PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
Logo of nihpaAbout Author manuscriptsSubmit a manuscriptNIH Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
 
J Am Geriatr Soc. Author manuscript; available in PMC Aug 1, 2013.
Published in final edited form as:
PMCID: PMC3419279
NIHMSID: NIHMS379498
Association Between the Part D Coverage Gap and Adverse Health Outcomes
Jennifer M. Polinski, ScD, MPH,1,2 William H. Shrank, MD, MSHS,1,2 Robert J. Glynn, PhD, ScD,1,2,3 Haiden A. Huskamp, PhD,2 M. Christopher Roebuck, MBA,4 and Sebastian Schneeweiss, MD, ScD1,2
1Division of Pharmacoepidemiology and Pharmacoeconomics, Brigham and Women’s Hospital and Harvard Medical School, Boston, MA
2Harvard Medical School, Boston, MA
3Division of Preventive Medicine, Brigham and Women’s Hospital, Boston, MA
4RxEconomics, Sparks, MD
Correspondence/requests for reprints to: Jennifer M. Polinski, ScD, MPH, Division of Pharmacoepidemiology and Pharmacoeconomics, Brigham and Women’s Hospital, 1620 Tremont Street, Suite 3030, Boston, MA 02120, (617) 278-0931; Fax (617) 232-8602, jpolinski/at/partners.org
Background
Part D coverage gap entry is associated with a two-fold increased rate of drug discontinuation among beneficiaries now fully responsible for drug costs. Reduced adherence to drugs has been associated with adverse outcomes. We evaluated whether coverage gap entry is associated with risk of death or hospitalization for cardiovascular outcomes.
Design
Prospective cohort study. Beneficiaries entered the study upon reaching the coverage gap spending threshold and were observed until an event, reaching the threshold for catastrophic coverage, or year’s end. Exposed patients were responsible for drug costs in the gap; unexposed patients received financial assistance. We matched 9,436 exposed patients to 9,436 unexposed patients based on propensity score (PS) or high-dimensional propensity score (hdPS).
Setting
Medicare Part D drug insurance.
Participants
303,978 Medicare beneficiaries aged 65+ in 2006 and 2007 with linked prescription and medical claims who enrolled in stand-alone Part D or retiree drug plans and reached the gap spending threshold.
Measurements
Rates of death and hospitalization for any of 5 cardiovascular outcomes, including acute coronary syndrome+revascularization (ACS), after reaching the coverage gap spending threshold were compared using Cox proportional hazards models.
Results
In PS-matched analyses, exposed beneficiaries had elevated but non-significant hazards of death (HR=1.25; 95% CI 0.98–1.59) and ACS (HR=1.16; 0.83–1.62) compared with unexposed patients. hdPS-matched analyses minimized residual confounding and confirmed results: death (HR=0.99; 0.78–1.24); ACS (HR=1.07; 0.81–1.41). Exposed beneficiaries were no more or less likely to experience other outcomes than were the unexposed.
Conclusions
During the short-term coverage gap period, having no financial assistance to pay for drugs was not associated with an increased risk of death or hospitalization for cardiovascular causes. However, long-term health consequences remain unclear.
Keywords: Medicare Part D, coverage gap, adverse health outcomes, cardiovascular disease, drug discontinuation
Between 2.9–3.8 million (11–14%) Medicare Part D beneficiaries reach the Part D coverage gap period each year and are responsible for 100% of drug costs.1, 2 Beneficiaries enter the coverage gap following a period of initial coverage ($2,250 in total drug spending in 2006) and remain in the coverage gap until out-of-pocket spending reaches a catastrophic coverage spending threshold ($3,600 in 2006) when drug cost-sharing is dramatically reduced, or until year’s end.3 While the 2010 Patient Protection and Affordable Care Act contains provisions that will gradually eliminate the coverage gap by 2020, there are no short-term solutions to eliminate beneficiaries’ exposure4 and efforts to repeal these reforms are underway. 5, 6 Recently, we documented 1.65-fold increased discontinuation rates and greater rates of poor adherence for both cardiovascular and all drugs among Medicare beneficiaries enrolled in stand-alone Part D plans who entered the Part D coverage gap and were responsible for 100% of their drug costs as compared to beneficiaries who received financial assistance during this time.7 Similar studies in the Medicare Advantage Part D population have produced analogous results.811 Other drug benefits with gaps in coverage (drug caps, high deductible health plans) have been associated with both increased drug discontinuation and subsequent increases in adverse health outcomes and death.1215 To date, the effect of the Part D coverage gap on clinically relevant endpoints is unknown but urgently needed—closing the Part D gap might be expedited if adverse health effects were observed.
In this study, we examined the impact of entering the coverage gap on health outcomes in the same nationwide cohort of elderly Medicare beneficiaries in which we observed increased drug discontinuation during the gap.7 We hypothesized that beneficiaries who were responsible for 100% of drug costs during the gap would be more likely to experience adverse health events as a result of increased drug discontinuation and reduced drug adherence during this time.
Study design and exposures
We conducted a prospective open cohort study in 2 cohorts of Medicare beneficiaries, one cohort eligible in 2005–2006 (Early Part D Cohort) and a second cohort eligible in 2006–2007 (Established Part D Cohort). Beneficiaries entered the study on the date when they reached the coverage gap spending threshold ($2,250 in 2006; $2,400 in 2007 in plan+beneficiary spending, Appendix Figure 1). Exposed beneficiaries were responsible for 100% of drug costs after reaching the coverage gap spending threshold (i.e., received no financial assistance) and unexposed otherwise. Beneficiaries enrolled in plans with generic drug coverage during the coverage gap were classified as exposed; these beneficiaries were removed in sensitivity analyses. In our dataset, no Part D plan offered both generic and branded drug coverage during the coverage gap. Beneficiaries were censored at the earliest date of an outcome, reaching the catastrophic coverage spending threshold (beneficiary out-of-pocket drug spending of $3,600 in 2006; $3,850 in 2007), or December 31.
Data source, study population
We studied fee-for-service Medicare beneficiaries aged≥65 with prescription drug coverage through one of 129 stand-alone Part D plan or a retiree drug plan in 2006 or 2007 that was administered by Caremark, a pharmacy benefits manager. Caremark drug claims were linked to Medicare Parts A, B, and enrollment data to obtain demographic, diagnostic, and healthcare utilization information.
Beneficiaries had ≥1 Medicare claim in both the baseline and study years. Because Part D started in 2006, Early Part D Cohort (2005–2006) beneficiaries only had prescription claims available for 2006. For these beneficiaries, we required continuous Caremark eligibility and ≥1 prescription claim in 2006. Established Part D Cohort (2006–2007) beneficiaries had continuous Caremark eligibility and ≥1 prescription claim in 2006 and in 2007. To ensure we had some drug information for all beneficiaries, we restricted the cohorts to beneficiaries reaching the coverage gap spending threshold ≥60 days after drug plan enrollment in the study year. Additionally, we restricted our cohorts to those who did not enter a nursing home or hospice during the baseline year nor in the 2 month “trigger period” prior to reaching the coverage gap spending threshold, as these admissions would be indicative of worsening health and might confound the exposure-outcome relationship.
We used an algorithm7 that considered plan enrollment and plan and beneficiaries’ out-of-pocket spending to categorize beneficiaries into four benefit types (Appendix Table 1). Part D-enrolled beneficiaries who did not receive financial assistance to pay for drugs (non-subsidy enrollees) reached the coverage gap spending threshold and were subsequently responsible for 100% of their drug costs in the coverage gap. These beneficiaries were classified as exposed. The remaining three groups, Part D full subsidy beneficiaries, Part D partial subsidy beneficiaries and retirees, all received financial assistance after reaching the coverage gap spending threshold and were classified as unexposed. Full subsidy beneficiaries were enrolled in Part D plans with a coverage gap; however, their low incomes (<$7,500 in 2006; <$7,620 in 2007) limited their per-prescription cost-sharing ≤$5 in 2006 or ≤$5.35 in 2007, even in the coverage gap period. Full subsidy beneficiaries are dually eligible for Medicare (usually based on age ≥65) and Medicaid (based on either low income or high medical costs relative to income (e.g., spend down)).16 Partial subsidy beneficiaries were enrolled in Part D plans with a coverage gap, but their incomes ($7,500–$11,500 in 2006; $7,620–$11,710 in 2007) limited cost-sharing ≤15% for each prescription, even in the coverage gap period. The final unexposed beneficiaries were of retirees enrolled in retiree (non-Part D) drug plans. None of the retiree drug plans had a coverage gap design, so cost-sharing for each prescription did not change when beneficiaries reached the coverage gap spending threshold. While the 3 beneficiary types in the unexposed group had different cost-sharing, none were responsible for 100% of drug costs after they reached the coverage gap spending threshold and thus provided an appropriate comparator group. To examine concerns that the unexposed group’s heterogeneity might bias results through unmeasured selection (confounding) factors, we conducted sensitivity analyses in which the unexposed group was limited to retirees only.
Covariate assessment and propensity score matching
A beneficiary’s characteristics influence Part D plan choice and cost-sharing obligations, and in turn, both patient and plan characteristics influence whether he will reach the coverage gap spending threshold and thus be responsible for all drug costs. To adjust for such ‘confounding by health system use’, a propensity score (PS) model assessed each beneficiary’s likelihood of exposure after reaching the coverage gap during two time periods: the baseline year and a “trigger period,” the 2 months prior to reaching the coverage gap spending threshold. Covariates assessed in both the baseline year and in the 2 month trigger period included diagnosis of dementia, cancer, COPD/emphysema, renal failure, end-stage renal disease, depression, HIV/AIDS, diabetes, atrial fibrillation (AF), hypercholesterolemia, hypertension, coronary artery disease, congestive heart failure (CHF), stroke, venous thromboembolism, myocardial infarction (MI), and/or acute coronary syndrome (ACS) with revascularization, Charlson comorbidity score,14 and number of office-based drug infusions, physician visits, and hospitalizations. Additional predictors included age, gender, race, region of the U.S., rural/urban residence, median household income from Census block data,17 time (in days) from plan enrollment to reaching the threshold, and total Medicare inpatient and outpatient spending in the baseline year. The number of unique drugs used,15 use of medications to prevent or treat MI/ACS/revascularization/CHF, stroke, and VTE, and total (plan + beneficiary) drug spending in the 2 months prior to cohort entry were added to the model. To explore the sensitivity of our results, we considered 2 alternative covariate assessment strategies: assessment in the 12 months; or 6 months prior to reaching the coverage gap spending threshold. Definitions and codes are in the Appendix. Each exposed beneficiary was PS- matched to one unexposed beneficiary using nearest neighbor matching.18
To further explore the sensitivity of our results to residual confounding, we secondly modeled a high-dimensional propensity score (hdPS). While the PS approach requires that candidate confounders be identified a priori, the hdPS approach identifies and selects large numbers of empirical covariates for adjustment based on their potential to causing bias. The hdPS’ goal is to minimize residual confounding due to measured covariates (often proxies for unmeasured factors) that were not originally included in the PS.19 To implement the hdPS, we first identified 10 domains from which covariates might be selected: drug use; plan cost-sharing details; and inpatient, outpatient, ambulatory, and skilled nursing facility diagnoses and procedures. Due to the likelihood of rare outcomes, we assessed each variable’s potential for confounding based solely on its association with exposure and screened candidate covariates so as to not include covariates that were potential instruments.19, 20 The first 400 of these empirically-identified covariates, together with all PS model covariates, were included in the final hdPS model.21 We then employed nearest neighbor 1:1 matching as before. To assess the discriminative ability of the PS and hdPS models, we calculated c statistics.
Drug discontinuations
To explore the plausibility of our hypothesis that reaching the coverage gap spending threshold led to increased drug discontinuation rates which in turn might result in adverse health outcomes, we calculated the time to discontinuation of drugs to treat cardiovascular disease, diabetes, dementia, depression, and rheumatoid arthritis (drug list in Appendix). Drugs available upon reaching the coverage gap spending threshold were considered for discontinuation, defined as >30 days when no days’ supply was available and no further fills were made during the coverage gap.7 If increased rates of drug discontinuation among the exposed began immediately after beneficiaries reached the coverage gap spending threshold, we might be confident that subsequent adverse health events might be associated with drug discontinuations and coverage gap exposure.
Outcomes
Our primary outcome was death from any cause. We then focused on cardiovascular-related outcomes, given the high prevalence of cardiovascular disease in the general population, the many drugs available to prevent and treat it, and the likelihood of events if drugs were indeed discontinued. We examined the number of emergency room visits and non-discretionary hospitalizations (hospitalizations within 24 hours of an emergency room visit) with a primary or secondary diagnosis of a cardiovascular condition. In both exposed and unexposed groups, we assessed rates of first hospitalization with a primary or secondary diagnosis code for ACS with revascularization; CHF; or AF and rates for two composite outcomes, death or hospitalization for MI or stroke; and hospitalization for MI or stroke. In sensitivity analyses, we defined hospitalization outcomes using the primary diagnosis code only and conducted separate analyses among beneficiaries who reached the coverage gap spending threshold before August (<236 days, the median time to reach the threshold) and August or after (≥ 236 days) to explore effect modification. Definitions and codes are in the Appendix.
Statistical analysis
We cross-tabulated beneficiaries’ characteristics at baseline by benefit group before and after matching. We produced Kaplan-Meier curves for time to discontinuation of drugs and calculated discontinuation risk for the exposed and unexposed at the median and interquartile range times spent in the coverage gap period. Each individual drug was assessed for potential discontinuation, so analyses were at the drug level. For both PS- and hdPS-matched cohorts, we used Cox proportional hazards models to estimate hazard ratios for death and each cardiovascular outcome. Poisson models for emergency room visits and non-discretionary hospitalizations used generalized estimating equation methods22 to account for repeated events among individuals and a scale parameter to account for overdispersion.23 After testing for effect measure modification by cohort using Wald’s tests and finding none, we conducted pooled cohort analyses that accounted for the correlation between beneficiaries present in both cohorts by using robust standard errors.23 The Human Subjects Committee at Brigham and Women’s Hospital approved this study. Data use agreements were in place with all data providers.
Among all 303,978 beneficiaries who reached the coverage gap in 2006 or 2007, 94,220 (31%) were enrolled in Part D plans while the remainder were enrolled in non-Part D retiree plans. Among enrollees in Part D plans, 9,436 (10%) received no financial assistance to pay for drugs during the coverage gap (Table 1). At least 79% had hypertension, and ≥35% had coronary artery disease. In the Early Part D cohort, 38% of non-subsidy enrollees had diabetes compared to 50% of full subsidy beneficiaries. Among retirees, 35% had coronary artery disease compared to 40% of non-subsidy enrollees. In the 2 months prior to cohort entry in 2006, non-subsidy beneficiaries used an average of 6 drugs, whereas partial subsidy beneficiaries used an average of 7 and full subsidy beneficiaries an average of 8. All beneficiaries reached the gap spending threshold late in the year. Non-enrollees reached the threshold last, after 229–240 days (~8 months).
Table 1
Table 1
Baseline characteristics of all 303,978 Medicare beneficiaries who reached the coverage gap spending threshold in 2006 or 2007.
The PS model had a c statistic of 0.72. Of the 9,436 exposed beneficiaries, 9,383 (99%) could be PS-matched. The hdPS model had an improved c statistic of 0.91; still, nearly all exposed patients, 9210 (98%), could be hdPS-matched. After PS-matching, measured covariate distributions were largely balanced between the exposed and unexposed groups (Table 2), and residual differences were minimal: for example, unexposed matched beneficiaries had $123 more Medicare Parts A and B spending in the baseline year than did exposed beneficiaries. hdPS-matched results were similar (data not shown).
Table 2
Table 2
Selected baseline characteristics* of multivariate propensity score 1:1 matched beneficiaries in the Early Part D cohort, 2006 and the Established Part D cohort, 2007.
Kaplan-Meier curves demonstrated the almost immediate difference in drug discontinuation rates after coverage gap entry (Figure 1). Beneficiaries spent a median 95 days (IQR: 49, 145) in the coverage gap, with exposed beneficiaries having discontinued 10.1% (IQR: 4.8%, 13.5%) of drugs by day 95, whereas unexposed had discontinued 6.1% (IQR: 2.8%, 8.8%).
Figure 1
Figure 1
Survival function estimates: drug discontinuation since reaching the coverage gap spending threshold
In the PS-matched pooled cohort, the exposed death rate was 25% greater than that of the unexposed, 5 versus 4 deaths/100 person-years (Table 3). Among the exposed, 73% of deaths occurred outside the hospital, whereas among the unexposed, 74% of deaths occurred outside the hospital. Exposed patients were 20% more likely to have hospitalizations for MI or stroke or to die than were the unexposed. In the hdPS-matched pooled cohort, differences in event rates for this outcome disappeared. Similarly, in the hdPS-matched cohort, the exposed death rate was no greater than that of the unexposed and exposed patients were no more likely to die outside of a hospital setting than were unexposed patients (73% versus 68%, p=0.31).
Table 3
Table 3
Number and rates of health services utilization and cardiovascular outcome events in the coverage gap period among beneficiaries who received no financial assistance (exposed) compared with beneficiaries who received financial assistance (unexposed) in (more ...)
In PS-matched analyses, exposed beneficiaries had elevated but non-significant increased hazards of death (HR=1.25, 95% CI, 0.98–1.59) (Table 4). Hazard ratios were nearly identical when location of death was considered: out-of-hospital death (HR=1.25, 0.94–1.65). Exposed beneficiaries were 28% more likely to die or have a hospitalization for MI or stroke (1.02–1.60) but no more or less likely than unexposed beneficiaries to have increased rates of non-discretionary hospitalizations (hospitalizations occurring within 24 hours of an emergency room visit, HR=0.65, 0.41–1.03) or emergency room visits (HR=0.87, 0.74–1.02) with cardiovascular diagnoses. In hdPS-matched analyses, the hazard of all cause death remained non-significant (HR=0.99, 0.78–1.24) as did the hazard of out-of-hospital death (HR=1.07, 0.81–1.41). In contrast to PS-matched results, in the hdPS-matched analysis, exposed beneficiaries were no more or less likely to die or have a hospitalization for MI or stroke (HR=1.04, 0.84–1.30). None of the analyses showed effect modification by pre-August versus August and after arrival to the spending threshold.
Table 4
Table 4
Hazard ratios of adverse health events in the coverage gap, comparing beneficiaries who received no financial assistance (exposed) with beneficiaries who received financial assistance (unexposed) in the coverage gap.
The separate sensitivity analyses which 1) excluded exposed beneficiaries with generic drug coverage during the coverage gap period, 2) limited the unexposed group to retirees only (excluded low-income beneficiaries with full or partial subsidies), 3) explored alternate covariate assessment periods, or 4) used only the primary diagnosis code for cardiovascular outcomes all produced analogous results.
In this study of health outcomes among beneficiaries who reached the Part D coverage gap spending threshold, we found that having no financial assistance to pay for drugs in the coverage gap was not associated with an elevated risk for death or hospitalization for cardiovascular outcomes during the 3 month median length coverage gap. Although exposure was associated with a 28% (2–61%) increased risk of the composite outcome death or hospitalization for MI or stroke during the gap in the main analysis, this effect became clinically meaningless in sensitivity analyses with finer control for measured confounders (HR=1.04, 0.84–1.30). Our findings of no increased risk of adverse health outcomes are the first to use linked medical and pharmacy claims to explore the clinical consequences of increased drug discontinuation and decreased drug adherence during the coverage gap.
Drug discontinuation and reduced adherence are associated with poor clinical outcomes,2428 and our results are inconsistent with studies that have examined the impact of drug benefit restrictions and poor medication adherence on health outcomes in other settings.13, 14, 24, 26, 27, 29 For example, increased prescription cost-sharing for Canadian elderly was associated with a 9% decrease in essential drug use and a 7% increased rate of adverse events, including death.14 The fact that we did not observe an increase in unintended health outcomes despite the documented increase in drug discontinuation during the coverage gap is likely due to several factors. While most prior studies that showed adverse health effects had follow-up of >1 year,8, 1214, 24, 26, 27, 29, 30 beneficiaries in our study reached the coverage gap relatively late in the year (median 236 days (~8 months)) and spent only a median 95 days in the gap period before they regained drug coverage at the beginning of the next calendar year. For the average beneficiary, the length of time spent in the coverage gap may be too short to experience any meaningful unintended health outcome.26, 27 This observation highlights the peculiar structure of the coverage gap: in contrast to other drug cost-sharing policies, the gap is a temporary, often brief period after which one regains insurance at the beginning of the next year. While it is possible to study adverse outcomes after beneficiaries exit the coverage gap and regain drug insurance, it is then more difficult to attribute that adverse event to coverage gap exposure. Secondly, on the absolute scale, coverage gap exposure was associated with the additional discontinuation of 15 drugs per 100 person-years,7 a relatively small effect in absolute terms; combined with the brief length of the gap period, we may not be able to observe sufficient numbers of events. Finally, our results may differ due to the characteristics of patients included in our sample. Patients were eligible for our study because they repeatedly filled prescriptions whose aggregate costs brought them to the coverage gap spending threshold. Persistence with medications is associated with other healthy behaviors that reduce the likelihood of adverse outcomes.31 At present, the long-term impact of the coverage gap on health outcomes remains unclear and merits further scrutiny.
Given the non-randomized design of our study, we cannot rule out the possibility of unmeasured confounding or residual confounding that would support alternate explanations. However, we did employ multiple approaches to reduce confounding, including restriction of the population to those with no nursing home or hospice admissions, PS-matching and hdPS-matching. In particular, the hdPS approach allows for extensive adjustment for covariates that may be confounders themselves and are often proxies for or correlated with unmeasured confounders, thus reducing residual confounding and minimizing the impact of unmeasured confounding.19 Compared to the PS model’s c-statistic of 0.72, the hdPS model c-statistic was 0.91, indicating better discriminative ability of this model and improved balance of measured covariates between groups. The advantages of the hdPS as a sensitivity analysis were most apparent in examining the hazard ratio for death or hospitalization for MI or stroke. Whereas the PS-matched result was marginally significant, HR=1.28 (1.02–1.60), the hdPS-matched result reduced residual confounding and showed non-significant differences between exposed and unexposed beneficiaries, HR=1.04 (0.84–1.30). Our study has robust external validity, as we examined the heterogeneous experiences of beneficiaries in 129 stand-alone Part D plans who reflect the larger population of Part D enrollees, 66% of whom were enrolled in stand-alone plans in 2010.27 In our study, 62% of beneficiaries were women, comparable to the 60% female stand-alone plan population and 52% of beneficiaries were aged 65–74, comparable to the 49% in the larger stand-alone population.32 Our study population was less racially diverse: 8.4% African American compared with 12% in the larger population.
Our results suggest no increased risk of death or hospitalization for other cardiovascular outcomes during the short-term coverage gap period, despite evidence of increased rates of drug discontinuation and reduced drug adherence among those fully exposed to drug costs during that time. These apparent null findings could be explained by the short follow-up period that reflects the Part D coverage gap’s peculiar structure, the small absolute effect of coverage gap exposure on drug discontinuation, and the small number of observed events. In contrast to the larger body of research evidence describing the deleterious effects of restricting drug insurance benefits on health outcomes, we find that the coverage gap might on average be just short enough to avoid short-term negative health outcomes. However, it remains unclear whether Part D coverage gap-related lapses in financial assistance to pay for drugs affect health outcomes in the long-term. Any drug cost-sharing policy which reduces essential medication use is problematic, and present knowledge deficits regarding the gap’s long-term effects should be considered as Part D beneficiaries’ financial responsibilities in the coverage gap change under the Affordable Care Act’s incremental reforms.4
Supplementary Material
Supp TableS1-S3&FigureS1
ACKNOWLEDGMENTS
The authors would like to thank Joyce Lii, MS for exceptional programming assistance.
Funding: National Institute on Aging T32 AG000158 (Dr. Polinski); National Institute of Mental Health R01 5U01MH079175-02 (Dr. Schneeweiss); National Heart Lung and Blood Institute K23 HL-090505 (Dr. Shrank), a Robert Wood Johnson Foundation Investigator Award in Health Policy Research (Dr. Huskamp), and a research grant from CVS Caremark.
Sponsor’s role: The sponsors had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; or preparation, review, or approval of the manuscript.
Footnotes
Conflict of interest:
Dr. Polinski is a consultant to Buccaneer Computer Systems and Service, Inc, a contractor for the Centers for Medicare and Medicaid Services. Within the past 5 years, Dr. Polinski’s spouse was employed as an engineer by DePuy Orthopaedics, a subsidiary of Johnson & Johnson, and had Johnson & Johnson stock totaling < $3100 in value. Dr. Shrank is a consultant to United Healthcare, which has a Part D business, but his consulting is unrelated. Dr. Shrank has received research funding from CVS Caremark, Aetna and Express Scripts, which also have Part D business. Dr. Glynn has worked on grants to the Brigham & Women's Hospital, his employer, from Astra Zeneca and Novartis related to the design, statistical monitoring, and analysis of clinical trials in the setting of cardiovascular drugs. Dr. Glynn also signed a consulting agreement to give a one-time Grand Rounds talk on comparative effectiveness research methods at Merck. At the time of the study, Mr. Roebuck was an employee of CVS Caremark. Dr. Schneeweiss is a paid member of the Scientific Advisory Board of HealthCore and a consultant to World Health Information Science Consultants, LLC. Dr. Schneeweiss is Principal Investigator of the Brigham and Women’s Hospital DEcIDE Center on Comparative Effectiveness Research funded by AHRQ and the DEcIDE Methods Center. Dr. Schneeweiss received funding through investigator-initiated grants awarded to his employer, Brigham and Women’s Hospital from Pfizer, Novartis, and Boehringer-Ingelheim. Opinions expressed here are only those of the authors and not necessarily those of the agencies.
Author contributions: Dr. Polinski was responsible for the conception and design, acquisition of data, analysis and interpretation of data, statistical analysis, drafting of the manuscript, and critical revision of the manuscript. Dr. Polinski had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Dr. Shrank participated in the design of the study, acquisition of data, analysis and interpretation of data, and critical revision of the manuscript. Drs. Huskamp and Glynn participated in the design of the study, analysis and interpretation of data, and critical revision of the manuscript. Mr. Roebuck participated in the acquisition of the data, analysis and interpretation of the data, and critical revision of the manuscript. Dr. Schneeweiss participated in the design of the study, analysis and interpretation of data, critical revision of the manuscript, and supervision.
1. Hoadley J, Hargrave E, Cubanski J, et al. [Accessed May 20, 2011];The Medicare Part D coverage gap: Costs and consequences in 2007. 2008 Available at: http://www.kff.org/medicare/upload/7811.pdf.
2. Powers C. Beneficiary experience. Paper presented at: Medicare Part D data symposium; October 30, 2008; Baltimore, M.D.. [Accessed May 20, 2011]. Available at: http://www.cms.gov/PrescriptionDrugCovGenIn/08_PartDData.asp.
3. Centers for Medicare and Medicaid Services. Medicare Prescription Drug Benefit; Final Rule, 42 CFR 70 no. 18. 2005
4. U.S. Congress. Patient Protection and Affordable Care Act. P.L. No. 111–148; 111th Congress, 2nd session. 2010
5. Ryan P. [Accessed April 22, 2011];The Path to Prosperity: Restoring America's Promise. 2011 Available at: http://budget.house.gov/UploadedFiles/PathToProsperityFY2012.pdf.
6. Kaiser Family Foundation. [Accessed April 22, 2011];Proposed changes to Medicare in the "Path to Prosperity.". 2011 Available at: http://www.kff.org/medicare/upload/8179.pdf.
7. Polinski JM, Shrank WH, Huskamp HA, et al. Changes in drug utilization during a gap in insurance coverage: An examination of the Medicare Part D coverage gap. PLoS Med. 2011;8:e1001075. [PMC free article] [PubMed]
8. Raebel MA, Delate T, Ellis JL, et al. Effects of reaching the drug benefit threshold on Medicare members' healthcare utilization during the first year of Medicare Part D. Med Care. 2008;46:1116–1122. [PubMed]
9. Schmittdiel JA, Ettner SL, Fung V, et al. Medicare Part D coverage gap and diabetes beneficiaries. Am J Manag Care. 2009;15:189–193. [PMC free article] [PubMed]
10. Schneeweiss S, Patrick AR, Pedan A, et al. The effect of Medicare Part D drug coverage on drug use and cost-sharing in seniors without prior drug benefits. Health Aff. 2009;28:w305–W316. [PMC free article] [PubMed]
11. Zhang Y, Donohue JM, Newhouse JP, et al. The effects of the coverage gap on drug spending: A closer look at Medicare Part D. Health Aff. 2009;28:w317–w325. [PMC free article] [PubMed]
12. Soumerai SB, Ross-Degnan D, Avorn J, et al. Effects of Medicaid drug-payment limits on admission to hospitals and nursing homes. N Engl J Med. 1991;325:1072–1077. [PubMed]
13. Hsu J, Price M, Huang J, et al. Unintended consequences of caps on Medicare drug benefits. N Engl J Med. 2006;354:2349–2359. [PubMed]
14. Tamblyn R, Laprise R, Hanley JA, et al. Adverse events associated with prescription drug cost-sharing among poor and elderly persons. JAMA. 2001;285:421–429. [PubMed]
15. Joyce GF, Goldman DP, Karaca-Mandic P, et al. Pharmacy benefit caps and the chronically ill. Health Aff. 2007;26:1333–1344. [PubMed]
16. Kaiser Commission on Medicaid and the Uninsured. Proposed Models to Integrate Medicare and Medicaid Benefits for Dual Eligibles: A Look at the 15 State Design Contracts Funded by CMS. http://www.kff.org/medicaid/upload/8215.pdf.
17. U.S. Census. [Accessed January 11, 2010];American Fact Finder. http://factfinder.census.gov/servlet/DatasetMainPageServlet?_program=DEC&_submenuId=&_lang=en&_ts=.
18. Austin PC. Some methods of propensity-score matching had superior performance to others: Results of an empirical investigation and Monte Carlo simulations. Biom J. 2009;51:171–184. [PubMed]
19. Schneeweiss S, Rassen JA, Glynn RJ, et al. High-dimensional propensity score adjustment in studies of treatment effects using health care claims data. Epidemiology. 2009;20:512–522. [PMC free article] [PubMed]
20. Rassen JA, Glynn RJ, Brookhart MA, et al. Observed performance of high-dimensional propensity score analyses of treatment effects in small samples. Am J Epidemiol. 2011;173:1404–1413. [PMC free article] [PubMed]
21. Rassen JA. [Accessed February 12, 2011];High Dimensional Pharmacoepidemiology toolbox. http://www.hdpharmacoepi.org/.
22. Liang KY, Zeger SL. Regression analysis for correlated data. Ann Rev Pub Health. 1993;14:43–68. [PubMed]
23. Glynn RJ, Buring JE. Ways of measuring rates of recurrent events. BMJ. 1996;312:364–367. [PMC free article] [PubMed]
24. Ho PM, Peterson ED, Wang L, et al. Incidence of death and acute myocardial infarction associated with stopping clopidogrel after acute coronary syndrome. JAMA. 2008;299:532–539. [PubMed]
25. Ho PM, Spertus JA, Masoudi FA, et al. Impact of medication therapy discontinuation on mortality after myocardial infarction. Arch Intern Med. 2006;166:1842–1847. [PubMed]
26. Rasmussen JN, Chong A, Alter DA. Relationship between adherence to evidence-based pharmacotherapy and long-term mortality after acute myocardial infarction. JAMA. 2007;297:177–186. [PubMed]
27. Ho PM, Magid DJ, Shetterly SM, et al. Medication nonadherence is associated with a broad range of adverse outcomes in patients with coronary artery disease. Am Heart J. 2008;155:772–779. [PubMed]
28. Ho PM, Tsai TT, Wang TY, et al. Adverse events after stopping clopidogrel in post–acute coronary syndrome patients circ. Cardio Outcomes Qual. 2010;3:303–308. [PubMed]
29. 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;166:1836–1841. [PubMed]
30. Heisler M, Langa KM, Eby EL, et al. The health effects of restricting prescription medication use because of cost. Med Care. 2004;42:626–634. [PubMed]
31. Dormuth CR, Patrick AR, Shrank WH, et al. Statin adherence and risk of accidents: A cautionary tale. Circulation. 2009;119:2051–2057. [PMC free article] [PubMed]
32. Centers for Medicare and Medicaid Services. Medicare and Medicaid Statistical Supplement, Table 14.4. [Accessed September 18, 2011];Health Care Fin Rev. 2008 Available at: http://www.cms.gov/medicaremedicaidstatsupp/lt/itemdetail.asp?itemid=CMS1216048.