Prior Authorization Requirements in Medicare Advantage and County Social Vulnerability

Abstract

Objective:

To assess whether Medicare Advantage (MA) beneficiaries in socially vulnerable counties face more restrictive prior authorization (PA) requirements.

Study Design:

Cross-sectional analysis of 2,945 U.S. counties in 2022.

Methods:

The primary outcome was the average percentage of Medicare-covered service categories requiring PA across all MA plans in a county, adjusted for service mix and plan enrollment, derived from the Centers for Medicare & Medicaid Services MA benefit, landscape, and enrollment files. Secondary outcomes included the county-level PA rates for selected service categories, the number of plans with more versus less restrictive PA policies, and average monthly premiums. The exposure was the county-level Social Vulnerability Index (SVI) subdomains (socioeconomic status, household characteristics, racial/ethnic minority status, housing and transportation), categorized into deciles. Unadjusted linear regression models were used, weighted by county MA enrollment.

Results:

Counties in the highest versus lowest SVI decile in the socioeconomic domain had 7.3 percentage points (pp) higher overall PA rates (p<0.001), with the largest gap for psychiatric services (16.9 pp, p<0.001). Conversely, average monthly premiums declined across socioeconomic deciles, from $29 to $8 (p<0.001). Total plan counts were stable across deciles, but the composition differed: the most vulnerable counties offered five fewer plans with less restrictive PA policies and four more plans with more restrictive policies (both p<0.001).

Conclusions:

MA beneficiaries in socially vulnerable counties face more restrictive PA requirements, especially for psychiatric services. These disparities may compound existing structural barriers and warrant consideration in future MA policy reforms.

Precis:

This study links county-level social vulnerability to variation in Medicare Advantage prior authorization policies.

Introduction

As of 2024, 54% of Medicare beneficiaries are enrolled in Medicare Advantage (MA).¹ MA is managed care provided by private insurers receiving a monthly capitated payment from Medicare for enrollees. Private insurers offer MA plans through contracts with the Centers for Medicare & Medicaid Services (CMS), and each contract can include multiple plans within a local market that spans a single county or a broader region. MA has grown in popularity for various reasons including offering extra benefits (e.g., vision, hearing, dental), low premiums, annual out-of-pocket cost caps, and aggressive marketing.² At the same time, MA plans use various utilization management tools to control costs, most notably restrictive provider networks and prior authorizations (PAs).³

PA is the process through which a provider obtains approval from an insurer before rendering certain medical services. MA plans use PA more frequently than other insurance programs (e.g., traditional Medicare or Medicaid).^4,5 Nearly all (99%) MA plans require PA for some services, especially relatively expensive services such as hospital and skilled nursing facility stays.⁶ Although PA is designed to reduce unnecessary services and waste, it might create barriers to access, delay necessary care, and increase unmet care needs.^7-10 However, the use of PA varies substantially across MA insurers and plans.^11,12 If individuals in different markets are exposed to and subsequently enroll in plans with varying PA requirements, PA might disproportionally impact certain groups, potentially exacerbating health disparities.

The Social Vulnerability Index (SVI), developed by the Centers for Disease Control and Prevention (CDC), is designed to help identify communities that are susceptible to adverse events such as infectious disease outbreaks or natural disasters. It incorporates a wide range of social determinants of health, including income, education, employment, population density, housing, and racial and ethnic compositions.¹³ Growing evidence shows that social vulnerability is associated with worse health outcomes including disability, comorbidity, and mortality,^14-16 as well as increased health care costs and use.¹⁷ Social vulnerability is also linked to individuals’ access to medical care, treatment adherence, and consistent engagement with the health care system.¹⁸ For instance, individuals in communities with lower education and income levels may lack health literacy or access to transportation, limiting their ability to obtain timely care.¹⁶

For MA beneficiaries living in more socially vulnerable counties, additional structural barriers may arise depending on the types of health plans available to them. Evidence has shown that racial and ethnic minorities are more likely to enroll in MA plans with lower star ratings,¹⁹ driven in part by limited access to high-quality plans in their local markets.²⁰ In addition, counties with higher social vulnerability tend to have fewer highly rated MA plans.^21,22 While these findings speak to disparities in plan quality, it is also plausible that MA plans enter socioeconomically disadvantaged markets with more restrictive PA requirements to control cost, given the relatively high health care spending among beneficiaries in these areas. However, to the best of our knowledge, no study has examined whether socially vulnerable counties are more likely to be served by MA plans with more restrictive PA policies.

This study aims to assess the relationship between county-level social vulnerability and the restrictiveness of PA policies in MA plans. Using multiple publicly available national datasets, we examine whether higher social vulnerability is associated with higher PA rates for Medicare-covered services among MA plans available in each county. Findings will contribute to understanding how MA plan design may reinforce or mitigate disparities in access to care.

Methods

Data and Sample.

The primary data source was CMS MA plan benefits data, submitted annually by plans during the bidding process and detailing PA requirements across service categories. We defined counties as local markets and aggregated MA benefits data at the county level using the CMS MA landscape file, which lists all plans available in each county as well as each plan’s monthly premiums and parent organization. These data were then linked to the CDC’s SVI and the MA enrollment file. All data were from 2022; benefits data were from the third quarter.

We excluded counties with incomplete information, resulting in 2,945 counties across all 50 U.S. states and the District of Columbia. Consistent with prior studies,^12,22 we excluded MA plans limited to specific populations or subject to different payment regulations, including Special Needs Plans, Medicare-Medicaid plans, Program of All-Inclusive Care for the Elderly plans, employer-sponsored plans, Part B-only plans, and cost plans.

Prior Authorization Rates.

The main outcome was the average PA rate for in-network Medicare-covered services across all MA plans in each county. We first calculated the percentage of service categories requiring PA for each plan using a set of binary indicators from the plan benefits data, where 1 indicated PA was required and 0 otherwise. We included 34 Medicare-covered service categories, excluding supplemental benefits to ensure consistency in the denominator across plans. Each service category was weighted by the corresponding Medicare fee-for-service (FFS) spending from CMS’s Provider Summary by Type of Services files, giving greater weights to more costly or commonly used services (e.g., inpatient hospital) relative to lower-cost or rarely used services (e.g., digital rectal exam). We assumed the relative distribution of spending across service categories between FFS and MA is comparable, making FFS spending a valid proxy for weighting. The list of service categories, relevant procedure codes used to define each category, corresponding FFS spending, and final weights is provided in eAppendix Table 1. We then averaged PA rates across all plans in each county, weighted by plan enrollment, to capture the average restrictiveness of plans to which enrollees are exposed.

We further examined county-level PA rates by service category, defined as the percentage of plans requiring PA for that service within a county, weighted by plan enrollment. Since there are 34 distinct service categories, we followed Neprash et al. to select four sentinel categories based on cost (inpatient acute hospital services), access concerns (psychiatric services), rapid growth in PA use (diagnostic procedures, tests, and lab services), and high prevalence of PA requirements (Part B drugs).¹²

Finally, we examined the distribution of MA plans in each county, including the total number of plans, and the number with less or more restrictive PA policies. Plans were classified as having less or more restrictive PA policies if their service mix- and enrollment-adjusted PA rates fell into the first (requiring PA for 0–80% of services, with a mean of 42%) or fifth (requiring PA for 99–100% of services, with a mean of 99%) national quintile, respectively.

Monthly premiums.

We calculated the average monthly consolidated premiums (Parts C and D) across all MA plans for each county, adjusted by plan enrollment.

Social Vulnerability Index.

Our primary exposure was the county-level SVI by subdomains (i.e., socioeconomic status, household characteristics, racial and ethnic minority status, and housing and transportation; eAppendix Table 2), measured using national percentile rankings and grouped into deciles, with the first representing the least vulnerable and the tenth the most.

Statistical Analysis.

We first described average PA rates and monthly premiums by SVI deciles, grouped into low (deciles 1–3), moderate (deciles 4–7), and high (deciles 8–10) vulnerability. Differences across groups were assessed using F-tests. Analyses were weighted by county MA enrollment so that larger counties contributed proportionally more. To visualize geographic variation and potential spatial correlation, we overlaid county-level SVI scores and average PA rates on a national map. For brevity, we present results based on the socioeconomic status SVI domain; findings for the other domains were consistent.

Next, we examined the association of county-level SVI deciles with average PA rates using linear regression models, weighted by county MA enrollment. Predicted values from the linear models were presented. We also conducted several secondary analyses. First, we repeated the analysis by service category to assess whether the association varied across services. Second, we examined the association of SVI with average monthly premiums to explore potential tradeoffs faced by beneficiaries. Third, we examined the association between SVI and plan availability (overall and by restrictiveness of PA policies) to evaluate whether the association was driven by specific types of plans. Finally, to determine whether these patterns reflect variations in PA policies within individual MA carriers, we examined the association between SVI and average PA rates at the carrier level, weighted by the number of enrollees per carrier.

Sensitivity Analyses.

We conducted five sensitivity analyses. First, we used the PA rate without adjustment for service mix, applying equal weights across the 34 service categories, since any PA requirement can introduce delays or deter care regardless of a service’s fiscal importance. Second, we adjusted our regression models by county-level characteristics from the Area Health Resources Files not captured by SVI but likely associated with variations in MA markets, such as total population, number of active doctors of medicine, hospital beds, and skilled nursing facility beds, and rurality. Third, we excluded counties in the lowest or highest 1 percent of population size (N=58) to examine whether our findings were driven by very small or large counties. Fourth, we stratified the sample by rurality to assess whether the association differed across metropolitan and non-metropolitan areas. Fifth, to account for market competition, we stratified the sample by counties with three or fewer versus more than three insurance carriers.

Results

Sample Characteristics.

Compared to less socioeconomically vulnerable counties (deciles 1–3), highly vulnerable counties (deciles 8–10) had MA plans with higher average PA rates, adjusted for service mix and plan enrollment (92% vs. 87%; both p<0.001; Table 1). This pattern held across service categories, with smaller differences for diagnostic procedures/tests/lab services (92% vs. 90%; p=0.036) and larger differences for psychiatric services (89% vs. 77%; p<0.001). Consistently, high-SVI markets had fewer permissive plans (6 vs. 9; p<0.001) and more restrictive plans (13 vs. 7; p<0.001). In contrast, average monthly premiums were lower in high-SVI counties ($9 vs. $29; p<0.001). A bivariate map showed that counties characterized by both high vulnerability and high PA rates were concentrated in Southern states (eAppendix Figure 1).

Table 1.

County-level prior authorization rates and other characteristics by Social Vulnerability Index deciles

	SVI deciles (socioeconomic domain)
	1–3 (low vulnerability)	4–7 (moderate vulnerability)	8–10 (high vulnerability)	P-value
Number of counties	885	1,173	887
MA plan PA rates
Overall (%), unadjusted	63.1 (10.9)	66.0 (8.6)	69.6 (6.2)	<0.001
Overall (%), service mixadjusteda	87.4 (9.8)	89.7 (7.3)	92.4 (5.6)	<0.001
By service category
Acute hospital services (%)	96.5 (10.4)	97.7 (7.3)	98.2 (5.2)	<0.001
Psychiatric services (%)	77.2 (24.2)	82.9 (21.3)	88.7 (16.6)	<0.001
Diagnostic procedures/tests/lab services (%)	90.2 (20.5)	92.3 (17.5)	91.5 (15.7)	0.036
Part B drugs (%)	97.3 (8.7)	98.6 (5.9)	99.2 (4.4)	<0.001
Plan availability
Total no. of MA plans	37.8 (14.5)	40.2 (16.4)	43.3 (16.7)	<0.001
No. of less restrictive plans (1st quintile of PA rates)b	8.5 (6.2)	7.5 (6.5)	5.7 (3.9)	<0.001
No. of more restrictive plans (5th quintile of PA rates)b	7.3 (4.0)	9.1 (7.9)	12.7 (10.4)	<0.001
Monthly consolidated premiums ($)	29.3 (22.8)	18.2 (16.5)	9.2 (9.9)	<0.001

NOTES: MA, Medicare Advantage; PA, prior authorization; SVI, Social Vulnerability Index.

^aThe service mix-adjusted PA rate is calculated as the percentage of service categories requiring PA, weighted by the fee-for-service Medicare payment for each category.

^bMore and less restrictive plans are defined as those in the fifth and first quintiles of service mix- and enrollment-adjusted PA rates across all plans nationwide, respectively. In more restrictive plans, 99% to 100% of service categories (mean=99%) require PA, relative to 0% to 80% (mean=42%) in less restrictive plans.

Sample means account for plan size and county MA enrollment, so that larger plans and counties contribute proportionally more. Values are means with standard deviations in parentheses. Differences across SVI deciles were assessed using F-tests.

Prior Authorization Rates.

The average PA rates increased most notably across the socioeconomic SVI domain, from 86% in the first decile to 94% in the tenth decile (Figure 1), a difference of 7.3 percentage points (pp; p<0.001; eAppendix Table 3). The household characteristics domain showed a smaller difference of 5.5 pp (p<0.001). Estimates for other domains did not show consistent patterns.

Figure 1. County-level average prior authorization rate by Social Vulnerability Index deciles.

NOTES: PA, prior authorization.

The outcome is the service mix-adjusted PA rate, calculated as the percentage of service categories requiring PA, weighted by the fee-for-service Medicare payment for each category. Rates are averaged across all plans in each county, with each plan weighted by its enrollment. Regressions are weighted by county-level MA enrollment so that larger counties contribute proportionally more. Point estimates represent predicted PA rates for each decile, with 95% confidence intervals shown as error bars. The dashed line indicates the predicted value for the least vulnerable counties.

Service Categories.

Consistent with the main findings, associations were generally strongest for the socioeconomic domain across service categories, except for diagnostic procedures/tests/lab services, where the household characteristics domain showed the largest increase in PA rates (82% to 96% from the first to tenth decile; p<0.001; eAppendix Figures 2-5). For psychiatric services, large differences were observed in both the socioeconomic domain (16.9 pp, 75% to 92%; p<0.001) and the household characteristics domain (17.3 pp, 73% to 90%; p<0.001). For acute hospital services and Part B drugs, which already have high PA use, increases in PA rates across socioeconomic SVI deciles were smaller (5.1 pp and 4.5 pp, respectively; both p<0.001).

Monthly Premiums.

Premiums showed a pattern opposite to PA rates. In the socioeconomic domain, average monthly premiums declined from $29 in the first decile to $8 in the tenth (p<0.001; Figure 2 and eAppendix Table 4). Similar declines were observed in the household characteristics and racial/ethnic minority domains, though the magnitudes were smaller ($19 and $9 lower, respectively; both p<0.001).

Figure 2. County-level mean monthly premium by Social Vulnerability Index deciles.

NOTES: The outcome is the average monthly consolidated premium (Parts C and D) across all plans in a county, weighted by the number of enrollees per plan. Regressions are weighted by county-level MA enrollment so that larger counties contribute proportionally more. Point estimates represent predicted premiums for each decile, with 95% confidence intervals shown as error bars. The dashed line indicates the predicted value for the least vulnerable counties.

Plan Availability.

For brevity, we present findings only for the socioeconomic domain. The total number of MA plans was stable across SVI deciles (Figure 3 and eAppendix Table 5). However, the composition by PA restrictiveness differed. Compared to the least vulnerable counties, the most vulnerable counties offered five fewer plans with more permissive PA policies (p<0.001) and four more plans with more restrictive policies (p<0.001).

Figure 3. Number of plans with more vs. less restrictive prior authorization policies per county by socioeconomic status Social Vulnerability Index deciles.

NOTES: MA, Medicare Advantage; PA, prior authorization.

Estimates are predicted number of MA plans (overall and by restrictiveness of PA policies) calculated from separate linear regression models. More and less restrictive plans are defined as those in the fifth and first quintiles of service mix-adjusted PA rates across all plans nationwide, respectively. In more restrictive plans, 99% to 100% of service categories (mean=99%) require PA, relative to 0% to 80% (mean=42%) in less restrictive plans. Regressions are weighted by county-level MA enrollment so that larger counties contribute proportionally more. Error bars indicate 95% confidence intervals.

Distributions by Insurer.

We observed substantial variation in PA policies across insurers. UnitedHealthcare, the largest by market share, applied the most restrictive policies, with 99% of service categories requiring PA across socioeconomic SVI deciles, whereas Kaiser Permanente applied a relatively lower PA rate (77%; eAppendix Table 6). Overall, the positive association between SVI and PA rates persisted at the insurer level (Figure 4). Notably, this association appears to be driven by smaller insurers.

Figure 4. Average Social Vulnerability Index and service mix-adjusted prior authorization rates across Medicare Advantage carriers.

NOTES: PA, prior authorization; SVI, Social Vulnerability Index.

Each dot represents an individual insurance carrier (e.g., Blue Cross Blue Shield of Michigan and Blue Cross Blue Shield of Minnesota are grouped under Elevance; WellCare and Lasso are grouped under Centene). Marker size reflects the number of enrollees, and only the seven largest carriers (accounting for 77% of Medicare Advantage enrollees in 2022) are labeled. The red line shows the best-fit regression line, weighted by total MA enrollment per carrier. The service mix-adjusted PA rate is calculated as the percentage of service categories requiring PA, weighted by the fee-for-service Medicare payment for each category.

Sensitivity Analyses.

Findings were robust to multiple sensitivity analyses: 1) using PA rates unadjusted for service mix, 2) adjusting for county characteristics, 3) excluding counties with very small or large populations, 4) stratifying by rurality, and 4) stratifying by three or fewer versus more than three insurance carriers in the county (see eAppendix Figures 6 through 12).

Discussion

This cross-sectional national study provides descriptive evidence that MA beneficiaries in more socially vulnerable counties faced more restrictive PA requirements, particularly for psychiatric services. These patterns reflect differences in plan availability and insurer-level policies, with fewer permissive plans, more restrictive plans, and generally stricter PA policies among insurers in high-SVI markets. The associations were concentrated in the socioeconomic domain, indicating that beneficiaries in counties with lower income and educational attainment may be disproportionately exposed to restrictive PA policies. We also observed that enrollees in high-SVI counties tended to face lower premiums, suggesting tradeoffs that beneficiaries may face when choosing plans.

The results highlight important and previously understudied disparities for MA beneficiaries. Higher SVI has been associated with reduced access to care, greater prevalence of chronic conditions such as cancer, cardiovascular disease, and Alzheimer’s disease, as well as increased health care spending driven by higher rates of postoperative complications, emergency department visits, and readmissions.^{16,17,24-26,18} This study demonstrates that, among MA beneficiaries, restrictive PA policies may compound existing disparities. In more vulnerable counties, both beneficiaries and providers may face greater administrative burdens in filing PA requests and navigating denials and appeals, and the risk of delayed or deferred care—each of which may contribute to poorer outcomes and greater downstream service use.^27,28

Of the four service categories examined in detail, psychiatric services showed the largest gap in average PA rates across SVI deciles. Psychiatric services also experienced rapid growth in PA use between 2009 and 2019.¹² Prior studies have documented substantial disparities in access to mental health services, care quality, and out-of-pocket costs among MA enrollees with mental health conditions, both within MA and compared to traditional Medicare.^29-32 Notably, psychiatrist networks in MA plans are much narrower than those in Medicaid managed care and Affordable Care Act Marketplaces.³³ This study suggests that, coupled with these barriers, variation in PA policies may further aggravate disparities in access to and use of psychiatric services for MA beneficiaries in more vulnerable counties.

These findings contribute to the limited body of literature on geographic variation in MA plans’ use of PA and its potential role in exacerbating disparities. Existing studies on PA have focused on characterizing its overall use^11,34 or assessing its effects on specific services—primarily prescription drugs,^5,35-37 and to a lesser extent, home health,¹⁰ dental care,³⁸ imaging,³⁹ and radiology.⁸ One related study documented significant geographic variation in the proportion of MA enrollees subject to PA and found that certain geographic characteristics (i.e., poverty rates, racial composition, rurality, and hospital market concentration) were associated with PA exposure, though these associations weakened from 2009 to 2019.¹² Our study expands on this study by using composite measures of social vulnerability that not only capture some of these characteristics but also incorporate a broader range of social determinants of health—including education, employment, insurance status, housing cost burden, language proficiency, household crowding, and transportation. This approach informs how PA policies vary across counties defined by multidimensional vulnerability, and how such variation could affect access for populations facing overlapping structural disadvantages. In addition, our study uses more recent data from 2022, which is important given the rapid evolution of the MA market over time.

With the increasing use of PA and growing calls for reform in MA,^7,12 it is critical to assess how PA affects socially vulnerable populations. Ongoing regulatory efforts—such as CMS’s proposed rule mandating a 7-day turnaround for standard PA decisions⁴⁰—may mitigate some adverse effects. Our findings suggest CMS should monitor not only overall PA burden but also whether requirements disproportionately affect enrollees in socially vulnerable areas. Policymakers could also consider incorporating social vulnerability into MA payment policy. For example, raising benchmarks in socially vulnerable counties could create stronger financial incentives for plans to participate in these markets and potentially offer more generous benefits, including more permissive PA practices. In parallel, given the importance of plan choice in MA, existing decision-support tools (e.g., CMS Plan Finder) and resources (e.g., State Health Insurance Assistance Programs) may consider incorporating PA requirements as a key plan attribute to help beneficiaries choose optimal coverage. This may be particularly valuable for beneficiaries in socially vulnerable counties, who often have greater care needs but limited health literacy and financial resources.

Limitations.

This study has several limitations. First, MA plan benefits data only indicate whether PA is required within broad service categories, without information on its frequency or specific services subject to these requirements. Second, the aggregated county-level analysis cannot determine whether PA is beneficial or harmful to individual beneficiaries. Despite criticisms that PA may increase administrative burden and delay necessary care, beneficiaries continue to select plans with these requirements,¹² likely weighing them against other plan attributes.²³ Indeed, we found that counties with higher social vulnerability—where PA requirements are more restrictive—also tended to have lower premiums, suggesting a tradeoff between access and cost. Individual-level data are needed to assess enrollment choices and how PA policies actually affect enrollees’ access and use. Third, since our analysis is purely descriptive, we cannot infer causality.

Conclusions

As MA enrollment continues to rise, it is critical to understand how PA policies shape care delivery, particularly for vulnerable populations. We found that higher social vulnerability was associated with higher PA rates across U.S. counties. These findings suggest that beneficiaries in socioeconomically disadvantaged areas may experience greater administrative burdens, care delays, and poorer outcomes due to more restrictive PA requirements of available plans. Further research is needed to examine how PA policies influence beneficiary plan selection and individual-level outcomes, including care experiences, service use, and health outcomes.

Takeaway points:

Prior authorization (PA) is widely used by Medicare Advantage (MA) plans to manage costs. This study found that MA plans in more socially vulnerable counties imposed PA on a higher share of Medicare-covered services, raising concerns about equitable plan design.

• MA plans in socially vulnerable counties impose more restrictive PA policies.

• These counties offer fewer plans with less restrictive PA policies and more plans with more restrictive policies.

• PA for psychiatric services shows the strongest association with social vulnerability.

• Enrollees in high-vulnerability counties tended to face lower premiums, suggesting tradeoffs beneficiaries may face when selecting plans.