Health Care Spending, Use, and Financial Hardship Among Traditional Medicare and Medicare Advantage Enrollees With Mental Health Symptoms

Abstract

Objective:

We examined the differences in health care spending and utilization, and financial hardship between Traditional Medicare (TM) and Medicare Advantage (MA) enrollees with mental health symptoms.

Design:

Cross-sectional study.

Participants:

We identified Medicare beneficiaries with mental health symptoms using the Patient Health Questionnaire-2 and the Kessler-6 Psychological Distress Scale in the 2015–2021 Medical Expenditure Panel Survey.

Measurements:

Outcomes included health care spending and utilization (both general and mental health services), and financial hardship. The primary independent variable was MA enrollment.

Results:

MA enrollees with mental health symptoms were 2.3 percentage points (95% CI: −3.4, −1.2; relative difference: 16.1%) less likely to have specialty mental health visits than TM enrollees with mental health symptoms. There were no significant differences in total health care spending, but annual out-of-pocket spending was $292 (95% CI: 152−432; 18.2%) higher among MA enrollees with mental health symptoms than TM enrollees with mental health symptoms. Additionally, MA enrollees with mental health symptoms were 5.0 (95% CI: 2.9−7.2; 22.3%) and 2.5 percentage points (95% CI: 0.8−4.2; 20.9%) more likely to have difficulty paying medical bills over time and to experience high financial burden than TM enrollees with mental health symptoms.

Conclusion:

Our findings suggest that MA enrollees with mental health symptoms were more likely to experience limited access to mental health services and high financial hardship compared to TM enrollees with mental health symptoms. There is a need to develop policies aimed at improving access to mental health services while reducing financial burden for MA enrollees.

INTRODUCTION

Mental health symptoms are a pervasive and costly challenge for Medicare beneficiaries.¹ Prior research finds that 22.7% have a serious mental illness, such as bipolar disorder, schizophrenia, or major depressive disorder, and 7.5% have other common mental health disorders, such as anxiety disorders, personality disorders, and posttraumatic stress disorder.² Moreover, Medicare beneficiaries with mental health symptoms spend significantly more on health care than those without mental health symptoms, with average annual spending on mental health services of $2,024 for those with serious mental illness and $343 for those with other common mental health disorders, compared to $189 for those without mental illness.²

Mental health services have the potential to improve well-being for Medicare beneficiaries with mental health symptoms, but this population often experiences limited access to adequate care and report poor experiences with health care. While near-universal access to Medicare coverage at age 65 improves access to care and health care use overall,^3–7 a recent study found that the use of mental health services decreased among adults with mental health symptoms who transitioned to Medicare.⁸ Furthermore, Medicare beneficiaries with depressive symptoms reported worse experiences with care across all aspects of care.⁹ There are a number of factors that may contribute to these problems. First, Medicare covers both outpatient and inpatient services as well as prescription drugs to treat mental health symptoms, but coverage for mental health services is limited. For example, Medicare does not cover some important mental health services, such as psychiatric rehabilitation, assertive community treatment, or peer support services.¹⁰ Second, there is a shortage of psychiatrists who accept Medicare.¹¹ In 2023, 7.7% of psychiatrists opted out of Medicare, compared to an overall opt-out rate of 1.1%. This is likely due to lower reimbursement rates compared to commercial insurance.¹²

However, little is known about whether the experiences of Medicare beneficiaries with mental health symptoms differ under two different Medicare programs: the federally administered Traditional Medicare (TM) program versus the privately run Medicare Advantage (MA) program. MA is a managed care alternative to TM, a fee-for-service program that reimburses providers for each medical procedure or treatment. Specifically, MA plans are paid on a capitated basis rather than for each service performed, potentially creating the incentive for more efficient care delivery. While prior work has found that MA may improve the efficiency of health care delivery without compromising quality,¹³ there is a concern that capitation may lead plans to underprovide care, particularly for patients with complex conditions like mental illness. A prior study found that MA enrollees with mental illness reported higher dissatisfaction with out–of–pocket expenses for medical care than TM enrollees with mental illness.¹⁴ To the best of our knowledge, however, research on whether MA enrollees with mental health symptoms receive less mental health services and face a higher financial burden is limited.

To address the gap in the literature, we examined the differences in health care spending and utilization (both general and mental health services), and financial hardship between TM and MA enrollees with mental health symptoms.

METHODS

Data

We employed a multiple-year, cross-sectional study design using data from the 2015 to 2021 Medical Expenditure Panel Survey (MEPS), which is a nationally representative survey of the US noninstitutionalized civilian population conducted by the Agency for Healthcare Research and Quality. Although the COVID-19 pandemic caused significant disruptions to field operations of federal surveys, including the MEPS, investigations indicate that the quality of the collected data has been minimally affected.¹⁵ MEPS collects data from two primary sources. The Household Component (HC) collects data from individual household members through survey questionnaires and the Medical Provider Component (MPC) collects data from a sample of health care providers to MEPS HC respondents. The HC data includes demographic, socioeconomic, and health characteristics. The MPC data includes dates of visits or services, types of health care services used, and diagnoses codes for medical encounters. For this study, in order to identify service use, medications, and expenditures specific to psychiatric diagnoses, we merged the following five datasets from MEPS: the full-year consolidated data files from the HC and the medical conditions files, the office-based medical provider visit files, the outpatient visits files, and the prescribed medicine files from the MPC.

Sample

Consistent with prior research,¹⁶ we identified Medicare beneficiaries (including those younger than 65 as well as those 65 years or older) with probable mental health symptoms in each annual cross section of the MEPS using the Patient Health Questionnaire (PHQ-2) and the Kessler 6 (K6) Psychological Distress Scale. These instruments are self-administered questionnaires that measure depressive symptoms and psychological distress, respectively. The PHQ-2 has strong sensitivity (87%) and specificity (78%) for major depressive disorder.¹⁷ Also, the K6 scale has strong sensitivity (90%) and specificity (89%) for a diagnosis of any mental illness.¹⁸ Individuals scoring 3 or higher on the PHQ-2 or 13 or higher on the K6 Psychological Distress Scale were defined as having mental health symptoms (major depression and psychological distress).

Outcomes

We constructed three types of primary outcomes: 1) health care spending, 2) health care utilization, and 3) financial hardship. While health care spending was measured as a continuous outcome, health care utilization and financial hardship were measured as binary outcomes. For health care spending, we included total spending and out-of-pocket spending. All spending measures were adjusted to 2021 dollars using the Personal Consumption Expenditures Price Index for health care. For health care utilization, we measured two types: general health care utilization and mental health service utilization. General health care utilization included five measures (inpatient care, outpatient care, emergency department visits, prescription drug use, and home health care [total and then further categorized into agency and nonagency use]). Mental health service utilization included three measures (outpatient mental health visits, specialty mental health visits, and psychotropic medication fills). Following prior research,¹⁶ outpatient mental health visits were identified as those in which the primary reason for the visit was to receive treatment for a mental health disorder classified as one of the following ICD codes: 291, 292, or 295−314 for ICD-9 codes and F31, F32, F34, and F39 for ICD-10 codes. Specialty mental health visits were distinguished from mental health visits in other settings (e. g., primary care) based on a survey item that asked if the primary reason for the visit was to receive treatment from a mental health professional, such as a psychiatrist, psychologist, counselor, or social worker, for a mental health disorder. Psychotropic medication fills were identified using the Multum Medi-Source Lexicon drug classification system. For financial hardship, we measured four measures.¹⁹ Objective financial hardship was based on the proportion of family income spent on out-of-pocket expenses (greater than 20% and greater than 40% of family income were defined as high and catastrophic financial burden, respectively).²⁰ Subjective financial hardship was defined as reporting a problem paying medical bills or paying medical bills over time. In addition to total spending and out-of-pocket spending for all services, we further analyzed these measures for mental health services (defined as outpatient mental health visits, specialty mental health visits, and psychotropic medication fills) and nonmental health services. We also examined service-specific spending across inpatient care, outpatient care, emergency department visits, prescription drug use, and home health care (agency and nonagency use).

Independent Variables

Our primary independent variable was MA enrollment, which was determined based on enrollment status in December. To adjust for differences in characteristics between TM and MA enrollees with mental health symptoms, we included the following individual-level characteristics in regression models: age, sex, race/ethnicity, employment, marital status, education, family income, health insurance, US census region of residence, chronic conditions, and self-reported health status (self-reported physical health, self-reported mental health, the Physical Component Summary score for the Short-Form 12 [PCS-12], the Mental Component Summary score for the Short-Form 12 [MCS-12]). Self-rated mental and physical health status were rated in five categories, which we categorized as good, very good, or excellent versus fair or poor. The SF-12 score is a standardized mental health measure scored on a scale of 0 to 100, with higher scores indicating better health.

Statistical Analysis

Prior research has found that MA enrollees are healthier than TM enrollees.²¹ This suggests that a direct comparison between the two groups could be potentially biased. To address this selective enrollment, we used a propensity score−based approach. This approach matches TM and MA enrollees with mental health symptoms based on observed characteristics, which helps to reduce the bias in our results. We chose inverse probability of treatment weighting (IPTW) as it is especially useful when the conventional approach of a multivariable regression to simultaneously control many confounders is not feasible.²² Following prior research,²³ we computed each individual’s propensity for enrolling in MA based on the individual-level characteristics described in the above using a logistic regression model with enrollment in MA as the outcome variable and then matched individuals in MA and TM using IPTW. Then, we estimated the sample characteristics between TM and MA enrollees with mental health symptoms before and after applying IPTW. Next, we estimated adjusted outcomes using regression models with IPTW while adjusting for the individual-level characteristics described in the above and year fixed effects. Health care spending was analyzed using a two-part model and health care utilization and financial hardship were analyzed using linear probability models. From the regression results, we calculated the mean adjusted values of the outcomes for TM and MA enrollees with mental health symptoms while holding constant all other variables except enrollment in MA versus TM, allowing us to compare the outcome of interest between TM and MA enrollees with mental health symptoms. We conducted the analysis for the entire population and stratified by dual eligible status as prior research shows that mental health symptoms are most experienced by low-income beneficiaries dually eligible for Medicare and Medicaid.¹⁰ We also conducted the analysis by limiting our sample to older adults. For all analyses, we used survey weights to adjust the sample characteristics to be representative of the Medicare population. We also accounted for the complex survey design in standard error estimation.

RESULTS

Our final sample included 3,910 Medicare beneficiaries with mental health symptoms, of whom 2,247 were enrolled in TM and 1,663 were enrolled in MA (Table 1). This represents 11.31% of the total Medicare population. The severity of mental health symptoms did not differ between TM and MA enrollees with mental health symptoms: 4.01 versus 4.03 for the PHQ-2 and 11.8 versus 12.1 for the K6 Psychological Distress Scale (Appendix Table A). Before applying IPTW, there were several differences in the sample characteristics between TM and MA enrollees with mental health symptoms. Particularly, MA enrollees with mental health symptoms were more likely to be individuals from minoritized backgrounds and to have low family incomes than TM enrollees with mental health symptoms. However, there were no statistically significant differences in the number of chronic conditions and self-reported health status. After applying IPTW, however, the differences were not substantial (Table 2). Specifically, the absolute differences in all characteristics fell within 1.0 percentage points. The results of our logistic regression analyses estimating IPTW are presented in Appendix Table B. Also, there was a substantial overlap in propensity scores between TM and MA enrollees with mental health symptoms (Appendix Figure).

TABLE 1.

Sample Characteristics of TM and MA Enrollees With Mental Health Symptoms, 2015–2021

Characteristic	TM (N = 2,247)	MA (N = 1,663)	p Value
Age, %			0.010
<65	36.4	31.8
65–74	34.4	36.8
75–84	20.1	21.2
85+	9.1	10.3
Female, %	53.7	61.0	0.000
Race/ethnicity, %			0.000
Non-Hispanic white	70.2	67.3
Hispanic	8.4	13.9
Non-Hispanic black	12.0	11.0
Non-Hispanic Asian	4.7	4.5
Non-Hispanic other or multiple	4.7	3.3
Employed, %	9.1	7.2	0.063
Married, %	39.5	40.6	0.570
Education, %			0.003
High school or lower	19.4	21.6
College graduate	52.6	54.0
Advanced degree	28.0	24.3
Family income, %			0.003
<199% of FPL	52.1	56.2
200%−399% of FPL	25.8	27.2
≥400% of FPL	22.1	16.6
Health insurance, %
Medicaid coverage	29.2	27.5	0.335
Any private coverage	39.0	19.6	0.000
US census region, %			0.000
Northeast	15.3	13.1
Midwest	23.8	17.9
South	42.1	45.0
West	18.8	24.0
Chronic conditions, %			0.337
0	18.2	18.0
1−3	52.1	49.6
4−6	27.3	29.2
7+	2.4	3.3
Self-reported health status
Self-rated good physical health, %	43.1	45.1	0.847
Self-rated good mental health, %	54.5	53.9	0.831
Physical component score, mean (SD)	32.2 (11.2)	31.6 (10.4)	0.564
Mental component score, mean (SD)	35.5 (9.0)	35.7 (9.8)	0.130

TM: traditional Medicare; MA: Medicare Advantage; FPL: federal poverty level.

TABLE 2.

Sample Characteristics of TM and MA Enrollees With Mental Health Symptoms After IPTW Application, 2015–2021

Characteristic	TM (N = 2,247)	MA (N = 1,663)	Differences, Percentage Points
Age, %
<65	36.4	31.8	−0.5
65–74	34.4	36.8	0.2
75–84	20.1	21.2	0.2
85+	9.1	10.3	0.2
Female, %	53.7	61.0	−1.0
Race/ethnicity, %
Non-Hispanic white	70.2	67.3	0.7
Hispanic	8.4	13.9	0.0
Non-Hispanic black	12.0	11.0	−0.7
Non-Hispanic Asian	4.7	4.5	−0.1
Non-Hispanic other or multiple	4.7	3.3	0.0
Employed, %	9.1	7.2	−0.2
Married, %	39.5	40.6	0.5
Education, %
High school or lower	19.4	21.6	−0.4
College graduate	52.6	54.0	0.5
Advanced degree	28.0	24.3	0.0
Family income, %
<199% of FPL	52.1	56.2	−0.8
200%−399% of FPL	25.8	27.2	0.6
≥400% of FPL	22.1	16.6	0.2
Health insurance, %
Medicaid coverage	29.2	27.5	−0.8
Any private coverage	39.0	19.6	−0.3
US census region, %
Northeast	15.3	13.1	0.2
Midwest	23.8	17.9	0.0
South	42.1	45.0	−0.6
West	18.8	24.0	0.6
Chronic conditions, %
0	18.5	17.9	−0.4
1−3	52.0	49.9	0.4
4−6	27.3	28.8	0.0
7+	2.3	3.3	0.2
Self-reported health status
Self-rated good physical health, %	43.1	45.1	−0.3
Self-rated good mental health, %	54.5	53.9	0.3
Physical component score, mean (SD)	32.2 (11.2)	31.6 (10.4)	0.6
Mental component score, mean (SD)	35.5 (9.0)	35.7 (9.8)	−0.2

TM: traditional Medicare; MA: Medicare Advantage; IPTW: inverse probability of treatment weighting; FPL: federal poverty level.

To address selective enrollment into MA, we used a propensity score–based approach. Specifically, we computed the inverse probability of treatment weighting (IPTW) as a propensity for enrolling in MA based on individual-level characteristics.

Our IPTW-adjusted analysis showed that MA enrollees with mental health symptoms were lesslikely to use specialty mental health visits and more likely to report financial hardship than TM enrollees with mental health symptoms (Table 3). Specifically, MA enrollees with mental health symptoms were 2.3 percentage points (95% CI: −3.4, −1.2; relative difference: 16.1%) less likely to have specialty mental health visits than TM enrollees with mental health symptoms. However, there were no statistically significant differences in general health care utilization and other mental health services utilization (outpatient mental health visit or psychotropic medication fill). There were also no statistically significant differences in total health care spending, but annual out-of-pocket spending was $292 (95% CI: 152, 432; relative difference: 18.2%) higher among MA enrollees with mental health symptoms than TM enrollees with mental health symptoms. This difference stemmed primarily from nonmental health services, which cost $302 (95% CI: 168, 436) more for MA enrollees (Appendix Table C). Annual out-of-pocket spending for mental health services was not necessarily higher among MA enrollees with mental health symptoms than TM enrollees with mental health symptoms. Notably, MA enrollees with mental health symptoms were 2.6 percentage points (95% CI: 1.5, 3.7; 100.0%) more likely to have nonagency home health care than TM enrollees with mental health symptoms. This finding was consistently observed in spending measures (Appendix Table D). Moreover, MA enrollees with mental health symptoms were 5.0 percentage points (95% CI: 2.9, 7.2; 22.3%) and 2.5 percentage points (95% CI: 0.8, 4.2; 20.9%) more likely to have family paying medical bills over time and experience high financial burden than TM enrollees with mental health symptoms, respectively. Our results were consistent when limited to older adults.

TABLE 3.

Differences in Health Care Spending and Use and Financial Hardship Between TM and MA Enrollees With Mental Health Symptoms, 2015–2021

	Unadjusted Values, %		Adjusted Estimatesa
Outcome	TM (N = 2,247)	MA (N = 1,663)	Absolute Differences, Percentage Points (95% CI)	Relative differences, %
Health care spending, $
Total	22,931 (569)	22,474 (327)	−583 (−1,519, 352)	−2.5
Out-of-pocket	1,605 (93)	1,872 (296)	292 (152, 432)	18.2
Health care utilization, %
General health care utilization
Inpatient care	24.98	26.42	1.6 (−3.5, 6.6)	6.4
Outpatient care	42.53	44.41	2.0 (−0.2, 4.2)	4.7
Emergency department	36.73	36.91	0.3 (−3.7, 4.2)	0.8
Prescription drug	95.41	96.30	1.0 (−0.5, 2.5)	1.0
Home health	22.46	23.72	1.2 (−3.4, 5.9)	5.3
Agency	21.01	21.55	0.5 (−4.0, 5.1)	2.4
Nonagency	2.60	5.15	2.6 (1.5, 3.7)	100.0
Mental health services utilization
Outpatient mental health visitb	21.20	19.16	−2.2 (−5.2, 0.9)	−10.4
Specialty mental health visitc	14.32	12.10	−2.3 (−3.4, −1.2)	−16.1
Psychotropic medication filld	50.45	51.34	1.0 (−3.1, 5.0)	2.0
Financial hardship, %
Family having problems paying medical bills	19.83	23.20	3.3 (−0.2, 6.8)	16.6
Family paying medical bills over time	22.47	27.46	5.0 (2.9, 7.2)	22.3
High financial burdene	11.96	14.41	2.5 (0.8, 4.2)	20.9
Catastrophic financial burdenf	6.44	9.58	3.2 (−0.6, 6.9)	49.7

TM: traditional Medicare; MA: Medicare Advantage.

^aAdjusted differences were estimated using regression models with IPTW while adjusting for individual-level characteristics and year fixed effects.

^bOutpatient mental health visits were identified as those in which the primary reason for the visit was to receive treatment for a mental health disorder (291, 292, or 295−314 for ICD-9 codes and F31, F32, F34, and F39 for ICD-10 codes).

^cSpecialty mental health visits were identified as those in which the primary reason for the visit was to receive treatment from a mental health professional, such as a psychiatrist, psychologist, counselor, or social worker, for a mental health disorder.

^dPsychotropic medication fills were identified using the Multum Medi-Source Lexicon drug classification system.

^eA high financial burden was defined as annual out-of-pocket health care expenses exceeding 20% of the annual postsubsistence income. Post-subsistence income was calculated based on nomograms of food cost from the US Bureau of Labor Statistics.

^fExpenses surpassing 40% of the annual postsubsistence income were categorized as a catastrophic financial burden, in line with the definition endorsed by the World Health Organization.

Our IPTW-adjusted analysis also found that these results were mainly driven by non−dual-eligible enrollees. For non−dual-eligible enrollees, MA enrollees with mental health symptoms were 3.0 percentage points (95% CI: −5.6, −0.3; relative difference: 24.9%) less likely to have specialty mental health visits than TM enrollees with mental health symptoms (Table 4). However, no statistically significant differences were found in general health care utilization and other mental health services utilization (outpatient mental health visit or psychotropic medication fill). There were also no statistically significant differences in total health care spending, but annual out-of-pocket spending was $205 (95% CI: 120, 290; 9.9%) higher among MA enrollees with mental health symptoms than TM enrollees with mental health symptoms. Furthermore, MA enrollees with mental health symptoms were 3.2 percentage points (95% CI: 0.9, 5.5; 12.9%) and 3.2 percentage points (95% CI: 1.5, 4.9; 25.8%) more likely to have family members paying medical bills over time and to experience high financial burden than TM enrollees with mental health symptoms, respectively. Notably, MA enrollees with mental health symptoms were 3.1 percentage points (95% CI: 2.0, 4.2; 108.8%) more likely to have nonagency home health care than TM enrollees with mental health symptoms. For dual-eligible enrollees, there were no statistically significant differences in health care spending and utilization as well as financial hardship (Table 5).

TABLE 4.

Differences in Health Care Spending and Use and Financial Hardship Between Non−Dual-Eligible TM and MA Enrollees With Mental Health Symptoms, 2015–2021

	Unadjusted Values, %		Adjusted Estimatesa
Outcome	TM (N = 1,484)	MA (N = 1,093)	Absolute Differences, Percentage Points (95% CI)	Relative differences, %
Health care spending, $
Total	20,231 (511)	19,694 (546)	−192 (−1,091, 708)	−0.9
Out-of-pocket	2,067 (162)	2,344 (398)	205 (120, 290)	9.9
Health care utilization, %
General health care utilization
Inpatient care	24.73	25.10	0.8 (−6.0, 7.6)	3.2
Outpatient care	43.20	45.78	2.3 (−0.8, 5.5)	5.3
Emergency department	35.51	34.91	−0.4 (−11.2, 10.3)	−1.1
Prescription drug	94.82	96.49	1.7 (−1.0, 4.3)	1.8
Home health	18.31	21.32	3.4 (0.0, 6.9)	18.6
Agency	16.61	19.66	3.4 (−0.6, 7.4)	20.5
Nonagency	2.85	5.57	3.1 (2.0, 4.2)	108.8
Mental health services utilization
Outpatient mental health visitb	18.93	16.25	−3.1 (−8.9, 2.8)	−16.4
Specialty mental health visitc	12.05	9.17	−3.0 (−5.6, −0.3)	−24.9
Psychotropic medication filld	49.94	50.52	0.5 (−3.5, 4.6)	1.0
Financial hardship, %
Family having problems paying medical bills	21.15	23.53	1.7 (−0.1, 3.5)	8.0
Family paying medical bills over time	24.83	28.98	3.2 (0.9, 5.5)	12.9
High financial burdene	12.42	16.03	3.2 (1.5, 4.9)	25.8
Catastrophic financial burdenf	6.44	10.35	3.6 (−0.2, 7.5)	55.9

TM: traditional Medicare; MA: Medicare Advantage.

^aAdjusted differences were estimated using regression models with IPTW while adjusting for individual-level characteristics and year fixed effects.

^bOutpatient mental health visits were identified as those in which the primary reason for the visit was to receive treatment for a mental health disorder (291, 292, or 295−314 for ICD-9 codes and F31, F32, F34, and F39 for ICD-10 codes).

^cSpecialty mental health visits were identified as those in which the primary reason for the visit was to receive treatment from a mental health professional, such as a psychiatrist, psychologist, counselor, or social worker, for a mental health disorder.

^dPsychotropic medication fills were identified using the Multum Medi-Source Lexicon drug classification system.

^eA high financial burden was defined as annual out-of-pocket health care expenses exceeding 20% of the annual postsubsistence income. Post-subsistence income was calculated based on nomograms of food cost from the US Bureau of Labor Statistics.

^fExpenses surpassing 40% of the annual postsubsistence income were categorized as a catastrophic financial burden, in line with the definition endorsed by the World Health Organization.

TABLE 5.

Differences in Health Care Spending and Use and Financial Hardship Between Dual-Eligible TM and MA Enrollees With Mental Health Symptoms, 2015–2021

	Unadjusted Values, %		Adjusted Estimatesa
Outcome	TM (N = 763)	MA (N = 570)	Absolute Differences, Percentage Points (95% CI)	Relative Differences, %
Health care spending, $
Total	29,492 (720)	29,796 (1,267)	−435 (−3,651, 2,781)	−1.5
Out-of-pocket	484 (34)	630 (63)	74 (−3, 151)	15.3
Health care utilization, %
General health care utilization
Inpatient care	25.60	29.58	4.4 (−3.5, 12.3)	17.2
Outpatient care	40.85	41.15	3.4 (−0.1, 6.9)	8.3
Emergency department	39.77	41.72	2.9 (−11.3, 17.2)	7.3
Prescription drug	96.87	95.84	−0.7 (−2.1, 0.7)	−0.7
Home health	32.84	29.46	−2.0 (−11.1, 7.0)	−6.1
Agency	32.01	26.09	−4.7 (−10.8, 1.5)	−14.7
Nonagency	1.98	4.14	2.2 (−1.0, 5.4)	111.1
Mental health services utilization
Outpatient mental health visitb	26.86	26.16	−0.2 (−3.4, 3.1)	−0.7
Specialty mental health visitc	20.01	19.14	0.2 (−3.6, 3.9)	1.0
Psychotropic medication filld	51.74	53.32	2.0 (−6.2, 10.2)	3.9
Financial hardship, %
Family having problems paying medical bills	16.53	22.40	4.1 (−5.3, 13.4)	24.8
Family paying medical bills over time	16.58	23.79	6.1 (−0.8, 13.0)	36.8
High financial burdene	10.82	10.53	0.3 (−6.2, 6.8)	2.8
Catastrophic financial burdenf	6.46	7.73	2.1 (−1.3, 5.5)	32.5

TM: traditional Medicare; MA: Medicare Advantage.

^aAdjusted differences were estimated using regression models with IPTW while adjusting for individual-level characteristics and year fixed effects.

^bOutpatient mental health visits were identified as those in which the primary reason for the visit was to receive treatment for a mental health disorder (291, 292, or 295−314 for ICD-9 codes and F31, F32, F34, and F39 for ICD-10 codes).

^cSpecialty mental health visits were identified as those in which the primary reason for the visit was to receive treatment from a mental health professional, such as a psychiatrist, psychologist, counselor, or social worker, for a mental health disorder.

^dPsychotropic medication fills were identified using the Multum Medi-Source Lexicon drug classification system.

^eA high financial burden was defined as annual out-of-pocket health care expenses exceeding 20% of the annual postsubsistence income. Post-subsistence income was calculated based on nomograms of food cost from the US Bureau of Labor Statistics.

^fExpenses surpassing 40% of the annual postsubsistence income were categorized as a catastrophic financial burden, in line with the definition endorsed by the World Health Organization.

DISCUSSION

Our analysis of a nationally representative sample found that MA enrollees with mental health symptoms used specialty mental health visits less often and were more likely to experience financial hardship than TM enrollees with mental health symptoms, confirming our hypothesis.

As a number of factors may have influence on the comparison between these groups, we employed propensity score matching with a rich array of covariates, including demographic, socioeconomic, and health characteristics. Indeed, the severity of mental health symptoms were similar between TM and MA enrollees. Thus, these findings suggest that MA enrollees with mental health symptoms face significant barriers to accessing affordable mental health care.

Prior research suggests that MA has the potential to improve the efficiency of care delivery without compromising care quality for some populations.¹³ Unlike TM, which pays providers on a fee-for-service basis, MA plans are paid a fixed amount per member each month. This can create an incentive for MA plans to be efficient in providing care, as observed among those with some conditions such as Alzheimer’s disease and related dementias, diabetes, and health failure.^24–26 For those with mental health symptoms, however, this incentive may lead to underuse of necessary services. Previous research indicates that there was less use of medical care provision among MA enrollees with mental illness compared to TM enrollees with mental illness.¹⁴ However, it was unclear whether this group also faced lower use of mental health care provision. In the current study, we did find that MA enrollees with mental health symptoms had lower use of specialty mental health visits than TM enrollees with mental health symptoms. This may be because MA plans have less financial incentive to provide mental health care. Specifically, payments to MA plans have traditionally been adjusted only minimally for clinical characteristics of enrollees, leading to underpayments for those with mental health symptoms who often have complex chronic physical health conditions.²⁷ Indeed, actual health care spending for people with mental illness is substantially higher than the general population.² Underprovision of services may be the result of MA plans limiting in-network providers. Research shows that MA plans had only 23% of the psychiatrists in a county and 36% of MA plans included less than 10% of psychiatrists in their county.²⁸ Notably, MA enrollees with mental health symptoms used nonagency home health care more frequently. This suggests that MA enrollees with mental health symptoms may rely on nonagency home health care for mental health services.

Our findings also suggest that MA enrollees with mental health symptoms face greater financial hardship than TM enrollees with mental health symptoms. Understanding the financial implications is critical as recent MA enrollment growth has been driven primarily by socially vulnerable populations, including individuals from minoritized backgrounds and dual-eligible beneficiaries.²⁹ This finding is consistent with prior research, which found that MA enrollees with mental illness reported higher dissatisfaction with out-of-pocket expenses for medical care than TM enrollees with mental illness.¹⁴ There may be several explanations for this finding. First, the cost of mental health services may be relatively high under MA. For example, MA enrollees paid an average of $9 more for mental health services than for other services in-network.³⁰ Second, MA enrollees may have limited access to in-network mental health providers (only 18.2% of mental health professionals were included in MA’s network) leading them to seek out-of-network mental health services, which can be more expensive.³⁰ Research shows that average out-of-network psychotherapy services were 5% higher under MA than TM, while average in-network prices for mental health services were 13% lower under MA than TM.³⁰ We also found that MA enrollees with mental health symptoms used outpatient mental health services less frequently than TM enrollees with mental health symptoms, but there were no differences in out-of-pocket spending for mental health services. This may suggest that per-visit costs for mental health services may be higher in MA than TM.

Our study informs important policy considerations, highlighting the need to improve mental health provider networks in MA. Medicare’s mental health coverage has substantial gaps, such as a 190-day limit on inpatient psychiatric hospice care, which can lead to high out-of-pocket costs for Medicare beneficiaries with chronic mental illness. Moreover, limited payments to mental health providers likely lead to insufficient provider availability, which could prevent all beneficiaries from receiving the care they need. These gaps may be even more pronounced in MA plans with narrow networks for mental health services. To improve access to in-network specialty mental health providers, the criteria for determining whether a network is adequate should be reviewed and updated. Evidence suggests that narrow physician networks are positively associated with star ratings, which implies that MA plans may use narrow networks to achieve higher star ratings by contracting with select physicians.³¹ To help Medicare beneficiaries identify plans with adequate networks, the Centers for Medicare & Medicaid Services could incentivize plans to make comprehensive, up-to-date provider directories available by incorporating measures of adequate provider capacity into their star rating systems.

Limitations

This study has several limitations. First, our sample was limited to the noninstitutionalized US population, excluding incarcerated people, nursing home residents, and those in residential treatment. Second, while we used measures validated as strongly predictive of mental health symptoms, the measures are self-reported, and respondents may underreport mental health problems due to stigma or privacy concerns. Relatedly, a range of support options outside of the healthcare system may exist for individuals experiencing mental health symptoms, and thus it is possible that this population may not necessarily perceive an unmet need for medical care. Third, while we used propensity score weighting to create comparable groups of TM and MA enrollees with mental health symptoms, unobserved differences in patient factors may have remained. Furthermore, the IPTW-based approach may not always be as efficient as than doubly-robust estimators, but prior research suggests comparable results despite some varying levels of uncertainty.²² Thus, our analysis is associational, and our results do not necessarily have a causal interpretation. Finally, prior research found that the MEPS underrepresented psychiatric inpatient utilization, with 66% fewer psychiatric admissions and 14% fewer nonpsychiatric admissions than a more comprehensive survey of admissions.³² This could lead to an underestimation of health care spending, use, and financial hardship among those with mental health symptoms.

CONCLUSION

Our analysis of a nationally representative sample found that MA enrollees with mental health symptoms used specialty mental health visits less often and experienced greater financial hardship than TM enrollees with mental health symptoms. This finding indicates that MA enrollees with mental health symptoms face significant barriers to accessing affordable mental health care. To address these barriers, policymakers should develop policies aimed at improving access to mental health services while reducing financial burden for MA enrollees.

Supplementary Material

Supplementary material associated with this article can be found in the online version at https://doi.org/10.1016/j.jagp.2024.01.014.

Highlights.

• What is the primary question addressed by this study?

  Are there differences in health care spending and utilization, and financial hardship between Traditional Medicare (TM) and Medicare Advantage (MA) enrollees with mental health symptoms?

• What is the main finding of this study?

  MA enrollees with mental health symptoms had fewer specialty mental health visits than TM enrollees with mental health symptoms.

  MA enrollees with mental health symptoms faced greater financial hardship than TM enrollees with mental health symptoms.

• What is the meaning of the finding?

  This finding indicates that MA enrollees with mental health symptoms face significant barriers to accessing affordable mental health care.

  To address these barriers, policymakers should develop policies aimed at improving access to mental health services while reducing financial burden for MA enrollees.