITE 2)lo-risk 1 7802+ MEDICARE -1.0658 0.0000 0.0000 0.0000 1.8278 0.0000 0 39099 954226089 80-89 WHITE 2)lo-risk 1 7802+ MEDICARE -1.0658 0.0000 0.0000 0.0000 0.0000 1.9839 0 39856 954266263 60-79 WHITE 2)lo-risk 1 7802+ MEDICARE -1.0658 0.0000 0.0000 0.0000 1.8278 0.0000 0 40258 954332524 20-39 WHITE 2)lo-risk 1 7802+ INSURANCE -1.0658 0.0000 2.3996 0.0000 0.0000 0.0000 0 40459 954403152 60-79 WHITE 2)lo-risk 1 7802+ INSURANCE -1.0658 0.0000 0.0000 0.0000 1.8278 0.0000 0 40498 954407233 80-89 WHITE 2)lo-risk 1 7802+ MEDICARE -1.0658 0.0000 0.0000 0.0000 0.0000 1.9839 0 40552 954413177 60-79 WHITE 2)lo-risk 1 7802+ MEDICARE -1.0658 0.0000 0.0000 0.0000 1.8278 0.0000 0 41808 954552065 80-89 WHITE 2)lo-risk 1 7802+ MEDICARE -1.0658 0.0000 0.0000 0.0000 0.0000 1.9839 0 42155 954591313 60-79 WHITE 2)lo-risk 1 7802+ MEDICARE -1.0658 0.0000 0.0000 0.0000 1.8278 0.0000 0 128 Appendix E. Likelihood of an epilepsy diagnosis conditional on satisfying the model parameters (n=Random Sample 87 records) Obs id1 agegp racegp comorbgp VISIT dxgp payer beta0 beta11 beta12 beta13 beta14 beta15 beta16 44838 955751246 80-89 WHITE 2)lo-risk 1 7802+ MEDICARE -1.0658 0.0000 0.0000 0.0000 0.0000 1.9839 0 45064 955874494 60-79 WHITE 2)lo-risk 1 7802+ MEDICARE -1.0658 0.0000 0.0000 0.0000 1.8278 0.0000 0 46455 959147161 40-59 WHITE 2)lo-risk 1 7802+ INSURANCE -1.0658 0.0000 0.0000 2.1972 0.0000 0.0000 0 47165 959418311 40-59 WHITE 2)lo-risk 1 7802+ MEDICARE -1.0658 0.0000 0.0000 2.1972 0.0000 0.0000 0 47202 959439666 80-89 WHITE 2)lo-risk 1 7802+ MEDICARE -1.0658 0.0000 0.0000 0.0000 0.0000 1.9839 0 47661 959676433 60-79 WHITE 2)lo-risk 1 7802+ INSURANCE -1.0658 0.0000 0.0000 0.0000 1.8278 0.0000 0 48159 959747803 80-89 WHITE 2)lo-risk 1 7802+ MEDICARE -1.0658 0.0000 0.0000 0.0000 0.0000 1.9839 0 48216 959753596 60-79 WHITE 2)lo-risk 1 7802+ MEDICARE -1.0658 0.0000 0.0000 0.0000 1.8278 0.0000 0 50337 954261897 40-59 BLACK 2)lo-risk 1 7802+ UNINSURED -1.0658 0.0000 0.0000 2.1972 0.0000 0.0000 0 50342 954267565 20-39 BLACK 2)lo-risk 1 7802+ UNINSURED -1.0658 0.0000 2.3996 0.0000 0.0000 0.0000 0 52030 952947189 40-59 BLACK 2)lo-risk 1 7803 INSURANCE -1.0658 0.0000 0.0000 2.1972 0.0000 0.0000 0 52041 954006697 40-59 BLACK 2)lo-risk 1 7803 INSURANCE -1.0658 0.0000 0.0000 2.1972 0.0000 0.0000 0 53490 955750418 60-79 BLACK 2)lo-risk 1 7803 MEDICARE -1.0658 0.0000 0.0000 0.0000 1.8278 0.0000 0 53595 955817050 60-79 BLACK 2)lo-risk 1 7803 MEDICARE -1.0658 0.0000 0.0000 0.0000 1.8278 0.0000 0 53772 955927239 0-9 BLACK 2)lo-risk 1 7803 INSURANCE -1.0658 0.0000 0.0000 0.0000 0.0000 0.0000 0 54153 959195727 40-59 BLACK 2)lo-risk 1 7803 MEDICARE -1.0658 0.0000 0.0000 2.1972 0.0000 0.0000 0 54723 959660436 40-59 BLACK 2)lo-risk 1 7803 MEDICARE -1.0658 0.0000 0.0000 2.1972 0.0000 0.0000 0 54764 959694110 60-79 BLACK 2)lo-risk 1 7803 MEDICARE -1.0658 0.0000 0.0000 0.0000 1.8278 0.0000 0 54956 959903570 40-59 BLACK 2)lo-risk 1 7803 MEDICARE -1.0658 0.0000 0.0000 2.1972 0.0000 0.0000 0 55628 950344027 20-39 WHITE 2)lo-risk 1 7802+ UNINSURED -1.0658 0.0000 2.3996 0.0000 0.0000 0.0000 0 55687 950697662 20-39 WHITE 2)lo-risk 1 7802+ UNINSURED -1.0658 0.0000 2.3996 0.0000 0.0000 0.0000 0 57018 956111353 40-59 WHITE 2)lo-risk 1 7802+ UNINSURED -1.0658 0.0000 0.0000 2.1972 0.0000 0.0000 0 57911 955563903 20-39 WHITE 2)lo-risk 1 7803 INSURANCE -1.0658 0.0000 2.3996 0.0000 0.0000 0.0000 0 59907 954251015 40-59 WHITE 2)lo-risk 1 7803 INSURANCE -1.0658 0.0000 0.0000 2.1972 0.0000 0.0000 0 60328 954432776 40-59 WHITE 2)lo-risk 1 7803 MEDICARE -1.0658 0.0000 0.0000 2.1972 0.0000 0.0000 0 129 Obs beta21 beta22 beta31 beta41 beta42 beta51 beta52 beta61 beta62 beta63 logits probit probit2 epilepgp3 38 0.4394 0.0000 0.8274 1.2732 0 0.0000 0.0000 0.9143 0.000 0.0000 4.3249 0.98694 98.69 1)Probable 207 0.0000 0.0000 0.8274 1.2732 0 0.0000 0.6168 0.9143 0.000 0.0000 3.2687 0.96334 96.33 1)Probable 678 0.0000 0.0000 0.8274 1.2732 0 1.5276 0.0000 0.0000 0.883 0.0000 2.5624 0.92840 92.84 1)Probable 1478 0.0000 0.0000 0.8274 0.0000 0 1.5276 0.0000 0.9143 0.000 0.0000 2.2035 0.90056 90.06 1)Probable 2323 0.0000 0.0000 0.0000 0.0000 0 0.0000 0.0000 0.9143 0.000 0.0000 1.7849 0.85630 85.63 1)Probable 2877 0.4394 0.0000 0.0000 0.0000 0 0.0000 0.6168 0.9143 0.000 0.0000 1.6075 0.83306 83.31 1)Probable 3462 0.4394 0.0000 0.8274 1.2732 0 0.0000 0.0000 0.0000 0.883 0.0000 1.4742 0.81369 81.37 1)Probable 3944 0.4394 0.0000 0.0000 0.0000 0 0.0000 0.0000 0.0000 0.883 0.0000 1.3100 0.78751 78.75 1)Probable 5495 0.4394 0.0000 0.0000 1.2732 0 0.0000 0.6168 0.9143 0.000 0.0000 0.9443 0.71997 72.00 1)Probable 6085 0.4394 0.0000 0.0000 1.2732 0 0.0000 0.6168 0.9143 0.000 0.0000 0.9443 0.71997 72.00 1)Probable 7885 0.4394 0.0000 0.0000 0.0000 0 0.0000 0.6168 0.0000 0.000 0.0000 0.6932 0.66668 66.67 1)Probable 9761 0.0000 0.0000 0.0000 0.0000 0 1.5276 0.0000 0.0000 0.000 0.0000 0.4618 0.61344 61.34 1)Probable 11066 0.4394 0.0000 0.0000 0.0000 0 0.0000 0.0000 0.9143 0.000 0.0000 0.2879 0.57148 57.15 1)Probable 11527 0.4394 0.0000 0.0000 0.0000 0 0.0000 0.0000 0.9143 0.000 0.0000 0.2879 0.57148 57.15 1)Probable 11587 0.4394 0.0000 0.0000 0.0000 0 0.0000 0.0000 0.9143 0.000 0.0000 0.2879 0.57148 57.15 1)Probable 12246 0.0000 0.0000 0.0000 0.0000 0 0.0000 0.6168 0.0000 0.000 0.0000 0.2538 0.56311 56.31 1)Probable 14522 0.0000 0.0000 0.8274 0.0000 0 0.0000 0.6168 0.9143 0.000 0.0000 0.0591 0.51477 51.48 1)Probable 14964 0.4394 0.0000 0.0000 1.2732 0 0.0000 0.6168 0.0000 0.883 0.0000 0.0300 0.50750 50.75 1)Probable 15886 0.4394 0.0000 0.0000 1.2732 0 0.0000 0.6168 0.0000 0.883 0.0000 0.0300 0.50750 50.75 1)Probable 16890 0.0000 0.0000 0.0000 0.0000 0 0.0000 0.0000 0.9143 0.000 0.0000 -0.1515 0.46220 46.22 1)Probable 17260 0.0000 0.0000 0.0000 0.0000 0 0.0000 0.0000 0.9143 0.000 0.0000 -0.1515 0.46220 46.22 1)Probable 17531 0.0000 0.0000 0.0000 0.0000 0 0.0000 0.0000 0.9143 0.000 0.0000 -0.1515 0.46220 46.22 1)Probable 17647 0.0000 0.0000 0.8274 0.0000 0 0.0000 0.0000 0.0000 0.000 0.0000 -0.2384 0.44068 44.07 1)Probable 18596 0.0000 0.0000 0.8274 0.0000 0 0.0000 0.0000 0.0000 0.883 0.0000 -0.2384 0.44068 44.07 1)Probable 18823 0.0000 0.0000 0.0000 0.0000 0 0.0000 0.6168 0.0000 0.000 0.5422 -0.2884 0.42840 42.84 1)Probable 19011 0.4394 0.0000 0.0000 0.0000 0 0.0000 0.6168 0.9143 0.000 0.0000 -0.3289 0.41851 41.85 1)Probable 20199 0.4394 0.0000 0.0000 0.0000 0 0.0000 0.6168 0.9143 0.000 0.0000 -0.3289 0.41851 41.85 1)Probable 20669 0.4394 0.0000 0.0000 0.0000 0 0.0000 0.6168 0.9143 0.000 0.0000 -0.3289 0.41851 41.85 1)Probable 130 Obs beta21 beta22 beta31 beta41 beta42 beta51 beta52 beta61 beta62 beta63 logits probit probit2 epilepgp3 21256 0.0000 0.0000 0.0000 1.2732 0 0.0000 0.6168 0.0000 0.883 0.0000 -0.4094 0.39906 39.91 2)Possible 23700 0.0000 0.0000 0.0000 1.2732 0 0.0000 0.6168 0.0000 0.883 0.0000 -0.4094 0.39906 39.91 2)Possible 24169 0.4394 0.0000 0.8274 0.0000 0 0.0000 0.6168 0.0000 0.000 0.0000 -0.4158 0.39752 39.75 2)Possible 25195 0.4394 0.0000 0.0000 1.2732 0 0.0000 0.6168 0.0000 0.000 0.5422 -0.5122 0.37468 37.47 2)Possible 26980 0.4394 0.0000 0.0000 0.0000 0 0.0000 0.0000 0.0000 0.000 0.0000 -0.6264 0.34833 34.83 2)Possible 27630 0.4394 0.0000 0.0000 0.0000 0 0.0000 0.0000 0.0000 0.000 0.0000 -0.6264 0.34833 34.83 2)Possible 28218 0.4394 0.0000 0.0000 0.0000 0 0.0000 0.0000 0.0000 0.000 0.0000 -0.6264 0.34833 34.83 2)Possible 29065 0.4394 0.0000 0.0000 0.0000 0 0.0000 0.0000 0.0000 0.883 0.0000 -0.6264 0.34833 34.83 2)Possible 29884 0.4394 0.0000 0.0000 0.0000 0 0.0000 0.0000 0.0000 0.000 0.0000 -0.6264 0.34833 34.83 2)Possible 32917 0.0000 0.0000 0.8274 0.0000 0 0.0000 0.6168 0.0000 0.883 0.0000 -0.8552 0.29834 29.83 3)Unlikely 33715 0.0000 0.0000 0.8274 0.0000 0 0.0000 0.6168 0.0000 0.883 0.0000 -0.8552 0.29834 29.83 3)Unlikely 34359 0.0000 0.0000 0.0000 1.2732 0 0.0000 0.6168 0.0000 0.000 0.5422 -0.9516 0.27856 27.86 3)Unlikely 35955 0.0000 0.0000 0.0000 0.0000 0 0.0000 0.0000 0.0000 0.883 0.0000 -1.0658 0.25620 25.62 3)Unlikely 36432 0.0000 0.0000 0.0000 0.0000 0 0.0000 0.0000 0.0000 0.883 0.0000 -1.0658 0.25620 25.62 3)Unlikely 36745 0.0000 0.0000 0.0000 0.0000 0 0.0000 0.0000 0.0000 0.883 0.0000 -1.0658 0.25620 25.62 3)Unlikely 37754 0.0000 0.0000 0.0000 0.0000 0 0.0000 0.0000 0.0000 0.883 0.0000 -1.0658 0.25620 25.62 3)Unlikely 39099 0.0000 0.0000 0.0000 0.0000 0 0.0000 0.0000 0.0000 0.883 0.0000 -1.0658 0.25620 25.62 3)Unlikely 39856 0.0000 0.0000 0.0000 0.0000 0 0.0000 0.0000 0.0000 0.883 0.0000 -1.0658 0.25620 25.62 3)Unlikely 40258 0.0000 0.0000 0.0000 0.0000 0 0.0000 0.0000 0.0000 0.000 0.0000 -1.0658 0.25620 25.62 3)Unlikely 40459 0.0000 0.0000 0.0000 0.0000 0 0.0000 0.0000 0.0000 0.000 0.0000 -1.0658 0.25620 25.62 3)Unlikely 40498 0.0000 0.0000 0.0000 0.0000 0 0.0000 0.0000 0.0000 0.883 0.0000 -1.0658 0.25620 25.62 3)Unlikely 40552 0.0000 0.0000 0.0000 0.0000 0 0.0000 0.0000 0.0000 0.883 0.0000 -1.0658 0.25620 25.62 3)Unlikely 41808 0.0000 0.0000 0.0000 0.0000 0 0.0000 0.0000 0.0000 0.883 0.0000 -1.0658 0.25620 25.62 3)Unlikely 42155 0.0000 0.0000 0.0000 0.0000 0 0.0000 0.0000 0.0000 0.883 0.0000 -1.0658 0.25620 25.62 3)Unlikely 44838 0.0000 0.0000 0.0000 0.0000 0 0.0000 0.0000 0.0000 0.883 0.0000 -1.0658 0.25620 25.62 3)Unlikely 45064 0.0000 0.0000 0.0000 0.0000 0 0.0000 0.0000 0.0000 0.883 0.0000 -1.0658 0.25620 25.62 3)Unlikely 46455 0.0000 0.0000 0.0000 0.0000 0 0.0000 0.0000 0.0000 0.000 0.0000 -1.0658 0.25620 25.62 3)Unlikely 47165 0.0000 0.0000 0.0000 0.0000 0 0.0000 0.0000 0.0000 0.883 0.0000 -1.0658 0.25620 25.62 3)Unlikely 131 Obs beta21 beta22 beta31 beta41 beta42 beta51 beta52 beta61 beta62 beta63 logits probit probit2 epilepgp3 47202 0.0000 0.0000 0.0000 0.0000 0 0.0000 0.0000 0.0000 0.883 0.0000 -1.0658 0.25620 25.62 3)Unlikely 47661 0.0000 0.0000 0.0000 0.0000 0 0.0000 0.0000 0.0000 0.000 0.0000 -1.0658 0.25620 25.62 3)Unlikely 48159 0.0000 0.0000 0.0000 0.0000 0 0.0000 0.0000 0.0000 0.883 0.0000 -1.0658 0.25620 25.62 3)Unlikely 48216 0.0000 0.0000 0.0000 0.0000 0 0.0000 0.0000 0.0000 0.883 0.0000 -1.0658 0.25620 25.62 3)Unlikely 50337 0.4394 0.0000 0.0000 0.0000 0 0.0000 0.0000 0.0000 0.000 0.5422 -1.1686 0.23711 23.71 3)Unlikely 50342 0.4394 0.0000 0.0000 0.0000 0 0.0000 0.0000 0.0000 0.000 0.5422 -1.1686 0.23711 23.71 3)Unlikely 52030 0.4394 0.0000 0.0000 0.0000 0 0.0000 0.6168 0.0000 0.000 0.0000 -1.2432 0.22388 22.39 3)Unlikely 52041 0.4394 0.0000 0.0000 0.0000 0 0.0000 0.6168 0.0000 0.000 0.0000 -1.2432 0.22388 22.39 3)Unlikely 53490 0.4394 0.0000 0.0000 0.0000 0 0.0000 0.6168 0.0000 0.883 0.0000 -1.2432 0.22388 22.39 3)Unlikely 53595 0.4394 0.0000 0.0000 0.0000 0 0.0000 0.6168 0.0000 0.883 0.0000 -1.2432 0.22388 22.39 3)Unlikely 53772 0.4394 0.0000 0.0000 0.0000 0 0.0000 0.6168 0.0000 0.000 0.0000 -1.2432 0.22388 22.39 3)Unlikely 54153 0.4394 0.0000 0.0000 0.0000 0 0.0000 0.6168 0.0000 0.883 0.0000 -1.2432 0.22388 22.39 3)Unlikely 54723 0.4394 0.0000 0.0000 0.0000 0 0.0000 0.6168 0.0000 0.883 0.0000 -1.2432 0.22388 22.39 3)Unlikely 54764 0.4394 0.0000 0.0000 0.0000 0 0.0000 0.6168 0.0000 0.883 0.0000 -1.2432 0.22388 22.39 3)Unlikely 54956 0.4394 0.0000 0.0000 0.0000 0 0.0000 0.6168 0.0000 0.883 0.0000 -1.2432 0.22388 22.39 3)Unlikely 55628 0.0000 0.0000 0.0000 0.0000 0 0.0000 0.0000 0.0000 0.000 0.5422 -1.6080 0.16687 16.69 3)Unlikely 55687 0.0000 0.0000 0.0000 0.0000 0 0.0000 0.0000 0.0000 0.000 0.5422 -1.6080 0.16687 16.69 3)Unlikely 57018 0.0000 0.0000 0.0000 0.0000 0 0.0000 0.0000 0.0000 0.000 0.5422 -1.6080 0.16687 16.69 3)Unlikely 57911 0.0000 0.0000 0.0000 0.0000 0 0.0000 0.6168 0.0000 0.000 0.0000 -1.6826 0.15675 15.68 3)Unlikely 59907 0.0000 0.0000 0.0000 0.0000 0 0.0000 0.6168 0.0000 0.000 0.0000 -1.6826 0.15675 15.68 3)Unlikely 60328 0.0000 0.0000 0.0000 0.0000 0 0.0000 0.6168 0.0000 0.883 0.0000 -1.6826 0.15675 15.68 3)Unlikely 132 Appendix F - Posttraumatic epilepsy incidence analysis SCTBIFR: Overall risk of developing posttraumatic epilepsy (PTE) within 3 years of discharge after TBI. Original cohort = 2118 persons. 12 individuals removed who did not answer epilepsy-related questions. Also removed 145 cases with pre-existing epilepsy. Analysis cohort = 1961 persons. Note: there were 208 individuals who were inadvertently not asked epilepsy-related question in year 2 or 3. 35 of the individuals not asked in year 2 did not participate in year 3, and 2 individuals were not asked the epilepsy-related questions in year 3. Years of interview participation (with years adjusted to whether asked epilepsy-related questions on 2nd and/or 3rd interviews) for those 1846 without any epilepsy: Cumulative Cumulative YEAR Frequency Percent Frequency Percent ƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒ FIRST 556 30.12 556 30.12 Participated only in 1st year SECON 280 15.17 836 45.29 Participated in 1st & 2nd years THIRD 1010 54.71 1846 100.00 Participated all 3 years Year of onset of 115 cases with PTE: Cumulative ONSET Frequency Percent Frequency Crude rates ƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒ YEAR 1 43 37.39 43 43/1961 = 2.2% of those at risk developed PTE in 1st year YEAR 2 44 38.26 87 44/1362 = 3.2% of those at risk developed PTE in 2nd year YEAR 3 28 24.35 115 28/1038 = 2.7% of those at risk developed PTE in 3rd year IF UNSURE OF YEAR OF ONSET, TOOK LATER YEAR TO BE CONSERVATIVE (INCLUDE MAX YEARS AT RISK). How years were adjusted for missing years of questions: Last year of interview participation of cases not developing epilepsy: Cumulative Cumulative YEAR Frequency Percent Frequency Percent ƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒ FIRST 521 28.22 521 28.22 SECON 313 16.96 834 45.18 THIRD 1012 54.82 1846 100.00 Remaining cases interviewed each year (include those w/o posttraumatic epilepsy at start of year, so ‘at risk’): Y1 = 1961 Y2 = 1012 + 313 + 28 + 44 =1397 Y3 = 1012 + 28 =1040 But: 35 individuals without epilepsy were not asked ESD questions in year 2, and did not participate in year 3. 2 individuals without epilepsy were not asked ESD questions in year 3. 133 So, adjusting for above, Individuals interviewed concerning new PTE each year (person-years at risk): Y1 = 1961 Y2 = 1397 – 35 = 1362 Y3 = 1040 – 2 = 1038 4361 Crude incidence rate over these 3 years (or incidence density): 115 with PTE/4361 person-yrs = 0.0264 person-yrs, or average of 2.6 cases per 100 person-years, or average of 2.6 cases per 100 people over a 1 year period. Using exponential formula: Year Δt # new cases Person-yrs ID(1/yr) approx. R [= 1 – exp(-ID*Δt)] 1 1 43 1961 .02193 .02169 2 1 44 1362 .03231 .03179 3 1 28 1038 .02697 .02661 3 year risk = 1 – exp(-Σ ID * Δt) = .07697 = 7.7 cases per 100 people over a 3 year period. The above formula is the exponential formula for relating rates and incidence proportions (Morgenstern H, Kleinbaum DG, Kupper LL. Measures of disease incidence used in epidemiologic research. International Journal of Epidemiology. 9(1):97-104;1980.) (Also see Rothman & Greenland, Modern Epidemiology, 2nd edition, 1998). However, it is subject to 3 assumptions: 1. the population is closed; 2. the event under study is inevitable (there is no competing risk); and 3. the number of events at each time event is a small proportion of the number at risk at that time (<0.1). I do not believe assumptions 1 (because members are lost to follow-up) and 2 (because the event is not inevitable, death is a competing risk) hold for this cohort. Regardless, it is extremely close to the other results. Using the Classical Life Table method (Szklo & Nieto, Epidemiology Beyond the Basics, 2000): the 3-year cumulative incidence (probability) would equal 115/(1961 – ½(599+324)) = 115/1499.5 = .0767 In this method, half of those lost to follow-up are subtracted from those at risk, assuming that they were, on average, at risk for only half that time period. Using the Kaplan-Meier method: if you take into account the 3 time periods (which is ‘iffy’ considering we don’t know exact onsets of epilepsy for some cases), the cumulative probability of epilepsy during 3 years is 1 – [(1918/1961)*(1318/1362)*(1010/1038)] = .0791 Incidence rate based on aggregate data: 115/[(1961+1038)/2] = 115/1499.5 = .0767 per person-3 years = .0256 per person-year = 2.6 per 100 person-years. Incidence rate per person-time: .0767 per person-3 years = .0256 per person-year = 2.6 per 100 person-years. Confidence intervals using tabulated values of 95% confidence limit factors for a Poisson-distributed variable*. Observed number of events on which estimate is based = 115. Lower limit factor for n=100 is 0.818, upper limit factor for n=100 is 1.22. Lower limit factor for n=120 is 0.833, upper limit factor for n=120 is 1.200. Interpolated lower limit factor for 115 is 0.818 + .75(0.833 – 0.818) = 134 0.82925. Interpolated upper limit factor for 115 is 1.22 - .75(1.22 – 1.20) = 1.205. So, confidence limits for the 115 count are as follows: lower = 115*0.82925 = 95.36; upper = 115*1.205 = 138.58, and the 95% confidence limits for the rate are: lower = 95.36/1499.5 = 0.0636 per person-3 years; upper = 138.58/1499.5 = 0.0924 per person-3 years, or 2.1 and 3.1 per 100 person-years. *Values from Haenszel W, Loveland DB, Sirken MG. Lung cancer mortality as related to residence and smoking histories, I. White males. Journal of the National Cancer Institute. 28:947-1001, 1962. I believe the most appropriate methods for this data are the Classical Life Table methods, or the incidence rate based on aggregate data (which are equivalent). So, incidence rate of PTE = 2.6 per 100 person-years, 95% CI 2.1 – 3.1 per 100 person-years. Determine risk stratifying by severity Mild = AIS_H of 2 = 21/[(765+411)/2] = 21/588 = 0.0357 per person-3 year = 0.0119 per person-year = 1.2 per 100 person-years LCL: .619*21 = 12.999; 12.999/588 = 0.0221 per person-3 year = 0.0074 per person-year = 0.7 per 100 person-years UCL: 1.53*21 = 32.13; 32.13/588 = 0.0546 per person-3 year = 0.0182 per person-year = 1.8 per 100 person-years Moderate = AIS_H of 3 = 16/[(322+165)/2] = 16/243.5 = 0.0657 per person-3 year = 0.0219 per person-year = 2.2 per 100 personyears LCL: 0.572*16 = 9.152; 9.152/243.5 = 0.0376 per person-3 year = 0.0125 per person-year = 1.3 per 100 person-years UCL: 1.62*16 = 25.92; 25.92/243.5 = 0.1064 per person-3 year = 0.0355 per person-year = 3.6 per 100 person-years Severe = AIS_H of 4, 5 = 78[(874+462)/2] = 78/668 = 0.1168 per person-3 year = 0.0389 per person-year = 3.9 per 100 person-years LCL: .785+[.80(.798-.785)] = 0.7954; .7954*78 = 62.0412; 62.0412/668 = 0.0929 per person-3 year = 0.0310 per person-year = 3.1 per 100 person-years UCL: 1.27-[.80(1.27-1.25)] = 1.254; 1.254*78 = 97.812; 97.812/668 = 0.1464 per person-3 year = 0.0488 per person-year = 4.9 per 100 person-years 135 Appendix G - Sampling plan for inpatient and emergency department charts Final Sampling Write-Up - Inpatient/ED For each abstraction year (2001-2002) X number of records will be abstracted annually. In each year, 28.3% of the records abstracted will have a diagnosis code of 345.0-345.9 (Epilepsy), 60% will have a diagnosis code of 780.3 (Seizure NOS), 10% will have a diagnosis code of 780.2 (Syncope and Collapse), and 1.7% will have a diagnosis code of 293.0 (Acute Delirium). The following is the process for drawing the stratified random sample. It is recommended that an unduplicated data set be used in order to have a person level sample. Unduplicated Data Set: 1. Select all records in a calendar year that are from a South Carolina hospital for from individuals who have a county of residence in SC. The individuals must be over the age of one when the encounter occurred (this was inadvertently changed to include those under 1 year of age). 2. Every individual is assigned a unique ID. The unique ID was developed with the purpose to allow person identification across different data systems. The estimated proportion of correctness for most data sets is 99.5% 3. Individuals will be sorted based on unique ID and then the first record in order of date of occurrence during the calendar year of interest will be pulled to create the sample universe. Since epilepsy is a chronic condition, taking the first record in a calendar year is a random method for creating the sample universe. Once the unduplicated data set is created there will be a number of variables added to make pulling the sample easier and more logical. Organizing the Data Set: 1. Based on both primary and secondary diagnosis code (i.e. 345.x, 780.3, 780.2, and 293.0) each record is assigned a group and could be assigned more than one group if the individual has multiple diagnosis codes in the same record. a. GroupA=345.x (epilepsy) b. GroupB=780.3 (Seizure NOS) c. GroupC=780.2 (Syncope and Collapse) d. GroupD=293.0 (Acute Delirium) IF (PDIAG=:'345') THEN GROUPA=1; IF (SDIAG1=:'345') THEN GROUPA=2; IF (SDIAG2=:'345') THEN GROUPA=3; IF (SDIAG3=:'345') THEN GROUPA=4; IF (SDIAG4=:'345') THEN GROUPA=5; IF (SDIAG5=:'345') THEN GROUPA=6; IF (SDIAG6=:'345') THEN GROUPA=7; IF (SDIAG7=:'345') THEN GROUPA=8; IF (SDIAG8=:'345') THEN GROUPA=9; IF (PDIAG=:'7803') THEN GROUPB=1; IF SDIAG1=:'7803' THEN GROUPB=2; IF SDIAG2=:'7803' THEN GROUPB=3; IF SDIAG3=:'7803' THEN GROUPB=4; IF SDIAG4=:'7803' THEN GROUPB=5; IF SDIAG5=:'7803' THEN GROUPB=6; IF SDIAG6=:'7803' THEN GROUPB=7; IF SDIAG7=:'7803' THEN GROUPB=8; IF SDIAG8=:'7803' THEN GROUPB=9; 136 IF PDIAG=:'7802' THEN GROUPC=1; IF SDIAG1=:'7802' THEN GROUPC=2; IF SDIAG2=:'7802' THEN GROUPC=3; IF SDIAG3=:'7802' THEN GROUPC=4; IF SDIAG4=:'7802' THEN GROUPC=5; IF SDIAG5=:'7802' THEN GROUPC=6; IF SDIAG6=:'7802' THEN GROUPC=7; IF SDIAG7=:'7802' THEN GROUPC=8; IF SDIAG8=:'7802' THEN GROUPC=9; IF PDIAG=:'2930' THEN GROUPD=1; IF SDIAG1=:'2930' THEN GROUPD=2; IF SDIAG2=:'2930' THEN GROUPD=3; IF SDIAG3=:'2930' THEN GROUPD=4; IF SDIAG4=:'2930' THEN GROUPD=5; IF SDIAG5=:'2930' THEN GROUPD=6; IF SDIAG6=:'2930' THEN GROUPD=7; IF SDIAG7=:'2930' THEN GROUPD=8; IF SDIAG8=:'2930' THEN GROUPD=9; RUN; 2. Once all the records have been assigned a group, they will be given another variable that will tell what type of record it is to help when drawing the sample. This is simply to make understanding the program clearer. IF GROUPA GE 1 THEN EPILEPSY=1; IF GROUPB GE 1 THEN NOS=1; IF GROUPC GE 1 THEN SYNC=1; IF GROUPD GE 1 THEN AD=1; RUN; 3. After every record is assigned the appropriate groupings (remember they can be in more than one group) there will be four different random samples drawn to match the sampling proportion presented in the grant. However, no record will be chosen more than once if that record falls in more than one group. DATA EPILEPSY; SET TABLE2; IF EPILEPSY=1; run; PROC FREQ DATA=EPILEPSY; tables LOCATE / outpct out=size; run; DATA SIZE; SET SIZE; _NSIZE_=ROUND((PERCENT/100)*850); RUN; PROC SORT DATA=EPILEPSY; BY LOCATE; RUN; PROC SORT DATA=SIZE; BY LOCATE; RUN; PROC SURVEYSELECT DATA=EPILEPSY METHOD=SRS OUT=OUT.UNEPSAMP SAMPSIZE=SIZE; STRATA LOCATE ; TITLE'UNDUPSAMP'; RUN; From an e-mail asking for clarification from ORS: ‘For a given sampling year (which is the Calendar year) there was a hierarchy for the diagnosis, but no hierarchy for primary vs. secondary diagnosis or inpatient vs outpatient.’ 137 Appendix H- Sampling plan for physician office charts Sampling Write-Up -- physician office visits Sample 1 – 360 (8:2 ratio of epilepsy:seizure) • Medicaid – 86 • Medicare – 175 • SHP – 99 Sample 2 – 61 (Still 8:2 ratio and exclude one neurology practice) • Medicaid – 17 • Medicare – 35 • SHP – 9 Sample 3 – 60 (All Medicare) Sample 4 – 101 (Keep 8:2 ratio and exclude Columbia and Charleston practices) • Kept all 81 epilepsy records left and drew additional 20 needed to keep 8:2 ratio • Medicaid – 16 • Medicare – 52 • SHP – 32 Epilepsy = 345.x, seizure = 780.3. Sampling excluded anyone who had had an inpatient or ED visit in 2001 or 2002. 138 ACKNOWLEDGMENT We express our heartfelt gratitude and appreciation to the following people for their participation in the South Carolina Epidemiological Studies of Epilepsy and Seizure Disorders. The South Carolina Department of Health and Environmental Control, especially Ms. Lou-Ann Carter Ms. Georgette Demian Ms. Lynnore Liggins Mr. Wesley Gravelle Ms. Harriet Range Ms. Beili Dong Dr. Jennifer Chiprich The Office of Research and Statistics, South Carolina Budget and Control Board, especially Mr. Andrew Pope Mr. Christopher Finney The Epilepsy Foundation of South Carolina, especially Ms. Christine Porter Ms. Barbara Brothers Department of Neurosciences and Clinical Neurophysiology Services, Medical University of South Carolina, especially Ms. Kelly Cavins Ms. Kris Topping Department of Physical Therapy, Colorado State University, Dr. Patricia Sample Center for Health Services and Policy Research, University of South Carolina, Ms. Lyn Phillips Division of Adult and Community Health, National Center for Chronic Disease Prevention and Health Promotion, Centers for Disease Control and Prevention, especially Ms. Rosemarie Kobau 139