The Prevalence of Type 2 Diabetes Mellitus in a Wisconsin Hmong Patient Population Kevin Koobmoov Thao MD Primary Care Research Fellow UW Department of Family Medicine Outline • • • • The Wisconsin Hmong Diabetes in the Hmong population Results Next Steps Population Totals • 33,791 Hmong living in Wisconsin according to 2000 US Census data - .63% of the states total population • 57.1% of the Hmong are under 18 years old and 2/3 are under 24 • Average size for Hmong family 6.4 persons • This population has experienced a 100% growth since the 1990 census. The Prevalence of Type 2 Diabetes Mellitus in a Wisconsin Hmong Patient Population Purpose: To compare the prevalence of diabetes in the Hmong subpopulation of the University of Wisconsin Department of Family Medicine ambulatory care population to non-Hispanic white patients. The UW Clinical-Public Health Data Exchange Pilot • Extraction of patient electronic medical records of the UW DFM clinic population from years 2007-2009 • Outpatient visits to UW DFM clinics documented with the EpicCare© EMR • Project Objective: To link patient electronic health records with public health databases to facilitate multidimensional investigations of population health. Multi-Level Modeling and Data Mining of Disease Risk, Disparity, and Health Outcome Quality Outcomes = Asthma Diabetes CVD / CHF Immunizations Obesity Hypertension Smoking Alcohol A1c level LDL Patient Factors + Age Gender Race/ethnicity Co-morbidities Medications Language Insurance Urban / Rural Clinician Factors + Age Gender Certifications Graduation date Years of practice Clinic Factors + Location Capabilities Processes Census Block Group HDL BP Hospitalizations Community Factors Census Block Group: Poverty Education level Built environment: Traffic Recreation / parks Safety / crime Psycho-demographics Restaurant mix Fast food sales Fresh fruit & vegetable sales / consumption Public Health Programs Health Care -Process factors (e.g, time to repeat follow-up) Electronic Health Record & Hospitalization Data Census / ESRI Data + PH Information Systems Study Data Selection • EMR extract data contains demographic and health information on 192,201 unique ambulatory care patients • 2.5 million clinical encounters • Patient Confidentiality was protected by removal of identifying information before extraction ▫ Name, exact birth date, SS#, exact address, HIV diagnosis information, Medical Record Numbers Population Selection by Race, Ethnicity and Language Race/Ethnicity Language Non-Hispanic White 157,526 (82.0%) Comparison Group Non-Hispanic White 157,526 (82.0%) Total patient Population 192,201 Selected Group Non-Hispanic Asian 5743 (2.99%) Hmong 611 (0.32%) Hmong 611 (0.32%) Variable Definitions • Race/Ethnicity/Language: coded from the EMR fields • Age: Obtained from the EMR and categorized into appropriate categories • Body Mass Index (BMI): calculated from the earliest weight and height measurements in the patients record ▫ ▫ ▫ ▫ ▫ Then classified into categories Normal weight (BMI<25) Overweight (BMI 25-30) Obese (BMI >30) BMI Missing • Health Insurance: Encoded from the EMR ▫ Commercial, No Insurance, Workers Compensation ▫ Medicare ▫ Medicaid Type 2 Diabetes Diagnosis Variables • International Classification of Disease 9th Revision (ICD-9) diagnosis codes ▫ 250.x0 and 250.x2 where x can be variable • Laboratory Values ▫ ▫ ▫ ▫ Fasting glucose >126 mg/dL x 2 2 hour Glucose Tolerance Test > 200 x 2 Random glucose > 200 x 2 Hgb 1 Ac > 6.5% • Medication list ▫ Medications listed under the classification “antidiabetes medication” in the EMR (excluding Metformin) Type 2 Diabetes Diagnosis Algorithm DM Type 2 was diagnosed if : 1. Both the encounter and diagnosis fields were consistent with diagnosis Or 2. Either the encounter or diagnosis field indicated a diagnosis and the diagnosis was confirmed by laboratory or medication list support of diabetes diagnosis Cases of inconsistency of ICD-9 codes within or between encounter and problem list fields were also addressed with another algorithm to determine type 2 diabetes diagnosis Characteristics of Hmong and non-Hispanic white populations Characteristic Number (percent of total) Hmong Non-Hispanic White 611 157,526 Average age years 30.4 + 0.97 37.4 + .05 Age range in years (percentage of total) 0-17 257 (42.6%) 30503 (19.4%) 18-54 227 (37.2%) 93374 (59.3%) 55-64 73 (12%) 17802 (11.3%) 65+ 54 (8.8%) 158747 (10%) 611 157526 40.4% 45.7% Total Sex % male Characteristics of Hmong and non-Hispanic white populations cont. Characteristic Mean BMI (kg/m2) Hmong Non-Hispanic White 24.0 + .34 26.9 + .02 BMI Category (percent of total) Underweight and Normal Weight 159 (26.0%) 42236 (26.8%) Overweight 106 (17.4%) 31599 (20.1%) 76 (12.4%) 32434 (20.6%) BMI Missing 270 (44.2%) 51257 (32.5%) Health Insurance (percent of total) Commercial/Workers comp/ No Insurance Medicaid 153 (25.0%) 130606 (82.9%) 413 (67.6%) 9987 (6.3%) 45 (7.36%) 16933 (10.8%) Obese Medicare UW DFM Data of Hmong Patients in Wisconsin UW DFM Data of Hmong Patients in Wisconsin Census 2000 Data on the Hmong of Wisconsin Crude Diabetes Prevalence Hmong Number with Diabetes Total Study Population Adults (age >18) 41 Diabetes Prevalence Non-Hispanic White Number with Diabetes Diabetes Prevalence χ^2 pvalue Odds Ratios 6.7% 7590 4.8% 0.029 1.4 41 11.6% 7583 6.0% <.001 2.1 Crude Diabetes Prevalence Hmong Non-Hispanic White Number with Diabetes Diabetes Prevalence Number with Diabetes Diabetes Prevalence χ^2 pvalue Odds Ratios Age Range 0-17 0 0.0% 7 0.0% 18-54 12 5.3% 2622 2.8% 55-64 19 26.0% 65+ 10 0.02 1.9 2176 12.2% <0.001 2.5 18.5% 2785 17.6% 0.856 1.1 7 4.4% 369 0.9% <.001 5.2 Over weight 15 14.2% 1160 3.7% <.001 4.3 Obese 11 14.5% 3920 12.1% 0.524 1.2 BMI Normal weight Multivariable Logistic Regression Analysis Non-Hispanic White (OR) Model 1 Model 2 1.0 1.0 Wald χ^2 P value 1.4 (1.0-2.0) 0.042 1.7 (1.2-2.5) 0.003 Hmong (OR) OR is the odds ratio for diabetes (95% CI). Model 1 is adjusted for age, sex, and insurance Model 2 is adjusted for age, sex, insurance and BMI Limitations 1. Selection Bias of the study population UW DFM ambulatory care population size large, but non-random sample of Wisconsin residents. Questions of generalizability. 2. Selection Bias of the Hmong sample Language field utilized for interpretive services. Unknown what proportion of Hmong are listing Hmong as language. Hmong ethnicity not an option for ethnicity coding. 3. Missing BMI data 44.2% and 32.5% of records were missing height and weight data to calculate BMI BMI missing category was created and included in statistical analysis Models including and excluding BMI examined Conclusion • This study supports previous study conclusions that health care providers should be aware of the increase risk for diabetes in the Hmong population (Her 2005, McCarty 2005). • Physicians should consider screening for glucose intolerance in the Hmong patient population starting at younger ages and lower BMI (McCarty 2005). • Further population based research should be conducted to evaluate the prevalence of diabetes in the Wisconsin Hmong population. Next Steps? More Epidemiology (miniSHOW?) Risk Factor Exploration • Diabetes Prevention ▫ Community Based Participatory Research ▫ Increase physical activity ▫ Improve nutrition • Diabetes Management ▫ Clinical effectiveness trials of culturally appropriate Diabetes education ▫ Improve diabetes self management education Acknowledgments MPH Program/Research Mentors MPH Preceptor: Lawrence Hanrahan PhD Director of Public Health Informatics Chief Epidemiologist Bureau of Health Information, Wisconsin Division of Public Health Research Mentor: Brian Arndt MD Faculty UWSMPH Department of Family Medicine MPH Capstone Committee Chair: John Frey MD Professor Department of Family Medicine Head of Community Engagement Institute for Clinical and Translational Research University of Wisconsin School of Medicine and Public Health Public Health Informatics Specialist: Aman Tandias MS Bureau of Health Information, Wisconsin Division of Public Health Theresa Guilbert MD Faculty UWSMPH Department of Pediatrics Barbara Duerst MS, RN MPH Associate Program Director UWSMPH UW Department of Family Medicine Mentors The work presented here was carried out while Kevin Thao was a Primary Care Research Fellow supported by a National Research Service Award (T32HP10010) from the Health Resources and Services Administration to the University Of Wisconsin Department Of Family Medicine Bruce Barrett MD, PhD Director of the Primary Care Research Fellowship Department of Family Medicine MaryBeth Plane PhD Director of DFM Research Services Department of Family Medicine Terry Little University Services Program Associate Hmong/Madison Community Mentors Fuechue Thao Public Health Clinic Aide Madison Dane County Public Health Susan Webb-Lukomski RN, BSN Madison Dane County Public Health References • • • • • • • • • • • Culhane-Pera, K., Peterson, K. a, Crain, a L., Center, B. a, Lee, M., Her, B., et al. (2005). Group visits for Hmong adults with type 2 diabetes mellitus: a pre-post analysis. Journal of health care for the poor and underserved, 16(2), 315-27. doi: 10.1353/hpu.2005.0030. Culhane-Pera, K. a, Her, C., & Her, B. (2007). "We are out of balance here": a Hmong cultural model of diabetes. Journal of immigrant and minority health / Center for Minority Public Health, 9(3), 179-90. doi: 10.1007/s10903-006-9029-3. Devlin, H., Roberts, M., Okaya, A., & Xiong, Y. M. (2006). Our lives were healthier before: focus groups with African American, American Indian, Hispanic/Latino, and Hmong people with diabetes. Health promotion practice, 7(1), 47-55. doi: 10.1177/1524839905275395. Franzen, L., & Smith, C. (2009a). Differences in stature , BMI , and dietary practices between US born and newly immigrated Hmong children q. Social Science & Medicine, 69(3), 442-450. Elsevier Ltd. doi: 10.1016/j.socscimed.2009.05.015. Franzen, L., & Smith, C. (2009b). Acculturation and environmental change impacts dietary habits among adult Hmong. Appetite, 52(1), 173-83. doi: 10.1016/j.appet.2008.09.012. Her, C., & Mundt, M. (2005). Risk prevalence for type 2 diabetes mellitus in adult Hmong in Wisconsin: a pilot study. WMJ : official publication of the State Medical Society of Wisconsin, 104(5), 70-7. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/16138520. Himes, J. H., Story, M., Czaplinski, K., & Dahlberg-Luby, E. (1992). Indications of early obesity in low-income Hmong children.pdf. American Journal of Diseases of Children, 146(1), 67-9. Koltyk, J. A. (1997). New Pioneers in the Heartland: Hmong Life in Wisconsin. Allyn & Bacon. Mccarty, D. J. (2005). Glucose intolerance in Wisconsin ’ s Hmong population. Wisconsin Medical Journal, 104(5), 13-15. Stang, J., Kong, A., Story, M., Eisenberg, M. E., & Neumark-Sztainer, D. (2007). Food and weight-related patterns and behaviors of Hmong adolescents. Journal of the American Dietetic Association, 107(6), 936-41. doi: 10.1016/j.jada.2007.03.003. University of Wisconsin and Applied Population Laboratory. (2002). Wisconsin ’ s Hmong Population. Thank You, Questions? “The ability to ask the right question is more than half the battle of finding the answer.” Thomas J. Watson Other Unexplored Risk Factors Obesity Risk Factors Related to Environmental Change Poor Dietary Habits Heart Disease Obesity Physical Inactivity Diabetes Cancer Obesity Risk Factors Related to Environmental Change Poor Dietary Habits Heart Disease Obesity Physical Inactivity http://kcortiz.photoshelter.com/gallery-image/FORCEDREBELLION-HMONG-CIA-VETERANS-OF-THE-SECRETWAR/G0000ddMEaqXj9SU/I0000iOTLjb2km_w Diabetes Cancer Diabetes in 5 minutes to the Hmong Chronic disease of insulin (kua fajsiv) Two types Risk Factors Age Ethnicity Obesity ○ Poor diet and physical inactivity Adverse Health Outcomes Treatments • One Type • Death Sentence • Risk Factors ▫ America ▫ Weather ▫ Anguish/Loss of Home ▫ Obesity Poor diet and physical inactivity • Adverse Health Outcomes • Herbs/Nothing Limitations Continued 1. 2. 3. Selection Bias of the study population 2009 BRFSS reported 80.8% of Americans had primary care providers and 81.65% were seen for routine health check up in the last two years The Wisconsin Family Health Survey, 2001-2005 indicates 92% of surveyed Wisconsin residents had a place of routine health care and 87% of Wisconsin Asians reported having a place for routine healthcare Unknown – Primary care utilization patterns of Hmong in Wisconsin Diabetes screening rates of Hmong in Wisconsin primary care clinics Selection Bias of the Hmong sample Unknown – Proportion of Hmong patients utilizing interpretive services Potential surname based analysis method possible, but not validated Missing BMI data Patient Race and Ethnicity Breakdown Ethnicity Frequency Percent Frequency Percent 2953 1.54% 2260 1.18% Missing Missing 7858 4.09% American Indian or Hispanic/Latino 1761 0.92% Alaska Native Not Hispanic or 171758 89.36% 5743 2.99% Latino Asian Patient Refuses Black or African 7584 3.95% to Answer 1050 0.55% American Native Hawaiian or 8582 4.47% 245 0.13% Unknown Other Pacific Islander 192201 100.00% 165700 86.21% Total White Patient Refuses to 1379 0.72% Answer 7529 3.92% Unknown Total 192201 100.00% Race Diabetes Diagnosis Algorithm Criteria diabetes_p diabetes_e diabetes_p or diabetes_e diabetes diabetes1_p diabetes1_e diabetes1_p or diabetes1_e diabetes1 diabetes2_p diabetes2_e diabetes2_p or diabetes2_e diabetes2 Patient Count Prevalence 9,788 5.09% 10,452 5.44% 11,483 5.97% 9,804 5.10% 678 0.35% 737 0.38% 828 0.43% 740 0.39% 8,975 4.67% 9,673 5.03% 10,605 5.52% 9,034 4.70% Cases where the patient has ICD 9 codes for both type 1 and type 2 For patients with both diabetes type 1 and type 2 ICD 9 codes, determine which is the most likely one to be correct using the following algorithm. Use 250.0x only (omit ICD 9 codes for diabetes complications) • Rationale: Some users may not have realized the diabetes complications have type-specific codes. Therefore the codes for diabetes complications are not reliable in resolving conflict between types. Look at the latest 3 entries only, using encounter date for encounter dx and entry date for problem list dx. • Rationale: Data entry errors decrease over time as users become more familiar with the system. Therefore we can expect the later entries to be more reliable. Patients may have been initially misdiagnosed and the diagnosis was later corrected. Take the majority of the latest 3 entries. If there is only one entry, then use that entry's dx. If there are two entries and they are of different types: if the dates are different, take the more recent one if the dates are the same, leave the type unspecified