Journal of Health Organization and Management
Assessing the relationship between patient satisfaction and clinical quality in an ambulatory setting
Tawnya Bosko Kathryn Wilson
Article information:
To cite this document:
Tawnya Bosko Kathryn Wilson , (2016),"Assessing the relationship between patient satisfaction and clinical quality in an
ambulatory setting", Journal of Health Organization and Management , Vol. 30 Iss 7 pp. Permanent link to this document:
http://dx.doi.org/10.1108/JHOM-11-2015-0181
Downloaded on: 10 September 2016, At: 06:58 (PT)
References: this document contains references to 0 other documents.
To copy this document: permissions@emeraldinsight.com
The fulltext of this document has been downloaded 2 times since 2016*
Downloaded by Cornell University Library At 06:58 10 September 2016 (PT)
Users who downloaded this article also downloaded:
(2016),"Leadership and management in UK medical school curricula", Journal of Health Organization and Management, Vol.
30 Iss 7 pp. -
Access to this document was granted through an Emerald subscription provided by emerald-srm:333301 []
For Authors
If you would like to write for this, or any other Emerald publication, then please use our Emerald for Authors service
information about how to choose which publication to write for and submission guidelines are available for all. Please
visit www.emeraldinsight.com/authors for more information.
About Emerald www.emeraldinsight.com
Emerald is a global publisher linking research and practice to the benefit of society. The company manages a portfolio of
more than 290 journals and over 2,350 books and book series volumes, as well as providing an extensive range of online
products and additional customer resources and services.
Emerald is both COUNTER 4 and TRANSFER compliant. The organization is a partner of the Committee on Publication
Ethics (COPE) and also works with Portico and the LOCKSS initiative for digital archive preservation.
*Related content and download information correct at time of download.
Downloaded by Cornell University Library At 06:58 10 September 2016 (PT)
Assessing the relationship between patient satisfaction and clinical quality in an ambulatory
environment
Introduction
Physician reimbursement and overall physician compensation are increasingly becoming
tied to quality based factors such as patient satisfaction results and performance on clinical
measures such as those provided by the Healthcare Effectiveness Data and Information Set
(HEDIS) and less based strictly on the volume of healthcare services provided (MGMA, 2014).
According to MGMA, in 2013 almost 6% of physician total compensation was linked to clinical
quality measures with another 2% of physician compensation tied to patient satisfaction
(MGMA, 2014). Other surveys have shown that up to 59% of physicians have at least some
portion of their compensation tied to patient satisfaction results (Zgierska, et al., 2014).
As the United States’ healthcare system continues to evolve from a reimbursement
system based on volume to one based on value, understanding the relationship between
physician quality metrics such as patient satisfaction and clinical quality metrics is extremely
important to successfully incentivizing physician behavior change. Consider two distinct parts of
quality – providing the clinically appropriate medical care as well as providing care in a manner
that results in a patient being satisfied. If patient satisfaction is based exclusively on or highly
correlated with physician clinical quality, then it is redundant and administratively costly to
measure and reimburse based on both; however, if patient satisfaction is measuring something
valued by patients but different from clinical quality, then incentivizing satisfaction may
increase demand and profits. In addition, different clinical quality measures may differentially
affect patient satisfaction. This paper uses a unique data set for physician hospital organization
in Northeast Ohio to assess the relationship between patient satisfaction and a variety of
clinical quality measures in an ambulatory setting to determine if there is significant overlap
between the two areas or if they are separate domains. Assessing this relationship will help to
determine whether there is congruence between different types of clinical quality performance
and patient satisfaction and therefore provide insight to appropriate financial structures for
physicians.
Patient satisfaction and patient preferences with their physicians have been extensively
studied. For example, Godager (2012) showed that overall, patients prefer physicians that are
similar to themselves in observable characteristics such as age and gender. Jackson, et al.
(2001) also found that patient satisfaction is directly influenced by patient age. Further,
Bertakis, et al. (2003) found that patients tend to be more satisfied with female physicians than
male. Vick and Scott (1998) showed that the most important attribute to patients was the
ability to talk to their doctor, and choosing their own treatment was the least important
element. In support of linking physician payment to patient satisfaction data, studies have
shown a relationship between patient perceptions of their physician and overall outcomes such
as adherence, satisfaction, trust, health status change and symptom resolution (Franks et al.,
2006). Zolnierek and DiMatteo (2009) showed the link between patient satisfaction and
Downloaded by Cornell University Library At 06:58 10 September 2016 (PT)
improved adherence to physician recommendations. On the other hand, researchers have
shown that patients often request elective services that offer limited benefit based on
marketing or other non-medically evident motives; and physicians often honor such requests to
improve patient satisfaction (Kravitz, et al., 2005). Research has also shown that in cases where
physicians’ compensation is more heavily tied to patient satisfaction, physicians are more likely
to order elective testing such as advanced imaging services for back pain (Pham, et al., 2009).
Most of the research on the relationship between patient satisfaction and overall
physician quality has focused on medical outcomes rather than clinical measures that are used
in physician incentives. One of the most significant recent studies showed that among those
patients with the highest patient satisfaction scores, there was a lower odds of visiting the
emergency department, a higher odds of inpatient admission, increased total expenditures
(relative to less satisfied patients), increased prescription drug expenditures and higher overall
mortality (Fenton, et al., 2012). Two studies, Schneider, et al. (2001) and Sequist et al. (2008),
do use HEDIS quality metrics in an ambulatory environment. However, both studies use health
plan data for the analysis, and look at patient experience measures (i.e. getting care quickly,
courtesy and respect of doctor’s office staff, communication with providers, etc.) rather than
overall patient satisfaction. Both studies find a relationship between patient experience and
clinical quality, though only modest (Sequist, et al, 2008) and variable across experience
composites (Schneider, et al, 2001). Doyle, et al. (2013) and Anhang-Price, et al. (2014) each
review the literature on the relationship between patient experience and either clinical
safety/effectiveness or quality respectively and each find the data to be mixed relative to the
relationship between outpatient or ambulatory satisfaction and certain quality indicators.
This paper extends the currently available research by studying the correlation between
overall patient satisfaction and a variety of clinical quality measures using a physician
organization’s quality data set and patient satisfaction results with a robust set of physician
characteristics and patient demographics. To our knowledge, this is the first study to use
physician organization data in this manner, primarily because of the relative infancy of quality
programs within provider organizations. Because of this, our study contributes to the literature
in multiple ways; most importantly, the data is directly from the physician medical record as
opposed to claims based data for measuring quality performance and it integrates physician
demographic information that is not available when using other data sets such as health plan
related information.
Conceptual Framework
The primary question of interest is how physician clinical quality, as measured by HEDIS
measures, is related to patient satisfaction scores. To allow the relationship to vary by types of
HEDIS measure (for example, HEDIS measures related to chronic care or HEDIS measures
related to prescription of antibiotics), a regression with patient satisfaction as the dependent
variable will be estimated separately for each of five HEDIS composite measures: vaccine,
preventative measures, prescription of generics, chronic conditions, and appropriate use of
antibiotics. For example, critics cite prescribing antibiotics as one of the major areas where
1
Downloaded by Cornell University Library At 06:58 10 September 2016 (PT)
there is asymmetric information, i.e. the patient desires to be prescribed an antibiotic to cure
their condition, but the doctor rather than the patient has the knowledge to know whether or
not an antibiotic is an effective treatment for the condition (Pilnick and Dingwall, 2011). And,
oftentimes taking an antibiotic can be detrimental to the patient’s health and to the general
health of the public (antibiotic resistant bacteria) (Ong, et al., 2007). Thus, the argument is that
if physicians are incentivized to improve patient satisfaction, they may inappropriately
prescribe antibiotics in order to keep patients happy, even though there are negative health
consequences; thus we may expect the HEDIS measures related to antibiotic prescription to be
negatively correlated with patient satisfaction (Ong, et al., 2007). In contrast, the prescribing of
generic prescription drugs may save the patient money and increase satisfaction (Pham, et al.,
2007). Previous literature does not provide as much insight into how physician performance on
clinical quality metrics for preventative measures, vaccination, and chronic condition will be
related to patient satisfaction. Given that patients with chronic conditions generally see a
physician more often, it is hypothesized that the relationship between the chronic condition
HEDIS measures and patient satisfaction will be stronger than between preventative or
vaccination and patient satisfaction.
The model will include a variety of physician and practice characteristics including
certification, doctor and practice volume, and physician demographic characteristics, as well as
demographic characteristics of the patient, such as age, gender, and insurance status. The
expected relationships between physician certification/practice variables and patient
satisfaction are ambiguous. The two variables measuring doctor volume of patients and
practice volume could be positively related to patient satisfaction if they are capturing
unobserved factors about the doctor or practice that cause it to have higher demand. In
contrast, a higher patient volume could be negatively related to patient satisfaction if it results
in patients having less time with the physician for any given appointment. If being an MD or
board certified is acting as a proxy for physician quality, then these would be expected to be
positively related with patient satisfaction; however, if the HEDIS measures are capturing the
quality of the physician then the hypothesis would be no residual relationship with satisfaction.
Data & Descriptive Statistics
In 2012 a regional hospital organization in northeast Ohio designed and implemented a
physician performance bonus program across all of its physician practices in efforts to begin
linking physician reimbursement to clinical quality and patient satisfaction metrics. The
organization consists of approximately 120 physicians across various specialties, with a strong
primary care base. We link the clinical quality database from calendar year 2013 with the
patient satisfaction database and physician demographics for all physicians in the practice. The
focus is patient satisfaction as expressed by patient ranking of their physician on a scale of 1-10,
10 being the highest, and its connection to physician performance on clinical quality metrics as
measured by individual physician performance on HEDIS measures. Every patient satisfaction
response received during calendar year 2013 was included so long as there was corresponding
clinical quality performance data for the designated physician. The unit of observation, then, is
2
Downloaded by Cornell University Library At 06:58 10 September 2016 (PT)
at the patient level with every patient satisfaction score representing an observation. The total
sample size is 2,808 observations.
Patient Satisfaction results include patient responses to the Clinician and Groups (“CG”)CAHPS survey received during the 2013 calendar year for the physicians. The CG-CAHPS survey
is a tool created by the Agency for Healthcare Research and Quality (“AHRQ”) that has been
shown to be valid and reliable (Dyer et al., 2012). There are many aspects to the CG-CAHPS
survey, certain questions focusing on the patient experience and others that measure patient
satisfaction. It is important to understand the difference between satisfaction and experience.
Generally, patient experience refers to aspects of the overall care experience such as how long
a patient waits to be seen, whether or not they were able to get an appointment as quickly as
they thought they needed, and how effectively their provider communicates with them (Bleich,
Ozaltin and Murray, 2009). Satisfaction, on the other hand, often refers to the patient’s overall
satisfaction with their care or physician, for example Keating, et al., 2002. On the CG-CAHPS
survey, patients were asked the following question as a measure of patient satisfaction: “Using
any number from 1 to 10, where 1 is the worst doctor possible and 10 is the best doctor
possible, what number would you use to rate this doctor?” In administration of the patient
satisfaction survey, the organization issued a survey to every patient with an email address on
file each time they had an office visit at one of the participating practices. For those patients
without an email address on file, a monthly random selection of patients was identified to
receive a hard-copy, mailed survey. Due to the nature of the survey process, those patients
that visit the doctor more frequently or see more doctors within the network would have more
opportunities to respond to a survey and may be represented in the data more than once.
Further, patients that responded more than once may have different results for each response.
During the time period, the response rate to the survey was 12%, which is lower than the
response rates in Sequist et al. (2008) and Scheider et al. (2001). Though this may be
considered a low response rate for surveys in general, relative to patient satisfaction in
healthcare, it is slightly above the average of 11% (Scaletta, 2015).
The clinical quality metrics database includes 26 HEDIS measures that were included in
the performance bonus program for calendar year 2013. The analytics team for the
organization would extract information from patients’ electronic health records on the quality
metrics on a monthly basis and calculate the physician’s aggregate score for each HEDIS
measure. While we as researchers did not have access to individual patient electronic medical
records, we have the aggregate scoring based on the electronic medical records of all the
patients in a physician’s practice. Therefore, it is not possible to determine whether the
individual physician met a certain clinical quality measure for the specific patient that
responded to the patient satisfaction question. Rather, for each composite group, the clinical
quality measure is the average across the relevant metrics in that composite group of the
percent of the time that the physician seen by that patient met the requirement for the
appropriate population based on the electronic medical records for all the physician’s patients.
Appendix Table 1 describes the HEDIS measures and indicates the composite group they
belong to. The measures were selected based on the organization’s contracts with the three
3
Downloaded by Cornell University Library At 06:58 10 September 2016 (PT)
major payers in their market. These measures represent the HEDIS measures that are
incorporated into the payer contracts as clinical quality metrics that they are accountable for
under the terms of their pay for performance program. Because the majority of the HEDIS
measures included in the organization’s performance bonus program apply exclusively to
primary care physicians, the study was limited to family medicine, internal medicine, pediatric
and gynecology physicians. The 26 HEDIS measures were divided into five groups: vaccine,
preventative measures, prescription of generics, chronic conditions, and appropriate use of
antibiotics. This allows for the relationship between patient satisfaction and physician clinical
quality to vary depending on the measure of clinical quality. Note that not all HEDIS measures
are relevant for all physician specialties; for example, childhood immunizations is relevant for
pediatrics or family practice but not gynecology.
Patient and physician demographic factors are linked with the patient satisfaction and
clinical quality metrics data. The physician demographic variables available include: physician
age, physician gender, physician specialty, physician degree (Allopathic with an M.D. or
Osteopathic with a D.O.), and physician board certification status. Physician productivity is
measured by physician work relative value unit (wRVUs), which provides information about
how busy the doctor is individually. Total visit volume by facility shows the size and volume of
an individual office location where the visit took place. Patient demographic variables include
age, gender, race/ethnicity, and insurance status (private commercial insurance, Medicare,
Medicaid, or self-pay). Previous studies that rely on health plan claims data have not included
physician and patient characteristics in their analysis.
Descriptive statistics for the variables used in the analysis are provided in Table 1. Note
that patient satisfaction is very high in the data, with the average satisfaction score 9.53 on a
10-point scale. While the range of satisfaction scores are from 1 to 10, only 2% of satisfaction
scores were 5 or lower, 2% were a 6 or 7, 7% were an 8, 14% were a 9, and 75% were a 10. The
high average score may be related to the fact that patients are evaluating their primary care
physician, someone they would have multiple interactions with over the course of years. For
the physician clinical quality measures, physicians are least likely to have the HEDIS metric met
for the preventative measures, such as patients having an annual preventative visit, biannual
mammograms for females ages 40-69, patients having an annual blood pressure measurement,
etc. On average doctors only have 62.1% of the relevant patient population meeting the
various requirements that compose the composite preventative metric; the range across
doctors is an average of 38.4% to 83.8% of the relevant patient population meeting the metrics.
Appropriate vaccine screening and prescribing of generics are the metrics most often met, with
doctors averaging 89.3% and 90.4% of their patients meeting the metric. Finally, appropriate
antibiotic prescribing and chronic care measures are met an average of 83.4% and 85.5%,
respectively. Figure 1 shows scatter plot graphs relating the patient satisfaction score and each
of the five composite HEDIS measures; the graphs indicate that despite the high number of
satisfaction scores of 10, there is variation in the scores and the HEDIS measures.
In considering the relationship between the physician clinical quality as measured by the
various HEDIS composite measures and patient satisfaction, it is instructive to first examine the
4
Downloaded by Cornell University Library At 06:58 10 September 2016 (PT)
relationship between the different HEDIS composite measures. Table 2 presents pairwise
correlations of the various clinical quality measures. While some measures, such as the
prescribing of generics and vaccination, are highly correlated (correlation coefficient=0.78),
other measures show little correlation with each other. For example, the prescribing of generic
prescriptions does not appear to be correlated with appropriate use of antibiotics
(correlation=0.02) or the preventative metrics (correlation=-0.01). The limited correlation
between some of the HEDIS metrics suggest that it may be important to design incentive
patterns that measure various dimensions of clinical quality separately.
Methodology
Regression analysis is used to examine the relationship between patient satisfaction and
physician clinical quality across a variety of HEDIS measures with an extensive set of physician
and patient variables also included in the model. In each regression, the patient satisfaction
score is the dependent variable and independent variables include the relevant HEDIS
composite measure, physician and practice volume variables, physician credential variables of
MD and board certified, physician age, physician gender, patient age, patient race, patient age,
and insurance status. Sequist, et al. (2008) and Schneider, et al. (2001) also used composite
measures, but they applied composites to the CAHPS survey responses in assessing overall
patient experience where this research focuses on the overall rating of the physician as a
measure of patient satisfaction.
Because the patient satisfaction variable is ordinal rather than cardinal, the models are
estimated using an ordered probit model. Essentially the model is estimating the probability of
having the score for each of the patient satisfaction scores (1 to 10). This eliminates the implicit
constraint from Ordinary Least Squares regression that a one-unit increase in satisfaction is the
same for all scores (i.e., an increase in patient satisfaction from 2 to 3 is the same as an increase
from 9 to 10) and that a multiple unit difference is that magnitude larger than a one unit
difference (i.e., the difference between 6 and 10 is 4 times the difference between 9 and 10).
The coefficient estimates from the ordered probit model indicate the sign of the coefficient, but
the magnitude of the effect must be calculated based on the density function. Therefore, in
addition to the coefficient estimate and standard error, the marginal effect of each variable on
the probability that the satisfaction score will be 10 is also reported. The models are estimated
using the STATA “oprobit” command with robust standard errors using “mfx,
predict(outcome(10))” to calculate the marginal effect.
Two sets of sensitivity analysis are also conducted. Given the high proportion of
satisfaction scores of 10 (75%), in addition to the ordered probit model described above we
estimate a probit model with a dependent variable that equals 1 if the satisfaction score is 10
and equals zero otherwise. Finally, we estimate the model using a dummy variable indicating if
the HEDIS measure is above the median of HEDIS scores, rather than actual score, to see if the
results are consistent when relative score (rather than absolute score) is considered.
Results
5
Downloaded by Cornell University Library At 06:58 10 September 2016 (PT)
The results of the regressions for each of the HEDIS composite measures is reported in
Table 3. Each column of the table reports the results for an ordered probit regression with
patient satisfaction as the dependent variable and the column heading indicating the physician
clinical quality variable used in the estimation. The appropriate prescribing of antibiotics,
prescribing generics or documenting medical necessity for brand name drugs, and appropriate
vaccination are all positively and statistically significantly related to patient satisfaction. For
antibiotics and generics the magnitude of effect is similar, associated with 33.4% and 30.5%
higher probability of a satisfaction score of 10; the magnitude of effect for vaccination is about
half the size but still is associated with a 16.3% higher probability of a score of 10. In contrast
to these positive effects, the coefficient estimate on the preventative composite is negative and
statistically significant at the 0.06 level; it is associated with a 14.9% lower probability of a
patient satisfaction score of 10. The chronic condition clinical quality measure is not related to
satisfaction, with a statistically insignificant coefficient estimate and small marginal effect.
The positive and significant coefficient on antibiotics is counter to the argument that
physicians will prescribe antibiotics requested by the patient even when the antibiotic is not
appropriate in order increase patient satisfaction. Using the HEDIS measures, closer adherence
to appropriate antibiotic prescribing is associated with higher, not lower, patient satisfaction.
For antibiotics, generics, and vaccination, the positive relationship between the HEDIS
measures and patient satisfaction suggests that for these areas the physician clinical quality and
patient satisfaction are capturing the same domain of overall physician quality.
In the case of preventive measures, physician clinical quality and patient satisfaction are
not only capturing different domains of overall physician quality, but are in conflict. Doctors
who have a higher percent of their patients getting appropriate preventative visits and
screenings have a lower patient satisfaction score. If both adherence to these measures and
patient satisfaction are priorities, then in designing a physician compensation scheme it will be
necessary to capture both domains. Similarly, while not in conflict, the lack of relationship
between patient satisfaction and the chronic condition clinical quality measures suggest that,
like with preventative, it will be necessary to design a compensation scheme that captures
these separate domains.
As previously mentioned, the results in the literature about the relationship between
patient satisfaction and clinical quality measures are mixed. Sequist, et al. (2008) found only
modest correlations between patient experience and clinical quality, ultimately determining
that they are separate, but somewhat related domains. Schneider, et al. (2001) did find a
correlation between certain patient experience composites and clinical quality of health plans.
While our research focuses on patient satisfaction with their physician as opposed to the
overall patient experience of care, our results are also mixed.
Three physician and patient demographic variables were shown to be significant
predictors of patient satisfaction rating across all of the specifications: physician age, physician
gender, and patient age. Older physicians have lower satisfaction scores, while male physicians
have significantly higher scores even controlling for the clinical quality measures. The other
physician and practice variables, while on occasion are statistically significant, do not show a
6
Downloaded by Cornell University Library At 06:58 10 September 2016 (PT)
consistent pattern; doctor volume, measured by wRVU, is negatively associated with patient
satisfaction but the effect is only statistically significant in some of the specifications. Older
patients tend to have a greater level of satisfaction with their physicians, but patient gender or
race do not have an effect. Similarly, compared to private insurance, those on Medicaid,
Medicare, or self pay do not have a statistically significant difference in satisfaction.
The results for patient age and patient satisfaction are similar to those Jackson et al.
(2001) found for patient experience. However, the results of higher patient satisfaction for
male physicians is contrary to Bertakis’s, et al. (2003) finding that patients tend to have higher
ratings of patient experience with female physicians than males. This difference in the
physician gender relationship with patient experience and patient satisfaction is an area where
further research is warranted.
In designing a fair compensation scheme that includes patient satisfaction, it is
important to understand how these patient and demographic variables may impact satisfaction
scores. If there are systematic differences in how younger or male physicians interact with
patients that results in higher patient satisfaction, then incorporating satisfaction into
compensation would be appropriate. However, the fact that age and gender variables, which
physicians have no control over, impact satisfaction even controlling for the clinical quality
metrics may reflect systematic bias. It is outside the scope of this research to be able to
determine if there are systematic differences in behavior by race and gender, but the strong
relationship suggests reason for caution in using patient satisfaction in a compensation package
unless the data is risk-adjusted for these characteristics.
The results for the sensitivity analysis are generally consistent with the base model.
Estimating the model as a probit with a dependent variable equal to 1 if the satisfaction score is
10, reported in Appendix Table 2, produces very similar results to the ordered probit model
(Table 3). The statistical significance and marginal effects for the HEDIS variables are consistent
with the exception of the preventative HEDIS measure; while this variable is statistically
significant at the 10% level in the ordered probit (Table 3), it is not statistically significant in the
probit (Appendix Table 2). This suggests that for the vaccination measure what is happening at
lower satisfaction scores is also important to capture; however, caution should be given to
reading too much into this result given that the original coefficient estimate was only
statistically significant at the 10% level. Estimating the model using relative HEDIS score (a
dummy variable indicating the physician’s score is above the median), reported in Appendix
Table 3, also results in the same sign and level of statistical significance for all of the HEDIS
measures except vaccination, which is no longer statistically significant. The coefficient
estimates for the physician age, physician gender, and patient age variables all remain
statistically significant in all of these alternative specifications.
Limitations
This study has several key limitations. First, the data for this study come from a single
hospital provider organization in Northeast Ohio and the generalizability to other settings is not
known without conducting studies in multiple, diverse settings. Second, the response rate for
7
Downloaded by Cornell University Library At 06:58 10 September 2016 (PT)
the satisfaction survey is only 12%. The potential bias of a low response rate depends on the
selection of who responds as well as the correlation between selection and the physician HEDIS
measures. For example, if patients who are very satisfied are more likely to complete the
survey but this is unrelated to physician HEDIS scores, then average satisfaction will appear
higher in the sample than it is in the population, reflected in a higher coefficient estimate on
the intercept, but the coefficient estimate on the HEDIS variable will be unbiased because there
is no correlation between unobserved selection factors and the HEDIS variables. If, however,
those satisfied patients who have doctors with high HEDIS scores are more likely to respond
than are low satisfaction patients of the same doctors (or more likely to respond than high
satisfaction patients of doctors with low HEDIS scores) then it will appear that the HEDIS
variables are more strongly related to patient satisfaction than in the population and the
coefficient will be biased upwards. Third, it is not possible to link whether the HEDIS quality
measure was met for the individual patient who is completing the satisfaction survey. Rather,
the physician clinical quality measure is the percent of the physician’s relevant patient
population that met the HEDIS measure. This would be less of a concern if the survey response
rate were high, as the sample who took the survey would be more likely to mirror the
population. If, however, the patients who had the HEDIS standard met for them are more likely
to complete the survey (regardless of the overall HEDIS score of the physician), then the ability
to measure the relationship between patient satisfaction and clinical quality measures would
be muted with a bias towards zero for the coefficient estimates on the HEDIS variables. Finally,
there is no way for the study to determine if there are systematic differences in physician
behavior by age or gender; therefore, it is not possible to distinguish if the significant
coefficient estimates on these variables are capturing systematic differences in physician
behavior or capturing patient bias.
Conclusion
This study uses a unique data set of HEDIS quality variables and patient satisfaction data
to examine the relationship between physician clinical quality and patient satisfaction. The
results suggest that different types of clinical quality measures are not necessarily correlated
and they may have different effects on patient satisfaction. While some aspects of clinical
quality, such as appropriate prescribing of antibiotics and generics as well as vaccination are
positively related to patient satisfaction, other measures, like preventative measures and
treatment of chronic conditions, are negatively related to satisfaction or not related at all. The
results suggest patient satisfaction and physician clinical quality may occupy the same domain
across some clinical quality measures but for other measures satisfaction and clinical quality are
unrelated or negatively related. Therefore, for some clinical quality metrics, it will be important
to separately compensate clinical quality and satisfaction. In addition, in designing a physician
compensation program, it is important to have a variety of clinical quality measures. Finally,
the strong relationship between the variables of physician age, physician gender, and patient
age and the level of patient satisfaction are important to consider when designing a physician
compensation package based on patient satisfaction.
8
9
Downloaded by Cornell University Library At 06:58 10 September 2016 (PT)
Downloaded by Cornell University Library At 06:58 10 September 2016 (PT)
References
Anhang-Price, R., Elliott, M., Zaslavsky, A., Hays, R., Lehrman, W., Rybowski, L., Edgman-Levitan, S. and
Cleary, P. (2014). “Examining the Role of Patient Experience Surveys in Measuring Health Care
Quality”. Medical Care Research Review, Volume 71, Number 5, pp. 522-554.
Bertakis, K.D., Franks, P. and Azari, R. (2003). “Effects of Physician Gender on Patient Satisfaction”.
Journal of the American Medical Women’s Association, Volume 58, Number 2, pp. 69-75.
Bleich, S., Ozaltin, E., and Murray, C. (2009). “How Does Satisfaction with the Healthcare System Relate
to Patient Experience?” Bulletin of the World Health Organization, Volume 87, Number 4.
Doyle, C., Lennox, L., Bell, D. (2013). “A Systematic Review of Evidence on the Links Between Patient
Experience and Clinical Safety and Effectiveness”. BMJ Open, Volume 3.
Dyer, N., Sorra, J.S., Smith, S.A., Cleary, P.D., and Hays, R.D. (2012). “Psychometric Properties of the
Consumer Assessment of Healthcare Providers and Systems Clinician and Group Adult Visit
Survey”, Medical Care, Vol. 50 No. 11 Suppl. 3, pp. S28-S34.
Fenton, J.J., Jerant, A.F., Bertakis, K.D., and Franks, P. (2012). “The Cost of Satisfation: a national study
of patient satisfaction, health care utilization, expenditures, and mortality”, Archives of Internal
Medicine, Vol. 172 No. 5, pp. 405-411
Franks, P. J., Jerant, A.F., Fiscella, K., Shields, C.G., Tancredi, D.J., and Epstein, R.M. (2006). “Studying
Physician Effects on Patient Outcomes: Physician Interactional Style and Performance on Quality
of Care Indicators”, Social Science & Medicine, Vol. 62 No. 2, pp. 422-432.
Godager, G. (2012). “Birds of a Feather Flock Together: A Study of Doctor-Patient Matching”, Journal of
Health Economics, Vol. 31 No. 1, 296-305.
Jackson, J., Chamberlin, J. and Kroenke, K. (2001). “Predictors of Patient Satisfaction”. Social Science and
Medicine, Volume 52, pp. 609-620.
Keating, N., Green, D., Kao, A., Gazmararian, J., Wu, V., and Cleary, P. (2002). “How are Patients’ Specific
Ambulatory Care Experiences Related to Trust, Satisfaction, and Considering Changing
Physicians?” Journal of General Internal Medicine, Volume 17, Issue 1.
Kravitz, R. L., Epstein, R.M., Feldman, M.D., Franz, C.E., Azari, R., Wilkes, M.S., Hinton, L., Franks, P.
(2005). “Influence of Patients' Requests for Direct-to-Consumer Advertised Antidepressants: A
Randomized Controlled Trial”, JAMA, Vol. 293 No. 16, pp. 1995-2002.
MGMA. (2014). Physician Compensation and Production Survey: 2014 Report Based on 2013 Data,
Englewood, CO: MGMA.
Ong, S., Nakase, J., Moran, G., Karras, D., Kuehnert, M., and Talan, D. (2007). “Antibiotic Use for
Emergency Department Patients with Upper Respiratory Infections: Prescribing Practices,
Patient Expectations and Patient Satisfaction”, Annals of Emergency Medicine, Volume 50, Issue
3, pp. 213-220.
Pham, H. H., Landon, B.E., Reschovsky, J.D., and Schrag, D. (2009). “Rapidity and Modality of Imaging for
Acute Low Back Pain in Elderly Patients”, Archives of Internal Medicine, Vol. 169 No. 10, pp. 972981.
Pham, H.H., Alexander, C., O’Malley, A. (2007). “Physician Consideration of Patients’ Out-of-pocket Costs
in Making Common Clinical Decisions”, JAMA Internal Medicine, Volume 167, No. 7, pp. 663668.
Pilnick, A. and Dingwall, A. (2011). “On the Remarkable Persistence of Asymmetry in Doctor/Patient
Interaction: A Critical Review”, Social Science and Medicine, Volume 72, Issue 9, pp. 1374-1382.
Scaletta, T. (2015, May 20). Timely Patient Satisfaction Surveys: No Longer and Option. Hospital Impact.
10
Downloaded by Cornell University Library At 06:58 10 September 2016 (PT)
Schneider, E.C., Zaslavsky, A.M., Landon, B.E., Lied, T.R., Sheingold, S., and Clearly, P.D. (2001) “National
Quality Monitoring of Medicare Health Plans”, Medical Care, Vol. 39 No. 12, pp. 1313-1325.
Sequist, T.D., Schneider, E.C., Anastario, M., Odigie, E., Marshall, R., Rogers, W., and Safran, D. (2008).
“Quality Monitoring of Physicians: Linking Patients’ Experiences of Care to Clinical Quality and
Outcomes”, Journal of General Internal Medicine, Vol. 23 No. 11, pp. 1784-1790.
Vick, S. and Scott, A. (1998). “Agency in Healthcare. Examining Patient's Preferences for Attributes of the
Doctor-Patient Relationship”, Journal of Health Economics, Vol. 17, pp. 587-605
Zgierska, A., Rabago, D., and Miller, M.M. (2014). “Impact of Patient Satisfaction Ratings on Physicians
and Clinical Care”, Patient Preference and Adherence, Vol. 8, pp. 837-446.
Zolnierek, K. B. and DiMatteo, M.R. (2009). “Physician Communication and Patient Adherence to
Treatment: A Meta-Analysis”, Medical Care, Vol. 47 No. 8, pp. 826-834.
11
Table 1: Summary Statistics
Downloaded by Cornell University Library At 06:58 10 September 2016 (PT)
Category
Dependent Variable
Physician Quality
Measures Based on %
of Time HEDIS Metric
Met
Physician and Office
Volume Measures
Physician Degree and
Certification
Physician Demographic
Variables
Patient Demographic
Variables
Patient Insurance
Status
Variable
Patient Satisfaction
Antibiotics
Chronic Care
Generics
Preventative
Vaccine
Doctor Volume (wRVU)
Practice Visit Volume
Allopathic (MD) = Yes
Board Certification = Yes
Provider Age
Provider Male
Patient Male
Patient Race Caucasian
Patient Age
Private Insurance
Insurance Self Pay
Medicaid
Medicare
Mean
9.53
0.8338
0.8550
0.9039
0.6212
0.8928
5331
16975
0.5754
0.9408
48.42
0.8034
0.4348
0.9764
57.68
0.6959
0.0185
0.0167
0.2688
Std. Dev.
1.077
0.1208
0.0881
0.0653
0.1146
0.1618
1446
6902
0.4943
0.2358
9.61
0.3974
0.4958
0.1515
15.99
0.4601
0.1348
0.1283
0.4434
Min
1
0.3333
0.5496
0.6480
0.3841
0.2460
697
2257
0
0
30
0
0
0
0
0
0
0
0
Max
10
1
1
0.9861
0.8379
1
8731
26322
1
1
78
1
1
1
98
1
1
1
1
Table 2: Pairwise Correlations of the Physician HEDIS Quality Measures
(level of statistical significance is in brackets below the correlation;
** indicates statistically significant at the .05 level)
Antibiotics
Chronic Care
Preventative
Generics
Vaccine
Antibiotics
1.0
0.19**
[0.00]
0.05**
[0.00]
0.02
[0.25]
0.12**
[0.00]
Chronic Care
Preventative
Generics
Vaccine
1.0
0.67**
[0.00]
0.33**
[0.00]
0.27**
[0.00]
1.0
-0.01
[0.68]
-0.01
[0.54]
12
1.0
0.78**
[0.00]
1.0
Table 3: Ordered Probit Regression with Patient Satisfaction as the Dependent Variable
(Column title indicates the HEDIS physician quality measure included as an independent variable in each
regression. Standard errors are in parenthesis and the marginal effect on the probability of a satisfaction
score of 10 is in brackets; *** indicates significant at the .01 level, ** at the .05 level, * at the .10 level)
Physician Clinical
Quality Variable
Downloaded by Cornell University Library At 06:58 10 September 2016 (PT)
Doctor Volume
(wRVU*1,000)
Practice Volume
* 10,000
Allopathic (MD)
Board Certified
Physician Age
Physician Male
Patient Male
Patient
Caucasian
Patient Age
Insurance is
Self Pay
Medicaid
Medicare
Log Likelihood
# of Observations
Antibiotics
1.062***
(0.287)
[0.334]
-0.062***
(0.022)
[-0.019]
0.021
(0.040)
[0.007]
0.053
(0.055)
[0.017]
-0.442***
(0.146)
[-0.119]
-0.010***
(0.003)
[-0.003]
0.405***
(0.077)
[0.137]
0.037
(0.051)
[0.011]
-0.032
(0.161)
[0.010]
0.010***
(0.002)
[0.003]
0.135
(0.168)
[0.041]
0.174
(0.198)
[0.052]
0.013
(0.067)
[0.004]
-2412
2,808
Preventative
-0.474*
(0.251)
[-0.149]
-0.032
(0.022)
[-0.010]
0.029
(0.041)
[0.009]
0.035
(0.057)
[0.011]
-0.113
(0.122)
[-0.034]
-0.012***
(0.003)
[-0.004]
0.327***
(0.079)
[0.109]
0.029
(0.051)
[0.009]
0.030
(0.159)
[0.009]
0.009***
(0.002)
[0.003]
0.096
(0.169)
[0.029]
0.176
(0.199)
[0.052]
0.021
(0.068)
[0.007]
-2418
2,808
Chronic Condition
0.089
(0.297)
[0.028]
-0.045**
(0.021)
[-0.014]
0.022
(0.040)
[0.007]
0.065
(0.055)
[0.021]
-0.109
(0.120)
[-0.033]
-0.011***
(0.003)
[-0.003]
0.361***
(0.079)
[0.121]
0.031
(0.051)
[0.010]
-0.023
(0.158)
[0.007]
0.010***
(0.002)
[0.003]
0.113
(0.169)
[0.034]
0.182
(0.199)
[0.053]
0.018
(0.068)
[0.006]
-2420
2,808
13
Generics
0.968***
(0.377)
[0.305]
-0.038*
(0.021)
[-0.013]
0.041
(0.041)
[0.013]
0.023
(0.057)
[0.007]
-0.122
(0.120)
[-0.037]
-0.010***
(0.003)
[-0.003]
0.321***
(0.077)
[0.107]
0.029
(0.051)
[0.009]
0.025
(0.159)
[0.008]
0.010***
(0.002)
[0.003]
0.120
(0.171)
[0.036]
0.173
(0.199)
[0.051]
0.017
(0.067)
[0.005]
-2417
2,808
Vaccination
0.489***
(0.180)
[0.163]
-0.035
(0.030)
[-0.012]
0.089
(0.057)
[0.030]
-0.033
(0.068)
[-0.011]
0.251
(0.196)
[0.089]
-0.013***
(0.003)
[-0.004]
0.342***
(0.088)
[0.119]
-0.002
(0.061)
[-0.001]
-0.010
(0.168)
[-0.003]
0.008***
(0.002)
[0.003]
0.236
(0.200)
[0.073]
0.316
(0.238)
[0.095]
-0.014
(0.081)
[0.005]
-1747
1,866
Downloaded by Cornell University Library At 06:58 10 September 2016 (PT)
Figure 1: Scatter Plot of Patient Satisfaction and the HEDIS Measures
14
Nephropathy screening, visit
w/nephrologist, ACEI/ARB
Comprehensive eye exam in
measurement year or a
negative retinal exam in prior
year
Children ages 3
months to 18 years
who were assessed
with URI
All diabetic members
All diabetic members
All diabetic members
All diabetic members
Treatment for
Children with Upper
Respiratory Infection
(“URI”)
Diabetes Care HbA1C
Diabetes Care - LDL
-C
Diabetes Care Nephropathy
Screening
Diabetes Care - Eye
Exam
Appropriate Testing
for Children with
Pharyngitis
15
LDL-C testing on all diabetic
patients
Patients should not be
dispensed a prescription for
antibiotic medication on or
within 3 days after the Index
Episode start date
Children 2–18 years of age
diagnosed with pharyngitis
and dispensed an antibiotic
must have a test for group A
streptococcus for the episode
Patients should not be
dispensed a prescription for
antibiotic medication on or
within three days after the
Index Episode start date
HbA1C testing on all diabetic
patients
Patients 18-64 who
had an outpatient
visit with any
diagnosis of acute
bronchitis (ICD 466)
Patients 2-18 years
of age with a
diagnosis of only
pharyngitis (ICD 462)
Appropriate Antibiotic
use with Acute
Bronchitis
Requirement
Population
Measure
Annually
Annually
Annually
Annually
Within 3
days of
Episode
Strep test
administered
in the 7-day
period
Within 3
days of
episode
Frequency
Appendix Table 1: HEDIS Clinical Quality Metrics
Specialty
Internal
Medicine, Family
Practice
Internal
Medicine, Family
Practice
Internal
Medicine, Family
Practice
Internal
Medicine, Family
Practice
Internal
Medicine, Family
Practice
Pediatrics,
Family Practice
Internal
Medicine,
Family Practice
Downloaded by Cornell University Library At 06:58 10 September 2016 (PT)
Chronic
Condition
Chronic
Condition
Chronic
Condition
Chronic
Condition
Antibiotic
Antibiotics
Antibiotics
Composite
Group
Patients with
persistent asthma.
Excludes members
with any history of
emphysema, chronic
obstructive
pulmonary disease,
cystic fibrosis, and
acute respiratory
failure.
Patients hospitalized
and discharged with
an AMI who do not
have a
contraindication to
beta blockers
Appropriate Asthma
Medicines
Beta Blocker after
Acute Myocardial
Infarction (“AMI”)
Patients who are
treated with diuretics
during the
measurement year
Patients with
ischemic vascular
disease or
discharged alive with
PCI, CABG, AMI
Patients who are
treated with
Anticonvulsants
during the
measurement year
Patients who are
treated with Digoxin
during the
measurement year
Annual Monitoring of
Persistent
Medications:
Diuretics
Annual Monitoring of
Persistent
Medications: Digoxin
Annual Monitoring of
Persistent
Medications:
Anticonvulsants
Lipid Screening Cardiac Conditions
16
Patients need prescription for
beta blocker for at least 6
months post discharge
Patients have at least one
serum potassium and either a
serum creatinine or a BUN
test during the measurement
year
Patients have at least one
serum potassium and either a
serum creatinine or a BUN
test during the measurement
year
Patients have at least 1 claim
for an asthma controller
medication
Patients have at least 1
serum drug measurement (for
the prescribed drug) during
the measurement year
LDL-C testing on all patients
Continually
for 6 month
period
Annually
Annually
Annually
Annually
Annually
Internal
Medicine, Family
Practice
Internal
Medicine, Family
Practice,
Pediatrics
Internal
Medicine, Family
Practice
Internal
Medicine, Family
Practice
Internal
Medicine, Family
Practice
Internal
Medicine, Family
Practice
Downloaded by Cornell University Library At 06:58 10 September 2016 (PT)
Chronic
Condition
Chronic
Condition
Chronic
Condition
Chronic
Condition
Chronic
Condition
Chronic
Condition
Patients 50-80
Pap smear performed
Females 21-64
Colorectal Cancer
Screening
Document BP
All members > 18
Blood Pressure
(“BP”) Measurement
Cervical Cancer
Screening
17
Fecal occult blood ([“FOBT”],
gFOBT, or iFOBT) test in
current year, or flexible
sigmoidoscopy in the past 5
years, or double contrast
barium enema within the past
5 years, or colonoscopy in the
past 10 years
Patient had mammogram
during year or within the past
year
Females 40-69
Breast Cancer
Screening
Prescribe medications that
come in generic form or
document medical necessity
of brand medication
Annual preventative visit
Patients need prescription for
beta blocker
All members > 18
Patients 18 years or
older who have been
diagnosed with heart
failure any time in the
past
All members
Annual Preventative
Visit
Generic Dispensing
Rate
Beta Blocker for
Heart Failure
Varies
Every 36
months
Annually
Bi-annually
Annually
N/A
Continually
Internal
Medicine, Family
Practice,
Gynecology
Internal
Medicine, Family
Practice
Internal
Medicine, Family
Practice
Internal
Medicine, Family
Practice,
Gynecology
All
All
Internal
Medicine, Family
Practice
Downloaded by Cornell University Library At 06:58 10 September 2016 (PT)
Preventative
Preventative
Preventative
Preventative
Preventative
Generic
Chronic
Condition
Patients turning 2 in
measurement period
Childhood
Immunizations –
Measles, Mumps,
Rubella (“MMR”)
Childhood
Immunizations Varicella
Patients turning 2 in
measurement period
All members 13+
All members 0-18
Tobacco Use Status
Well Child Visits
Osteoporosis
Management
Patients age 67
years old older who
do not have a
diagnosis of
glaucoma or
glaucoma suspect
anytime in the past
Females 67+ who
suffered a fracture
Glaucoma Screening
18
VZV immunization
Perform bone mineral density
or prescribe Rx for
osteoporosis within 12
months before or 6 months
after a fracture
Document tobacco use status
Age 0-1: At least 5 visits, age
2-18: At least 1 visit
MMR immunization
Patients need glaucoma
screening from an optometrist
or ophthalmologist
On or before
2nd birthday
On or before
2nd birthday
Annually
Annually
12 months
before - 6
months after
a fracture
Every 2
years
Pediatrics,
Family Practice
All
Pediatrics,
Family Practice
Pediatrics,
Family Practice
Internal
Medicine, Family
Practice,
Gynecology
Internal
Medicine, Family
Practice
Downloaded by Cornell University Library At 06:58 10 September 2016 (PT)
Vaccine
Vaccine
Preventative
Preventative
Preventative
Preventative
Appendix Table 2: Probit Regression with Patient Satisfaction=10 as the Dependent Variable
(Column title indicates the HEDIS physician quality measure included as an independent variable in each
regression. Standard errors are in parenthesis and the marginal effect is in brackets;
*** indicates significant at the .01 level, ** at the .05 level, * at the .10 level)
Physician Clinical
Quality Variable
Downloaded by Cornell University Library At 06:58 10 September 2016 (PT)
Doctor Volume
(wRVU*1,000)
Practice Volume
* 10,000
Allopathic (MD)
Board Certified
Physician Age
Physician Male
Patient Male
Patient
Caucasian
Patient Age
Insurance is
Self Pay
Medicaid
Medicare
Log Likelihood
Antibiotics
1.196***
(0.298)
[0.375]
-0.065***
(0.023)
[-0.020]
0.019
(0.042)
[0.006]
0.070
(0.058)
[0.022]
-0.505***
(0.155)
[-0.132]
-0.010***
(0.003)
[-0.003]
0.381***
(0.080)
[0.128]
0.014
(0.054)
[0.004]
0.057
(0.172)
[0.018]
0.009***
(0.002)
[0.003]
0.066
(0.191)
[0.020]
0.184
(0.210)
[0.054]
0.043
(0.071)
[0.014]
-1537
Preventative
-0.394
(0.262)
[-0.124]
-0.034
(0.023)
[-0.010]
0.012
(0.043)
[0.004]
0.059
(0.060)
[0.019]
-0.134
(0.123)
[-0.041]
-0.011***
(0.003)
[-0.004]
0.300***
(0.082)
[0.100]
0.007
(0.054)
[0.002]
0.054
(0.170)
[0.017]
0.009***
(0.002)
[0.003]
0.030
(0.193)
[0.009]
0.187
(0.212)
[0.055]
0.050
(0.071)
[0.155]
-1544
Chronic Condition
0.076
(0.319)
[0.024]
-0.045**
(0.022)
[-0.014]
0.018
(0.042)
[0.006]
0.084
(0.058)
[0.027]
-0.129
(0.122)
[-0.039]
-0.010***
(0.003)
[-0.003]
0.330***
(0.079)
[0.111]
0.008
(0.054)
[0.003]
0.049
(0.170)
[0.016]
0.009***
(0.002)
[0.003]
0.043
(0.193)
[0.014]
0.192
(0.212)
[0.056]
0.047
(0.070)
[0.015]
-1545
19
Generics
1.153***
(0.414)
[0.363]
-0.037*
(0.022)
[-0.012]
0.037
(0.043)
[0.012]
0.034
(0.060)
[0.011]
-0.144
(0.122)
[-0.043]
-0.010***
(0.003)
[-0.003]
0.284***
(0.079)
[0.095]
0.006
(0.054)
[0.002]
0.049
(0.171)
[0.016]
0.009***
(0.002)
[0.003]
0.056
(0.195)
[0.017]
0.182
(0.212)
[0.054]
0.046
(0.071)
[0.014]
-1542
Vaccination
0.567***
(0.197)
[0.189]
-0.032
(0.031)
[-0.011]
0.027
(0.060)
[0.009]
-0.016
(0.071)
[-0.005]
0.252
(0.200)
[0.089]
-0.012***
(0.003)
[-0.004]
0.300***
(0.094)
[0.104]
-0.015
(0.065)
[-0.005]
0.005
(0.178)
[0.002]
0.007***
(0.002)
[0.002]
0.210
(0.216)
[0.066]
0.330
(0.249)
[0.099]
0.008
(0.086)
[0.025]
-1084
Downloaded by Cornell University Library At 06:58 10 September 2016 (PT)
Appendix Table 3: Ordered Probit Regression with Patient Satisfaction as the Dependent Variable
and a Relative HEDIS Measure of Physician Quality
(Column title indicates the HEDIS physician quality measure included as an independent variable in each
regression as a dummy variable that equals 1 if the physician’s HEDIS score is above the median score.
Standard errors are in parenthesis and the marginal effect on the probability of a satisfaction score of 10
is in brackets; *** indicates significant at the .01 level, ** at the .05 level, * at the .10 level)
Antibiotics Preventative Chronic Condition
Generics
Vaccination
Physician Clinical
0.125***
0.010
-0.035
0.156***
0.154***
Quality Score is
(0.053)
(0.057)
(0.057)
(0.068)
(0.064)
Above the Median
[0.039]
[0.003]
[-0.011]
[0.049]
[0.051]
Doctor Volume
-0.029
-0.023
-0.020
-0.009
-0.022
(wRVU*1,000)
(0.021)
(0.022)
(0.021)
(0.021)
(0.030)
[-0.009]
[-0.007]
[-0.006]
[-0.003]
[-0.007]
Practice Volume *
0.076***
0.083***
0.087***
-0.041
0.083
10,000
(0.034)
(0.034)
(0.034)
(0.040)
(0.060)
[0.024]
[0.026]
[0.027]
[-0.013]
[0.028]
Allopathic (MD)
0.021
0.030
0.027
-0.083
-0.031
(0.055)
(0.057)
(0.055)
(0.070)
(0.068)
[0.007]
[0.009]
[0.007]
[-0.026]
[-0.010]
Board Certified
-0.127
-0.060
-0.043
-0.052
0.262
(0.123)
(0.124)
(0.125)
(0.121)
(0.196)
[-0.038]
[-0.018]
[-0.013]
[-0.016]
[0.093]
Physician Age
-0.007***
-0.008***
-0.008***
-0.009***
-0.013***
(0.003)
(0.003)
(0.003)
(0.003)
(0.003)
[-0.002]
[-0.002]
[-0.003]
[-0.003]
[-0.004]
Physician Male
0.236***
0.194***
0.178***
0.193***
0.334***
(0.070)
(0.073)
(0.069)
(0.066)
(0.088)
[0.077]
[0.62]
[0.057]
[0.062]
[0.115]
Patient Male
0.010
0.005
0.003
0.008
-0.006
(0.051)
(0.051)
(0.051)
(0.051)
(0.061)
[0.003]
[0.002]
[0.001]
[0.002]
[-0.002]
Patient Caucasian
0.028
0.023
0.022
0.026
-0.025
(0.159)
(0.158)
(0.157)
(0.158)
(0.168)
[0.009]
[0.007]
[0.007]
[0.008]
[-0.008]
Patient Age
0.010***
0.010***
0.010***
0.010***
0.007***
(0.002)
(0.002)
(0.002)
(0.002)
(0.002)
[0.003]
[0.003]
[0.003]
[0.003]
[0.002]
Insurance is
0.080
0.069
0.064
0.060
0.248
Self Pay
(0.162)
(0.164)
(0.164)
(0.164)
(0.193)
[0.024]
[0.021]
[0.019]
[0.018]
[0.077]
Medicaid
0.112
0.109
0.104
0.089
0.339
(0.190)
(0.191)
(0.191)
(0.191)
(0.239)
[0.033]
[0.033]
[0.031]
[0.027]
[0.101]
Medicare
-0.023
-0.021
-0.019
-0.016
0.019
(0.067)
(0.067)
(0.067)
(0.069)
(0.081)
[-0.007]
[-0.007]
[-0.006]
[-0.005]
[0.006]
Log Likelihood
-2514
-2518
-2517
-2515
-1747
20