Health Care Utilization, Care Satisfaction, and Health Status for Medicare Advantage and Traditional Medicare Beneficiaries With and Without Alzheimer Disease and Related Dementias

Key Points español 中文 (chinese)

Question  Are there differences in health care utilization, care satisfaction, and health status among US Medicare beneficiaries with Alzheimer disease and related dementias enrolled in Medicare Advantage vs traditional Medicare?

Findings  This cohort study of 47 100 Medicare beneficiaries found that Medicare Advantage beneficiaries with Alzheimer disease and related dementias had lower health care utilization rates than did traditional Medicare beneficiaries with Alzheimer disease and related dementias, especially for medical practitioner visits. Overall, there were no differences in care satisfaction and health status.

Meaning  These findings suggest that Medicare Advantage plans may achieve lower health care utilization through high efficiency of care rather than underprovision of care.

Importance  Compared with traditional Medicare (TM) fee-for-service plans, Medicare Advantage (MA) plans may provide more-efficient care for beneficiaries with Alzheimer disease and related dementias (ADRD) without compromising care quality.

Objective  To determine differences in health care utilization, care satisfaction, and health status for MA and TM beneficiaries with and without ADRD.

Design, Setting, and Participants  A cohort study was conducted of MA and TM beneficiaries with and without ADRD from all publicly available years of the Medicare Current Beneficiary Survey between 2010 and 2016. To address advantageous selection into MA plans, county-level MA enrollment rate was used as an instrument. Data were analyzed between July 2019 and December 2019.

Exposures  Enrollment in MA.

Main Outcomes and Measures  Self-reported health care utilization, care satisfaction, and health status.

Results  The sample included 47 100 Medicare beneficiaries (25 900 women [54.9%]; mean [SD] age, 72.2 [11.4] years). Compared with TM beneficiaries with ADRD, MA beneficiaries with ADRD had lower utilization across the board, including a mean of −22.3 medical practitioner visits (95% CI, −24.9 to −19.8 medical practitioner visits), −2.3 outpatient hospital visits (95% CI, −3.6 to −1.1 outpatient hospital visits), −0.2 inpatient hospital admissions (95% CI, −0.3 to −0.1 inpatient hospital admissions), and −0.1 long-term care facility stays (95% CI, −0.2 to −0.1 long-term care facility stays). A similar trend was observed among beneficiaries without ADRD, but the difference was greater between MA and TM beneficiaries with ADRD than between MA and TM beneficiaries without ADRD (mean, −15.0 medical practitioner visits [95% CI, −18.7 to −11.3 medical practitioner visits], −1.7 outpatient hospital visits [95% CI, −3.0 to −0.3 outpatient hospital visits], and −0.1 inpatient hospital admissions [95% CI, −1.0 to 0.0 inpatient hospital admissions]). Overall, no or negligible differences were detected in care satisfaction and health status between MA and TM beneficiaries with and without ADRD.

Conclusions and Relevance  Compared with TM beneficiaries, MA beneficiaries had lower health care utilization without compromising care satisfaction and health status. This difference was more pronounced among beneficiaries with ADRD. These findings suggest that MA plans may be delivering health care more efficiently than TM, especially for beneficiaries with ADRD.

Caring for people with Alzheimer disease and related dementias (ADRD) will generate substantial costs to the US health care system. Both the number of individuals with ADRD and the associated costs are projected to increase over time. As of 2010, there were 4.5 million US individuals with ADRD, and that number is expected to increase to 13.2 million in 2050.¹ In addition, mean per-person Medicare costs for beneficiaries with ADRD were estimated to be $23 497 in 2011, more than triple the mean $7223 costs for Medicare beneficiaries without ADRD.^2,3 Total costs (including health care, long-term care, and hospice services) for Medicare beneficiaries with ADRD are projected to increase from $172 billion in 2010 to $1.1 trillion in 2050.³ Such a dramatic increase in the costs associated with ADRD will pose a substantial burden to the US federal government.

Managed care provides opportunities to reduce the growth rate of health care costs. Medicare provides a managed care option, the Medicare Advantage (MA) program, which allows beneficiaries to enroll in private insurance plans rather than in traditional fee-for-service Medicare (TM). There are several differences between MA and TM, but perhaps the most important is that MA practitioners are paid on a capitated basis rather than for each service performed. Capitation creates the incentive for practitioners to be efficient in their approach to care because their revenue is fixed prospectively.⁴ The MA plans use various techniques to control health care utilization, such as restricted practitioner networks, prior authorization, and utilization review, as well as investing in preventive services, care coordination, and chronic disease management.^5-9

There is evidence that MA plans tend to enroll beneficiaries who are healthier than average, and comparisons that use beneficiaries with similar health profiles have found lower health service utilization among MA beneficiaries than among TM beneficiaries.^5-8 These results have been attributed, in part, to improved care coordination, chronic condition management, provision of low-intensity care, and transitions to less-expensive care settings in MA plans. In addition, compared with TM beneficiaries, MA beneficiaries had lower hospital readmission rates,^6,7,10 better clinical quality outcomes,^11,12 better patient experiences,^11,13 and lower mortality rates.^6,14 These findings support the notion that care coordination and management strategies among MA plans have the potential to improve the efficiency of care delivery without compromising care quality.

Within the literature addressing the role of MA plans in providing lower utilization with quality comparable to that of TM, we did not find any reference to the impact of MA plans among individuals with ADRD. However, there is evidence suggesting that care delivery and health care utilization are inefficient for TM beneficiaries with ADRD. A large proportion of health care utilization for beneficiaries with ADRD is associated with transitions to high-cost settings, such as an inpatient setting or skilled nursing facility,^15-17 some of which have been shown to be unnecessary or preventable.^18-21 Moreover, MA plans may make targeted improvements in the care management of beneficiaries with ADRD because of the growing volume of ADRD beneficiaries enrolled. Previous research²² found that after a new ADRD diagnosis, TM beneficiaries were more likely to switch to MA plans, whereas MA beneficiaries were more likely to stay in MA plans.

To address this gap, we examined health care utilization, care satisfaction, and health status among MA and TM beneficiaries with ADRD. We compared our findings with those of a similar analysis among beneficiaries without ADRD to address the association of MA enrollment with the outcomes.

We used the Medicare Current Beneficiary Survey and the Geographic Variation Public Use File. The Medicare Current Beneficiary Survey provides a nationally representative sample of the Medicare population with a 4-year follow-up. The data provide individual-level information on demographic characteristics, socioeconomic characteristics, health care utilization, care satisfaction, and health status. The Geographic Variation Public Use File provides county-level MA enrollment rates. Our analysis uses all publicly available data from 2010 to 2016. The 2014 Medicare Current Beneficiary Survey data were never released.

This study was approved by the University of Pennsylvania’s institutional review board and received a waiver of informed consent and HIPPA authorization because the data were deidentified. This study follows the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cohort studies.

We included Medicare beneficiaries aged 65 years or older with 12 months of continuous enrollment in MA or TM. We excluded those whose original eligibility was attributable to disability or end-stage kidney disease and those who died. We then identified the following 4 groups: MA beneficiaries with ADRD, TM beneficiaries with ADRD, MA beneficiaries without ADRD, and TM beneficiaries without ADRD. We identified ADRD cases through the beneficiary or proxy survey responses to the following question: “Has a doctor ever told you that you had Alzheimer disease or dementia?”

Our outcomes were self-reported health care utilization, care satisfaction, and health status. First, we assessed utilization for each of the following 9 types of service: inpatient hospital admission, outpatient hospital visit, medical practitioner visit, home health visit, hospice stay, short-term facility stay (eg, skilled nursing facility), long-term care facility stay (eg, nursing home), prescription drug purchase measured as a single purchase of a single drug in a single container, and dental visit. Self-reported utilization for TM beneficiaries undergoes extensive validation using Medicare claims data and has generally been found to be accurate.^23,24 Second, we assessed the extent to which beneficiaries were satisfied with their plans in terms of care quality, out-of-pocket costs, access to specialists, follow-up after initial treatments, and physician’s concern for overall health. Satisfaction was measured in 4 levels: very dissatisfied, dissatisfied, satisfied, or very satisfied. Finally, we assessed self-reported general health status compared with same-age people and overall health status compared with 1 year ago. General health status compared with same-age people was measured in 5 levels: poor, fair, good, very good, or excellent. Overall health status compared with 1 year ago was measured in 2 levels: worse health vs same or better health. A higher value indicates better care satisfaction or health status.

Our key independent variables were 12-month enrollment in MA, the presence of ADRD, and its interaction term. To control for differences in sample characteristics among MA and TM beneficiaries, we included the following variables: age, sex, race/ethnicity, education level, income, Medicare and Medicaid dual eligibility, marital status, indicator for living with someone, residence in a metropolitan area, US Census region of residence, comorbidity, number of limitations on activities of daily living, and year.

Previous research^25-28 has found that healthy beneficiaries are more likely to enroll in MA than TM, suggesting that advantageous selection would invalidate a direct comparison between MA and TM beneficiaries. To address selection bias, we used an instrumental variable approach, using county-level MA enrollment rate as an instrument for the individual-level decision to enroll in MA plans. We calculated the county-level MA enrollment rate as the share of Medicare beneficiaries (aged ≥65 years) enrolled in MA plans.

We estimated sample characteristics and outcomes and tested unadjusted differences between MA and TM beneficiaries with and without ADRD. We used χ² tests for categorical variables and analysis of variance for continuous variables. Next, we performed a 2-stage least-squares regression model. In the first stage, we obtained the estimated likelihood of enrolling in MA plans while accounting for advantageous selection into MA plans according to the county-level MA enrollment rates. In the second stage, we estimated the association between estimated enrollment in MA plans from the first stage and the outcomes of interest. To assess whether the instrument was strong, we tested the association with MA enrollment and then examined F statistics, where a value greater than 10 traditionally indicates a strong instrument.²⁹ To assess whether the instrument was valid, we examined the association between the instrument and measured confounders because we cannot directly assess the association between the instrument and unmeasured confounders. Both stages adjusted for the aforementioned control variables and adjusted the SEs for clustering within county.

Using the marginal effects estimated from the 2-stage least-squares regression model, we estimated the mean values of the outcomes for MA beneficiaries with ADRD, TM beneficiaries with ADRD, MA beneficiaries without ADRD, and TM beneficiaries without ADRD, respectively. We then performed postestimation tests to estimate the differences in the outcomes between MA and TM beneficiaries with and without ADRD, respectively. We conducted several sensitivity analyses. First, we reexamined our analysis by using state-level MA enrollment rates, because there may be some concern about the validity of the county-level MA enrollment rate as an instrument. Second, we adjusted the SEs for clustering within individual and county. In our primary analysis, we treated the Medicare Current Beneficiary Survey data for each year as an independent annual cross-sectional survey, even though some beneficiaries were included in the data over the course of multiple years. We used survey weights to adjust sample characteristics to be representative of the Medicare population. All analyses were conducted using Stata statistical software version 16.0 (StataCorp). All P values were from 2-sided tests, and results were deemed statistically significant at P < .05. Data were analyzed between July 2019 and December 2019.

Our sample included 47 100 Medicare beneficiaries (25 900 women [54.9%]; mean [SD] age, 72.2 [11.4] years) (Table 1). We identified 1006 MA beneficiaries with ADRD, 1841 TM beneficiaries with ADRD, 14 880 MA beneficiaries without ADRD, and 29 373 TM beneficiaries without ADRD. The MA and TM beneficiaries with ADRD had similar demographic characteristics (mean [SD] age, 77.14 [11.0] vs 77.56 [12.0] years; 610 [60.6%] vs 1089 [59.2%] women; and 454 [45.1%] vs 811 [44.1%] married). However, there were differences in sample characteristics between MA and TM beneficiaries without ADRD in terms of comorbidities, especially hardening of the arteries (1333 [9.0%] vs 2923 [10.0%] beneficiaries), hypertension (10 519 [70.7%] vs 19 919 [67.8%] beneficiaries), cancer (5043 [33.9%] vs 10 797 [36.8%] beneficiaries), rheumatoid arthritis (2445 [16.4%] vs 4424 [15.1%] beneficiaries), osteoporosis (3381 [22.7%] vs 6333 [21.6%] beneficiaries), asthma or chronic obstructive pulmonary disease (2929 [19.7%] vs 6234 [21.2%] beneficiaries), diabetes (4502 [30.3%] vs 7760 [26.4%] beneficiaries), mental illness (1052 [7.1%] vs 2977 [10.1%] beneficiaries), and depression (3813 [25.6%] vs 8168 [27.8%] beneficiaries) (Table 1). Despite these differences in comorbid conditions, MA beneficiaries without ADRD were not necessarily healthier than TM beneficiaries without ADRD.

Our unadjusted analysis showed that MA beneficiaries with ADRD had fewer inpatient hospital admissions (mean [SD], 0.3 [0.6] vs 0.5 [1.0] inpatient hospital admissions), outpatient hospital visits (mean [SD], 3.5 [9.6] vs 6.2 [10.2] outpatient hospital visits), medical practitioner visits (mean [SD], 17.3 [19.0] vs 39.7 [39.0] medical practitioner visits), and long-term care facility stays (mean [SD], 0.1 [0.4] vs 0.3 [1.1] long-term facility stays) than TM beneficiaries with ADRD, but they had more prescription drug purchases (mean [SD], 65.5 [54.9] vs 58.5 [52.4] prescription drug purchases) (Table 2). A similar result was found among beneficiaries without ADRD (mean [SD], 44.2 [47.8] vs 41.5 [46.5] prescription drug purchases).

There were only modest differences in care satisfaction (mean [SD] care satisfaction scores, 3.8 [0.7] vs 3.8 [0.7] for quality of medical care; 3.6 [0.7] vs 3.6 [0.8] for out-of-pocket costs for medical care; 3.5 [1.0] vs 3.5 [1.1] for available care by specialists; 3.3 [1.2] vs 3.4 [1.2] for follow-up after initial treatment; and 3.8 [0.7] vs 3.8 [0.7] for physician’s concern for overall health) and health status (mean [SD] health status score, 3.2 [0.9] vs 3.2 [0.9] for general health status compared with same-age people; and 0.8 [0.4] vs 0.8 [0.4] for overall health status compared with 1 year ago) between MA and TM beneficiaries without ADRD (Table 2). It should be noted that approximately 55% of responses for individuals with ADRD relied on proxy responses.

We found that the mean county-level MA enrollment rate (0.009% [95% CI, 0.008%-0.010%] for health care utilization except for prescription drug purchase; 0.010% [95% CI, 0.009%-0.011%] for prescription drug purchase; 0.009% [95% CI, 0.008%-0.011%] for quality of medical care; 0.009% [95% CI, 0.008%-0.010%] for out-of-pocket costs for medical care; 0.009% [95% CI, 0.008%-0.010%] for available care by specialists; 0.009% [95% CI, 0.008%-0.010%] for follow-up after initial treatment; 0.008% [95% CI, 0.009%-0.011%] for physician’s concern for overall health; 0.009% [95% CI, 0.008%-0.011%] for general health status compared with same-age people; and 0.009% [95% CI, 0.008%-0.010%] for overall health status compared with 1 year ago) was a strong and valid instrument. Greater MA enrollment was associated with a higher likelihood of enrolling in MA plans, and F statistics were greater than 10 (F score range, 184.56-341.04) (Table 3). Also, most individual-level control variables were balanced across values of the instrument.

Our instrumental variable analysis showed that MA beneficiaries with ADRD had lower levels of health care utilization than TM beneficiaries with ADRD, including a mean of −22.3 medical practitioner visits (95% CI, −24.9 to −19.8 medical practitioner visits), −2.3 outpatient hospital visits (95% CI, −3.6 to −1.1 outpatient hospital visits), −0.2 inpatient hospital admissions (95% CI, −0.3 to −0.1 inpatient hospital admissions), and −0.1 long-term care facility stays (95% CI, −0.2 to −0.1 long-term care facility stays) (Table 4). There were no statistically significant differences in home health visits, short-term facility stays, prescription drug purchases, and dental visits. Similar trends were observed among beneficiaries without ADRD, in that MA beneficiaries had fewer medical practitioner visits (mean, −15.0 medical practitioner visits; 95% CI, −18.7 to −11.3 medical practitioner visits), outpatient hospital visits (mean, −1.7 outpatient hospital visits; 95% CI, −3.0 to −0.3 outpatient hospital visits), and inpatient hospital admissions (mean, −0.1 inpatient hospital admissions; 95% CI, −0.1 to 0.0 inpatient hospital admissions) than TM beneficiaries. In addition, MA beneficiaries without ADRD had a mean of 19.4 more prescription drug purchases (95% CI, 10.4 to 28.5 prescription drug purchases) than TM beneficiaries without ADRD.

Our instrumental variable analysis also showed that, overall, there were no statistically significant differences in care satisfaction and health status between MA and TM beneficiaries with ADRD (except for satisfaction on physician’s concern for overall health; mean score, 0.1; 95% CI, 0.0 to 0.1) and without ADRD (except for general health status compared with same-age people; mean score, −0.1; 95% CI, −0.2 to −0.1) (Table 5). Results are robust to using state-level MA enrollment rates as an instrument (eTable 1 in the Supplement) and clustering within individual and county (eTable 2 and eTable 3 in the Supplement).

In an analysis of a nationally representative sample of the Medicare population, we found that compared with TM beneficiaries with ADRD, MA beneficiaries with ADRD had lower health care utilization rates, particularly for medical practitioner visits. A similar trend was observed among beneficiaries without ADRD, but the magnitude of the difference in health care utilization was larger between beneficiaries with ADRD than between beneficiaries without ADRD. On the other hand, no or marginal differences were detected in care satisfaction and health status between MA and TM beneficiaries with and without ADRD.

We observed that MA and TM beneficiaries with ADRD had similar demographic and health characteristics. We also found that there were differences in sample characteristics between MA and TM beneficiaries without ADRD, but this does not necessarily indicate that healthier beneficiaries were more likely to enroll in MA than TM. These results are consistent with the more recent literature,⁶ which finds little evidence to suggest that MA plans still enroll healthier beneficiaries than TM. The similar sample characteristics of beneficiaries with ADRD are of particular interest because previous research²² has found that beneficiaries have increasingly enrolled in MA plans after receiving a diagnosis of ADRD. This may reflect the preference of beneficiaries with ADRD for MA plans because MA plans have the flexibility to provide enhanced services for complex and high-need patients through coordinated care that addresses the medical, behavioral, and social aspects of the disease.

We also found that MA beneficiaries had fewer medical practitioner visits, outpatient hospital visits, and inpatient hospital admissions than TM beneficiaries, and these differences were more pronounced among beneficiaries with ADRD than beneficiaries without ADRD. The largest decrease was in medical practitioner visits. Medical practitioner visits are of particular interest because they measure individual events for a variety of medical services, equipment, and supplies, possibly reflecting a high intensity of care. Hence, a higher number of medical practitioner visits among TM beneficiaries compared with MA beneficiaries may indicate inefficient care delivery in TM associated with a lack of incentive to control utilization and coordinate care. Furthermore, the fee-for-service payment system under TM may incentivize more face-to-face visits, but MA plans have greater flexibility in the methods for delivering the care. For example, MA plans have provided additional telehealth services as a supplemental benefit, enabling MA enrollees to have access to care without going to their practitioners. Further decreases in medical practitioner visits among MA enrollees are expected starting in 2020, when MA plans will be able to include telehealth as a basic government-funded benefit.^30,31 This is particularly relevant to beneficiaries with ADRD, who tend to have more-frequent transitions and require care coordination.^32,33 Another notable finding is that MA beneficiaries with ADRD had fewer inpatient hospital admissions than TM beneficiaries with ADRD. Although the magnitude of the difference in inpatient hospital admissions between MA and TM beneficiaries with ADRD was modest, lower inpatient hospital admissions among MA beneficiaries with ADRD are notable because hospitalizations may adversely affect the health status of beneficiaries with ADRD by increasing the risk of nosocomial infections, falls, and cognitive decline.^34,35

We detected no differences in care satisfaction between MA and TM beneficiaries with or without ADRD. This finding provides evidence suggesting that MA plans may not tailor benefit packages to selectively attract healthy beneficiaries, leading to decreased advantageous selection over time.^6,36,37 However, there is evidence showing that advantageous selection has been decreased but not eliminated; specifically, 11% and 2% of MA beneficiaries voluntarily switched to another MA plan or TM, respectively.³⁸ In particular, switching to TM was high among MA beneficiaries with high needs and high costs.^26-28,39,40 High disenrollment rates were partly attributable to poor patient experience.⁴¹

There were no or negligible differences in health status between MA and TM beneficiaries with or without ADRD. This finding suggests that lower health care utilization among MA beneficiaries may not be attributable to under-provision of care and, thus, not come at the cost of poorer care quality. Rather, MA plans may achieve lower health care utilization through high efficiency of care. This study contributes to the growing literature showing that TM lacks a direct financial incentive to control utilization, which could lead to excess care provision that does not improve patient outcomes.^6,10,42 Previous research³⁴ found that MA beneficiaries had increased inpatient utilization and total charges by 60% and 50%, respectively, when they were forced out of MA plans because of plan exit. However, the increases in utilization and charges were not associated with any measurable reduction in hospital quality or patient mortality.³⁴

This study has several limitations. First, our variables may be subject to self-reporting errors. Although self-reported utilization for MA beneficiaries was not validated, this is less likely to affect our findings because self-reported utilization for TM beneficiaries has been found to be accurate on the basis of validation using Medicare claims data. Second, our findings for beneficiaries with ADRD may be confounded by proxy response because approximately 55% of them relied on proxy response, although there is not a differential proxy response rate by MA vs TM. Third, we did not detect differences in patient satisfaction, and this could be associated with sample size. Fourth, we found that MA and TM beneficiaries had similar comorbidities characteristics. However, comorbidities might not be equal across MA and TM because of aggressive diagnostic coding in MA plans.^43,44 Fifth, previous research²⁷ has found that MA beneficiaries disenrolled from their plans after health shocks. Requiring 12-month continuous enrollment in MA or TM to ensure accurate health plan attribution may lead to some selection on care satisfaction. Sixth, because of the coarse measurements available, we could not account for the severity of ADRD.

Compared with TM beneficiaries, MA beneficiaries had lower rates of health care utilization without compromising care satisfaction and health status, particularly among beneficiaries with ADRD. These findings suggest that MA plans may be more efficient than TM at delivering health care for beneficiaries with ADRD.

Accepted for Publication: February 6, 2020.

Published: March 30, 2020. doi:10.1001/jamanetworkopen.2020.1809

Open Access: This is an open access article distributed under the terms of the CC-BY License. © 2020 Park S et al. JAMA Network Open.

Corresponding Author: Sungchul Park, PhD, Department of Health Management and Policy, Dornsife School of Public Health, Drexel University, 3215 Market St, Nesbitt Hall, Philadelphia, PA, 19104 (smp462@drexel.edu).

Author Contributions: Dr Park had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Park, Fishman, Larson, Coe.

Acquisition, analysis, or interpretation of data: Park, White, Fishman, Coe.

Drafting of the manuscript: Park, Fishman.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Park, White, Coe.

Obtained funding: Fishman, Larson, Coe.

Administrative, technical, or material support: Larson, Coe.

Supervision: Fishman, Coe.

Conflict of Interest Disclosures: Dr Larson reported receiving royalties from UpToDate outside the submitted work. No other disclosures were reported.

Funding/Support: This work was supported by grant R01 AG049815 from the National Institutes of Health and grant AG U01 0006781 from the National Institute of Aging, National Institutes of Health.

Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.