The Impact Of Medicare’s Mental Health Cost-Sharing Parity On Use Of Mental Health Care Services

Abstract

Before implementation of cost-sharing parity in Medicare, beneficiaries faced higher cost sharing for mental health services than for other medical services. The Medicare Improvements for Patients and Providers Act of 2008 phased in cost-sharing reductions in Medicare for outpatient mental health services in the period 2010–14. Using data for 2006–15 from the Medical Expenditure Panel Survey and difference-in-differences analyses, we assessed whether this reduction in mental health cost sharing was associated with changes in specialty and primary care outpatient mental care visits and psychotropic medication fills. We compared people with Medicare and those with private insurance before and after parity implementation. Medicare beneficiaries’ use of psychotropic medication increased after the implementation of cost-sharing parity, but we did not detect a change in visits. Changes in the use of psychotropic medications were greater among people with probable serious mental illness and among Medicare beneficiaries who did not report having supplemental coverage. The increased medication use could signal improvements in mental health care access among Medicare beneficiaries, especially among the subgroups most likely to benefit from the policy change.

Improving the affordability and generosity of insurance coverage for mental health care compared to physical health care has been a major policy focus for many years. The Medicare Improvements for Patients and Providers Act (MIPPA) of 2008 eliminated Medicare’s limitation on coverage for outpatient mental health treatment, which had set the coinsurance rate at 50 percent for mental health visits compared with 20 percent for most other services. For beneficiaries without supplemental coverage from Medicaid, the act gradually reduced cost sharing to 45 percent in 2010, 40 percent in 2011 and 2012, 35 percent in 2013, and 20 percent in 2014 and afterward. While addressing potential overuse in mental health service utilization,^1,2 historically higher cost sharing could have hindered access to mental health services for Medicare beneficiaries with mental illness or substance use disorder.

Mental health treatment cost-sharing policies in Medicare are relevant given the high rates of mental illness among Medicare beneficiaries and poor access to mental health care in the US. At least one in five Medicare beneficiaries lives with a mental health disorder: Estimates range from 20 percent to 35 percent.^3,4 In the general US population, only 30–50 percent of people with serious mental illness receive any mental health care services annually.^5,6 Little is known about how mental health care use is affected by variations in cost sharing among Medicare beneficiaries.

Medicare was largely exempted from parity legislation before MIPPA. For example, Medicare was exempted from the Paul Wellstone and Pete Domenici Mental Health Parity and Addiction Equity Act of 2008, which mandated parity in treatment limits between mental health and physical health treatment in group health and Medicaid managed care plans, starting in 2010 for most plans. Likewise, before MIPPA, many states and the Federal Employees Health Benefits Program had enacted policies that promoted parity among commercially insured populations that did not apply to Medicare beneficiaries. Those parity policies were associated with decreased out-of-pocket spending for people with bipolar disorder, major depression, and adjustment disorder but not with any increases in overall mental health care expenditures in the program.⁷ Other studies have found that reductions in out-of-pocket spending led to increased mental health service use.^8,9

Within the Medicare program, the only prior evidence on the potential effects of parity is from Medicare Advantage. Before MIPPA, Medicare Advantage plans’ benefit policies varied in terms of their cost sharing for mental health and physical health visits, but only 20 percent of plans had equivalent cost sharing for mental health and physical health visits.¹⁰ Beneficiaries who chose plans with cost-sharing equivalence had greater use of clinically appropriate mental health services following a psychiatric hospitalization, compared to those who chose plans with greater cost sharing for mental than for physical health.¹⁰ No prior study has evaluated the association between parity policies and use of mental health care among traditional fee-for-service Medicare beneficiaries. The effects of mental health parity within a fee-for-service population could differ from those within a population covered by managed care plans, since fee-for-service beneficiaries are generally not subject to utilization management by insurance plans, which can restrict use.¹¹ On the other hand, many beneficiaries have supplemental coverage—through employer-sponsored, Medigap, or Medicare Advantage plans—that provide partial or full coverage of coinsurance. Thus, the impact of the policy change on out-of-pocket spending could be more muted than the statutory change from 50 percent to 20 percent would suggest.

Using data from the nationally representative Medical Expenditure Panel Survey (MEPS), we evaluated the association between the implementation of MIPPA and changes in use of mental health care for Medicare beneficiaries not dually eligible for Medicaid, comparing trends over time among them with trends among the privately insured with similar income levels. We hypothesized that the reductions in Medicare cost sharing would lead to larger increases in mental health care utilization for Medicare beneficiaries, particularly those with probable serious mental illness.

Study Data And Methods

Data

We used data for 2006–15 from MEPS, which provides in-depth, comprehensive information on health care use and expenditures—including details on outpatient and office-based visits and prescription drug use—for a nationally representative sample of the noninstitutionalized US population. Respondents provide information about their sociodemographic and clinical characteristics, health care use, and expenditures in the MEPS–Household Component. Respondents’ medical care provider and pharmacy records in the MEPS–Medical Provider Component provide supplementary information about health care utilization and expenditures.

To identify the association between changes in coinsurance under MIPPA and changes in rates of mental health treatment, we compared adult Medicare beneficiaries with household incomes greater than 200 percent of the federal poverty level (hereafter, “Medicare beneficiaries”) to privately insured adults with household incomes greater than 200 percent of poverty (hereafter, “the privately insured”). For the comparison group, we focused on people with employer-sponsored coverage, excluding people covered by nongroup insurance plans to minimize confounding associated with the increases in nongroup insurance coverage and generosity due to the implementation of major provisions of the Affordable Care Act (ACA) in 2014.

We focused on Medicare beneficiaries with incomes above 200 percent of poverty because lower-income beneficiaries could be eligible for Medicaid coverage of Medicare cost sharing (for example, through the Qualified Medicare Beneficiary program) and therefore would not have experienced any change in mental health care cost sharing associated with MIPPA. In secondary analyses, we further classified Medicare beneficiaries as having Medicare coverage only or Medicare plus supplemental coverage—which could include individual Medigap, Medicare Advantage, or employer plans.

We limited our primary analytic sample to people with probable serious mental illness, defined as having heightened depressive symptoms (a score of at least 3 on the Patient Health Questionnaire [PHQ]-2) or severe psychological distress (a score of at least 13 on the Kessler K6 Psychological Distress Scale). The K6 scale has a sensitivity of 90 percent and specificity of 89 percent for mental illness,¹² as defined by the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, and it can distinguish cases from noncases with consistency across sociodemographic subsamples.¹³ The PHQ-2¹⁴ scale for detecting depression demonstrates strong sensitivity (87 percent) and specificity (78 percent) for major depressive disorder among an ethnically diverse outpatient population.¹⁵

We examined whether the variation in coinsurance was associated with changes in household-reported mental health care treatment during the past year. We classified mental health care treatment into four categories: any mental health treatment (outpatient visit or psychotropic medication fill), any outpatient mental health visit (including visits with specialty mental health providers or primary care providers), any specialty mental health visit (with a psychiatrist, psychologist, counselor, or social worker), and any psychotropic medication fill. Mental health visits included visits for a disorder covered by International Classification of Diseases, Ninth Revision (ICD-9), codes 291, 292, or 295–314.¹⁶ These diagnoses were assigned based on MEPS respondent reports of the reason or reasons for a visit. Another method of identifying specialty mental health visits was if the respondent reported the visit to be for “psychotherapy or mental health counseling.” This methodology has been shown to have high sensitivity (88 percent) to provider reports of treatment for mental health disorders.¹⁷ We identified fills of psychotropic medication based on the Multum Medisource Lexicon drug classification system.¹⁸

We adjusted for MEPS respondents’ sex, age, race/ethnicity (using Census Bureau definitions for individuals that identify as non-Hispanic white, non-Hispanic black, Hispanic, and non-Hispanic Asian), self-reported mental and physical health (poor or fair versus good, very good, or excellent), and education (less than high school graduate, high school graduate, any college, or college graduate). Because of the high rates of comorbidity between physical ailments and mental disorders, we also included the presence of chronic illnesses (none, one, or two or more) from a list of eleven conditions that all MEPS respondents were asked about.^19,20 Finally, we adjusted for K6 and PHQ-2 scores.

Analytic Strategy

We first assessed unadjusted trends in the use of any mental health treatment, any specialty mental health visits, and psychotropic medication among people with probable serious mental illness from 2006 to 2015, comparing Medicare beneficiaries to the privately insured. Next, we estimated multivariable linear probability models of the association between insurance category, time, and mental health care treatment, adjusting for clinical need and sociodemographic characteristics. To identify the association of the change in policy with the mental health treatment outcome, we estimated difference-in-differences models that compared pre- to post policy changes (2006–09 versus 2010–15) in mental health treatment between the Medicare beneficiaries and privately insured (for a further description of these models, see the online appendix).²¹

We conducted two sets of secondary analyses. First, we reestimated models among the general population, which included both people with and without probable serious mental illness. Second, to identify the differential impact of cost-sharing parity on Medicare beneficiaries with supplemental private coverage, we reestimated models that compared Medicare beneficiaries who had additional private coverage (n = 15,721) with Medicare-only beneficiaries (n = 4,849) and the privately insured (n = 78,280).

We conducted sensitivity analyses to assess robustness. First, because parity was phased in gradually, we assessed the year-by-year impact of cost-sharing parity on mental health care utilization, reestimating models and comparing 2009 with each pre- and postpolicy year (2006–08 and 2011–15, with the 2010 phase-in year removed). Second, we reestimated models on people with at least a mild need for mental health treatment (widening inclusion criteria to people with a PHQ-2 score of 2 or more or a K6 score of 5 or more). Third, to ensure that our findings concerning psychotropic medication fills were not driven by Medicare Part D’s expanded coverage of barbiturates and benzodiazepines in 2013, we reestimated our regression models for any psychotropic medication fill, excluding these classes of drugs. Fourth, to determine if there were changes in the mean number of mental health care visits per patient associated with the cost-sharing reduction, we estimated linear regression models that examined changes in the annual number of such visits. Fifth, to address the concern that MEPS underestimates visits that address mental health, we conducted a sensitivity analysis that assessed the impact of cost-sharing parity on primary care provider visits with psychotropic medication management—a common but likely underreported type of mental health visit. A primary care provider visit with psychotropic medication management was operationalized as any visit to a primary care provider (a family, general, or internal medicine provider) with or without a linked mental health diagnosis, accompanied by at least one psychotropic medication fill, and no specialty mental health provider visits in the year of study.

Limitations

Our study had several limitations. First, it was limited to Medicare and private insurance beneficiaries with family incomes greater than 200 percent of poverty in the noninstitutionalized population of the United States—excluding incarcerated people, those in nursing homes, and those in residential treatment.

Second, it is possible that MEPS underestimates visits that address mental health concerns, given that it relies on household reports. Respondents could underreport mental health problems due to stigma or privacy concerns.²² They might also be more likely to recall or report other nonmental health reasons for visits, particularly if a visit covered a range of medical concerns. We conducted sensitivity analyses (described above) to address this limitation.

Finally, other influential policies or events could have affected the mental service use of the privately insured at the same time as MIPPA was enacted, thus confounding our estimate of the impact of the cost-sharing policy. One such potential confounder is the Mental Health Parity and Addiction Equity Act, but prior studies have demonstrated that it had little or no impact on mental health service use^7,23 and that much of the effects of parity had already been realized through the implementation of state parity laws.²⁴ We leveraged this finding in the current study by using the sample with private group insurance as the comparison group. Increased managed care penetration and other changes in private insurance markets also could have confounded the comparison with the privately insured. However, we found that the privately insured faced only a small reduction in the percentage of specialty mental health care paid out of pocket (from 35 percent in 2008–09 to 31 percent in 2011–15, according to MEPS data), so the evidence linking managed care penetration to mental health services use demonstrated little, if any, change.^25,26

Study Results

Exhibit 1 displays the characteristics of our sample population.

Exhibit 1:

Characteristics of Medicare beneficiaries and adults with private employer-sponsored health insurance, 2006–15

	Medicare	Private
Sample size	20,570	78,280
Dependent variables (%)
Any mental health treatment	27.92**	17.28
Any specialty mental health visit	4.18**	4.72
Any outpatient mental health visit	8.19	7.64
Any psychotropic medication fill	26.17**	14.61
Sex (ref: male) (%)
Female	51.60**	50.54
Age in years (ref: less than 64) (%)
65 or more	93.30**	35.77
Race/ethnicity (ref: non-Hispanic white) (%)
Non-Hispanic Black	5.98**	9.13
Hispanic	4.29**	9.45
Non-Hispanic Asian	4.08**	2.87
Education (ref: less than high school graduate) (%)
High school graduate	32.75**	23.33
Any college	23.90**	27.77
College graduate	31.22**	42.28
Mental health (mean score)
K6	3.05**	2.57
PHQ-2	0.65**	0.46
Mental health (ref: excellent, very good, or good) (%)
Fair or poor	9.17**	3.43
Physical health (ref: excellent, very good, or good) (%)
Fair or poor	18.27**	6.82
Chronic health conditions (mean number)	2.88**	0.99

SOURCE Authors’ analysis of data for 2006–15 from the Medical Expenditure Panel Survey. NOTES The sample population was restricted to people with incomes at or above 200 percent of the federal poverty level. Rates and means were weighted to be representative of the noninstitutionalized US population. Each reference group is the complement to the sum of the subcategories in each group with multiple subcategories, so the value for each reference group is 100 minus the sum of the values of the other categories, expressed as a percentage. K6 is the Kessler K6 Psychological Distress Scale. PHQ-2 is the Patient Health Questionnaire–2.

^**p < 0.05

Use Of Mental Health Care

The proportion of people who received any mental health treatment (visits or medications) increased from 48.1 percent in 2006 to 54.1 percent in 2009 and 56.9 percent in 2015 among Medicare beneficiaries, compared to an increase among the privately insured from 44.0 percent in 2006 to 49.5 percent in 2009 and 47.3 percent in 2015 (exhibit 2). Among Medicare beneficiaries, the proportion receiving any specialty mental health visits changed from 15.9 percent in 2006 to 13.4 percent in 2009 and 16.7 percent in 2015, compared to changes from 16.8 percent in 2006 to 19.2 percent in 2009 and 21.0 percent in 2015 among the privately insured (exhibit 3). Rates of any psychotropic medication fills in the past year for Medicare beneficiaries increased from 44.4 percent in 2006 to 46.1 percent in 2009 and 54.7 percent in 2015, compared to changes from 38.2 percent in 2006 to 44.8 percent and 2009 to 40.8 percent in 2015 among the privately insured (exhibit 3).

Exhibit 2.

Percentages of Medicare beneficiaries and adults with private employer-sponsored health insurance with any mental health service use, 2006–15

SOURCE Authors’ analysis of data for 2006–15 from the Medical Expenditure Panel Survey. NOTE The sample population was restricted to people with incomes at or above 200 percent of the federal poverty level.

Exhibit 3.

Percentages of Medicare beneficiaries and adults with private employer-sponsored health insurance with any use of specialty mental health service or psychotropic medication, 2006–15

SOURCE Authors’ analysis of data for 2006–15 from the Medical Expenditure Panel Survey. NOTE The sample population was restricted to people with incomes at or above 200 percent of the federal poverty level.

In multivariate analyses we found that, when compared with the privately insured, Medicare beneficiaries with probable serious mental illness experienced a 6.0-percentage-point increase from before to after implementation of cost-sharing parity in having any mental health treatment (visits or medications) and an 8.2-percentage-point increase in having any psychotropic medication fills in the past year (exhibit 4). (For the full regression model output, see appendix table 1).²¹ There were no significant changes over time in the use of any outpatient mental health visits (including primary and specialty care) or any specialty mental health visits for Medicare beneficiaries versus the privately insured.

Exhibit 4:

Adjusted difference-in-differences estimates of percentage-point changes from 2006–09 to 2011–15 for Medicare beneficiaries with mental health treatment, compared to adults with private employer-sponsored health insurance

	Any mental health treatment	Any outpatient mental health visit	Any specialty mental health visit	Any psychotropic medication fill
Beneficiaries with probable SMI	6.03**	−0.54	−3.12	8.16***
All beneficiaries	2.43**	−0.24	−0.35	2.53**
Beneficiaries with probable SMI, by supplemental coverage status
Without supplemental coverage	9.34**	−1.99	−3.57	10.99**
With supplemental coverage	3.61	−0.31	−2.97	6.03
All beneficiaries, by supplemental coverage status
Without supplemental coverage	2.31	0.06	−0.29	2.02
With supplemental coverage	2.31**	−0.33	−0.33	2.52**

SOURCE Authors’ analysis of data for 2006–15 from the Medical Expenditure Panel Survey. NOTES The implementation of mental health treatment cost sharing parity in Medicare per MIPPA occurred in 2010. Significance is after correction for multiple comparisons. SMI is serious mental illness.

^**p < 0.05

^***p < 0.01

Secondary Analyses

When we expanded our sample to the general population (including people without as well as those with probable serious mental illness), we found that Medicare beneficiaries had a significant increase in having any mental health care visit after implementation of the MIPPA, relative to the privately insured: a change of 2.4 percentage points after adjustment for other covariates. Results were similar for filling any psychotropic medication (exhibit 4). (For the full regression model output, see appendix table 2).²¹ There were no significant differences among this sample in the use of any outpatient mental health visits or any specialty mental health visits.

In a secondary analysis among people with probable serious mental illness that compared Medicare beneficiaries with supplemental coverage and those with Medicare only to the privately insured, having any mental health treatment and filling any psychotropic medication increased 9.3 percentage points and 11.0 percentage points, respectively, for Medicare-only beneficiaries (exhibit 4). (For the full regression model output, see appendix exhibit 3).²¹ There were no significant changes in the use of any specialty mental health treatment or outpatient mental health visits in either Medicare group. When we ran the same analysis among the general population, we found that the results trended in a similar direction (exhibit 4). (For the full regression model output, see appendix exhibit 4).²¹

In a sensitivity analysis that evaluated year-by-year changes, we found that—similar to our main findings—mental health cost-sharing parity was associated with a significant yearly increase in any psychotropic medication fill in 2012, 2013, and 2015 in the population with probable serious mental illness and in 2012–15 in the general population; p < 0.05) (appendix exhibits 5a and 5b).²¹ In another sensitivity analysis, we used an expanded definition of mental illness that included people with mild distress or depressive symptoms. We found that Medicare beneficiaries had a significant increase in any psychotropic medication fill, mirroring one of our main findings (appendix exhibit 6).²¹ In another sensitivity analysis, which excluded prescription fills for barbiturates and benzodiazepines from our dependent variable—any psychotropic medical fill—the results did not deviate from our main findings (appendix exhibit 7).²¹ We also assessed the impact of cost-sharing parity on the number of mental health visits and found no significant differences between Medicare beneficiaries and the privately insured in any of the outcomes (appendix exhibit 8).²¹ Lastly, a sensitivity analysis that addressed potential underreporting of mental health visits found that parity was associated with a 14.0 percent increase in the group with probable serious mental illness and with a 6.8 percent increase in the general population in having a primary care provider visit with psychotropic medication management (appendix exhibits 9a and 9b, respectively).²¹

Discussion

MIPPA reduced cost sharing substantially for outpatient mental health services for Medicare beneficiaries. We found that the reduction from 50 percent to 20 percent coinsurance was associated with significant increases in the overall use of mental health treatment among Medicare beneficiaries compared with the privately insured, primarily through increases in the use of psychotropic medication. We detected no measurable change in the use of outpatient visits for mental health services.

Although the parity policy targeted cost sharing for outpatient mental health services, we did not detect a change in the use of primary care or specialty mental health visits. In contrast, the parity policy did not apply directly to medications, where we did find an effect. Arguably, the intent of the policy was to increase access to treatment, which was realized in terms of increases in psychotropic medication use.

A strength of this study’s data source is that MEPS measures need with validated measures of psychological symptoms and distress, an improvement over claims or encounter data in which diagnoses are likely to be endogenous to use. Our findings hold among the population with probable serious mental illness and the general population, but we found larger effects among the population with probable serious mental illness.

Another strength of the MEPS data is the ability it gave us to identify Medicare beneficiaries with supplemental coverage, such as through Medigap or employer supplements. When we limited the sample to those with unsupplemented fee-for-service Medicare benefits (people with Medicare only), we also found positive effects of parity on overall mental health treatment use, driven primarily by increases in use of psychotropic medications. Effects were larger in this subsample of beneficiaries, compared with people with Medicare and supplemental private insurance—which is consistent with the hypothesis that parity implementation would have greater impact on out-of-pocket spending for the group with unsupplemented benefits.

Our findings are in contrast to those of the RAND health insurance experiment, which found that random assignment to health plans with reduced cost-sharing levels was associated with a significant increase in mental health visits.¹ The timing of the RAND study differed substantially from that of the current study, with the latter set in a period of greater reliance on medications in mental health care. Our findings do correspond to those of more recent studies: One found that parity in the Federal Employees Health Benefits Program resulted in only a minimal effect on mental health visits and costs;²⁷ and another found that in a national sample, cost-sharing reductions had only minimal effects on outpatient mental health visits.⁹

While there was no effect of Medicare cost-sharing reductions on outpatient mental health visits, this policy did have an effect on increased psychotropic medication use. This is counterintuitive, given that MIPPA focused on coinsurance for mental health office visits and not on medication copayments. One possible explanation is that cost-sharing reductions increased visits with primary care providers in which psychotropic medications were prescribed, but these visits were not recalled by the respondent as being related to mental health. This could have happened if a person who received routine treatment for physical health conditions (such as high cholesterol, diabetes, and arthritis) also received a prescription for a psychotropic medication (for example, an antidepressant). There is some evidence to support this explanation in our study: When we relaxed the requirement for the visit to be associated with a mental health diagnosis code, we found a positive effect of parity on having a primary care provider visit with psychotropic medication management.

This increase in psychotropic medication use among Medicare beneficiaries without a simultaneous increase in outpatient mental health visits is in accordance with other studies of practice patterns that have found an increase in the number of prescriptions provided to patients without a psychiatric diagnosis or receipt of specialty mental health care. For example, in an analysis of MarketScan claims data, Ilse Wiechers and colleagues found that 69 percent of people ages 50–64 years who were prescribed a psychotropic medication did not have a visit with a psychiatric diagnosis within the past year.²⁸ Similarly, Ramin Mojtabai and Mark Olfson found an increase in the percentage of people without a diagnosed mental disorder who received antidepressant medication: a change from 59 percent in 1996 to 72 percent in 2007.²⁹ Our findings of increased overall psychotropic medication use among Medicare beneficiaries mirror accelerated trends of increased psychotropic medication use in the US more generally.³⁰ The increase that we found persisted even after we excluded benzodiazepines and barbiturates—psychotropic medications that were covered by Medicare Part D beginning in January 1, 2013.³¹ In light of the increase in drug use without a concomitant change in visits, further work is needed to assess whether there is adequate outpatient monitoring of individuals’ psychotropic medication use.

Our findings that parity did not increase mental health visits also suggest that, even in fee-for-service environments, where utilization management is less common than in managed care environments, there may be other, nonfinancial barriers to receiving mental health treatment (for example, barriers related to transportation, the availability of providers, or community- or person-level stigma).^32,33 There could also have been crowd-out of mental health care for Medicare beneficiaries associated with increases in demand for care by people newly insured under the ACA’s expansion of eligibility for Medicaid and individual insurance Marketplaces. This could be of particular concern for Medicare beneficiaries who seek specialty mental health services because the number of specialty mental health providers is dwindling,³⁴ and few such providers accept patients with public insurance.³⁵

Conclusion

We found that the MIPPA provision that reduced mental health treatment cost sharing led to an increase in the overall use of mental health treatment and psychotropic medication use for Medicare beneficiaries with incomes greater than 200 percent of poverty, compared to their privately insured counterparts. This adds to the evidence about the impact of cost sharing on mental health services. Future studies are warranted to assess the effects of cost-sharing policies on more detailed measures of treatment quality and outcomes.