Abstract
Objectives:
To address concerns that post-acute cost-sharing may deter high-need beneficiaries from participating in Medicare Advantage (MA) plans, the Centers for Medicare and Medicaid Services (CMS) have capped cost-sharing for skilled nursing facility services (SNF) in MA plans since 2011. This study examines whether SNF use, inpatient use, and plan disenrollment changed following stricter regulations in 2015 that required most MA plans to eliminate or substantially reduce cost-sharing for SNF care.
Design:
Difference-in-differences retrospective analysis from 2013–2016.
Setting:
MA plans.
Participants:
3.1 million MA members in 320 plans with mandatory cost sharing reductions and 261 plans without such reductions.
Measurements:
Mean monthly number of SNF admissions, SNF days, hospitalizations, and plan disenrollees per 1,000 members
Results:
Mean total cost-sharing for the first 20 days of SNF services decreased from $911 to $104 in affected plans. Relative to concurrent changes in plans without mandated cost-sharing reductions, plans with mandatory cost-sharing reductions experienced no significant differences in the number of SNF days per 1,000 members (adjusted between-group difference: 0.4 days per 1,000 members [95% CI: −5.2, 6.0, p=0.89]), small decreases in the number of hospitalizations per 1,000 members (adjusted between-group difference: 0.6 admissions per 1,000 members [95% CI: −1.0, −0.1, p=0.03]), and small decreases in the number of SNF users who disenrolled at year end (adjusted between-group difference: −16.8 disenrollees per 1,000 members [95% CI: −31.9, −1.8, p=0.03]).
Conclusions:
Mandated reductions in SNF cost-sharing may have curbed selective disenrollment from MA plans without significantly increasing use of SNF services.
Introduction
Medicare Advantage (MA) plans currently enroll more than one-third of Medicare beneficiaries,1 but concerns persist about coverage for MA members in poor health.2 Compared to healthier enrollees, MA members with high-cost service use more frequently switch to traditional Medicare, suggesting potential dissatisfaction with MA coverage.3–6 Although MA plans may offer more benefits than traditional Medicare, like fitness memberships, MA members may have higher out-of-pocket costs for skilled nursing facility (SNF) and hospital care than traditional Medicare beneficiaries.7 While traditional Medicare has no cost-sharing for the first 20 days of SNF care, MA members’ estimated out-of-pocket costs for a 20-day SNF stay averaged $467-$687.7 Higher out-of-pocket costs for SNF services in MA may prompt beneficiaries who need these services to seek traditional Medicare coverage, thus perpetuating favorable selection in which sicker beneficiaries exit MA plans and healthier beneficiaries remain enrolled.
The Affordable Care Act included measures to promote greater parity in SNF cost-sharing between MA and traditional Medicare. Accordingly, the Centers for Medicare and Medicaid Services (CMS) implemented new caps on MA plans’ SNF cost-sharing in 2011.8 Cost-sharing caps for the first 20 days of a SNF stay varied depending on plans’ out-of-pocket limits. CMS capped copayments at $50 per day in 2011 for plans with higher out-of-pocket limits (>$3400), then lowered this cap to $0 by 2015. Daily copayment caps for plans with lower out-of-pocket limits (≤ $3400) decreased from $100 in 2011 to $40 by 2015. Since 2011, MA plans can charge no more than traditional Medicare for days 21 and beyond of SNF services.
CMS’ regulations may financially protect sicker beneficiaries and provide incentives for post-acute users to remain enrolled in MA plans. However, lowering SNF cost-sharing may also increase SNF use and related Medicare spending. If SNF services substitute for inpatient care, increased SNF use could reduce the demand for inpatient services. Prior evaluations of SNF cost-sharing produced mixed findings: SNF use declined when MA plans voluntarily increased copayments7 but neither SNF or inpatient use changed when CMS first implemented cost-sharing caps in 2011.8 CMS announced that plans may gain more flexibility to increase SNF cost-sharing,9 raising questions about the impact of the most recent SNF cost-sharing reductions.
The primary objective of this difference-in-differences study is to examine whether mandatory cost-sharing reductions were associated with changes in SNF use. Secondary objectives included examining changes in inpatient use and plan disenrollment after the mandatory SNF cost-sharing reductions.
Methods
Study Data
Data sources included MA plan benefit information, the Medicare Master Beneficiary Summary File (MBSF), Minimum Data Set (MDS), and MedPAR for years 2012–2016. Plan benefit information identified plans subject to cost-sharing caps. The MBSF detailed beneficiaries’ MA participation, Medicaid enrollment, and demographic characteristics. To measure SNF use, we analyzed admission and discharge dates from MDS assessments, which Medicare or Medicaid-certified nursing homes complete for every patient regardless of payer source.10 MedPAR inpatient data identified MA discharges from most hospitals (approximately 92% of all Medicare discharges11).
Population
We identified MA plans that operated throughout the study period (2012–2016). A MA organization’s contract may include multiple plans with different cost-sharing requirements; we counted each plan within a contract as a separate unit. Within selected plans, our primary study population for utilization outcomes included continuously enrolled plan members. We require members to remain enrolled in the same study plan from January 2012 until the study period ended (December 2016), death, or dual-eligible participation that included Medicaid cost-sharing coverage. Sicker beneficiaries, who are more like to use SNF services, may have been more likely to join plans after SNF copayments were lowered. By examining continuously enrolled beneficiaries who joined plans before the policy change, we could focus on whether utilization changes reflected price sensitivity to SNF cost-sharing among existing members rather than changes in selective entry into study plans.
However, this approach will be biased towards finding increased use of SNF services if SNF users were more likely to remain in study plans after SNF cost-sharing reductions. To assess whether selective disenrollment biased findings, we examined plan disenrollment over the study period among all members enrolled in study plans in January 2012. This outcome is also important because the cost-sharing caps aimed to reduce selective disenrollment from MA among sicker beneficiaries.
Because we identified MA enrollment via Part D plan identifiers, we excluded plans without Part D benefits. We also excluded plans that limited enrollment to specific populations, such as Special Needs Plans. Among the 3,918 plans operating in 2013, we excluded 1,609 plans that did not operate for the entire study period, 1,548 plans with limited enrollment, 4 plans with missing SNF cost-sharing information, 161 plans without Part D, and 3 plans with no identified members in the MBSF (Appendix Exhibit 1).
Among 5.1 million beneficiaries enrolled in selected plans in January 2012, we excluded those under age 21, residing outside the 50 states or the District of Columbia, or dual-eligible with Medicaid cost-sharing coverage (n=370,328). After excluding 27,570 members from 45 plans with fewer than 100 members, the final cohort included 581 study plans with 3.1 million continuously enrolled members for analyses of SNF and inpatient use. Analyses of disenrollment included those members, plus 1.6 million members who disenrolled from study plans after January 2012.
Main Exposure
The main exposure was whether a plan had mandatory SNF cost-sharing reductions between 2013 and 2015. For the study period’s first two years, plans with out-of-pocket limits between $3,401 and the maximum allowable limit of $6,700 could charge no more than $50 per day for the first 20 SNF days. CMS lowered this cap to $25 per day in 2014 and $0 per day in 2015. Plans with out-of-pocket limits $3,400 or below could charge no more than $100 per day in 2012 and 2013. This cap dropped to $50 and $40 per day in 2014 and 2015, respectively.
If a plan’s SNF copayments in 2013 exceeded the 2015 cap, we identified the plan as having mandatory cost-sharing reductions. This group included plans that had any SNF cost-sharing in 2013 for the first 20 days and an out-of-pocket limit between $3,401 – $6,700 plus plans with more than $800 in total SNF cost-sharing for the first 20 days and an out-of-pocket limit no greater than $3,400.
Main Outcomes
Because cost-sharing requirements varied at a plan level, we created plan-month level measures of outcomes. The main utilization outcomes were the number of SNF days (up to 100 days per admission), SNF admissions, inpatient days, and inpatient admissions (including readmissions) per month per 1,000 plan members. SNF and inpatient admissions were assigned to admission month; days were assigned to month of use. For example, a stay spanning December to February counted as 1 December admission and the respective number of days in December, January, and February. Analyses of inpatient and SNF days excluded the first three study months as we did not observe days from admissions that started at the end of 2011 and extended into January - March 2012.
To examine disenrollment among all January 2012 study plan members, we created a plan-level measure of plan exit: the number of plan members each month who exited their plans in the subsequent month divided by the total number of plan members. We scaled this measure to report it per 1,000 plan members. This measure excluded beneficiaries (n=133,003) in counties where plans no longer offered coverage. Like utilization analyses, beneficiaries were censored in disenrollment analyses after gaining Medicaid cost-sharing coverage. Due to MA lock-in rules, most MA plan exits occur at year-end. Accordingly, we modeled results at a month-level separately for disenrollment at the end of December (1 observation per plan per year) and disenrollment for months January through November (11 observations per plan per year). Disenrollment between December 2016 and January 2017 was not observed so December results are only analyzed through December 2015.
Additional Measures
We described plan members’ age, sex, race, and any income-based financial assistance for Part D coverage (the Part D Low Income Subsidy program) or Medicare premiums (dual-eligible beneficiaries with limited Medicaid benefits). Plan-level characteristics included premiums, out-of-pocket limit amounts, total cost-sharing for a twenty-day SNF stay, and total cost-sharing for a six-day inpatient stay.
Analysis Approach
We performed difference-in-differences analyses to compare changes in SNF use, inpatient use, and disenrollment rates among plans with and without mandatory cost-sharing reductions. Caps were fully implemented by January 1, 2015. We excluded the year 2014 from analyses as a transition year. We created two indicator variables, one identifying years after cap reductions (2015 and 2016) and another identifying plans with mandatory cost-sharing reductions. The interaction of these variables is the difference-in-differences estimate of how much of the change in each outcome can be attributed to mandatory cost-sharing reductions. All monthly plan rate outcomes were modeled using generalized linear models weighted by the number of plan members per month with standard errors clustered at the plan level. Plan fixed effects controlled for time-invariant differences across MA plans. A linear monthly trend variable adjusted for contemporaneous changes in outcomes over time. Calendar month fixed effects adjusted for seasonal changes. In analysis of December disenrollment, each plan only had one observation per year so we included an annual trend variable instead of calendar month fixed effects.
This approach assumed that plans with mandatory cost-sharing reductions would have had similar changes in outcomes before and after the policy change as plans without mandatory cost-sharing reductions if the policy change had not happened. While we cannot directly test this assumption, we tested whether plans with and without mandatory cost-sharing reductions had similar trends in outcomes in the two years prior to 2014. These models included an indicator variable for plans with mandatory cost-sharing changes, a linear monthly trend variable, an interaction between these variables, and plan and calendar month fixed effects. A statistically insignificant interaction term would indicate no difference in outcome trends between plan comparison groups prior to the policy change.
Sensitivity analyses
To understand whether populations with more frequent SNF use or limited financial resources were more affected by cost-sharing reductions, we performed stratified exploratory analyses for several groups: beneficiaries age 80 and above, beneficiaries with financial assistance for Part D coverage or Medicare premiums, beneficiaries with prior inpatient or SNF use in the past six months, and beneficiaries with inpatient admissions in the past year for joint replacement, hip fracture, or sepsis. These diagnoses are the three most common diagnoses for MA beneficiaries discharged to a SNF. For populations that required retrospective claims data for identification, we analyzed observations from July 2012 onwards (for recent SNF or inpatient use in past 6 months) or January 2013 onwards (for joint replacement, hip fracture, or sepsis in past 12 months). To explore whether eliminating copayments had a larger influence on outcomes, we separately analyzed results for plans that reduced copayments to $0 versus $50 per day. We also assessed whether our results changed when the study population was not limited to continuously enrolled beneficiaries. That analysis included all individuals enrolled in a study plan in January or their first month of Medicare benefits during any study year.
Finally, to assess if the impact of the policy varied over time, we estimated an event study for each utilization outcome with quarter-by-year fixed effects and interaction terms for each quarter and the indicator for plans with mandatory reductions.
Results
Characteristics of Treatment and Control Plans
Our study included 320 plans with mandatory SNF cost-sharing reductions and 261 plans without mandatory reductions, because their SNF cost-sharing levels in 2013 already met the caps effective in 2015 (Table 1). These plans had 1,294,368 and 1,608,487 members as of January 2013, respectively. Over three-quarters (78.0%) of beneficiaries in plans with mandatory cost-sharing reductions had out-of-pocket limits between $3,401 and $6,700 compared to 16.0% of beneficiaries in plans without mandatory reductions. From 2013 to 2015, average total cost-sharing for the first 20 SNF days dropped from $911 to $104 in plans with mandatory cost-sharing reductions and from $263 to $86 in plans without mandated reductions (Figure 1). Over the same years, total cost-sharing for a 6-day inpatient stay and premiums increased modestly in both plans with and without mandatory SNF cost-sharing reductions but plans without mandatory SNF cost-sharing reductions had larger increases in out-of-pocket limits (Appendix Exhibit 2).
Table 1:
Characteristics of Medicare Advantage plan members for plans with and without mandatory skilled nursing facility cost-sharing reductions, January 2013
| Plans with Mandatory Cost-Sharing Reductions | ||
|---|---|---|
| Yes | No | |
| Number of plans | 320 | 261 |
| Number of beneficiaries | 1,294,368 | 1,608,487 |
| Age category (%): Under age 65 | 8.7 | 5.6 |
| Age 65–79 | 61.5 | 61.4 |
| Age 80 and above | 29.8 | 33.0 |
| Sex (%): Male | 43.9 | 42.7 |
| Female | 56.1 | 57.3 |
| Race or ethnicity (%): White | 80.3 | 73.2 |
| African American | 9.6 | 6.6 |
| Hispanic | 6.8 | 14.2 |
| Asian American or other | 3.3 | 6.0 |
| Has Part D Low-Income Subsidy or partial Medicaid (premium coverage) (%): No | 90.1 | 92.4 |
| Yes | 9.9 | 7.6 |
| Census Region (%): New England | 7.0 | 3.1 |
| Mid-Atlantic | 20.1 | 11.2 |
| East North Central | 13.1 | 11.4 |
| West North Central | 4.8 | 4.8 |
| South Atlantic | 20.8 | 11.0 |
| East South Central | 5.3 | 0.2 |
| West South Central | 8.8 | 11.7 |
| Mountain | 7.4 | 9.2 |
| Pacific | 12.8 | 37.4 |
| Out-of-pocket limit category (%): Mandatory (Max $6,700) | 78.0 | 16.0 |
| Voluntary (Max $3,400) | 22.0 | 84.0 |
| Monthly plan premium, Mean $ (IQR) | 17 (0, 33) | 36 (0, 72) |
| Annual out-of-pocket limit, Mean $ (IQR) | 4,897 (3,700, 6,700) | 3,451 (3,400, 3400) |
| Total SNF cost-sharing, Days 1–20 Mean $ (IQR) | 911 (650, 1,000) | 263 (0, 500) |
| Total Inpatient cost-sharing, Days 1–6 Mean $ (IQR) | 1,111 (750, 1500) | 706 (200, 1,140) |
Note: Authors’ analysis of Master Beneficiary Summary File and data on Medicare Advantage plans’ benefit packages. Plans with mandatory cost-sharing reductions had to decrease copayments for the first 20 days of a skilled nursing facility stay between 2013 and 2015 to comply with reduced caps from the Centers for Medicare and Medicaid Services. SNF: Skilled nursing facility; IQR: Interquartile range
Figure 1. Expected cost-sharing for first 20 days of skilled nursing facility stay by out-of-pocket limit amount and mandatary cost-sharing reductions.
Note: Authors’ analysis of Master Beneficiary Summary File and data on Medicare Advantage plans’ benefit packages. Plans with mandated cuts had to decrease copayments for the first 20 days of a skilled nursing facility stay between 2013 and 2015 to comply with reduced caps from the Centers for Medicare and Medicaid Services. By 2015 plans with out-of-pocket limits $3,400 or lower could charge no more than $800 in total cost-sharing and plans with out-of-pocket limits between $3,401 and $6,700 could charge no cost-sharing. For this graph, plans are categorized based on their 2013 out-of-pocket limits and cost-sharing amounts. Mean cost-sharing estimates are weighted by the number of plan enrollees each January.
Changes in SNF and Inpatient Use
In difference-in-differences analyses, there was no significant change in SNF use associated with mandatory SNF cost-sharing reductions (Table 2, Appendix Exhibits 4 and 5). Prior to the policy change, plans with mandatory reductions averaged 131.3 SNF days per month per 1,000 members. In adjusted analyses, this rate of SNF days did not significantly increase after cost-sharing reductions (0.9, 95% CI [−3.5, 5.4], p=0.68). Further, the difference-in-differences estimate of 0.4 (95% CI −5.2, 6.0, p=0.89) demonstrates that SNF use in plans with mandatory cost-sharing reductions did not increase compared with concurrent trends in plans without such reductions. Both groups of plans also experienced similar changes in the rate of SNF admissions (difference-in-differences estimate: 0.04 (95% CI −0.2, 0.1, p=0.65). Pre-policy trends in these outcomes did not significantly differ between plans with and without mandatory cost-sharing reductions (Appendix Exhibit 3).
Table 2:
Difference-in-differences estimates for change in skilled nursing facility and inpatient use per month per 1,000 members
| Plans with Mandatory Cuts (n=320) | Plans without Mandatory Cuts (n=261) | Adjusted Difference in Differences (95% Confidence Interval) | |||||
|---|---|---|---|---|---|---|---|
| Before 2014 | After 2015 | Adjusted Difference (95% Confidence Interval) | Before 2014 | After 2015 | Adjusted Difference (95% Confidence Interval) | ||
| Skilled nursing facility days | 131.3 | 132.5 | 0.9 (−3.5, 5.4) | 118.8 | 119.5 | 0.5 (−3.7, 4.8) | 0.4 (−5.2, 6.0) |
| Skilled nursing facility admissions | 5.3 | 5.5 | 0.0 (−0.2, 0.2) | 5.1 | 5.3 | 0.1 (−0.1, 0.2) | −0.0 (−0.2, 0.1) |
| Inpatient admissions | 19.0 | 18.4 | −0.6 (−1.1, −0.2)** | 17.9 | 17.8 | −0.1 (−0.5, 0.3) | −0.6 (−1.0, −0.1)* |
Note: Authors’ analysis of Master Beneficiary Summary File, Minimum Data Set records, MedPAR files, and data on Medicare Advantage plans’ benefit packages. Plans with mandatory cost-sharing reductions had to decrease copayments for the first 20 days of a skilled nursing facility stay between 2013 and 2015 to comply with reduced caps from the Centers for Medicare and Medicaid Services.
p<0.05;
p<0.01.
Mandatory SNF cost-sharing reductions were associated with significantly fewer inpatient admissions. Prior to 2014, plans with and without mandatory cost-sharing reductions averaged 19.0 and 17.9 inpatient admissions per month per 1,000 members, respectively (Table 2, Appendix Exhibit 5). Plans with mandatory cost-sharing reductions had significant declines in the monthly number of inpatient admissions per 1,000 members compared with plans without cost-sharing reductions (difference-in-differences estimate: −0.6 [95% CI −1.0, −0.1], p=0.03). This decline represents a 3.2% reduction compared to baseline inpatient use among plans with mandatory reductions. Because plans with and without mandatory reductions had significantly different trends prior to 2014 for the rate of inpatient days (Appendix Exhibit 3), no difference-in-differences estimate is presented for that outcome.
In stratified analyses that examined populations with greater use of SNF services or limited financial resources, mandatory cost-sharing reductions were not associated with significant changes in SNF use for almost all groups examined (Appendix Exhibits 6 – 10). Difference-in-difference estimates for inpatient admissions were negative for most populations and statistically significant for some populations, including beneficiaries in plans where reductions eliminated SNF cost-sharing for the first 20 days.
Among beneficiaries with any enrollment in study plans during the study period (including those who disenrolled or joined after January 2012), the difference-in-differences estimate suggests that mandatory reductions were significantly associated with an increase in number of SNF days (5.4, 95% CI: 1.4, 9.4, p=0.01).
The event study analysis indicated no substantial differences from main findings for SNF days and admissions. Relative differences in inpatient admission rates were negative but not statistically significant after the policy change (Appendix Exhibit 11).
Changes in Disenrollment
All difference-in-differences estimates for plan disenrollment at the end of December were negative, but only the results for beneficiaries with SNF use in the past 6 months were statistically significant at a 5% level (Figure 2, Table 3, Appendix Exhibit 12). In plans with mandatory reductions, the mean number of recent SNF users leaving a plan at the end of the year declined from 93.0 to 66.1 beneficiaries per 1,000 beneficiaries. In adjusted analyses, this decrease was larger than the concurrent change observed in the plans without mandatory reductions (difference-in-differences estimate: −16.8 [95% CI: −31.9, −1.8], p =0.03 or a 18% reduction relative to baseline disenrollment rates among plans with mandatory reductions). Difference-in-difference estimates were not significant for plan disenrollment during January through November.
Figure 2. Year-end plan exit rates for plans with and without mandatory cost-sharing reductions.
Mean monthly number of beneficiaries who left plan per 1,000 members
Note: Authors’ analysis of Master Beneficiary Summary File, Minimum Data Set records, MedPAR files, and data on Medicare Advantage plans’ benefit packages. Plans with mandatory cost-sharing reductions had to decrease copayments for the first 20 days of a skilled nursing facility stay between 2013 and 2015 to comply with reduced caps from the Centers for Medicare and Medicaid Services.
Table 3.
Difference-in-differences estimates for changes in plan disenrollment per month per 1,000 members
| Plans with Mandatory Cuts (n=320) | Plans without Mandatory Cuts (n=261) | Adjusted Difference in Differences (95% Confidence Interval) | |||||
|---|---|---|---|---|---|---|---|
| Before 2014 | After 2015 | Adjusted Difference (95% Confidence Interval) | Before 2014 | After 2015 | Adjusted Difference (95% Confidence Interval) | ||
| December | |||||||
| All | 103.8 | 71.1 | −18.7 (−40.0, 2.5) | 53.0 | 40.9 | −4.5 (−10.3, 1.3) | −14.2 (−36.2, 7.8) |
| Hospital use past 6 months | 97.0 | 66.1 | −19.6 (−38.1, −1.1)* | 49.1 | 39.8 | −2.9 (−8.5, 2.8) | −16.7 (−36.0, 2.7) |
| SNF use past 6 months | 93.0 | 66.1 | −18.0 (−31.9, −4.1)* | 46.4 | 40.6 | −1.2 (−7.1, 4.7) | −16.8 (−31.9, −1.8)* |
| January – November | |||||||
| All | 3.0 | 2.6 | 0.5 (0.2, 0.9)** | 2.5 | 2.4 | 0.8 (0.4, 1.2)*** | −0.2 (−0.5, 0.0) |
| Hospital use past 6 months | 4.7 | 4.9 | 0.2 (−0.3, 0.7) | 3.8 | 3.9 | 0.1 (−0.4, 0.6) | 0.1 (−0.3, 0.4) |
| SNF use past 6 months | 9.2 | 9.3 | 0.2 (−0.8, 1.1) | 6.7 | 6.8 | −0.1 (−0.9, 0.8) | 0.2 (−0.4, 0.9) |
Note: Authors’ analysis of Master Beneficiary Summary File, Minimum Data Set records, MedPAR files, and data on Medicare Advantage plans’ benefit packages. Models focused on disenrollment at the end of December analyzed data from December 2012 – December 2015 and included an annual time trend variable and plan fixed effects. Models focused on disenrollment at the end of January-November months analyzed data from January 2012 – November 2016 (omitted December) and included a time trend variable and plan and calendar month fixed effects. To have a sufficient claims lookback period, we perform these stratified analyses using observations from July 2012 onwards for recent SNF or inpatient use in past 6 months).
Discussion
Stricter CMS regulations that eliminated or substantially lowered SNF cost-sharing in MA plans were not associated with increases in overall SNF use among continuously enrolled members. Plans with mandatory reductions experienced slight decreases in the monthly inpatient admission rate relative to plans that already had lower SNF cost-sharing. Although mandatory cost-sharing reductions were not associated with reductions in disenrollment rates among all plan members, disenrollment rates at year-end decreased among members with recent SNF use.
Our findings are consistent with previous null estimates in studies that examined whether MA post-acute cost-sharing changes influenced service use.8, 12 Compared to an earlier analysis that found no changes in SNF use after 2011 mandatory reductions in MA SNF cost-sharing,8 this study has several advantages, including more detailed inpatient and enrollment data. The 2011 SNF cost-sharing caps coincided with new cost-sharing caps in other sectors, including requirements for plans to offer an overall out-of-pocket limit. This study focused on a period with fewer concurrent CMS changes in cost-sharing limits, allowing more precise identification of the effects of limiting SNF cost-sharing in MA plans.
Our null findings differ from the larger literature demonstrating reduce use with greater cost-sharing for three potential reasons. First, SNF services may be unique because providers strongly influence referral and discharge decisions, thereby potentially diminishing the role of out-of-pocket costs. The RAND Health Insurance Experiment found that service use decreased with greater cost-sharing but this effect did not persist after conditioning on whether patients had already initiated a treatment episode.13, 14 This null effect is particularly relevant to SNF use, which usually follows hospitalizations. Second, we examined the effect of lower cost-sharing caps, not the direct effect of lower cost-sharing. Several factors may have attenuated differences in SNF trends between plans with or without mandatory cost-sharing reductions, including voluntary SNF cost-sharing reductions among plans that already met requirements and plans switching to higher out-of-pocket limit categories when faced with mandatory cost-sharing reductions. Third, MA plans may use managed care strategies, such as network restrictions and prior authorization,15 to counteract any potential increases in SNF use associated with lower cost-sharing.
The only significant increase in SNF use associated with lower cost-sharing was observed among beneficiaries with enrollment at any point in study plans. Most of this population joined plans after 2012, so increased SNF use could reflect several factors: price sensitivity to SNF cost-sharing, selective plan entry among beneficiaries more likely to use SNF services after caps were lowered, and differences in member characteristics over time. If this policy increased the likelihood that SNF users join MA plans, then in future years SNF use may increase in the MA program.
The mechanism underlying the modest declines in inpatient admissions (including readmissions) associated with mandatory SNF cost-sharing reductions are not clear, but one possibility is that SNF use may substitute for inpatient care. Because MA plans can waive the three-day hospitalization requirement for SNF services,16 MA members may have been more likely to be admitted directly to SNFs with lower SNF cost-sharing instead being hospitalized before the SNF stay. Evidence from traditional Medicare suggests that reductions in SNF length of stay due to greater cost-sharing after day 20 are associated with increased inpatient readmissions.17
Our findings should reassure policymakers about the consequences of capping cost-sharing for SNF care in MA plans. Despite concerns that lower cost-sharing may increase SNF use and associated Medicare spending, we found that SNF use did not increase among existing MA plan members. Reducing post-acute cost-sharing was intended to counteract favorable selection in MA plans, where patients with more complex health care needs are more likely to exit these plans. Prior to mandatory cost-sharing reductions, plans with higher SNF cost-sharing levels had greater disenrollment rates compared to plans with lower cost-sharing levels, but this gap narrowed among beneficiaries with SNF use after the policy change, suggesting the policy may have achieved its intended effects.
Some potential unanticipated consequences of capping SNF cost-sharing should also be considered. Cost-sharing for SNF services declined, but over the same time plans with and without mandatory SNF cost-sharing reductions raised premiums and inpatient cost-sharing amounts. Out-of-pocket limits increased, especially in plans without mandatory cost-sharing reductions. Such measures may distribute higher out-of-pocket costs across a broader share of the MA population instead of concentrating costs among SNF users.
Our study has several limitations. Difference-in-differences analyses assume that plans with mandatory cost-sharing changes would have experienced the same changes in outcomes as comparison plans if there had been no policy change. We cannot test this assumption but plans had parallel trends in study outcomes prior to the policy change. Members’ observed characteristics, and probably unobserved characteristics like comorbidity, differed between plans with and without mandatory cost-sharing changes. These characteristics may influence health care use and disenrollment but will only bias our difference-in-differences estimates if the distribution or effect of these characteristics was inconsistent over the study period. The results from stratified analyses should be considered exploratory because some significant relationships may exist by chance based on the number of comparisons made. Further, we cannot confirm that MA plans paid for all observed SNF days since we do not have MA claims data.
When CMS required further reductions in SNF cost-sharing in MA plans, the plans with mandatory cost-sharing reductions experienced no relative changes in SNF use but did have slight decreases in inpatient use compared to plans that already had low cost-sharing levels. Our findings suggest that the current caps in SNF cost-sharing may have reduced SNF out-of-pocket costs for beneficiaries without increasing use of SNF care.
Supplementary Material
Acknowledgments:
The authors would like to thank Zilu Zhou for assistance with data analysis. An earlier version of this study has been presented at the 2019 ASHEcon Conference in Washington, DC.
Funding:
This study was funded by the National Institutes on Aging (grants R01AG044374-03, P01AG027296-09 and K01AG058700). Dr. Thomas’ effort was supported by the US Department of Veterans Affairs Health Services Research and Development Service (CDA 14-422). The views expressed in this article are those of the authors and do not necessarily reflect the position or policy of the National Institutes of Health, Department of Veterans Affairs, or US government.
Footnotes
Disclosures of potential conflicts of interest: Dr. Keohane reports spousal income from TriStar Health. The investigators report no other conflicts of interest.
References
- 1.Jacobson GA, Damico A, Neuman T. A Dozen Facts About Medicare Advantage. 2018. Available at: https://www.kff.org/medicare/issue-brief/a-dozen-facts-about-medicare-advantage/. Accessed May 15, 2019.
- 2.Neuman P, Jacobson GA. Medicare advantage checkup. N Engl J Med 2018;379:2163–2172. [DOI] [PubMed] [Google Scholar]
- 3.Rahman M, Keohane L, Trivedi AN, et al. High-Cost Patients Had Substantial Rates Of Leaving Medicare Advantage And Joining Traditional Medicare. Health Aff 2015;34:1675–1681. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Meyers DJ, Belanger E, Joyce N, et al. Analysis of Drivers of Disenrollment and Plan Switching Among Medicare Advantage Beneficiaries. JAMA Intern Med 2019 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.McWilliams JM, Hsu J, Newhouse JP. New risk-adjustment system was associated with reduced favorable selection in medicare advantage. Health Aff (Millwood) 2012;31:2630–2640. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Ankuda CK, Ornstein KA, Covinsky KE, et al. Switching Between Medicare Advantage And Traditional Medicare Before And After The Onset Of Functional Disability. Health Aff 2020;39:809–818. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Keohane LM, Grebla RC, Rahman M, et al. First-dollar cost-sharing for skilled nursing facility care in medicare advantage plans. BMC Health Serv Res 2017;17:611. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Keohane LM, Rahman M, Thomas KS, et al. Effects of Caps on Cost Sharing for Skilled Nursing Facility Services in Medicare Advantage Plans. J Am Geriatr Soc 2018;66:992–997. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Centers for Medicare and Medicaid Services. Advance Notice of Methodological Changes for Calendar Year (CY) 2020 for Medicare Advantage (MA) Capitation Rates, Part C and Part D Payment Policies and 2020 Draft Call Letter. 2019. Available at: https://www.cms.gov/Medicare/Health-Plans/MedicareAdvtgSpecRateStats/Downloads/Advance2020Part2.pdf. Accessed February 28, 2020.
- 10.Intrator O, Hiris J, Berg K, et al. The residential history file: studying nursing home residents’ long-term care histories. Health Serv Res 2011;46:120–137. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Huckfeldt PJ, Escarce JJ, Rabideau B, et al. Less Intense Postacute Care, Better Outcomes For Enrollees In Medicare Advantage Than Those In Fee-For-Service. Health Aff 2017;36:91–100. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Li Q, Keohane LM, Thomas K, et al. Association of Cost Sharing With Use of Home Health Services Among Medicare Advantage Enrollees. JAMA Intern Med 2017;177:1012–1018. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Manning WG, Newhouse JP, Duan N, et al. Health insurance and the demand for medical care: evidence from a randomized experiment. Am Econ Rev 1987;77:251–277. [PubMed] [Google Scholar]
- 14.Aron-Dine A, Einav L, Finkelstein A. The RAND health insurance experiment, three decades later. Journal of Economic Perspectives 2013;27:197–222. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Gadbois EA, Tyler DA, Shield RR, et al. Medicare Advantage control of postacute costs: Perspectives from stakeholders. The American journal of managed care 2018;24:e386. [PMC free article] [PubMed] [Google Scholar]
- 16.Grebla RC, Keohane L, Lee Y, et al. Waiving The Three-Day Rule: Admissions And Length-Of-Stay At Hospitals And Skilled Nursing Facilities Did Not Increase. Health Aff 2015;34:1324–1330. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Werner RM, Konetzka RT, Qi M, et al. The impact of Medicare copayments for skilled nursing facilities on length of stay, outcomes, and costs. Health Serv Res 2019;54:1184–1192. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.