Title: Changes in prescription fills for diabetes after the implementation of Affordable Care Act Medicaid expansions

Authors: Rebecca Myerson 1 MPH, PhD; Tianyi Lu1 BA, MA; Elbert Huang2 MD, MPH, FACP Affiliations: 1 University of Southern California School of Pharmacy and Schaeffer Center for Health Policy and Economics, Los Angeles, CA, 90007, USA 2 Department of Medicine, University of Chicago, Chicago, IL, 60637, USA Corresponding author: Rebecca Myerson University of Southern California School of Pharmacy Schaeffer Center for Health Policy and Economics Verna and Peter Dauterive Hall Office 414E 635 Downey Way Los Angeles, CA, 90089, USA Email: [email protected]

Abstract Importance: Diabetes is a top contributor to avoidable burden of disease in the United States due to its high prevalence and incomplete detection and treatment. The Affordable Care Act (ACA) Medicaid expansions expanded insurance coverage in low-income adults, a group at elevated risk of untreated diabetes. Objective: Determine whether Medicaid expansions after the ACA were associated with increased pharmaceutical treatment of diabetes. Design: Differences-in-differences analysis of over 1 billion diabetes prescriptions fills from January 2008 to December 2015, in states that did vs. did not expand Medicaid after the Affordable Care Act. Setting: All states of the United States. Participants: Residents of the United States. Exposures: Residing in a state where Medicaid was expanded after the ACA. Main outcomes and measures: The primary outcome was diabetes prescription fills per 1,000 population. Secondary outcomes were insulin prescription fills and biguanide prescription fills per 1,000 population; these are the first-line pharmaceutical treatments for insulin-dependent and non-insulin-dependent diabetes, respectively. Outcomes were calculated annually by patient age group, sex, and state. We measured changes in these outcomes associated with ACA Medicaid expansions using multivariable differences in differences regression models, adjusting for patient age and sex, temporal trends, state characteristics, and changes in economic conditions on the state-level. Results: Prior to the ACA, trends in diabetes prescription fills were not significantly different between states with vs. without subsequent Medicaid expansions. Expansions were associated with significant increases in diabetes prescription fills (34.6 per 1,000 population, 95% CI 5.8 to 63.5), significant increases in both insulin and biguanides. We found no significant increase of diabetes prescription fills for patients aged 65-69, an age group eligible for Medicare. Age groups with higher prevalence of diabetes prior to the ACA, such as people aged 55-59, showed larger increases in treatment after Medicaid expansions. Based on the clinical literature, we predict that these increases in treatment with insulin and biguanides could prevent over 11,000 and 6,000 diabetes related health events, respectively, over the next 10 years. Conclusions and Relevance: Affordable Care Act Medicaid expansions are associated with significant increases in diabetes prescription fills. These changes are predicted to have significant health consequences. The findings are relevant to ongoing policy discussions.

Introduction Expanding access to prescription medications for diabetes is critical for improving population health in the United States. Diabetes is a top source of years of life lost in the United States, both as a top-15 cause of death and as a risk factor for heart disease, the top cause of death (Danaei et al. 2009; Heron 2013; U.S. Burden of Disease Collaborators 2013). Many of the complications of diabetes can be prevented by effective primary care, including glucose-lowering drugs as needed (Farley et al. 2010; Huang, Meigs, and Singer 2001; Kelly et al. 2009; Ray et al. 2009; UK Prospective Diabetes Study group 1998b). Yet, not all people with diabetes receive the medications they need (Hill et al. 2011; Gakidou et al. 2011; Roehrig and Daly 2015). The number of adults with diagnosed diabetes who did not report taking diabetes medication reached 3.1 million shortly after the Affordable Care Act was passed, and was on an increasing trend in prior years (Centers for Disease Control and Prevention (CDC), 2012). Medicaid expansions after the Affordable Care Act (ACA) provided free insurance to adults with incomes less than 138% poverty level, thereby lowering out-of-pocket drug costs for a population at elevated risk for untreated diabetes (Decker et al. 2013). The earliest expansions started in January 2014; 30 states plus Washington, D.C. expanded Medicaid in 2014 or 2015. Studies have found that implementation of ACA Medicaid expansions was associated with an increase in the numbers of Medicaid prescriptions per enrollee and a drop in cost-related prescription nonadherence (Mulcahy, Eibner, and Finegold 2016; Kennedy and Wood 2016; Wen, Borders, and Druss 2016). Less is known about the effect on prescription drug utilization for specific chronic conditions such as diabetes. The most direct evidence shows an increase in diabetes prescriptions filled using Medicaid insurance after ACA Medicaid expansions, with limited crowd-out from other types of insurance during the first 15 months (Ghosh, Simon, and Sommers 2017). We are not aware of any studies that measure the total change in diabetes prescription fills during the first 24 months of the ACA Medicaid expansions. We are also not aware of any studies that report data by type of diabetes medication, a critical determinant of potential health effects. Although federal funding to support the Medicaid expansions is decided on an annual basis, the clinical literature suggests that the health impacts of improved diabetes treatment with insulin or biguanide medication would be detectable after 10 years (UK Prospective Diabetes Study group 1998b). To inform ongoing policy discussions, this paper assesses the impact of ACA Medicaid expansions on care for diabetes using prescription medications, and maps these changes to potential health effects if gains in treatment are maintained. We first present estimates of the total change in diabetes prescription fills per 1,000 associated with Medicaid expansions, as well as changes by type of medication, sex, and age. We use the estimates by age to conduct multiple checks of the data. We then combine these data with estimates from the clinical literature to provide a rough calculation of the possible health effects for people with diabetes going forward.

Methods Study design We used a quasi-experimental differences-in-differences design to distinguish changes in diabetes prescription fills related to ACA Medicaid expansions from temporal trends. In this method, trends in diabetes prescription fills before vs. after Medicaid expansions (first difference) are compared in states with vs. without ACA Medicaid expansions (second difference). January 2008 through December 2013 comprised the pre-intervention period and January 2014 through December 2015 comprised the post-intervention period. We defined ACA Medicaid expansion states as those states that expanded Medicaid in 2014 (27 states plus Washington D.C.) or 2015 (three states: Indiana, Montana and Pennsylvania), and classified the other 20 states as non-expansion or control states. Data We measured fills of diabetes prescriptions using IMS Health’s Xponent database. IMS Health gathers an 80% convenience sample of prescription fills and uses patented methods to estimate the remaining 20%. As of 2011, the IMS Health Xponent sample included information from about 38,000 pharmacies and captured sales in mass merchandisers, independent and food-store pharmaceuticals, mail service pharmacy outlets, and long-term care facilities. The data cover all purchases at included pharmacies including cash-only transactions; individuals enter the data as soon as they fill their first prescription. Our sample included 8 years of data, comprising over 1 billion diabetes prescription fills for patients aged 20-69. These data include 420 million prescription fills from states that did not expand Medicaid after the ACA and 591 million prescription fills from ACA Medicaid expansion states. We had information on patients’ age group, sex, the USC code of medication filled, the month and year of the fill, as well as the county and state of the fill. Prescriptions for insulin, a treatment for Type 1 and insulin-dependent Type 2 diabetes, and prescriptions for biguanides, a first-line treatment for non-insulin-dependent Type 2 diabetes, were identified based on the USC codes. Patients’ ages were classified using the following age groups: 20-25, 26-29, 30-34, 35-44, 45-54, 55-59, 60-64, and 65-69. Sex was classified as male or female. We paired the data on prescription fills by age group, sex, state, and year with the population estimates for that age-sexstate-year from the intercensal population data (U.S Census Bureau 2016). We also combined these data with annual unemployment rates for each state from the Bureau of Labor Statistics (Bureau of Labor Statistics 2017). We used the 2013-2014 National Health and Nutrition Examination Survey (NHANES) data to calculate the prevalence of diabetes just prior to and during the Medicaid expansions for each of our age groups of interest (National Center for Health Statistics 2008).

To facilitate calculations of the possible health effects of increased treatment, we extracted the number of patients needed to treat with insulin vs. biguanides to prevent a diabetes-related clinical event from the existing literature. We extracted these data from reports of randomized, controlled clinical trials (UK Prospective Diabetes Study group 1998a, 1998b). We also extracted the average number of insulin prescriptions filled per year among patients using insulin, as well as the average number of biguanide prescriptions filled per year among patients using biguanides, using the 2014 Medical Expenditure Panel Survey (MEPS) data (Agency for Healthcare Research and Quality 2014). Statistical analysis We used the differences-in-differences method to model changes in diabetes prescription fills associated with ACA Medicaid expansions. To account for the fact that the number of prescriptions filled mechanically increases or decreases when a location experiences in-migration or out-migration, we used negative binomial models where current population was the exposure variable (U.S Census Bureau 2016). Our primary outcome was total diabetes prescriptions filled per 1,000 population. Secondary outcomes included biguanide and insulin prescriptions filled per 1,000 population. We calculated these outcomes by year, age group, sex, and state. We clustered standard errors on the state level to account for the correlation of data for different age-sex groups within the same state, correlation of data from the same state over time, and the state-level nature of the Medicaid expansions. The validity of the difference-in-differences method rests on the assumption that in the absence of the policy intervention of interest, trends in counties with different policy interventions would have remained parallel. Although this assumption is not testable, parallel trends prior to the policy intervention provide evidence of the assumption’s plausibility. As such, we tested for parallel trends in Medicaid expansion vs. non-expansion states prior to the ACA, using data for years 2008 to 2013, for each model presented. We addressed possible residual confounding in several ways. First, year indicator variables were included in the model to control for year-specific shifts, such as changes in the economy. Second, state indicator variables were included to control for state-level characteristics that were unbalanced but remained fixed over time across the study period. Third, we adjusted for the age group and sex of the person filling the prescription. Fourth, we accounted for the fact that local trends in unemployment can determine the size of the population eligible for Medicaid, by adjusting for annual state-level unemployment rates. As a data check, we examined whether the prevalence of diabetes by age group was associated with changes in prescription fills after Medicaid expansions. To accomplish this, we stratified the models by age group and compared with the prevalence of diabetes calculated using NHANES

data from 2013-2014. We incorporated survey design variables to account for the complex, multi-stage sampling design of the NHANES survey. All analyses were performed using Stata/MP version 14. Results Table 1 presents baseline characteristics of states that did vs. did not expand Medicaid for adults under 138% federal poverty level during our sample period. Prescription fills for diabetes showed an increasing trend prior to the ACA, in both Medicaid expansion states and nonexpansion states. See Figure 1. For each outcome and age group, we failed to reject the null hypothesis that trends in our outcome of interest were similar in Medicaid expansion and nonexpansion states prior to the ACA. See Table S4 and Error! Reference source not found.. ACA Medicaid expansions were associated with increases in prescription fills for diabetes among adults aged 30-64 by 34.6 per 1,000 (95% CI 5.8 to 63.5). Men and women experienced similar, statistically significant increases. Biguanides accounted for the majority of diabetes prescriptions prior to the ACA, and accounted for the majority of the additional prescriptions associated with Medicaid expansions. However, insulin prescription fills showed a larger percentage increase after Medicaid expansions. See Table 2. The significant relationship between Medicaid expansions and diabetes prescription fills disappeared as expected for adults aged 65-69. Because these adults had just aged into eligibility for Medicare insurance, eligibility for Medicaid insurance was expected to have a smaller impact. See Table 3. Among age groups below the Medicare eligibility threshold, age groups with higher prevalence of diabetes experienced larger increases in treatment after Medicaid expansions. For example, 19 out of 100 people aged 55-59 had diabetes in 2013-2014 compared to only 4 out of 100 people aged 30-34, according to the NHANES data. Accordingly, people aged 55-59 were more likely than people aged 30-34 to fill additional diabetes prescriptions after Medicaid expansions. See Figure 2. The correlation between changes in prescription fills for diabetes after Medicaid expansions and diabetes prevalence was 0.27 (p<0.05). Discussion This study analyzed the associations between Medicaid expansions and prescription fills for diabetes medications per 1,000 population by age, sex, and type of medication. Our analysis leverages a large commercial dataset that captures over 1 billion diabetes prescription fills in the United States from January 2008 to December 2015. Rates of diabetes prescription fills were already increasing nationally prior to the first Medicaid expansions. Subsequently, populations in Medicaid expansion states experienced a significant additional increase in prescription fills for diabetes medications compared to populations in

states without Medicaid expansions. Both men and women experienced a significant increase in diabetes prescription fills. Patterns across age groups provided evidence of robustness. Age groups with higher prevalence of diabetes showed larger increases in diabetes prescription fills after Medicaid expansions. In addition, changes in prescription fills after Medicaid expansions were statistically significant for patients aged 60-64 but not statistically significant for patients aged 65-69, an age group that is eligible for Medicare insurance. Insulin showed a larger proportional increase in treatment than biguanides. Insulin is not yet available as a generic medication, resulting in relatively high out of pocket costs for insulin compared to other diabetes drugs (Greene and Riggs 2015). In many cases, Medicaid patients face no cost sharing for prescription drugs. Therefore, the higher out of pocket costs of insulin for patients without Medicaid insurance could relate to the larger impact of Medicaid insurance on use of insulin. In the absence of longer-term follow-up data, changes in prescriptions fills for insulin and biguanides after the ACA Medicaid expansions can be informative about possible health effects of the expansions for people with diabetes. In a back of the envelope calculation, our results would imply that about 1.9 million additional insulin prescriptions and 2.6 million additional biguanides prescriptions are filled in Medicaid expansion states annually as a result of ACA Medicaid expansions. Based on the average insulin and biguanide prescription fills per treated patient in the most recent MEPS data, this would translate to 450,000 and 218,000 additional patients treated annually with these two therapies. Over 10 years, treating these additional patients could prevent over 11,000 and 6,000 diabetes-related clinical events, respectively (UK Prospective Diabetes Study group 1998a, 1998b). Diabetes-related clinical events include blindness, heart attack, stroke, renal failure, amputation, retinopathy, blindness, cataract extraction, or mortality, including sudden death from hyperglycemia or hypoglycemia. Our study has a number of strengths. A key barrier to analyzing the effects of policy changes on treatment of specific conditions by patient demographics is the difficulty in obtaining sufficiently large sample sizes in self-reported survey data. We are able to overcome this barrier by using the most comprehensive prescription sales data available in the United States. Because these data are collected at the time that prescriptions are filled, our data are also highly timely, providing 8 months of additional follow-up compared to existing studies. Use of sales data also side-steps issues of self-report bias and permit us to correctly identify the time of treatment. Second, although Medicaid expansion states differ from non-expansion states in some respects, population level factors that differ between comparison groups do not bias the results in a difference-in-differences analysis as long as trends between the comparison groups would have remained parallel in the absence of an intervention. We estimated prior trends using 6 years of pre-expansion data, and found that trends were similar across populations with versus without Medicaid expansions prior to the ACA. We additionally adjusted for state and year indicator

variables, patient age and sex, and annual changes in unemployment on the state-level to address residual confounding. Our study has limitations. First, we were not able to track individuals over time. Instead, we rely on repeated observations at the age-sex-state level across different years. Second, patient race was not reported in the IMS Health data and therefore was not included in our analyses. Third, prescription fills that occurred outside the IMS sampling frame were not directly observed. To address this problem, we used estimates provided after imputation of incomplete data using IMS Health’s patented formula. Fourth, we evaluate the association between Medicaid expansions and diabetes prescription fills in Medicaid expansion states, also known as a treatment on the treated effect. Accordingly, our findings may not generalize to a nationwide Medicaid expansion. Finally, ours is an observational study and we cannot rule out that changes over time unrelated to Medicaid expansions account for our results. Conclusions Medicaid expansions after the ACA were associated with an accelerated increase in treatment of diabetes using prescription medications for adults under age 65. This increase occurred for both men and women. If gains in treatment are maintained over time, the observed increases in treatment with insulin and biguanides in Medicaid expansion states could prevent over 17,000 adverse health events attributable to diabetes over the next 10 years.

Table 1. Baseline characteristics of populations in states that did vs. did not expand Medicaid under the ACA State-level average State-level in non-expansion average in states expansion states Prevalence of diagnosed diabetes in 2010

p-value of difference

7.958

9.6

.058

6,161,336

5,887,206

.89

Mortality in 2010, per 100,000 people

829.565

826.595

.933

Proportion male in 2010

49.196

49.464

.246

Proportion Hispanic in 2010

11.532

9.105

.397

Proportion black in 2010

10.2

12.55

.463

Proportion over age 65 in 2010

13.526

12.835

.151

Population in 2010

Note: Diagnosed diabetes prevalence numbers were obtained from the Centers for Disease Control and Prevention. Mortality numbers were obtained from Centers for Disease Control and Prevention (CDC)/National Center for Health Statistics (NCHS), National Vital Statistics System. Proportion male/Hispanic/black/over age 65 numbers are obtained from 2010 Census Briefs.

Table 2. Additional change in diabetes prescription fills per 1,000 population in Medicaid expansion states after Medicaid expansions, age 30-64 Prescriptions per 1,000 at baselinea All By sex Men Women By type of medication Insulin Biguanide

596.2

Change after Medicaid expansions: Difference in differences estimateb 34.6* (5.8 to 63.5)

624 569

36.4* (7.6 to 65.3) 33.2* (3.6 to 62.8)

132.3 285.6

15.7** (9 to 22.4) 21** (5.7 to 36.4)

95% confidence intervals are in parentheses. * denotes significant at 0.05 level. ** denotes significant at 0.01 level. a

Data from Medicaid expansion states in 2013, the last year prior to the ACA Medicaid expansions Estimates are adjusted for year indicator variables, state indicator variables, patient’s age group and gender as well as current state-level unemployment rates. b

Table 3. Additional change in diabetes prescription fills per 1,000 population in Medicaid expansion states after Medicaid expansions, by age group A. Total diabetes prescriptions

Age 20-25 Age 26-29 Age 30-34 Age 35-44 Age 45-54 Age 55-59 Age 60-64 Age 65-69

Prescriptions per 1,000 prior to Medicaid expansionsa 63.4 67.6 141.4 299 643.1 990.2 1205.4 1401.6

Change after Medicaid expansions: Difference in differences estimateb 6.1** (2.2 to 10.1) 5.5** (1.4 to 9.6) 11** (3.8 to 18.1) 9.1 (-7.2 to 25.4) 26.1 (-4.6 to 56.8) 63.9** (22.5 to 105.3) 57.5* (2.8 to 112.1) 41.3 (-11.8 to 94.5)

B. Insulin prescriptions

Age 20-25 Age 26-29 Age 30-34 Age 35-44 Age 45-54 Age 55-59 Age 60-64 Age 65-69

Prescriptions per 1,000 prior to Medicaid expansionsa 37.3 29.6 50.4 76.6 138.1 206.6 249.1 293.8

Change after Medicaid expansions: Difference in differences estimateb 4.8** (2.6 to 7) 3.8** (1.9 to 5.7) 6.6** (4 to 9.3) 8** (3.5 to 12.5) 14.9** (6.9 to 22.8) 24.8** (15.2 to 34.5) 20.2** (8.4 to 32) 9.2 (-1.5 to 19.8)

C. Biguanide prescriptions

Age 20-25 Age 26-29 Age 30-34 Age 35-44 Age 45-54 Age 55-59 Age 60-64

Prescriptions per 1,000 prior to Medicaid expansionsa 21.8 29.9 66.9 147.7 313.8 470.6 559.3

Change after Medicaid expansions: Difference in differences estimateb 2.2 (-.1 to 4.6) 2.6* (.3 to 4.9) 6.1** (2.3 to 9.9) 6 (-2.6 to 14.7) 18.2* (1.8 to 34.6) 36.4** (13.6 to 59.2) 37.2* (7.7 to 66.7)

Age 65-69

625.3

29.4 (-.3 to 59)

95% confidence intervals are in parentheses. * denotes significant at 0.05 level. ** denotes significant at 0.01 level. a

Data from Medicaid expansion states in 2013, the last year prior to the ACA Medicaid expansions Estimates are adjusted for year indicator variables, state indicator variables, patient’s age group and gender as well as current state-level unemployment rates. b

Figure 1 Prescription fills for patients aged 30-64, in Medicaid expansion vs. non-expansion states A. Diabetes medications as a whole

B. Diabetes medication by type

The vertical lines indicate 2013, the last year before the ACA Medicaid expansions.

Figure 2 Changes in prescription fills after ACA Medicaid expansions, and prevalence of diabetes just prior to and during Medicaid expansions (NHANES data, 2013-2014)

60

55-59

40

60-64

20

45-54

30-34

35-44

0

20-25 26-29

0

5

20 15 10 Diabetes prevalence per 100 population

25

Works Cited Agency for Healthcare Research and Quality. 2014. “Medical Expenditure Panel Survey.” http://www.ahrq.gov/research/data/meps/index.html. Bureau of Labor Statistics. 2017. “Local Area Unemployment Statistics Home Page.” Accessed April 8. https://www.bls.gov/lau/. “CDC - Number of Adults by Diabetes Medication Status - Treating Diabetes - Data & Trends - Diabetes DDT.” Accessed April 12, 2017. https://www.cdc.gov/diabetes/statistics/meduse/fig1.htm. Danaei, Goodarz, Eric L Ding, Dariush Mozaffarian, Ben Taylor, Jürgen Rehm, Christopher J L Murray, and Majid Ezzati. 2009. “The Preventable Causes of Death in the United States: Comparative Risk Assessment of Dietary, Lifestyle, and Metabolic Risk Factors.” PLoS Medicine 6 (4): e1000058. doi:10.1371/journal.pmed.1000058. Decker, Sandra L., Deliana Kostova, Genevieve M. Kenney, and Sharon K. Long. “Health Status, Risk Factors, and Medical Conditions Among Persons Enrolled in Medicaid vs Uninsured LowIncome Adults Potentially Eligible for Medicaid Under the Affordable Care Act.” JAMA 309, no. 24 (June 26, 2013): 2579–86. doi:10.1001/jama.2013.7106. Farley, Thomas, Mehul Dalal, Farzad Mostashari, and Thomas Frieden. 2010. “Deaths Preventable in the U.S. by Improvements in Use of Clinical Preventive Services.” American Journal of Preventive Medicine 38 (6): 600–609. doi:10.1016/j.amepre.2010.02.016. Gakidou, Emmanuela, Leslie Mallinger, Jesse Abbott-Klafter, Ramiro Guerrero, Salvador Villalpando, Ruy Lopez Ridaura, Wichai Aekplakorn, et al. 2011. “Management of Diabetes and Associated Cardiovascular Risk Factors in Seven Countries: A Comparison of Data from National Health Examination Surveys.” Bulletin of the World Health Organization 89 (3): 172–83. doi:10.2471/BLT.10.080820. Ghosh, Ausmita, Kosali Simon, and Benjamin Sommers. 2017. “The Effect of State Medicaid Expansions on Prescription Drug Use: Evidence from the Affordable Care Act.” w23044. Cambridge, MA: National Bureau of Economic Research. http://www.nber.org/papers/w23044.pdf. Greene, Jeremy A., and Kevin R. Riggs. 2015. “Why Is There No Generic Insulin? Historical Origins of a Modern Problem.” New England Journal of Medicine 372 (12): 1171–75. doi:10.1056/NEJMms1411398. Heron, Melonie. 2013. “Deaths: Leading Causes for 2010.” National Vital Statistics Reports: From the Centers for Disease Control and Prevention, National Center for Health Statistics, National Vital Statistics System 62 (6): 1–96. Hill, Steven C., G. Edward Miller, and Merrile Sing. “Adults with Diagnosed and Untreated Diabetes: Who Are They? How Can We Reach Them?” Journal of Health Care for the Poor and Underserved 22, no. 4 (November 13, 2011): 1221–38. doi:10.1353/hpu.2011.0149.Huang, E. S., J. B. Meigs, and D. E. Singer. 2001. “The Effect of Interventions to Prevent Cardiovascular Disease in Patients with Type 2 Diabetes Mellitus.” The American Journal of Medicine 111 (8): 633–42. Kelly, Tanika N., Lydia A. Bazzano, Vivian A. Fonseca, Tina K. Thethi, Kristi Reynolds, and Jiang He. 2009. “Systematic Review: Glucose Control and Cardiovascular Disease in Type 2 Diabetes.” Annals of Internal Medicine 151 (6): 394–403. Kennedy, Jae, and Elizabeth Geneva Wood. 2016. “Medication Costs and Adherence of Treatment Before and After the Affordable Care Act: 1999–2015.” American Journal of Public Health 106 (10): 1804–7. doi:10.2105/AJPH.2016.303269. Mulcahy, Andrew W., Christine Eibner, and Kenneth Finegold. 2016. “Gaining Coverage Through Medicaid Or Private Insurance Increased Prescription Use And Lowered Out-Of-Pocket Spending.” Health Affairs 35 (9): 1725–33. doi:10.1377/hlthaff.2016.0091.

National Center for Health Statistics. 2008. “National Health and Nutrition Examination Survey Data.” U.S. Department of Health and Human Services, Centers for Disease Control and Prevention,. https://wwwn.cdc.gov/nchs/nhanes/. Ray, Kausik K, Sreenivasa Rao Kondapally Seshasai, Shanelle Wijesuriya, Rupa Sivakumaran, Sarah Nethercott, David Preiss, Sebhat Erqou, and Naveed Sattar. 2009. “Effect of Intensive Control of Glucose on Cardiovascular Outcomes and Death in Patients with Diabetes Mellitus: A MetaAnalysis of Randomised Controlled Trials.” The Lancet 373 (9677): 1765–72. doi:10.1016/S0140-6736(09)60697-8. Roehrig, C., and M. Daly. 2015. “Prevalence Trends For Three Common Medical Conditions: Treated And Untreated.” Health Affairs 34 (8): 1320–1323. doi:10.1377/hlthaff.2015.0283. UK Prospective Diabetes Study group. 1998a. “Effect of Intensive Blood-Glucose Control with Metformin on Complications in Overweight Patients with Type 2 Diabetes (UKPDS 34).” Lancet (London, England) 352 (9131): 854–65. ———. 1998b. “Intensive Blood-Glucose Control with Sulphonylureas or Insulin Compared with Conventional Treatment and Risk of Complications in Patients with Type 2 Diabetes (UKPDS 33).” The Lancet 352 (9131): 837–53. doi:10.1016/S0140-6736(98)07019-6. U.S. Burden of Disease Collaborators. 2013. “The State of US Health, 1990-2010: Burden of Diseases, Injuries, and Risk Factors.” JAMA : The Journal of the American Medical Association 310 (6): 591–608. doi:10.1001/jama.2013.13805. U.S Census Bureau. 2016. “Intercensal Population Estimates: County Level.” Accessed September 27. https://www.census.gov/popest/data/intercensal/. Wen, Hefei, Tyrone F. Borders, and Benjamin G. Druss. 2016. “Number Of Medicaid Prescriptions Grew, Drug Spending Was Steady In Medicaid Expansion States.” Health Affairs 35 (9): 1604–7. doi:10.1377/hlthaff.2015.1530.

Tables for eSupplement Table S4: Differences in trends in diabetes prescription fills prior to ACA Medicaid expansions (2008-2013): adults aged 30-64

All By sex Men Women By type of medication Biguanide Insulin

Difference in prior trends between expansion and non-expansion states b -2.9

p-value

-0.7 -4.8

0.91 0.42

-3.8 5.7

0.61 0.23

0.63

Data are adjusted for year indicator variables, state indicator variables, patient’s age group and gender as well as current state-level unemployment rates. b

* denotes significant at 0.05 level. ** denotes significant at 0.01 level.

Table S5. Differences in trends in diabetes prescription fills prior to ACA Medicaid expansions (2008-2013), by age group A. Total diabetes prescriptions

Age 20-25 Age 26-29 Age 30-34 Age 35-44 Age 45-54 Age 55-59 Age 60-64 Age 65-69

Difference in prior trends between expansion and nonexpansion states b 3.9 -7.2 0.3 -7.7 -5.5 2 -4.5 3.2

p-value

0.51 0.28 0.96 0.22 0.37 0.75 0.52 0.6

B. Insulin prescriptions

Age 20-25 Age 26-29 Age 30-34 Age 35-44 Age 45-54 Age 55-59 Age 60-64 Age 65-69

Difference in prior trends between expansion and nonexpansion states b 10.2 -4.6 9.3 6 2.5 10.4 -0.7 2.7

p-value

0.09 0.47 0.12 0.23 0.65 0.04 0.89 0.55

C. Biguanide prescriptions

Age 20-25 Age 26-29 Age 30-34 Age 35-44 Age 45-54 Age 55-59 Age 60-64

Difference in prior trends between expansion and nonexpansion states b -1.3 1.1 -1.2 -10.3 -4.8 2 -5

p-value

0.89 0.92 0.88 0.2 0.52 0.78 0.55

Age 65-69

4.8

0.51

* denotes significant at 0.05 level. ** denotes significant at 0.01 level.