The Part D coverage gap reform in 2011 improved adherence to diabetes medications in the coverage gap.
To investigate the impact of Part D coverage gap reform on diabetes medication adherence.
Retrospective data analysis based on pharmacy claims data from a national pharmacy benefit manager.
We used a difference-in-difference-indifference method to evaluate the impact of coverage gap reform on adherence to diabetes medications. Two cohorts (2010 and 2011) were constructed to represent the last year before Affordable Care Act (ACA) reform and the first year after reform, respectively. Each patient had 2 observations: 1 before and 1 after entering the coverage gap. Patients in each cohort were divided into groups based on type of gap coverage: no coverage, partial coverage (generics only), and full coverage.
Following ACA reform, patients with no gap coverage and patients with partial gap coverage experienced substantial drops in copayments in the coverage gap in 2011. Their adherence to diabetes medications in the gap, measured by percentage of days covered, improved correspondingly (2.99 percentage points, 95% confidence interval [CI] 0.49-5.48, P = .019 for patients with no coverage; 6.46 percentage points, 95% CI 3.34-9.58, P <.0001 for patients with partial coverage). Patients with full coverage also had lower copayments in the gap in 2011. However, their adherence did not increase (—0.13 percentage point, P = .8011).
In the first year of ACA coverage gap reform, copayments in the gap decreased substantially for all patients. Patients with no coverage and patients with partial coverage in the gap had better adherence in the gap in 2011.
Am J Manag Care. 2013;19(4):308-316The Affordable Care Act mandated that Part D coverage gap be shrunk starting from 2011. The reform had an important impact on adherence to diabetes medications in the gap.
Medicare Part D was implemented in 2006 to provide prescription drug benefits for Medicare beneficiaries. The implementation of Part D has led to many positive changes: reduced out-of-pocket expenses, increased medication fills, improved adherence, decreased medical spending, and fewer avoidable hospitalizations.1-10 One controversial aspect of the Part D benefit design is the coverage gap, commonly referred to as the “donut hole.” Patients with a Part D plan that has a standard defined benefit have a copayment of 25% on prescription drugs after they pay the initial deductible. After their total pharmacy spending reaches the coverage gap limit ($2830 in 2010), patients have a 100% copayment for their drugs until they reach the catastrophic limit ($6440 in 2010). In the catastrophic stage patients pay 5% or $2.60 per prescription.
The lack of coverage in the donut hole raises concern that the increase in copayment may have an adverse effect on drug utilization. Research has shown that reaching the coverage gap decreases medication fills11-14 and lowers adherence to essential medications.15,16 In response to these concerns, the Affordable Care Act (ACA), passed in March 2010, includes a provision to shrink the coverage gap. Starting in 2011, the copayment for brand drugs in the coverage gap was reduced 50%, and the copayment for generic drugs was reduced 7%. The copayments for brand drugs and generic drugs in the gap will be further reduced after 2012. In the year of 2020, the donut hole will be completely closed with a 25% copayment for both brand and generic drugs.17,18
The closure of the donut hole represents another significant change for Medicare Part D. A solid understanding of the latest reform is critical for the design of Part D benefits. This research contributes by investigating the impacts of the 2011 coverage gap reform on adherence to diabetes medications. We focused on diabetes because it is a highly prevalent disease among Medicare beneficiaries. It is estimated that in 2010, the prevalence rates for people aged 65 to 74 years and >75 years were 20.7% and 18.9%, respectively. The diabetes prevalence rates for the 2 groups are expected to become 30.1% and 32.7% in 2050.19 Effective management of diabetes is vital to the success of Medicare.
There are 4 types of Part D plans: standard defined benefit, actuarially equivalent, basic alternative, and enhanced alternative. Of the 4 types of plans, the actuarially equivalent and basic alternative plans are similar to the standard defined benefit plan in coverage. Patients with an enhanced alternative plan can pay a higher premium for better coverage, such as having coverage in the gap and/or covering more drugs than standard defined benefit plans.
The year 2011 is the first year of coverage gap reform mandated by the ACA. In 2011, patients reaching the gap received a 50% copayment reduction for brand drugs and a 7% copayment reduction for generic drugs. For patients whose Part D plans provided coverage in the gap, discounts were applied to their copayments directly. For example, suppose a patient who had coverage while in the gap in 2010 paid a 25% copayment for both brand and generic drugs in the coverage gap. In 2011, this same patient would pay a 12.5% copayment for brand drugs and a 23.25% copayment for generic drugs in the coverage gap.
Data for this research came from pharmacy claim data from MedImpact HealthCare Systems Inc’s book of business. Med- Impact is a national pharmacy benefit manager. The research data included a 2010 cohort and a 2011 cohort. The 2010 cohort represents the last year before ACA reform and the 2011 cohort represents the first year after ACA reform.
Patients in both cohorts were required to be continuously enrolled in the same health plan for 2 years: 1 year before the cohort year and the cohort year. Patients with diabetes were identified as having at least 2 diabetes drug claims in both the cohort year and the year before the cohort year. It was important to require patients to have used diabetes drugs before the cohort year so that we had a full year to measure adherence in the cohort year. The inclusion of 1 year of data before the cohort year also was important to construct a comorbidity index. All patients had to be at least 65 years old in their cohort year. Patients eligible for a low-income subsidy were removed because their copayment structures are different.
Each data cohort included patients with no coverage, partial coverage (ie, generic coverage only), and full coverage in the gap. In this research, patients with partial and full coverage all came from enhanced alternative plans. Patients with no coverage came from standard defined benefit, actuarially equivalent, basic alternative, and enhanced alternative plans. Theoretically, patients in actuarially equivalent, basic alternative, and enhanced alternative plans could have coverage gap limits different from those of standard defined benefit plans. However, that was not the case in this research. We checked and confirmed that all patients in the data set had the same coverage gap limits.
These identification steps generated 20,709 patients in the 2010 cohort and 20,212 patients in the 2011 cohort. We selected only patients who entered the coverage gap but did not enter the catastrophic level to make it easier to compare adherence before the gap with adherence in the gap.12,13,16 This exclusion led to 6828 patients in the 2010 cohort and 6124 patients in the 2011 cohort.
The outcome variable was adherence to diabetes medications. Diabetes is a long-term and progressive disease. Adherence to diabetes medication is critical for maintaining glycemic control and slowing down disease progression. Poor adherence can lead to inadequate glycemic control, higher probabilities of hospitalization, higher medical cost, and higher mortality.20-27
Adherence is measured by proportion of days covered (PDC), defined as the number of days covered by at least 1 diabetes medication divided by the length of the study period based on medication fill date and days of supply. The PDC methodology has been used in many studies.16,20,25,26,28 The value of PDC is bounded between 0.0 and 1.0.
This research used a difference-in-difference-in-difference method to identify the impact of coverage gap reform on adherence to diabetes medications. In this model, year 2010 data and year 2011 data were used to represent the periods before and after coverage gap reform, respectively. Each patient in a cohort year had 2 observations: one before the coverage gap and the other in the coverage gap. Patients with no coverage or partial coverage were the treatment group; patients with full coverage were the control group. In a straightforward difference- in-difference-in-difference model, the effect of coverage gap reform was captured by the interaction of coverage gap, year 2011, and the type of coverage in the coverage gap.
One problem of a straightforward difference-in-difference-in-difference model is that the outcome variable, PDC, was bounded between 0 and 1. In this research, the mean PDC value was around 80% and the distribution was highly skewed. This can be problematic in an ordinary least squares regression. One way to address this issue is to transform the outcome variable as the change of PDC after reaching the coverage gap, defined as PDC in the coverage gap minus PDC before the gap. After the transformation, the model can be written as:
Y = b0 + NoCov*b1 + GenCov*b2 + Y2011*b3 + NoCov*Y2011*b4 + GenCov*Y2011*b5 + e,
where Y equals PDC in the coverage gap minus PDC before the coverage gap for each patient, NoCov is the dummy variable indicating no coverage in the gap, GenCov is the dummy variable indicating partial coverage in the gap, and Y2011 is the dummy variable indicating the 2011 cohort.
In this model b4 and b5 are the coefficients of interest. They represent the change in adherence in the coverage gap in response to coverage gap reform for patients with no coverage and partial coverage, respectively.
Through the transformation, each patient’s idiosyncratic propensity to be adherent was removed. Patients’ unchanged characteristics such as age, sex, and geographical location were also removed. As a consequence, the inclusion of covariates such as age and sex was not necessary. The main model was run without covariates. We included covariates in the model as part of the robustness analysis. Control variables included age, sex, geographical location, whether insulin was used in the last year, and comorbidities. Information on patient demographics was obtained from patients’ eligibility files. Information on the use of insulin was available from the previous year’s pharmacy claim file. Comorbidities were measured by RxRisk, a comorbidity measure based on pharmacy claims in the prior year.29 Each RxRisk category indicates whether a patient took certain classes of medications (eg, those for hypertension and hyperlipidemia) in the previous year. Only RxRisk categories with prevalence rates higher than 5% were included in the study.
One notable difference between our outcome variable and the comparable measure in the Centers for Medicare & Medicaid Services (CMS) Five-Star Quality Rating System for Medicare Advantage Plans is that the CMS star rating uses only oral diabetes medications in its calculation. This research uses all diabetes medications including insulin. A robustness check was performed at the end of the study by removing patients with a history of insulin use from the model. This robustness check was important also because patients receiving insulin likely had either advanced type 2 diabetes or type 1 diabetes. These patients can be different from patients who do not use insulin, who tend to be patients in early stage of type 2 diabetes.
An alternative model was to use a dichotomous variable of adherence as the outcome variable, with a value of 1 assigned to patients having >80% PDC. This dichotomous variable of adherence is used frequently in adherence studies.20,25,26, 28 One problem of using a dichotomous outcome variable is that the interpretation of interaction terms in a logit/probit model is not straightforward. The interaction effect varies by patients. It can be positive or negative depending on many factors.30,31 This research uses a continuous outcome variable for ease of interpretation.
A generalized estimating equation was used to model the effect of coverage gap reform. The model was adjusted for health plan clustering effect to account for potential similarities of patients within the same health plan. The whole statistical analysis was performed with SAS 9.3 (SAS Institute, Cary, North Carolina).
presents the descriptive statistics. For the total sample, the average age was 74.17 years (SD 6.71 years), the average number of days in the coverage gap was 114.21 (SD 62.29 days), 49.9% of the patients were male, and 42.9% of the patients used insulin in the previous year. Patients with no coverage, partial coverage, and full coverage in the gap accounted for 27.2%, 14.0%, and 58.7% of the sample, respectively. Hypertension and hyperlipidemia were the 2 most common comorbidities, with both prevalence rates higher than 85%.
Patients in the 2010 and 2011 cohorts were similar in many aspects: age, days in the donut hole, percent male, percentage of patients using insulin, and comorbidities. There were 2 notable differences. The first difference was the distribution of coverage types. The percentage of patients without coverage was 23.1% in the 2010 cohort and 31.5% in the 2011 cohort. At the same time, the percentage of patients with full coverage decreased from 62.1% in the 2010 cohort to 55.4% in the 2011 cohort. The second difference was geographical location. The percentage of patients in the East decreased from 13.1% in the 2010 cohort to 0.5% in the 2011 cohort. The differences in coverage types and geographical locations were all statistically significant (Table 1).
The shows that reaching the coverage gap had a large impact on patients without coverage. For patients without coverage in 2010, copayments (all copayment data were normalized into 30 days of supply in this study) for diabetes medications increased from $19.20 before the coverage gap to $85.30 in the coverage gap, a 344% increase. Patients with partial coverage experienced a similar increase in copayment (320%) after reaching the coverage gap. In contrast, patients with full coverage experienced moderate changes, from $22.60 before the coverage gap to $25.60 in the coverage gap.
The coverage gap reform had an immediate effect on patients’ copayment in the coverage gap. In 2011, the average copayment for diabetes medications for patients without coverage in the gap rose from $20 before the gap to $54.60 in the gap. The percent increase (173%) was much lower than the percent increase in 2010 (344%). Patients with partial coverage experienced a 143% increase in average copayment in the gap in 2011, also much lower than the 320% increase in 2010. Patients with full coverage experienced a copayment reduction from $25.50 to $18.60, a 27% decrease.
breaks down copayments according to whether they were for brand and generic drugs. The major driver for the copayment change in the gap was the copayment change for brand drugs. In 2010, the copayment for brand diabetes drugs increased from $30.60 to $151.50 (a 395% increase) for patients without coverage who entered the coverage gap. Patients with partial coverage had a 419% increase in copayment for brand drugs after reaching the coverage gap. In 2011, patients from all groups had a substantial decrease in brand drug copayments in the coverage gap compared with 2010. The change in generic drug copayment in the gap from 2010 to 2011 was less than $2.
Corresponding to the reduction in copayment in the coverage gap, patients without coverage and those with partial coverage had improvements in adherence in the gap. In 2010, patients with no coverage and patients with partial coverage had a reduction of 10.5 and 10.4 percentage points in PDC, respectively, after reaching the gap. In 2011, after the coverage gap reform, the PDC drop in the coverage gap for patients with no coverage was 7.8 percentage points. The PDC decrease in the gap for patients with partial coverage was 4.5 percentage points in 2011, comparable to the decrease for patients with full coverage (Table 2). All 3 groups had decreased adherence after reaching the coverage gap, probably a reflection of a natural decrease in adherence over time.32
Table 2 confirms the importance of using PDC change in the gap as the dependent variable. The distribution of PDC was heavily truncated. Before the gap, PDCs across different groups were generally more than 80%, with the SD around 18 percentage points. In the gap, PDCs were generally around 75%, with the SD around 30 percentage points. Mean PDC plus 1 SD was usually bigger than 1.0, the maximum value for PDC. After the transformation, mean PDC changes plus or minus 3 SDs were generally within the lower and higher bound value (-1.0, 1.0).
Multivariate Model Analysis
shows results from the main model. Compared with patients with full coverage, patients with no coverage and patients with partial coverage had a substantial drop in adherence when they reached the gap in 2010. The PDCs dropped 7.41 percentage points (95% confidence interval [CI] -9.64, -5.18; P <.0001) and 6.70 percentage points (95% CI -9.31, -4.10; P <.0001) in the coverage gap, respectively.
The copayment decrease in 2011 had a large impact on adherence to diabetes medications. PDC in the coverage gap in 2011 increased by 2.99 percentage points (95% CI 0.49-5.48, P = .019) for patients with no coverage (b4) and 6.46 percentage points (95% CI 3.34-9.58, P <.0001) for patients with partial coverage (b5). Patients with full coverage experienced no change of adherence in the year 2011 (b3 -0.35, P = .482).
present the results of a robustness check analysis. In Table 4, part A, we included covariates such as age, sex, and history of using insulin in the model. As expected, the results were very close to the results in Table 3, our main model. We also tested the robustness of the results by removing patients who used insulin in the previous year (Table 4, part B) and by removing patients who lived in the East (Table 4, part C). The main model findings were robust to both changes.
Finally we examined whether improvements in adherence for patients with no coverage and partial coverage in 2011 were enough to offset the decrease in adherence before the ACA reform. To answer this question, we ran the same model in Table 3 by removing the 2010 cohort. After the removal, b3, b4, and b5 were dropped due to collinearity. If the improvements in adherence could offset the previous decrease, we should have observed that b1 and b2 were insignificant. Table 4, part D shows that b1 was strongly negative (-4.05, P = .001) while b2 (-0.18, P = .924) was negative but not significant.
Consistent with previous findings, we found that copayments rose substantially in the coverage gap for patients without full coverage before the coverage gap reform in 2011.11-16 Patients’ adherence to diabetes medications in the gap dropped as copayments increased.15,16 These findings confirm the importance of reforming the coverage gap to improve adherence to diabetes medications.
This research shows that the coverage gap reform had an immediate impact on patients’ copayments for diabetes medications. Compared with coverage gap copayments in 2010, copayments in the coverage gap for patients with no coverage and patients with partial coverage dropped 36% and 48%, respectively, in 2011. Patients with full coverage also benefited from the coverage gap reform. Their average copayments in the coverage gap in 2011 decreased by 27% compared with copayments in the gap in 2010. The major driving force for the copayment decrease was the 50% discount in brand drugs. The copayment change for generic drugs was small.
Corresponding to the copayment decrease, adherence to diabetes medications in the coverage gap improved in 2011 as intended, especially for patients with no coverage and patients with partial coverage. Patients with full coverage had no response to the copayment reduction in the coverage gap. Their adherence did not change even though their copayment also decreased by 27%.
Table 3 shows that adherence improvement in the coverage gap for patients with partial coverage (6.46, P <.0001) was greater than the improvement for patients with no coverage (2.99, P = .019). This difference is interesting because one major criticism of the coverage gap reform is that it favors coverage of brand drugs. In 2011, Part D reduced copayments of brand drugs by 50% but only reduced copayments for generic drugs by 7% in the coverage gap. One concern about the disparity of coverage is that it could discourage the use of less expensive generic drugs and lead to higher costs.33 An examination of patients with no coverage and patients with partial coverage shows that these 2 groups of patients had very similar adherence levels and copayment levels both before and in the coverage gap in 2010. In 2011, these 2 groups also had similar levels of copayment before the coverage gap in 2011 ($20 for no coverage and $18 for patients with partial coverage). After entering the coverage gap in 2011, patients with partial coverage did have a lower mean copayment in the coverage gap in 2011 ($43.70 vs $54.60 for patients with no coverage, Table 2). Patients with partial coverage also had a smaller decrease of adherence levels in the coverage gap (Table 2).
It is unclear to what degree the disparity between brand and generic drug copayments can explain copayment differences and adherence differences between patients with no coverage and patients with partial coverage in the gap in 2011. If balanced coverage of both brand drugs and generic drugs can improve adherence and reduce costs in comparison with the current policy, there may be a need for quicker closure of the generic drug coverage gap. More research is needed in this area.
This research has several limitations. First, it was based on data from 1 pharmacy benefits manager. The population profile could be different from that of the national population. It is unclear whether our findings can be generalized to the national population. Second, this research did not contain medical data or laboratory data. We cannot show whether or not the improved adherence translated into improved glycemic control. We could not divide patients into type 1 or
type 2 diabetes patients. The effect of improved adherence on hospitalization risks was also unclear. Third, this research relied on pharmacy claim data to measure adherence to insulin. This method had its limitations because the dose of insulin needs to be adjusted frequently based on several factors such as weight and response. If a patient adjusts his or her insulin dose substantially after filling the drug prescription, the measurement of adherence to insulin in this research may be inaccurate. Fourth, use of PDC may have overestimated the actual adherence. The PDC method does not diffferentiate between whether a patient needs to take a single drug or needs to take 2 or more drugs simultaneously. If a patient is supposed to take 2 drugs together but just takes 1 drug, the PDC calculation will regard this patient as adherent. This will lead to an overestimate of the actual adherence. Fifth, some physicians may give patients free samples to help them when they reach the coverage gap. We have no data on the use of free samples and it is unclear how the free samples might have influenced the research findings. Sixth, we limited our study to patients who entered the coverage gap but not the catastrophic level. Patients who entered the catastrophic level of coverage might have difference responses to the coverage gap reform. Finally, this research was limited to diabetes only. The impact of coverage gap reform on other chronic illnesses may be different. More research is needed in the future to conduct a complete cost-effectiveness analysis of the coverage gap reform.
This research is the first to investigate the impact of Part D coverage gap reform under the ACA. Our findings generally support the effectiveness of ACA Part D coverage gap reform. This research demonstrates that coverage reform in its first year had an immediate and direct effect on copayments for diabetes medications. Patients with no coverage and patients with partial coverage had improved adherence in the gap in 2011 compared with adherence in the gap in 2010. At the same time, this reform is far from complete. Patients with no coverage and patients with partial coverage still experienced higher copayments in the coverage gap. In addition, patients with no coverage still had significantly reduced adherence in the coverage gap in 2011 compared with adherence before the coverage gap. Because the coverage gap will remain until 2020, health plans still need to address medication adherence issues in the gap in the next several years.Acknowledgments
We gratefully acknowledge Emmet Keeler, the editors, and the 2 anonymous reviewers for comments that improved the paper substantially. An early version of this study was presented at the CMS 2012 Medicare Prescription Drug Benefits Symposium; March 20-21, 2012; Hunt Valley, Maryland. We thank the symposium participants for useful comments.
Author Affiliations: From MedImpact Healthcare Systems Inc (FZ, BVP, LB), San Diego, CA.
Author Disclosures: The authors (FZ, BVP, LLB) report no relationship or financial interest with any entity that would pose a conflict of interest with the subject matter of this article.
Authorship Information: Concept and design (FZ, LLB); acquisition of data (FZ); analysis and interpretation of data (FZ, BVP); drafting of the manuscript (FZ); critical revision of the manuscript for important intellectual content (FZ, BVP, LLB); statistical analysis (FZ); administrative, technical, or logistic support (BVP, LLB); and supervision (BVP, LLB).
Address correspondence to: Feng Zeng, PhD, MedImpact Healthcare Systems Inc, 10680 Treena St, San Diego, CA 92131. E-mail: fzeng@medimpact .com.1. Lichtenberg FR, Sun SX. The impact of Medicare Part D on prescription drug use by the elderly. Health Aff (Millwood). 2007;26(6): 1735-1744.
2. Ketcham JD, Simon KI. Medicare Part D’s effects on elderly patients’ drug costs and utilization. Am J Manag Care. 2008;14(11)(suppl): SP14-SP22.
3. Safran DG, Strollo MK, Guterman S, Li A, Rogers WH, Neuman P. Prescription coverage, use and spending before and after Part D implementation: a national longitudinal panel study. J Gen Intern Med. 2010; 25(1):10-17.
4. Schneeweiss S, Patrick AR, Pedan A, et al. The effect of Medicare Part D coverage on drug use and cost sharing among seniors without prior drug benefits. Health Aff (Millwood). 2009;28(2):w305-w316.
5. Yin W, Basu A, Zhang JX, Rabbani A, Meltzer DO, Alexander GC. The effect of the Medicare Part D prescription benefit on drug utilization and expenditures. Ann Intern Med. 2008;148(3):169-177.
6. Madden JM, Graves AJ, Ross-Degnan D, Briesacher BA, Soumerai SB. Cost-related medication nonadherence after implementation of Medicare Part D, 2006-2007. JAMA. 2009;302(16):1755-1756.
7. Zhang Y, Lave JR, Donohue JM, Fischer MA, Chernew ME, Newhouse JP. The impact of Medicare Part D on medication adherence among older adults enrolled in Medicare-Advantage products. Med Care. 2010;48(5):409-417.
8. Zhang Y, Donohue JM, Lave JR, O’Donnell G, Newhouse JP. The effect of Medicare Part D on drug and medical spending. N Engl J Med. 2009;361(1):52-61.
9. McWilliams JM, Zaslavsky AM, Huskamp HA. Implementation of Medicare Part D and nondrug medical spending for elderly adults with limited prior drug coverage. JAMA. 2011;306(4):402-409.
10. Afendulis CC, He Y, Zaslavsky AM, Chernew ME. The impact of Medicare Part D on hospitalization rates. Health Serv Res. 2011;46(4): 1022-1038.
11. Raebel MA, Delate T, Ellis JL, Bayliss EA. Effects of reaching the drug benefit threshold on Medicare members’ healthcare utilization during the first year of Medicare Part D. Med Care. 2008;46(10): 1116-1122.
12. Zhang Y, Donohue JM, Newhouse JP, Lave JR. The effects of the coverage gap on drug spending: a closer look at Medicare Part D. Health Aff (Millwood). 2009;28(2):w317-w325.
13. Sun SX, Lee KY, Aruru M. Examining Part D beneficiaries’ medication use in the doughnut hole. Am J Pharm Benefits. 2009;1(1):19-28.
14. Hales JW, George S. How the doughnut hole affects prescription fulfillment decisions involving cardiovascular medications for Medicare Part D enrollees. Manag Care. 2010;19(12):36-44.
15. Fung V, Mangione CM, Huang J, et al. Falling into the coverage gap: Part D drug costs and adherence for Medicare Advantage prescription drug plan beneficiaries with diabetes. Health Serv Res. 2010; 45(2):355-375.
16. Gu Q, Zeng F, Patel BV, Tripoli LC. Part D coverage gap and adherence to diabetes medications. Am J Manag Care. 2010;16(12):911-918.
17. Patient Protection and Affordable Care Act. HR 3590. Pub L No 111- 148, 124 Stat 119.
18. The Henry J. Kaiser Family Foundation. Explaining Health Care Reform: Key Changes to the Medicare Part D Drug Benefit Coverage Gap. Focus on Health Reform. http://www.kff.org/healthreform/upload/8059.pdf. Published March 2010. Accessed January 28, 2013.
19. Narayan KM, Boyle JP, Geiss LS, Saaddine JB, Thompson TJ. Impact of recent increase in incidence on future diabetes burden: U.S., 2005-2050. Diabetes Care. 2006;29(9):2114-2116.
20. Pladevall M, Williams LK, Potts LA, Divine G, Xi H, Lafata JE. Clinical outcomes and adherence to medications measured by claims data in patients with diabetes. Diabetes Care. 2004;27(12):2800-2805.
21. Krapek K, King K, Warren SS, et al. Medication adherence and associated hemoglobin A1c in type 2 diabetes. Ann Pharmacother. 2004; 38(9):1357-1362.
22. Schectman JM, Nadkarni MM, Voss JD. The association between diabetes metabolic control and drug adherence in an indigent population. Diabetes Care. 2002;25(6):1015-1021.
23. Grant R, Adams AL, Trinacty CM, et al. Relationship between patient medication adherence and subsequent clinical inertia in type 2 diabetes glycemic management. Diabetes Care. 2007;30(4): 807-812.
24. Hepke KL, Martus MT, Share DA. Costs and utilization associated with pharmaceutical adherence in a diabetic population. Am J Manag Care. 2004;10(2, pt 2):144-151.
25. Ho PM, Rumsfeld JS, Masoudi FA, et al. Effect of medication nonadherence on hospitalization and mortality among patients with diabetes mellitus. Arch Intern Med. 2006;166(17):1836-1841.
26. Lau DT, Nau DP. Oral antihyperglycemic medication nonadherence and subsequent hospitalization among individuals with type 2 diabetes. Diabetes Care. 2004;27(9):2149-2153.
27. Sokol MC, McGuigan KA, Verbrugge RR, Epstein RS. Impact of medication adherence on hospitalization risk and healthcare cost. Med Care. 2005;43(6):521-530.
28. Pharmacy Quality Alliance (PQA). Executive Update on Medication Quality Measures in Medicare Part D Plan Ratings 2013. PQA Measures Used By CMS in the Star Ratings. http://pqaalliance.org/measures/cms.asp. Published 2013. Accessed January 28, 2013.
29. Fishman PA, Goodman MJ, Hornbrook MC, et al. Risk adjustment using automated ambulatory pharmacy data: the RxRisk model. Med Care. 2003;41(1):84-99.
30. Norton EC, Wang H, Ai CR. Computing interaction effects and standard errors in logit and probit models. Stata Journal. 2004;4:154-167.
31. Ai CR, Norton EC. Interaction terms in logit and probit models. Econ Lett. 2003;80(1):123-129.
32. Benner JS, Glynn RJ, Mogun H, Neumann PJ, Weinstein MC, Avorn J. Long-term persistence in use of statin therapy in elderly patients. JAMA. 2002;288(4):455-461.
33. Shrank WH, Choudhry NK. Time to fill the doughnuts—health care reform and Medicare Part D. N Engl J Med. 2011;364(7):598-601.