Key Points
Question
Do rates of low-value care differ between traditional Medicare (TM) and Medicare Advantage (MA), and, if so, what elements of insurance design are associated with the differences?
Findings
In this cross-sectional study of 2 470 199 Medicare beneficiaries, those enrolled in MA received 9.2% fewer low-value services than those in TM (23.1 vs 25.4 total low-value services per 100 beneficiaries). The MA beneficiaries in health maintenance organizations and those in primary care organizations reimbursed within advanced value-based payment models had the lowest rates of low-value care.
Meaning
The study results suggest that low-value care is less common in MA than TM, with elements of insurance design present in MA associated with fewer low-value services.
Abstract
Importance
Low-value care in the Medicare program is prevalent, costly, potentially harmful, and persistent. Although Medicare Advantage (MA) plans can use managed care strategies not available in traditional Medicare (TM), it is not clear whether this flexibility is associated with lower rates of low-value care.
Objectives
To compare rates of low-value services between MA and TM beneficiaries and explore how elements of insurance design present in MA are associated with the delivery of low-value care.
Design, Setting, and Participants
This cross-sectional study analyzed beneficiaries enrolled in MA and TM using claims data from a large, national MA insurer and a random 5% sample of TM beneficiaries. The study period was January 1, 2017, through December 31, 2019. All analyses were conducted from July 2021 to March 2022.
Exposures
Enrollment in MA vs TM.
Main Outcomes and Measures
Low-value care was assessed using 26 claims-based measures. Regression models were used to estimate the association between MA enrollment and rates of low-value services while controlling for beneficiary characteristics. Stratified analyses explored whether network design, product design, value-based payment, or utilization management moderated differences in low-value care between MA and TM beneficiaries and among MA beneficiaries.
Results
Among a study population of 2 470 199 Medicare beneficiaries (mean [SD] age, 75.6 [7.0] years; 1 346 777 [54.5%] female; 229 107 [9.3%] Black and 2 126 353 [86.1%] White individuals), 1 527 763 (61.8%) were enrolled in MA and 942 436 (38.2%) were enrolled in TM. Beneficiaries enrolled in MA received 9.2% (95% CI, 8.5%-9.8%) fewer low-value services in 2019 than TM beneficiaries (23.1 vs 25.4 total low-value services per 100 beneficiaries). Although MA beneficiaries enrolled in health management organization and preferred provider organization products received fewer low-value services than TM beneficiaries, the difference was largest for those enrolled in health management organization products (2.6 fewer [95% CI, 2.4-2.8] vs 2.1 fewer [95% CI, 1.9-2.3] services per 100 beneficiaries, respectively). Across primary care payment arrangements, MA beneficiaries received fewer low-value services than TM beneficiaries, with the largest difference observed for MA beneficiaries whose primary care physicians were reimbursed within 2-sided risk arrangements.
Conclusions and Relevance
In this cross-sectional study of Medicare beneficiaries, those enrolled in MA had lower rates of low-value care than those enrolled in TM; elements of insurance design present in the MA program and absent in TM were associated with reduction in low-value care.
This cross-sectional study compares rates of low-value services between Medicare Advantage and traditional Medicare beneficiaries and explores how elements of insurance design present in Medicare Advantage are associated with the delivery of low-value care.
Introduction
Low-value care (tests, treatments, and procedures that provide little to no clinical benefit) is a widespread and costly source of waste, inefficiency, and potential harm in the US health care system.1,2 Within the Medicare program, as many as a third of beneficiaries receive low-value services annually.2,3,4 Despite considerable attention, the incidence of low-value care in Medicare has remained stable during the past 2 decades.5,6
Research on low-value care in the Medicare program has primarily focused on patient, physician, and delivery system characteristics associated with higher rates of low-value services.4,7,8,9 Efforts to reduce low-value care have similarly focused on raising awareness among these constituencies.10 Less is known about how elements of insurance design, such as coverage policies, benefit design, and payment contract types are associated with the provision of low-value care.
Comparing low-value care in Medicare Advantage (MA) and traditional Medicare (TM) represents an opportunity to explore the association between insurance design and low-value care. Unlike TM, MA plans have a greater ability to create networks of clinicians (network design); develop insurance products that incentivize in-network care, place a focus on primary care, and promote the efficient use of specialty services (product design); and review tests, treatments, and procedures for medical necessity and appropriateness before remitting payment (utilization management).11,12 Compared with TM, MA plans have also more rapidly adopted value-based payment models,13 which have been associated with reductions in low-value care.7 Consequently, we hypothesized that Medicare beneficiaries enrolled in MA would experience lower rates of low-value services than those enrolled in TM.
To investigate this hypothesis, we used data from a large, national MA insurer and a national sample of TM beneficiaries to compare rates of low-value care among MA and TM beneficiaries in 2019. We also explored whether elements of insurance design moderated the association between MA enrollment and the receipt of low-value care.
Methods
Data and Study Population
The sample of MA beneficiaries was drawn from enrollees in plans that were offered by a large, national insurer. The TM beneficiaries were identified from a random 5% national sample with Part A and B coverage. We limited the sample to beneficiaries 65 years or older as of January 1, 2019, with continuous health plan enrollment from 2017 through 2019, or until death or hospice enrollment in 2019. Continuous coverage from 2017 through 2018 was required to capture diagnoses used to determine eligibility for low-value services. We excluded beneficiaries with claims-based evidence of end-stage kidney disease during the preindex period and those dually enrolled in Medicare and Medicaid. From the MA sample, we also excluded beneficiaries whose clinicians delegated claims processing to a third party and those contractually excluded from research. From the TM sample, we also excluded beneficiaries with coordination of benefits, which occurs when beneficiaries have more than 1 insurance policy (eg, TM and an employer-sponsored plan), resulting in incomplete claims visibility.
This study was reviewed by the Humana Healthcare Research Human Subject Protection Office and deemed not human participants research; therefore, informed consent was waived. The study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines for cross-sectional studies.
Low-Value Care Measures
We identified 26 low-value services relevant to the Medicare population using adapted versions of previously described and widely-used claim-based measures originally developed by Schwartz et al.2,7,8,14 We grouped these measures in several, nonmutually exclusive ways: by clinical category (cancer screening, diagnostic and preventive testing, preoperative testing, imaging, cardiovascular testing and procedures, and other surgeries),2,7,8 whether they were typically ordered by primary care physicians (PCPs) (PCP-driven) or specialists (specialist-driven)7,8 and whether they were subject to preauthorization (PA) by the MA plan in 2019.15 Additional information on measure specification and classification is available in eTable 1 and eTable 2 in the Supplement.
Covariates
Patient-level covariates included age (65-74, 75-84, or 85-97 years), sex, race and ethnicity, original reason for Medicare entitlement, Charlson comorbidity index (CCI) score,16 core-based statistical area (CBSA), and county-level population density. Because each low-value service did not apply to every beneficiary, we constructed binary indicator variables for whether a beneficiary qualified for the potential receipt of each low-value service (eTable 1 in the Supplement). Race and ethnicity were assessed according to the US Centers for Medicare & Medicaid (CMS) beneficiary race code, which reflects data reported to the Social Security Administration, and categorized as Black, White, Other (including Asian, Hispanic, North American Native, and Other), and Unknown. Demographic characteristics were assessed as of January 1, 2019, and CCI scores were assessed using claims from 2018.
For each MA beneficiary, we used enrollment data to determine whether the beneficiary was enrolled in a health maintenance organization (HMO) or preferred provider organization (PPO) product. For MA beneficiaries enrolled in HMO products, we used contract data to categorize beneficiaries according to the payment model for their attributed primary care organization using the following taxonomy: fee-for-service (FFS), including pay-for-performance programs; shared savings with upside-only financial risk (upside-only risk); and shared savings with upside and downside financial risk (2-sided risk) (eMethods in the Supplement).
Analytic Approach
We compared demographic and clinical characteristics between beneficiaries enrolled in TM and those enrolled in MA using standardized mean differences. We then calculated unadjusted rates of low-value services among MA and TM beneficiaries in 2019 and compared these rates using t tests.
To estimate the association between insurance type and the receipt of low-value services, we constructed a series of ordinary least-squares regression models. The primary analysis estimated a composite measure of the total number of low-value services. We also modeled the number of low-value services within each clinical category, as well as each of the 26 low-value services individually. The primary independent variable for all models was insurance type (MA or TM). The other independent variables were age, sex, race and ethnicity, original reason for Medicare entitlement, CCI score, and population density. Models included CBSA fixed effects to account for differences in the geographic distribution of MA and TM beneficiaries in the study population. For the primary analysis estimating the total number of low-value services, we also included qualification flags for each low-value service measure as additional independent variables to account for potential differences in beneficiary eligibility across low-value service measures. For all other analyses, we restricted the populations to beneficiaries who were eligible for the specific low-value services. To account for beneficiaries that died or entered hospice in 2019, we weighted observations and annualized the number of low-value services by beneficiary months.
As noted previously, we hypothesized several mechanisms by which differences in insurance design between MA and TM might be associated with differences in the rate of low-value services. To investigate the role of network design, we stratified the MA population by product type, changing the primary exposure variable to one with 3 values: TM, MA PPO, and MA HMO. Because HMO products incentivized in-network utilization more than PPO plans, we hypothesized enrollment in HMO products would be associated with comparatively fewer rates of low-value services.
This stratification also allowed us to investigate the role of product design. Not only do HMO products incentivize in-network utilization, they also tend to require that beneficiaries select a PCP to manage their care, and encourage these PCPs to take accountability for referrals to specialists. As a result, we hypothesized that enrollment in HMO products would be associated with disproportionately fewer specialist driven low-value services. To test this hypotheses, we extended the previously described HMO vs PPO analysis by stratifying the low-value services into those that were PCP driven and specialist driven, modeling each separately.
To investigate the role of value-based payment, we stratified the MA population by primary care payment model, changing the primary exposure variable to one with 4 values: TM, MA FFS, MA upside-only risk, and MA 2-sided risk. Value-based payment models create accountability for cost and quality outcomes, which we hypothesized would be associated with comparatively fewer rates of low-value services. For this analysis, the MA population was restricted to beneficiaries enrolled in HMO products, because the MA plan largely implemented value-based payment models within HMO products. Finally, to investigate the role of utilization management, we stratified the low-value services according to whether they were subject to PA by the MA plan in 2019 with the hypothesis that those measures subject to PA would be comparatively less prevalent in MA than TM.
For all analyses, we reported regression-adjusted rates of low-value services per 100 beneficiaries per year, as well as adjusted differences between MA and TM beneficiaries. Adjusted rates were estimated using least-squares means, and adjusted differences represented the associated marginal effects. We calculated 95% CIs using robust standard errors.
Analyses were performed using SAS Enterprise Guide, version 8.3 (SAS Institute), from July 2021 to March 2022. All hypothesis tests were 2-sided, with P < .05 indicating significance.
Sensitivity Analyses
We conducted several sensitivity analyses to test the robustness of the study findings. First, we used an additional approach to adjust for observable differences between MA and TM beneficiaries by estimating the propensity to enroll in MA and assigning weights to each observation in the total low-value services model using the inverse probability of treatment weighting (IPTW) approach.17 Second, we used an alternative comorbidity specification, replacing the CCI with the CMS Hierarchical Conditions Category (CMS-HCC) score. Third, to address differences in documentation patterns between MA and TM that could bias comorbidity adjustment,18 we calculated the degree to which CMS-HCC scores in the MA cohort would need to be decreased such that we would observe no significant difference in low-value care between MA and TM beneficiaries. Fourth, to address differences in beneficiary risk and health that may not be captured by demographic characteristics and claims-based comorbidity measures, we followed an approach similar to Curto et al19 in which we predicted CBSA-level, comorbidity-adjusted 2019 mortality separately for the MA and TM beneficiaries in the sample and then used predicted mortality in place of CCI score as a covariate in the model of total low-value services. Fifth, in another attempt to address potential bias from patient selection into MA plans, we stratified beneficiaries according to quintiles of county-level MA penetration in 2019 and modeled total low-value services within each strata with the hypothesis that if significant selection effects existed, they would vary by MA penetration. This analysis also allowed us to explore whether imbalances in the geographic distribution of MA and TM beneficiaries might bias the ascertainment of the overall association between MA enrollment and low-value services. Finally, to address the possibility that lower rates of low-value services in MA could represent indiscriminate utilization differences between the 2 programs, we estimated the association between MA enrollment and appropriate breast cancer screening,20 a high-value service. Additional details on sensitivity analyses are provided in the eMethods in the Supplement.
Results
The study population comprised 1 527 763 MA beneficiaries and 942 436 TM beneficiaries (see eFigure 1 in the Supplement for the attrition diagram). Compared with TM beneficiaries, MA beneficiaries were younger, more likely to be Black, and more likely to have originally qualified for Medicare based on disability (Table 121).
Table 1. Baseline Demographic and Clinical Characteristics.
| Characteristic | No. (%) | SMDa | |
|---|---|---|---|
| Medicare Advantage (n = 1 527 763) | Traditional Medicare (n = 942 436) | ||
| Age category, y | |||
| 65-74 | 780 927 (51.1) | 469 906 (49.9) | 0.025 |
| 75-84 | 585 499 (38.3) | 339 791 (36.1) | 0.047 |
| ≥85 | 161 337 (10.6) | 132 739 (14.1) | 0.107 |
| Sex | |||
| Female | 828 238 (54.2) | 518 539 (55.0) | 0.016 |
| Male | 699 525 (45.8) | 423 897 (45.0) | 0.016 |
| Raceb | |||
| Black | 177 019 (11.6) | 52 088 (5.5) | 0.218 |
| White | 1 286 503 (84.2) | 839 850 (89.1) | 0.145 |
| Other | 47 981 (3.1) | 33 999 (3.6) | 0.026 |
| Unknown | 16 260 (1.1) | 16 499 (1.8) | 0.058 |
| Geographic region | |||
| Northeast | 36 090 (2.4) | 164 852 (17.5) | 0.523 |
| Midwest | 356 404 (23.3) | 220 935 (23.4) | 0.003 |
| South | 1 039 136 (68.0) | 380 811 (40.4) | 0.577 |
| West | 96 133 (6.3) | 175 838 (18.7) | 0.381 |
| Population density | |||
| Rural | 41 479 (2.7) | 25 758 (2.7) | 0.001 |
| Suburban | 265 514 (17.4) | 147 104 (15.6) | 0.048 |
| Urban | 1 220 770 (79.9) | 769 574 (81.7) | 0.044 |
| Original reason for Medicare entitlement | |||
| Age, ≥65 y | 1 326 507 (86.8) | 866 193 (91.9) | 0.165 |
| Disability | 201 218 (13.2) | 75 594 (8.0) | 0.168 |
| ESKD | 34 (<0.1) | 361 (<0.1) | 0.025 |
| Disability and ESKD | 4 (<0.1) | 288 (<0.1) | 0.024 |
| Charlson comorbidity index score, mean [SD] | 1.19 [1.7] | 1.22 [1.7] | 0.022 |
| Product type | |||
| HMO | 696 241 (45.6) | NA | NA |
| PPO | 776 815 (50.8) | NA | NA |
| Other | 54 707 (3.6) | NA | NA |
| Primary care payment arrangementc | |||
| 2-Sided risk | 332 854 (47.8) | NA | NA |
| Upside only risk | 261 806 (37.6) | NA | NA |
| Fee-for-service | 101 581 (14.6) | NA | NA |
Abbreviations: ESKD, end-stage kidney disease; HMO, health maintenance organization; NA, not applicable; PPO, preferred provider organization; SMD, standardized mean difference.
We considered SMDs greater than 0.10 to reflect meaningful differences between groups.21
Race was assessed according to the US Centers for Medicare & Medicaid Services beneficiary race code, which reflects data reported to the Social Security Administration. The Other category includes the following races: Asian, Hispanic, North American Native, and Other.
Primary care payment model was restricted to beneficiaries enrolled in HMO products, because the Medicare Advantage plan largely implements value-based payment models within HMO products.
In adjusted analyses, we found that MA beneficiaries received 9.2% (95% CI, 8.5%-9.8%) fewer low-value services in 2019 than TM beneficiaries (23.1 vs 25.4 total low-value services per 100 beneficiaries) (Table 2). The MA beneficiaries had lower rates of low-value services across all 6 clinical categories, ranging from 5.0% (95% CI, 3.1%-6.8%) fewer low-value cardiovascular tests and procedures to 40.0% fewer (95% CI, 38.2%-41.8%) low-value cancer screening services. At the individual service level, 14 measures were significantly less common in MA, and 4 were significantly less common in TM (Table 2). Unadjusted results were similar (eTable 3 in the Supplement).
Table 2. Adjusteda Differences in Rates of Low-Value Care for Medicare Beneficiaries Enrolled in Medicare Advantage and Traditional Medicare.
| Annual count | Medicare Advantage | Traditional Medicare | Absolute difference (95% CI) | Relative difference % (95% CI) | P value |
|---|---|---|---|---|---|
| All low-value services (per 100 beneficiaries) b | |||||
| All low-value services | 23.07 | 25.39 | –2.33 (–2.50 to –2.15) | –9.2 (–9.8 to –8.5) | <.001 |
| Low-value services by clinical category (per 100 eligible beneficiaries) c | |||||
| Cancer screening | 1.67 | 2.78 | –1.11 (–1.16 to –1.06) | –40.0 (–41.8 to –38.2) | <.001 |
| Diagnostic and preventive testing | 4.97 | 6.49 | –1.53 (–1.64 to –1.42) | –23.5 (–25.2 to –21.8) | <.001 |
| Preoperative testing | 14.17 | 15.05 | –0.88 (–1.27 to –0.48) | –5.8 (–8.5 to –3.2) | <.001 |
| Imaging | 13.65 | 15.17 | –1.52 (–1.66 to –1.39) | –10.0 (–10.9 to –9.1) | <.001 |
| Cardiovascular testing and procedures | 2.93 | 3.08 | –0.15 (–0.21 to –0.10) | –5.0 (–6.8 to –3.1) | <.001 |
| Other surgeries | 0.25 | 0.36 | –0.11 (–0.14 to –0.08) | –30.2 (–39.6 to –20.8) | <.001 |
| Individual low-value services (per 100 eligible beneficiaries) c | |||||
| Cancer screening | |||||
| Cancer screening for patients with CKD receiving dialysis | 0.02 | 0.04 | –0.01 (–0.03 to 0.00) | –37.3 (–78.1 to 4.0) | .08 |
| Cervical cancer screening for women older than 65 y | 1.76 | 3.18 | –1.42 (–1.48 to –1.35) | –44.5 (–46.6 to –42.5) | <.001 |
| Colorectal cancer screening for patients older than 85 y | 0.06 | 0.07 | –0.01 (–0.04 to 0.02) | –14.6 (–52.9 to 24.0) | .46 |
| PSA testing for men older than 75 y | 0.46 | 0.73 | –0.27 (–0.33 to –0.22) | –37.3 (–44.9 to –29.7) | <.001 |
| Diagnostic and preventive testing | |||||
| Bone mineral density testing at frequent intervals | 0.08 | 0.08 | 0.01 (–0.03 to 0.04) | 6.6 (–37.6 to 50.7) | .77 |
| Homocysteine testing for cardiovascular disease | 0.15 | 0.25 | –0.10 (–0.12 to –0.08) | –39.8 (–46.9 to –32.8) | <.001 |
| Hypercoagulability testing for patients with DVT | 0.74 | 1.05 | –0.31 (–0.43 to –0.18) | –29.2 (–41.0 to –17.5) | <.001 |
| PTH measurement for patients with stage 1-3 CKD | 16.31 | 21.89 | –5.58 (–5.95 to –5.20) | –25.5 (–27.2 to –23.8) | <.001 |
| Preoperative testing | |||||
| Preoperative chest radiography | 10.16 | 10.75 | –0.59 (–0.91 to –0.27) | –5.5 (–8.4 to –2.5) | <.001 |
| Preoperative echocardiography | 3.12 | 3.02 | 0.09 (–0.08 to 0.26) | 3.1 (–2.5 to 8.8) | .28 |
| Preoperative PFTs | 0.57 | 0.85 | –0.28 (–0.37 to –0.2) | –33.2 (–43.0 to –23.5) | <.001 |
| Preoperative stress testing | 2.17 | 2.22 | –0.05 (–0.20 to 0.10) | –2.2 (–8.8 to 4.4) | .51 |
| Imaging | |||||
| CT of the sinuses for uncomplicated acute rhinosinusitis | 2.80 | 3.36 | –0.56 (–0.71 to –0.41) | –16.6 (–21.1 to –12.1) | <.001 |
| Head imaging in the evaluation of syncope | 16.88 | 17.46 | –0.58 (–1.13 to –0.04) | –3.3 (–6.5 to –0.20) | .04 |
| Head imaging for uncomplicated headache | 29.75 | 28.70 | 1.05 (0.52 to 1.58) | 3.4 (1.8 to 4.9) | <.001 |
| Electroencephalogram for headaches | 0.26 | 0.31 | –0.05 (–0.11 to 0.00) | –17.0 (–35.3 to 1.1) | .07 |
| Back imaging for patients with nonspecific low back pain | 11.31 | 10.94 | 0.37 (0.20 to 0.54) | 3.7 (1.8 to 5.5) | <.001 |
| Screening for carotid artery disease in asymptomatic adults | 5.40 | 5.84 | –0.44 (–0.52 to –0.37) | –7.6 (–8.9 to –6.3) | <.001 |
| Screening for carotid artery disease for syncope | 5.61 | 5.94 | –0.33 (–0.58 to –0.08) | –5.6 (–9.8 to –1.3) | .01 |
| Cardiovascular testing and procedures | |||||
| Stress testing for stable coronary disease | 10.35 | 10.71 | –0.36 (–0.56 to –0.17) | –3.4 (–5.2 to –1.6) | <.001 |
| PCI with balloon angioplasty or stent placement for stable CAD | 1.29 | 1.01 | 0.28 (0.21 to 0.35) | 27.7 (20.5 to 34.9) | <.001 |
| Kidney artery angioplasty or stenting | 0.03 | 0.03 | 0 (0 to 0.01) | 7.2 (–17.0 to 31.7) | .56 |
| Carotid endarterectomy in asymptomatic patients | 0.07 | 0.06 | 0.01 (0 to 0.02) | 22.7 (2.9 to 42.4) | .02 |
| IVC filters for the prevention of pulmonary embolism | 0.07 | 0.08 | –0.01 (–0.02 to 0) | –14.6 (–24.9 to –4.2) | .01 |
| Other surgeries | |||||
| Vertebroplasty/kyphoplasty for osteoporotic vertebral fractures | 1.60 | 1.74 | –0.14 (–0.3.0 to 0.03) | –8.0 (–17.6 to 1.6) | .10 |
| Arthroscopic surgery for knee osteoarthritis | 0.09 | 0.15 | –0.06 (–0.08 to –0.04) | –40.1 (–51.9 to –27.9) | <.001 |
Abbreviations: CAD, coronary artery disease; CKD, chronic kidney disease; CT, computed tomography scan; DVT, deep vein thrombosis; IVC, inferior vena cava; PCI, percutaneous coronary intervention; PFTs, pulmonary function tests; PSA, prostate-specific antigen; PTH, parathyroid hormone.
Adjusted for beneficiary age, sex, race and ethnicity, original reason for Medicare entitlement, Charlson Comorbidity Index score, population density, and core-based statistical area.
Counts for total low-value services also adjust for eligibility for each of the low-value services included in the overall count.
Counts for low-value services by clinical category and individual low-value services are limited to beneficiaries eligible for the service.
The MA beneficiaries enrolled in HMO and PPO products received fewer low-value services than TM beneficiaries, but the difference was largest for those enrolled in HMO products (2.6 fewer [95% CI, 2.4-2.8] vs 2.1 fewer [95% CI, 1.9-2.3] services per 100 beneficiaries, respectively) (Table 3). Compared with MA beneficiaries enrolled in PPO products, those enrolled in HMO products had 13.0% fewer (95% CI, 11.4%-14.6%) specialist-driven services per 100 beneficiaries compared with 2.8% fewer (95% CI, 1.6%-4.0%) PCP-driven services.
Table 3. Adjusteda Rates of Low-Value Services in Traditional Medicare and Medicare Advantage, With Medicare Advantage Beneficiaries Stratified by Product Type.
| Characteristic | Per 100 beneficiaries | |||||
|---|---|---|---|---|---|---|
| Adjusted rates (95% CI) | Adjusted differences | |||||
| TM | MA HMO (95% CI) | MA PPO (95% CI) | MA HMO vs TM | MA PPO vs TM | MA HMO vs MA PPO | |
| Total low-value services | 25.39 (25.22-25.57) | 22.80 (22.58-23.01) | 23.28 (23.09-23.47) | –2.59b | –2.11b | –0.48b |
| PCP-driven low value services | 15.95 (15.79-16.11) | 13.57 (13.4-13.73) | 13.96 (13.82-14.10) | –2.38b | –1.99b | –0.39b |
| Specialist-driven low value services | 10.61 (10.46-10.76) | 8.19 (8.04-8.34) | 9.42 (9.28-9.55) | –2.42b | –1.19b | –1.23b |
Abbreviations: HMO, health maintenance organization; MA, Medicare Advantage; PPO, preferred provider organization; TM, traditional Medicare; PCP, primary care physician.
Adjusted for beneficiary age, sex, race and ethnicity, original reason for Medicare entitlement, Charlson comorbidity index score, population density, core-based statistical area, and eligibility for each of the low-value services included in the overall count.
Statistical significance at the P < .05 level.
Across primary care payment arrangements, MA beneficiaries received fewer low-value services than TM beneficiaries (Table 4). The largest difference was observed for MA beneficiaries whose PCPs were reimbursed within 2-sided risk arrangements (difference of 3.0 fewer [95% CI, 2.7-3.3] services per 100 beneficiaries). Among MA beneficiaries, those whose PCPs were reimbursed within 2-sided risk arrangements had the lowest rates of low-value care. There was no significant difference between FFS and upside-only risk.
Table 4. Adjusteda Rates of Low-Value Services in Traditional Medicare and Medicare Advantage, With Medicare Advantage Beneficiariesb Stratified by Primary Care Payment Model.
| Services | Per 100 beneficiaries | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Adjusted rates (95% CI) | Adjusted differences | ||||||||
| TM | MA FFS | MA upside-only risk | MA 2-sided risk | MA FFS vs TM |
MA upside-only risk vs TM | MA 2-sided risk vs TM | MA upside-only risk vs MA FFS | MA 2-sided risk vs MA FFS | |
| Total low-value services | 25.39 (25.22-25.57) | 22.92 (22.53-23.30) | 23.01 (22.72-23.32) | 22.38 (22.05-22.71) | –2.47c | –2.37c | –3.01c | 0.09 | –0.55c |
Abbreviations: FFS, fee for service; HMO, health maintenance organization; MA, Medicare Advantage; TM, traditional Medicare.
Adjusted for beneficiary age, sex, race and ethnicity, original reason for Medicare entitlement, Charlson comorbidity index score, population density, core-based statistical area, and eligibility for each of the low-value services included in the overall count.
Population restricted to beneficiaries enrolled in HMO products, because the MA plan largely implements value-based payment models within HMO products.
Statistical significance at the P < .05 level.
When stratifying the low-value services according to whether they were subject to PA by the MA plan, we found that MA beneficiaries received fewer of both sets of services, but that the difference was larger for those services not subject to PA (9.1% fewer [95% CI, 8.3%-9.9%] vs 25.3% fewer [95% CI, 24.0%-26.5%], respectively). Complete regression results are provided in eTable 4 in the Supplement.
Sensitivity Analyses
Using inverse probability of treatment weighting to adjust for differences between the cohorts, using an alternative comorbidity specification, and adjusting for predicted mortality did not meaningfully change the sign or magnitude of the association between MA enrollment and the total count of low-value services. We found that even when CMS-HCC scores for MA enrollees were deflated by 99%, the observed association remained significant. There was no meaningful difference in the association between MA enrollment and low-value care across quintiles of county-level MA penetration. Finally, we found that appropriate breast cancer screening, a measure of high-value care, was slightly more common in MA beneficiaries compared with TM beneficiaries (odds ratio, 1.008 [95 % CI, 1.004-1.012]). eTable 5 in the Supplement presents complete results from the sensitivity analyses.
Discussion
In this cross-sectional study of low-value care within the Medicare program, we found that MA beneficiaries received 9.2% fewer low-value services than TM beneficiaries in 2019 (23.1 vs 25.4 total low-value services per 100 beneficiaries). The MA beneficiaries enrolled in HMO products and those attributed to primary care organizations reimbursed within advanced value-based payment models had the lowest rates of low-value care.
Despite considerable research and policy attention to low-value care in TM, to our knowledge comparatively little is known about low-value care in the MA program. A study by Park et al5 used Medical Expenditure Panel Survey data to compare rates of low-value services in MA and TM and found no statistically differences for 10 of the 11 measures evaluated, which contrasts with our finding of less low-value care among MA beneficiaries. This difference could be explained by the fact that our data reflect the experience of a specific MA insurer, the data sources used to construct the low-value care measures (Medical Expenditure Panel Survey vs claims), the period studied (2006-2015 vs 2019), or the specific low-value care measures used (only 6 measures overlapped). It is also possible that a relatively small study population in the prior study (11 677 TM beneficiaries and 5164 MA beneficiaries) resulted in insufficient power to detect differences between MA and TM.
One unique element of our study was its exploration of the specific elements of insurance design that might moderate the delivery of low-value care within MA. By using enrollment and plan data from the MA insurer, we were able to better understand potential mechanisms underlying the observed association between MA enrollment and lower rates of low-value care. When stratifying the MA population by product type (HMO or PPO), we found that both groups of beneficiaries received fewer low-value services than TM beneficiaries, but that the difference was larger for beneficiaries enrolled in HMO products and largest among specialist driven services. Together, these findings suggest that elements of HMO product design, such as an increased incentive to seek in-network care and an accountable primary care relationship, moderate the association between MA enrollment and low-value care. When stratifying the MA population by primary care payment model, we found that low-value care was less prevalent across MA payment arrangements, but that MA beneficiaries attributed to primary care organizations reimbursed within advance value-based payment models (ie, 2-sided risk) had the lowest rates of low-value care. These findings are consistent with a prior analysis of value-based payment and low-value care in TM7 and suggest that payment reform may be an important lever in reducing low-value care. We did not find evidence that PA moderated the association between MA enrollment and low-value care.
Our findings have several implications for efforts to reduce low-value care. Within the TM program, in which low-value care has remained prevalent despite policy and research attention,5,6 it may be possible to leverage elements of insurance design to encourage the deadoption of low-value services. One opportunity is to expand the penetration of value-based payment models, an approach consistent with the strategy recently articulated by the Center for Medicare and Medicaid Innovation.22 There may also be opportunities to test approaches to create preferred networks, similar to what was originally proposed in the now-terminated Geographic Direct Contracting Model.23 Outside of the TM program, where MA, Medicaid, and commercial plans have more flexibility to experiment with approaches to insurance design, there is an opportunity to refine and test how to optimize network design, product design, utilization management, value-based payment, and other elements of insurance design to reduce the prevalence of low-value care.
Limitations
This study has several limitations. First, we focused on a specific set of low-value care measures observable in claims. Although these measures are well-studied, prevalent, and germane to the Medicare population,2,7,8 they do not account for all low-value care; therefore, our findings may not generalize to other measures sets or specifications. Second, although our composite measure of low-value care captures the aggregate burden of low-value services included in the analysis, it is weighted towards the most prevalent measures. Third, our study was limited to MA beneficiaries enrolled in a single (albeit large and national) MA plan. While this limits the generalizability of our findings to other plans, it did allow us to use data from the plan to better understand how elements of insurance design may moderate the association between MA enrollment and low-value care. Fourth, the stratified analysis of primary care payment model may have been biased toward the null to the extent that we did not account for value-based payment models in TM or possible spillover effects of value-based payment models in MA on TM beneficiaries. Fifth, while we assessed the potential effect of network design indirectly by leveraging the incentive to seek in-network care present in HMO products, we were not able to directly assess how network design in MA moderated the receipt of low-value care. Future research could focus specifically on this issue, comparing rates of low-value care among physicians in and out of network with MA plans. Finally, the study results may be subject to patient selection into MA on unobserved variables. However, it is not clear in what direction this potential bias would be expected to affect rates of low-value services, and the results were robust to several sensitivity analyses focused on possible confounding.
Conclusions
In this cross-sectional study of Medicare beneficiaries, we found that MA beneficiaries received fewer low-value services than TM beneficiaries, especially among MA beneficiaries enrolled in HMO products and those attributed to primary care organizations reimbursed within advanced value-based payment models.
eMethods. Supplemental Methods
eTable 1. Detail on Low-Value Care Measures
eTable 2. Sub-Group Classifications for Low-Value Care Measures
eFigure 1. Participant Flow through the Study
eTable 3. Unadjusted Rates of Low-Value Services
eTable 4. Full Regression Results
eTable 5. Summary of Results from Sensitivity Analyses
eTable 6. Baseline Demographic and Clinical Characteristics Pre- and Post-IPTW
References
- 1.Shrank WH, Rogstad TL, Parekh N. Waste in the US health care system: estimated costs and potential for savings. JAMA. 2019;322(15):1501-1509. doi: 10.1001/jama.2019.13978 [DOI] [PubMed] [Google Scholar]
- 2.Schwartz AL, Landon BE, Elshaug AG, Chernew ME, McWilliams JM. Measuring low-value care in Medicare. JAMA Intern Med. 2014;174(7):1067-1076. doi: 10.1001/jamainternmed.2014.1541 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Carter EA, Morin PE, Lind KD. Costs and trends in utilization of low-value services among older adults with commercial insurance or Medicare Advantage. Med Care. 2017;55(11):931-939. doi: 10.1097/MLR.0000000000000809 [DOI] [PubMed] [Google Scholar]
- 4.Colla CH, Morden NE, Sequist TD, Schpero WL, Rosenthal MB. Choosing wisely: prevalence and correlates of low-value health care services in the United States. J Gen Intern Med. 2015;30(2):221-228. doi: 10.1007/s11606-014-3070-z [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Park S, Jung J, Burke RE, Larson EB. Trends in use of low-value care in traditional fee-for-service Medicare and Medicare Advantage. JAMA Netw Open. 2021;4(3):e211762-e211762. doi: 10.1001/jamanetworkopen.2021.1762 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Mafi JN, Reid RO, Baseman LH, et al. Trends in low-value health service use and spending in the US Medicare fee-for-service program, 2014-2018. JAMA Netw Open. 2021;4(2):e2037328-e2037328. doi: 10.1001/jamanetworkopen.2020.37328 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Schwartz AL, Chernew ME, Landon BE, McWilliams JM. Changes in low-value services in year 1 of the Medicare pioneer accountable care organization program. JAMA Intern Med. 2015;175(11):1815-1825. doi: 10.1001/jamainternmed.2015.4525 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Schwartz AL, Jena AB, Zaslavsky AM, McWilliams JM. Analysis of physician variation in provision of low-value services. JAMA Intern Med. 2019;179(1):16-25. doi: 10.1001/jamainternmed.2018.5086 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Schpero WLMN, Morden NE, Sequist TD, Rosenthal MB, Gottlieb DJ, Colla CH. For selected services, Blacks And Hispanics more likely to receive low-value care than Whites. Health Aff (Millwood). 2017;36(6):1065-1069. doi: 10.1377/hlthaff.2016.1416 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Colla CH, Mainor AJ, Hargreaves C, Sequist T, Morden N. Interventions aimed at reducing use of low-value health services: a systematic review. Med Care Res Rev. 2017;74(5):507-550. doi: 10.1177/1077558716656970 [DOI] [PubMed] [Google Scholar]
- 11.Schwartz AL, Brennan TA, Verbrugge DJ, Newhouse JP. Measuring the scope of prior authorization policies: applying private insurer rules to Medicare Part B. JAMA Health Forum. 2021;2(5):e210859-e210859. doi: 10.1001/jamahealthforum.2021.0859 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Schwartz AL, Chen Y, Jagmin CL, et al. Coverage denials: government and private insurer policies for medical necessity in Medicare. Health Aff (Millwood). 2022;41(1):120-128. doi: 10.1377/hlthaff.2021.01054 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.The Mitre Corporation . APM measurement: progress of alternative payment models. Accessed November 15, 2021. https://hcp-lanorg/workproducts/apm-methodology-2019pdf
- 14.Charlesworth CJ, Meath THA, Schwartz AL, McConnell KJ. Comparison of low-value care in Medicaid vs Commercially Insured Populations. JAMA Intern Med. 2016;176(7):998-1004. doi: 10.1001/jamainternmed.2016.2086 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Humana . Medicare Advantage and dual Medicare-Medicaid plans preauthorization and notification list. Accessed November 15, 2021. https://docushare-webappsexternalpioneerhumanacom/Marketing/docushare-app?file=3483311
- 16.Quan H, Sundararajan V, Halfon P, et al. Coding algorithms for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Med Care. 2005;43(11):1130-1139. doi: 10.1097/01.mlr.0000182534.19832.83 [DOI] [PubMed] [Google Scholar]
- 17.Austin PC. An introduction to propensity score methods for reducing the effects of confounding in observational studies. Multivariate Behav Res. 2011;46(3):399-424. doi: 10.1080/00273171.2011.568786 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Medicare Payment Advisory Commission . Report to the Congress: Medicare payment policy. Accessed November 15, 2021. https://www.medpac.gov/wp-content/uploads/import_data/scrape_files/docs/default-source/reports/mar21_medpac_report_to_the_congress_sec.pdf
- 19.Curto V, Einav L, Finkelstein A, Levin J, Bhattacharya J. Health care spending and utilization in public and private Medicare. Am Econ J Appl Econ. 2019;11(2):302-332. doi: 10.1257/app.20170295 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.National Committee for Quality Assurance . Breast cancer screening. Accessed November 15, 2021. https://www.ncqa.org/hedis/measures/breast-cancer-screening
- 21.Austin PC. Using the standardized difference to compare the prevalence of a binary variable between two groups in observational research. Commun Stat Simul Comput. 2009;38(6):1228-1234. doi: 10.1080/03610910902859574 [DOI] [Google Scholar]
- 22.US Centers for Medicare & Medicaid Services . CMS white paper on CMS Innovation Center’s strategy. Accessed November 15, 2021. https://innovation.cms.gov/strategic-direction
- 23.US Centers for Medicare & Medicaid Services . Geographic direct contracting model. Accessed November 15, 2021. https://innovation.cms.gov/innovation-models/geographic-direct-contracting-model
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
eMethods. Supplemental Methods
eTable 1. Detail on Low-Value Care Measures
eTable 2. Sub-Group Classifications for Low-Value Care Measures
eFigure 1. Participant Flow through the Study
eTable 3. Unadjusted Rates of Low-Value Services
eTable 4. Full Regression Results
eTable 5. Summary of Results from Sensitivity Analyses
eTable 6. Baseline Demographic and Clinical Characteristics Pre- and Post-IPTW