Cost-Effectiveness of Medicare Drug Plans in Schizophrenia and Bipolar Disorder

, , , ,
The American Journal of Managed Care, February 2013, Volume 19, Issue 2

In Medicare Part D, generic drug coverage was cost saving compared with no coverage in bipolar disorder and schizophrenia while improving health outcomes.


Medicare Part D has a drug coverage gap, which imposes risks for discontinuing medications, particularly in mental health disorders where drug costs are high. However, some beneficiaries have generic drug coverage in the gap.


To examine the health outcomes and cost-effectiveness of generic-drug coverage compared with no gap coverage in patients with bipolar disorder and schizophrenia.

Study Design:

Markov model—based cost-effectiveness analysis using identical hypothetical cohorts to examine drug coverage strategies.


The incremental cost-effectiveness of Part D coverage strategies was estimated, using differences in medical costs and quality-adjusted life years between plans. Coverage strategy—specific costs and hospitalization rates were obtained from 2007 Medicare data, adjusted for age, sex, race, and health status.


When comparing generic-only coverage with no gap coverage, generic-only coverage cost less and was more effective than no gap coverage, due mainly to lower hospitalization rates. In sensitivity analyses, generic-only coverage continued to be favored over no gap coverage unless generic coverage costs increased >3% in bipolar disorder and >5% in schizophrenia; generic coverage in the gap was also favored in probabilistic sensitivity analyses.


In Medicare Part D, generic drug coverage was cost saving compared with no coverage in bipolar disorder and schizophrenia while improving health outcomes. Policy makers and insurers might consider generic-only coverage, rather than no gap coverage, to both conserve healthcare resources and improve health.

(Am J Manag Care. 2013;19(2):e55-e63)Optimal drug benefit design can decrease drug and nondrug spending while also improving health outcomes.

  • The Medicare Part D prescription drug benefit has a coverage gap, which has a greater impact on patients with bipolar disorder and schizophrenia.

  • A cost-effectiveness analysis of Part D plans is unconventional but can inform managed care decision making.

  • Generic drug coverage in the gap is cost saving compared with no gap coverage (the standard Part D design) and improves health outcomes.

Medicare Part D offers prescription drug coverage for Medicare beneficiaries. Since the program’s inception in 2006, many enrollees have benefited from improved drug coverage and increased medication use.1-3 However, a major concern is the large coverage gap in the standard Part D design, where beneficiaries pay 100% of medication costs out-of-pocket. About one-third of all Medicare beneficiaries enter this coverage gap each year,4,5 and once there, they often reduce medication use,4,6-8 which may lead to increases in hospitalization and medical spending.

The coverage gap is an even larger concern for Medicare beneficiaries with severe mental disorders such as bipolar disorder and schizophrenia. First, they are much more likely to enter the gap: 62% of aged beneficiaries with bipolar disorder and 56% of those with schizophrenia entered the gap in 2007.9 Second, if they discontinue psychotropic medications, they may relapse to more severe episodes and require psychiatric hospitalization.10 Third, they experience high rates of comorbid chronic physical conditions such as heart disease and diabetes, which can be exacerbated by untreated mental illness and increase morbidity, medical spending, and mortality.11

The standard Part D benefit in 2007 included 4 phases: (1) an initial $265 deductible; (2) a period in which beneficiaries paid 25% of drug costs between $265 and $2400; (3) a coverage gap in which they paid 100% of costs between $2401 and $3850, where they reached their total out-of-pocket spending catastrophic limit; and (4) a catastrophic coverage period where they paid 5% of costs.12 Although the standard Part D benefit includes these 4 phases, some companies offering Part D drug plans modified the design and offered either “actuarially equivalent” or enhanced plans. In 2007, for example, 72% of stand-alone Part D plans had the standard coverage gap, 27% offered coverage for generic drugs used in the gap, and fewer than 1% offered coverage for both brandname and generic drugs.

In addition, the Medicare Part D drug benefit includes substantial premium and cost-sharing subsidies for Medicare beneficiaries with low incomes, which come directly from the federal government. Each year, roughly 50% of beneficiaries enrolled in Part D plans receive a low-income subsidy (LIS).13 Due to more generous benefits, those with LIS are not exposed to the coverage gap even when their pharmacy spending reached the coverage gap threshold.14 Enrollment in LIS is not due to self-selection, because only beneficiaries who qualify due to low income can choose LIS and most eligible beneficiaries do enroll in LIS. For this reason, this group is very different from non-LIS beneficiaries.

Our objective in this analysis was to examine differences in health outcomes and costs between coverage groups in patients with bipolar disorder and schizophrenia. We were most interested in differences between the no-gap-coverage and the generic-coverage groups, because plan self-selection could be an issue and policies governing the plans’ availability and affordability could affect health costs and outcomes for beneficiaries entering the coverage gap. For example, sicker patients could more often select the more generous generic coverage, while patients in the less generous no-gap-coverage plans might defer care more often and have worse health outcomes as a result. Beneficiaries in those plans are also likely to be more comparable to each other than to beneficiaries in LIS plans, where prominent differences in income and education level occur. In the primary analysis, we evaluated the costeffectiveness of generic coverage compared with no coverage. In a secondary analysis, we also included the LIS plan to complete the comparisons.

METHODSData Source

We obtained demographic and enrollment information, plan benefits, prescription drug events, and medical claims for Medicare beneficiaries with schizophrenia and/ or bipolar disorder (100% of non-dual eligibles and a 20% random sample of dual eligibles) who were continuously enrolled in Part D plans in 2007. We identified patients with each condition based on the existence of 2 or more inpatient or outpatient claims for schizophrenia (International Classification of Diseases, Ninth Revision, code 295) or bipolar disorder (codes 296.0, 296.1, 296.4-2.96.8, 301.11, 301.13) between January 1, 2007, and December 31, 2007. Because their treatment patterns might differ, we separately evaluated those who qualified for Medicare due to disability (denoted hereafter as disabled) or due to age (denoted as aged). The study design was approved by the institutional review board at the University of Pittsburgh.


Health outcomes included mortality and all-cause hospitalization rates. Cost outcomes were drug costs and total medical costs. The drug costs included Part D plan payment before rebates, beneficiary out-of-pocket spending, and subsidy amount. The total medical costs included inpatient, outpatient, and physician services as well as drug costs.

Study Groups and Adjustments

To control for differences in beneficiary characteristics between groups, all outcomes (hospitalization, mortality, and costs) were adjusted for age, sex, race, number of Elixhauser comorbidities, and the Centers for Medicare & Medicaid Services Hierarchical Condition Category (CMS-HCC) scores; for pharmaceutical spending, we used the analogous prescription drug hierarchical condition category (CMS-RxHCC) scores.15 CMS-HCC and RxHCC scores are the beneficiary risk adjusters used by CMS to adjust payment to plans for medical and pharmacy costs, respectively.

The adjustment was conducted in 2 steps. First, we ran separate regressions for each outcome, controlling for drug coverage study group (eg, no coverage vs generic coverage) and the covariates listed above. Second, we calculated the predicted values based on beneficiaries’ drug coverage group and the averages for the covariates across coverage groups. More precisely, in the first stage, we ran a logistic regression for binary outcomes such as mortality; for costs, we ran an ordinary least squares regression to estimate the predicted value for each drug group. Some outcomes were approximately normally distributed, but inpatient spending had a large concentration of beneficiaries with no expenditures and some beneficiaries with very large expenditures. If our goal had been to obtain individual-level adjusted costs conditional on covariates, it would have been necessary to use a mixed-model approach (eg, a 2-part model or zeroinflated gamma model) to address the problem of skewed data with a large number of zero expenditures. However, our goal here was to estimate the predicted value for each group using the national average of the covariates; ordinary least squares regression gave us efficient estimators of the conditional expectation function. In addition, Buntin and Zaslavsky suggested that all these estimators give very similar results empirically.16 We ran this 2-stage adjustment for each outcome, separately for disabled and aged beneficiaries.

Self-selection could be a factor in the makeup of the no-coverage-gap and generic-only-coverage groups; patients who anticipate entering the coverage gap may chose the more generous generic- coverage plan. Because of this choice, we postulated more illness and healthcare expenditures in patients with generic coverage, which could bias the analysis against generic coverage compared with no gap coverage.

Markov Model

To estimate the incremental cost-effectiveness of drug coverage strategies, we constructed a 3-state Markov model comprised of outpatient therapy, hospitalization, and death. We used TreeAge Pro Suite 2009 software (TreeAge Software, Williamstown, Massachusetts) to build the model. Over the model’s 1-year time horizon and 12 monthly Markov cycles, hospitalization and death rates were calculated using Medicare utilization data. In this population, all patients were taking medications and few switched plans. In the Markov model, these characteristics were reflected in the assumptions that all cohort members were on at least 1 medication and that all cohort members remained in the same coverage group throughout the model time horizon. Beneficiaries could be hospitalized 1 or more times per year, based on the hospitalization probability.

We analyzed separate models for bipolar disorder and schizophrenia, as well as for disabled and aged beneficiaries. In the primary analyses, we used identical hypothetical condition—specific cohorts to test differences in costs and outcomes between no gap coverage and generic-only coverage for all cohort members; the secondary analysis examined all 3 coverage strategies (ie, no gap coverage, generic-only coverage, and LIS).

Costs were total drug and nondrug costs from the Medicare perspective, incorporated into the model as constant monthly costs. Hospitalization costs were not considered separately, but instead as a component of nondrug medical costs. Because it was a 1-year model, discounting was not performed.17

Table 1

The effectiveness term was quality-adjusted life years (QALYs), accounting for quality-of-life differences between coverage groups using utility weights. QALYs are the product of a health state’s utility and time spent in that state. Utilities measure preference for health states on a scale where 0 equals death and 1 equals perfect health. Utility weights for bipolar disorder were derived from SF-36 values18; schizophrenia utilities were Euroqol utility values19 (). In the base case analysis, we assumed no differences in outpatient utility between coverage plans; thus, differences in effectiveness between plans were based on differences in hospitalization rates.

Utilities for hospitalized patients with bipolar disorder were calculated as a weighted average, based on the relative likelihood (as observed in the referenced study) of uncontrolled mania or depression and their respective utility weights.18 Utilities for hospitalized schizophrenia patients were based on average pretreatment utilities.19 Hospitalized patients were assumed to have decreased utility throughout the month of their hospitalization to account for decreased utility prehospitalization and posthospitalization, as well as during their hospital stay. To account for the possibility of hospitalizations of differing duration or condition severity, we performed sensitivity analyses where hospitalization utility was varied widely. In the base case, we assumed that uncontrolled illness had the same utility, hospitalized or not, possibly biasing against strategies that result in fewer hospitalizations.

Sensitivity Analysis

In sensitivity analyses, we examined the effects of individual variation of all parameters. Utilities were varied over ranges shown in Table 1; costs were varied from 90% to 110% of the base case values, and the absolute risks of hospitalization were varied ±5% from base case values. We also performed a probabilistic sensitivity analysis, varying all parameters simultaneously over distributions for 10,000 model iterations. Beta distributions were assigned to utilities, normal distributions to the 5% hospitalization risk variation, and gamma distributions to the 90% to 110% cost multipliers. Given the Medicare database size (180,270 patients), these ranges were broader than observed 95% confidence intervals, and we used these broader ranges to test the robustness of our analysis.

Utility ranges examined in the sensitivity analysis were based on standard deviations from the schizophrenia study.19 Individuals with more generous insurance coverage have decreased anxiety about coverage, which can lead to better health compared with those with less generous coverage.20 In base case analyses, we assumed no differences in outpatient utility between coverage strategies; however, we relaxed this assumption in a separate sensitivity analysis.

Cost-Effectiveness Criterion

There is no US cost-effectiveness criterion; however, in general, interventions costing less than $100,000 per QALY gained are considered economically reasonable, while those costing much more than this figure are felt to be an expensive use of healthcare resources.21,22 A criterion of $50,000 per QALY gained is commonly cited but poorly justified.21,23 In this study, we used a cost-effectiveness criterion of $100,000 per QALY gained.

RESULTSMedicare Data Analysis

Table 2

Appendix A

Appendix B

The Medicare Part D database for bipolar disorder and schizophrenia contained 180,270 patients, 87,747 (48.7%) with bipolar disorder and 92,523 (51.3%) with schizophrenia; among all patients, 14.6% had no gap coverage, 7.1% had generic coverage, and 78.4% had LIS. Adjusted hospitalization rates, drug costs, and total medical costs used in the model are shown in . Hospitalization rates, calculated as the hospitalization frequency divided by the number of beneficiaries, were highest in the no-coverage group for both aged and disabled beneficiaries with either disorder. Medication costs were substantially higher in LIS beneficiaries, but total medical costs were more similar between groups. No statistically significant differences in death rates between coverage plans were seen, with <1% annual mortality among disabled recipients and approximately 3% annual mortality among aged recipients with either disorder. The appendices show study group descriptive statistics () and parameter values input into the model (). When adjustment for census region was added to the prior adjustment, relative differences between strategies were essentially unchanged.

Cost-Effectiveness Analysis

Table 3

shows the primary cost-effectiveness analysis results, comparing no gap coverage with generic-only coverage and ordering strategies by cost.17 In both bipolar disorder and schizophrenia, generic coverage in the gap dominated (less costly, more effective) the no-gap-coverage strategy. In 1-way sensitivity analyses, results were sensitive only to variation of total medical costs. In bipolar disorder, no coverage was favored over generic-only coverage if costs associated with no coverage decreased or generic-only costs increased by >3% in either the aged or disabled recipients when using a $100,000 per QALY gained criterion. In schizophrenia, no coverage was favored if its costs decreased >6% in the disabled and >5% in the aged. Individual variation of other parameter values did not lead to no gap coverage being favored.

In probabilistic sensitivity analyses, the likelihood that generic coverage would be favored at a $100,000 per QALY gained threshold varied from 62% in aged patients with bipolar disorder to 81% in disabled patients with schizophrenia.

A separate sensitivity analysis examined the effect of coverage plans having differing quality-of-life utility values. If no gap coverage had lower outpatient utility values, then generic coverage was more strongly favored. If generic coverage had lower outpatient utility values, generic coverage was still favored unless it decreased >20% (data not shown).

In a secondary analysis, we considered LIS coverage in our model, comparing it with the other 2 coverage strategies. Compared with generic medication coverage, LIS gained 0.0005 to 0.0014 QALYs (or about 4 to 12 hours), costing about $800,000 or more per QALY gained for disabled or aged beneficiaries with either disorder (data not shown).


The Medicare coverage gap has significant effects on beneficiary health-seeking behaviors. Plans differ in ways of managing the coverage gap, raising questions about whether some coverage schemes are more worthwhile than others. Using Medicare data, we found that generic medication coverage within the coverage gap was less expensive than having no gap coverage in patients with schizophrenia or bipolar disorder, while also improving health outcomes as measured by fewer hospitalizations. Thus, these data suggest that no drug coverage in the coverage gap is more expensive than generic drug coverage due to greater nondrug costs (largely mediated by greater hospitalization risk), leading to decreased quality of life for patients with no gap coverage. We acknowledge that beneficiaries might select plans with generic-only coverage because they anticipate they might use more drugs and enter the coverage gap. Because of plan self-selection, beneficiaries with generic-only coverage might be sicker than those in the no-coverage group, thus biasing against generic-only coverage (ie, making it seem less effective). However, despite plan selfselection and the possibility of bias, we found that a genericonly coverage strategy was favored over no gap coverage, with this finding being robust to variation in sensitivity analyses.

Our adjustment procedures help mitigate differences in measured factors between the generic-coverage and no-gapcoverage groups. We postulate that major differences in unmeasured factors (eg, socioeconomic factors, access to care) are less common in these coverage groups than in LIS plan beneficiaries. Hence, comparisons between no gap coverage and generic coverage are more likely to be robust from that standpoint. Generic-coverage plans might also have protocols in place to encourage generic drug use during pregap periods; this aspect of those plans could be driving observed cost differences. However, if this does occur, the case favoring generic coverage would only be strengthened.

In a secondary analysis, we found that drug costs for LIS beneficiaries were substantially higher than those for beneficiaries with no gap coverage or generic-only coverage. These cost differences, along with relatively small differences in effectiveness between plans, resulted in large incremental cost-effectiveness ratios for LIS coverage compared with generic-only plans. That could be because LIS patients are sicker or use more drugs because of full coverage; the lack of adjustment for socioeconomic factors in our model is likely a factor as well. For this reason, we consider analyses that include LIS coverage less robust than those comparing the other 2 coverage strategies. Our analysis has other limitations. We used a simple model where utility decrements occurred only with hospitalization and utility values came from other sources. This model might have been too simple to capture the complexity of schizophrenia and bipolar disorder, and the possible differential effects of drug coverage plans on patients’ quality of life. To increase the flexibility of our analysis, particularly in regards to utility weights, which are not available in Medicare data, we did not use net benefit regression, a strictly statistical approach to estimate cost-effectiveness.24,25 Instead, using regression techniques, we derived parameter values and input them into a decision model to estimate incremental cost-effectiveness ratios. Finally, in a sensitivity analysis, we examined the effects of utility differences between plans, given evidence of stress associated with lack of insurance coverage,20 but whether this effect applies in Medicare patients or whether differences occur due to coverage plans themselves is unclear.

Our analysis does have the advantage of using the full Medicare data set, heightening the generalizability of our findings, and we adjusted for demographic and comorbidity differences between coverage groups. In addition, the use of decision analysis modeling techniques allowed us to test the robustness of results by systematically varying parameter values and pointing out where thresholds exist.

In conclusion, using Medicare data, we found that generic medication coverage in the Medicare Part D coverage gap was cost saving compared with no gap coverage in patients with bipolar disorder and schizophrenia. Under the current provisions of the Affordable Care Act, the coverage gap will be gradually filled until 2020. For example, starting from 2012, patients will pay 50% for brand-name drugs filled in the coverage gap, but they have to pay 86% for generic drugs filled in the gap. Each year the percentage patients pay will decrease until it becomes 25% in 2020. Based on our findings, policy makers and insurers should consider a faster decrease of the percentage paid by patients for generic drugs as a means to improve the health of patients who do not qualify for LIS plans while conserving healthcare resources.Author Affiliations: From Department of Medicine, University of Pittsburgh School of Medicine (KJS, BLR), Pittsburgh, PA; Department of Health Policy and Management, Graduate School of Public Health, University of Pittsburgh (SHB, YZ), Pittsburgh, PA; Department of Psychiatry, School of Medicine and Department of Behavioral and Community Health Science, School of Public Health (CFR), Pittsburgh, PA.

Funding Source: National Institute of Mental Health RC1 MH088510, Agency for Healthcare Research and Quality R01 HS018657, University of Pittsburgh Central Research Development Fund.

Author Disclosures: Dr Reynolds reports receiving grants from the National Institutes of Health, Forest Labs, Bristol-Myers Squibb, and Pfizer. The other authors (KJS, SHB, BLR, YZ) 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 (KJS, YZ); acquisition of data (SHB, YZ); analysis and interpretation of data (KJS, SHB, CFR, BLR, YZ); drafting of the manuscript (KJS, SHB, CFR, BLR, YZ); critical revision of the manuscript for important intellectual content (KJS, CFR, BLR, YZ); statistical analysis (SHB); provision of study materials or patients (XXX); obtaining funding (YZ); administrative, technical, or logistic support (YZ); and supervision (YZ).

Address correspondence to: Yuting Zhang, PhD, Department of Health Policy and Management, University of Pittsburgh, 130 De Soto St, Crabtree Hall A664, Pittsburgh, PA 15261. E-mail: 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.

2. 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.

3. 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.

4. 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.

5. Centers for Medicare & Medicaid Services. Prescription drug coverage— general information. GenIn/09_ProgramReports.asp. Published 2010. Accessed February 3, 2011.

6. 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.

7. 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.

8. Schmittdiel JA, Ettner SL, Fung V, et al. Medicare Part D coverage gap and diabetes beneficiaries. Am J Manag Care. 2009;15(3): 189-193.

9. Zhang Y, Baik S, Zhou L, Reynolds CF, Lave JR. Medicare Part D’s coverage gap and mental health. Presented at: 164th Annual Meeting of the American Psychiatric Association; May 14-18, 2011; Honolulu, HI.

10. Kupfer DJ, Horner MS, Brent DA, et al, eds. Oxford American Handbook of Psychiatry. 1st ed. New York: Oxford University Press; 2008.

11. Jiang W, Alexander J, Christopher E, et al. Relationship of depression to increased risk of mortality and rehospitalization in patients with congestive heart failure. Arch Intern Med. 2001;161(15):1849-1856.

12. Henry J. Kaiser Family Foundation. The Medicare Prescription Drug Benefit. Fact Sheet. Published November 2006. Accessed September 8, 2011.

13. Centers for Medicare & Medicaid Services. 2010 enrollment information. Published 2010. Accessed November 22, 2010.

14. Kaiser Family Foundation. Low-Income Assistance Under the Medicare Drug Benefit. Fact Sheet. 7327_04.pdf. Published February 2008. Accessed September 8, 2011.

15. Centers for Medicare & Medicaid Services. 2011 RxHCC model software. asp#TopOfPage. Published 2011. Accessed March 12, 2011.

16. Buntin MB, Zaslavsky AM. Too much ado about two-part models and transformation? Comparing methods of modeling Medicare expenditures. J Health Econ. 2004;23(3):525-542.

17. Gold MR, Siegel JE, Russell LB, Weinstein MC, eds. Cost-Effectiveness in Health and Medicine. New York: Oxford University Press; 1996.

18. Calvert NW, Burch SP, Fu AZ, Reeves P, Thompson TR. The costeffectiveness of lamotrigine in the maintenance treatment of adults with bipolar I disorder. J Manag Care Pharm. 2006;12(4):322-330.

19. Davies LM, Barnes TR, Jones PB, et al. A randomized controlled trial of the cost-utility of second-generation antipsychotics in people with psychosis and eligible for clozapine. Value Health. 2008;11(4):549-562.

20. Lave JR, Keane CR, Lin CJ, Ricci EM, Amersbach G, La Vallee CP. The impact of lack of health insurance on children. J Health Soc Policy. 1998;10(2):57-73.

21. Braithwaite RS, Meltzer DO, King JT Jr, Leslie D, Roberts MS. What does the value of modern medicine say about the $50,000 per qualityadjusted life-year decision rule? Med Care. 2008;46(4):349-356.

22. Laupacis A, Feeny D, Detsky AS, Tugwell PX. How attractive does a new technology have to be to warrant adoption and utilization? Tentative guidelines for using clinical and economic evaluations. CMAJ. 1992;146(4):473-481.

23. Ubel PA, Hirth RA, Chernew ME, Fendrick AM. What is the price of life and why doesn’t it increase at the rate of inflation? Arch Intern Med. 2003;163(14):1637-1641.

24. Hoch JS, Briggs AH, Willan AR. Something old, something new, something borrowed, something blue: a framework for the marriage of health econometrics and cost-effectiveness analysis. Health Econ. 2002;11(5):415-430.

25. Hoch JS, Rockx MA, Krahn AD. Using the net benefit regression framework to construct cost-effectiveness acceptability curves: an example using data from a trial of external loop recorders versus Holter monitoring for ambulatory monitoring of “community acquired” syncope. BMC Health Serv Res. 2006;6:68.