RETIREMENT RESEARCH February 2018, Number 18-3 HOW DO PRESCRIPTION DRUGS AFFECT THE USE OF OTHER HEALTH SERVICES? By Gal Wettstein* Introduction Medicare Part D and Health Over the past decade, the availability of prescrip- Care Usage tion drugs has increased, particularly for the elderly. Medicare Part D expanded coverage to include pre- Medicare has provided health insurance to Americans scription drugs, and the Affordable Care Act (ACA) ages 65 and over since 1966. However, in its first 40 enhanced Part D’s coverage. While lowering the cost years, the program did not generally provide prescrip- of prescription drugs would obviously encourage more tion drug insurance.1 The Medicare Prescription use of medications, the implications of such changes Drug, Improvement and Modernization Act of 2003, for the rest of the health care market are less clear. which became effective in 2006, expanded Medicare to The answer depends on whether drugs are “substi- cover prescription drugs through the Part D program. tutes” for other care or “complements.” Drugs could In practice, only a quarter of those gaining Part D be substitutes if they prevent deterioration in health coverage acquired new drug insurance; three quar- conditions that would otherwise require more inten- ters used Part D to replace drug coverage from other sive care, such as surgery. However, in many ways, sources.2 Individuals simply replacing drug coverage drugs may be complements to other care, adding were likely not affected much by Part D, so they are value to other tools in the clinician’s toolbox. To ex- excluded from this analysis. For those who did gain plore this issue, this brief examines the use of health coverage, Part D’s generosity will increase further in services before and after the introduction of Part D. the near future, as the ACA reduces costs for those The discussion proceeds as follows. The first who spend more than a moderate amount on pre- section provides background on Part D and summa- scriptions.3 rizes previous reesearch on how it affects the use of It is already well-established that drug insurance other health services. The second section explains the in general, and Part D specifically, leads to greater use data and methodology used in this study. The third of prescription drugs.4 However, with respect to drug section shows the main results. The final section coverage’s impact on other health services, the evi- concludes that broadening the availability of drugs dence has been mixed; some estimates show drug in- increases the use of office-based health care, with a surance increased use while others show a decline or possible decline in the use of inpatient facilities. no effect.5 This analysis expands upon these previous studies of Part D by considering a broader population * Gal Wettstein is a research economist with the Center for Retirement Research at Boston College. 2 Center for Retirement Research and explicitly separating hospital use from office vis- To answer the question of how drugs interact with its, recognizing that these services may interact with other health care, the analysis compares the change in increased drug availability in different ways. health care use at age 65 before and after 2006. The change at age 65 before 2006 captures the effect of aging another year, as well as eligibility for Medicare Data and Methods Parts A and B. The change after 2006 includes all those same effects, plus the effect of acquiring drug The data are from the Medical Expenditure Panel coverage through Part D. The difference in those Survey (MEPS). The MEPS provides information on changes isolates the effect of Part D coverage on us- health care use, health insurance, health status, and age of the health services of interest. demographic characteristics for the U.S. civilian, In addition to the simple comparison of the noninstitutionalized population. The focus here is on change in usage at 65 before and after 2006, the analy- individuals ages 60-70.6 sis also includes a regression to control for personal As mentioned above, some individuals are exclud- characteristics that could affect medical spending. ed from the analysis because they would not stand The equation is: to gain new drug insurance: those with Medicaid, Tricare, or Veterans Administration (VA) coverage are Health care expenditures = ƒ (Over65 x Post2006, age, excluded because they would have had drug insurance year, personal characteristics). before Part D. Also excluded are those below age 65 who had Medicare coverage (likely due to disability). The same equation is estimated for three kinds of Since private insurance almost always covers prescrip- health expenditures: drugs, office visits, and inpatient tion drugs, individuals who had private coverage were facilities.10 The main independent variable in the excluded as well.7 Figure 1 shows that only about 23 regression identifies whether the individual was over percent of the original full sample could have been 65 and observed post 2006. The coefficient on this affected by Part D.8 variable simply compares the change in the depen- dent variable at age 65, before and after 2006. If this coefficient is positive, it would mean increased drug Figure 1. Percentage of Sample Ages 60-70 by coverage led to increased spending on the category Insurance Coverage being considered. The control variables are age, year, gender, marital In this analysis, status, education, and health status, including self- 23% reported health, and the presence of specific health conditions.11 To compare the effects of Part D overall to the effects only on those with poorer health, the regression is also estimated separately for those with Medicare pre-65, below average health.12 6% Medicaid, Results Privately insured, 10% 57% Military, Before turning to the regression results, it may be 4% useful to first look at whether the raw data indicate any potential connection between Part D and spend- Note: See endnote 9. ing on the various health services. Consistent with Source: Author’s estimates from Agency for Healthcare the literature, the analysis finds an increase in total Research and Quality, Medical Expenditure Panel Survey drug expenditures after age 65 once Part D was (MEPS) (2000-2005 and 2007-2009). implemented. Figure 2a on the next page depicts a similar increase in spending on office-based visits. Issue in Brief 3 Figure 2. Annual Spending per Person on Health Services by Age, Pre- and Post-Part D, 2009 Dollars a. Office-Based Visits b. Inpatient Facilities $2,000 $2,000 Pre-Part D Pre-Part D Post-Part D Post-Part D $1,500 $1,500 $1,000 $1,000 $500 $500 $0 $0 59 61 63 65 67 69 59 61 63 65 67 69 Age Age Note: To reduce sampling error, ages represent two-year age groups. Source: Author’s estimates from the 2000-2005 and 2007-2009 MEPS. In contrast, Figure 2b shows a small, but less notice- cal significance be easily assessed. These limitations able, reduction in total annual spending on inpatient are addressed in the regression analysis, with the facilities. Drug expenditures increased after age 65 by main results displayed in Figure 3. $486 more after 2006 relative to before 2006; office- The regression results indicate that those who based visits increased by $407; and inpatient facility gained drug coverage through Part D significantly expenditures declined by $127.  increased their expenditures on both drugs and While the figures clearly suggest that Part D may office-based services, particularly for those whose have had an impact on other health services, the data self-reported health is below average. This result is are merely descriptive. The displayed amounts do not consistent with the idea that drugs complement phy- control for individual characteristics, nor can statisti- sicians’ office visits by making them more valuable to Figure 3. Estimated Effect of Part D on Health Spending, Overall and for Individuals in Below Average Health, 2009 Dollars 432.4 Drug expenditures 654.9 443.9 Office-based visit expeditures 730.7 -387.8 Overall Inpatient facility expenditures -207.8 Below average health -$500 $0 $500 $1,000 Note: Solid bars indicate statistically significant results. Full results are available in Appendix Tables 1 and 2. Source: Author’s estimates from the 2000-2005 and 2007-2009 MEPS. 4 Center for Retirement Research individuals who can now afford their prescriptions.13 Endnotes Meanwhile, inpatient facility use appears to have de- clined, although the reductions did not reach statisti- 1 Medicare did cover inpatient drugs through Medi- cal significance. Such reductions are consistent with care Part A. Furthermore, Medigap and HMO plans the notion that better access to drugs substitutes for covering drugs existed, but provided limited insur- inpatient care by reducing severe health conditions. ance for high premiums and were chosen by only a small minority of those eligible. Conclusion 2 Engelhardt and Gruber (2011). The availability of drugs to the elderly is expected to 3 Part D’s standard benefit in 2006 was a $250 increase over the coming years, as the coverage from deductible, followed by spending up to $2,250 in Part D becomes more robust with its expansion under which individuals were responsible for 25 percent the ACA. Will this reduce spending on other health of costs. After that, there was a coverage gap (the care, or will it make that health care even more valu- “donut hole”): beneficiaries were responsible for all able, increasing expenditures as the toolkit available drug costs between $2,250 and $5,100. The ACA will to clinical practitioners expands? gradually eliminate this donut hole by 2020. Above The analysis above provides a nuanced answer: it the $5,100 threshold, beneficiaries received “cata- depends on the type of health care service. The analy- strophic” coverage, where they paid 5 percent of every sis shows that expanded access to drugs increased the additional dollar of spending. The cutoffs of these elderly’s use of office-based services, suggesting drugs ranges are adjusted annually based on average drug and these other services are complements. On the spending per beneficiary. inpatient expenditure side, the estimates are consis- tent with prescription drugs substituting for inpatient 4 For example, see Engelhardt and Gruber (2011). care, making the use of these services less necessary as people get treated before their conditions become 5 For example, see Gaynor, Li, and Vogt (2007); Gold- more severe. However, this result is ultimately incon- man, Joyce, and Zheng (2007); Zhang et al. (2009); clusive, perhaps because 2009 – the last year in the Liu et al. (2011); and Kaestner and Khan (2012). analyzed sample – could be too soon to expect Part D to have a large impact on such outcomes. Future 6 The data span the years 2000-2009. Data from 2010 research should continue to focus on this issue. onward are not examined due to changes in health insurance markets with the implementation of the Affordable Care Act. Data from 2006 are not included as it was the transition year in which Part D was introduced. 7 This approach relies on the assumption that individuals who chose to drop private insurance as a result of Part D’s introduction would not have had dif- ferent health care usage patterns than those who were not insured to begin with. Previous work suggests that sicker individuals may leave employment, and correspondingly private coverage, disproportionately due to Part D’s availability (Wettstein 2017). To ac- count for this possibility, the regressions in this analy- sis control for self-reported health and a variety of specific health conditions. Excluding these controls leaves the results virtually unchanged; thus selection on health due to the exclusion of individuals with private insurance does not seem to drive the results. Unreported results show that self-reported health and the incidence of specific diagnoses does not change within the group lacking private coverage due to Part D eligibility. Issue in Brief 5 8 The 23 percent included in the sample are con- References sistent with the estimate in Engelhardt and Gruber (2011) that three quarters of those gaining Part D Agency for Healthcare Research and Quality, Medical coverage used it to replace coverage from other Expenditure Panel Surveys. 2000-2005 and 2007- sources. Individuals with missing data on certain 2009. Rockville, MD. variables used in the analysis are also excluded, leaving a sample of 5,546 observations. The MEPS Engelhardt, Gary V. and Jonathan Gruber. 2011. samples individuals five times over two years. In this “Medicare Part D and the Financial Protection of brief, the survey is treated as a repeated cross-section, the Elderly.” American Economic Journal: Economic aggregated within calendar year. Standard errors are Policy 3(4): 77-102. adjusted to reflect that the sample includes only 3,666 unique individuals. Gaynor, Martin, Jian Li, and William B. Vogt. 2007. “Substitution, Spending Offsets, and Prescription 9 Individuals are assigned a source of health insur- Drug Benefit Design.” Forum for Health Economics ance coverage hierarchically in the following order: and Policy 10(2): 1-33. 1) Medicare coverage before age 65; 2) Medicaid; 3) military coverage, including Tricare, CHAMPUS, and Goldman, Dana P., Geoffrey F. Joyce, and Yuhui VA; and 4) private health insurance. If none of these Zheng. 2007. “Prescription Drug Cost Sharing: sources of coverage is listed, the individual is includ- Associations with Medication and Medical Utiliza- ed in the analysis. tion and Spending and Health.” Journal of the American Medical Association 298(1): 61-69. 10 All dollar amounts are inflated by the Consumer Price Index to 2009 dollars. Kaestner, Robert and Nasreen Khan. 2012. “Medicare Part D and Its Effect on the Use of Prescription 11 The conditions are diabetes, asthma, high blood- Drugs and Use of Other Health Care Services of pressure, coronary heart disease, angina, heart attack, the Elderly.” Journal of Policy Analysis and Manage- other heart disease, stroke, and emphysema. ment 31(2): 253-279. 12 Health is reported on a 5-point scale from 1 Liu, Frank Xiaoqing, G. Caleb Alexander, Stephanie Y. (excellent) to 5 (poor). In this analysis, below average Crawford, A. Simon Pickard, Donald Hedeker, and health is defined as a 3 or more. Surrey M. Walton. 2011. “The Impact of Medicare Part D on Out-of-Pocket Costs for Prescription 13 This finding reflects a number of potential mecha- Drugs, Medication Utilization, Health Resource nisms: first, one generally needs to see a doctor to get Utilization, and Preference-Based Health Utility.” a prescription, so increasing prescription drug use Health Services Research 46(4): 1104-1123. mechanically increases office visits. Second, once one is taking a medication, one might need to see a Wettstein, Gal. 2017. “Does Public Health Insurance doctor for follow-up. Third, some conditions can only Affect How Long People Work?” Issue in Brief be treated with medication, and individuals might 17-12. Chestnut Hill, MA: Center for Retirement forgo doctor visits if they know they cannot afford that Research at Boston College. medication. Fourth, increasing use of medication may lead to adverse drug effects that require further Zhang, Yuting, Julie M. Donohue, Judith R. Lave, care. Finally, once individuals see a doctor for any of Gerald O’Donnell, and Joseph P. Newhouse. 2009. the preceding reasons, they could be diagnosed with “The Effect of Medicare Part D on Drug and Medi- previously undetected conditions that require addi- cal Spending.” The New England Journal of Medi- tional treatment. cine 361: 52-61. APPENDIX Issue in Brief 7 Appendix Table 1. Estimated Effect of Part D on Health Spending, Full Sample, 2009 Dollars Variables Drug exp. Office visit exp. Inpatient exp. Over65xPost2006 432.4*** 443.9** -387.8 (158.5) (188.5) (648.6) Woman 162.8** 288.9*** -83.32 (80.29) (96.95) (264.1) Marital status Widowed 78.46 -23.38 -515.2** (121.6) (134.1) (250.6) Divorced -178.7* -129.5 555.8 (92.41) (108.0) (469.4) Separated -152.7 -131.0 -791.2** (230.3) (242.7) (334.4) Never married -323.4* 238.2 -599.7* (181.5) (328.0) (357.9) Health controls Yes Yes Yes Constant 2,450*** -361.0 6,073** (530.6) (655.6) (2,384) Year fixed effects Yes Yes Yes Age fixed effects Yes Yes Yes Education fixed effects Yes Yes Yes Observations 5,546 5,546 5,546 R-squared 0.226 0.098 0.069 Notes: Robust standard errors in parentheses. Statistically significant at 10-percent (*), 5-percent (**), or 1-percent level (***). Source: Author’s estimates from the 2000-2005 and 2007-2009 MEPS. 8 Center for Retirement Research Appendix Table 2. Estimated Effect of Part D on Health Spending, for Those with Below Average Health, 2009 Dollars Variables Drug exp. Office visit exp. Inpatient exp. Over65xPost2006 654.9*** 730.7*** -207.8 (224.4) (256.7) (889.2) Woman 115.9 208.8 -285.1 (112.8) (130.2) (389.3) Marital status Widowed 178.0 15.36 -580.0* (160.0) (176.3) (315.3) Divorced -207.6 -199.2 845.7 (138.4) (147.0) (681.9) Separated -357.0 -358.5 -1,565*** (321.2) (333.1) (363.5) Never married -428.7** 165.1 -1,079*** (197.8) (392.4) (372.1) Constant -316.2 -462.8 -661.7 (271.2) (303.4) (973.6) Year fixed effects Yes Yes Yes Age fixed effects Yes Yes Yes Education fixed effects Yes Yes Yes Observations 4,361 4,361 4,361 R-squared 0.057 0.046 0.018 Notes: Robust standard errors in parentheses. Statistically significant at 10-percent (*), 5-percent (**), or 1-percent level (***). Source: Author’s estimates from the 2000-2005 and 2007-2009 MEPS. RETIREMENT RESEARCH About the Center Affiliated Institutions The mission of the Center for Retirement Research The Brookings Institution at Boston College is to produce first-class research Syracuse University and educational tools and forge a strong link between Urban Institute the academic community and decision-makers in the public and private sectors around an issue of criti- cal importance to the nation’s future. To achieve Contact Information Center for Retirement Research this mission, the Center sponsors a wide variety of Boston College research projects, transmits new findings to a broad Hovey House audience, trains new scholars, and broadens access to 140 Commonwealth Avenue valuable data sources. Since its inception in 1998, the Chestnut Hill, MA 02467-3808 Center has established a reputation as an authorita- Phone: (617) 552-1762 tive source of information on all major aspects of the Fax: (617) 552-0191 retirement income debate. E-mail: crr@bc.edu Website: http://crr.bc.edu The Center for Retirement Research thanks AARP, BlackRock, Inc., The Blackstone Group L.P., The Capital Group Companies, Inc., J.P. Morgan Asset Management, MassMutual Financial Group, Prudential Finan- cial, Sage Advisory Services, Ltd. Co., State Street, and TIAA Institute for support of this project. © 2018, by Trustees of Boston College, Center for Retirement Research. All rights reserved. Short sections of text, not to exceed two paragraphs, may be quoted without explicit permission provided that the authors are identified and full credit, including copyright notice, is given to Trustees of Boston College, Center for Retirement Research. The research reported herein was supported by the Center’s Partnership Program. The findings and conclusions expressed are solely those of the author and do not represent the views or policy of the partners, Boston College, or the Center for Retirement Research.