Association of daily copayments with use of hospital care among medicare advantage enrollees

Cost sharing is a common technique utilized by health insurers to “share” a portion of an enrollee’s health expenditures with the enrollee. This often takes the form of a payment at the point of service (co-payment) or payment for a fixed percentage of the cost of a given health service (co-insurance). In the hospital setting, this could also be a lump sum payment at admission (a deductible), or a payment for each day in the hospital (a per diem) [1, 2].

The Medicare program has used cost sharing in various forms since its inception in 1965. Medicare enrollees are responsible for 20% coinsurance for physician visits and large inpatient deductibles for hospital admissions, with no cap on out-of-pocket spending. The role of cost sharing and its clinical and economic effects in the Medicare program are topics of ongoing health policy debate. A study of the commercially insured found substantial increases in hospital cost sharing from 2009 to 2013 [3].

There is relatively little evidence to guide policymakers about the impact of greater cost sharing in the Medicare program. The landmark RAND Health Insurance Experiment, a randomized trial of cost sharing in health care, found that persons with higher coinsurance rates used less care and had lower spending than those with more generous insurance [4]. However, the RAND experiment ended in 1982 and excluded the elderly, limiting its generalizability to contemporary Medicare beneficiaries. The imposition of an inpatient deductible in the United Mine Workers Health Plan in 1977 was associated with a 45% decline in the probability of having a hospitalization [5]. Recent studies demonstrate that in response to increased outpatient copayments, Medicare beneficiaries reduced their use of outpatient services, but made greater use of hospital care [6, 7]. To our knowledge, studies of the Medicare population related to incentives and cost sharing in the hospital have been limited to the Medigap program which provides supplemental insurance to cover inpatient deductibles and other copayments. These studies also find increased inpatient utilization for those experiencing decreased cost sharing through supplemental Medigap insurance [8‐11]. There are important differences between Medicare Advantage and Medigap. Most Medigap policies eliminate inpatient cost sharing altogether, which is a much larger change than our study and presumably less relevant for Medicare Advantage, purchasing Medigap reduces or eliminates cost sharing for many other services besides inpatient care, and the effect of cost sharing may differ in Medicare Advantage given supply-side, managed care constraints that are not found in traditional Medicare. Recent studies related to Medicare and Medicare Advantage focused on prescription drug use and adherence [12, 13], skilled nursing facility utilization [14], or Medicare Advantage enrollment [15]. The lack of inpatient utilization studies in the broader Medicare fee-for-service and Medicare Advantage programs is an important gap since hospital costs are the largest component of Medicare spending and the Medicare Part A deductible is the largest single out-of-pocket expense in the traditional Medicare benefit structure, $1340 in 2018 [16]. Further, as there has been increasing policy interest in raising cost sharing in the traditional Medicare program to control spending, one strategy has been to reduce the generosity of supplemental plans and expose enrollees to first-dollar cost sharing, the findings in this study would help to inform the debate [17].

Medicare Advantage plans, which currently enroll 31% of all beneficiaries, have experimented with changes to inpatient cost sharing [18]. A common change has been to eliminate the deductible and impose a daily copayment for each day of hospital care [19]. In 2010, “virtually all Medicare Advantage plans, 94%, required enrollees to share in the costs of inpatient care. 81% percent imposed copayments, 2% imposed coinsurance, and 11% used both. Among Medicare Advantage plans charging copayments for inpatient care, 79% charged a copayment per day, 16% charged a copayment per stay, and 5% charged both copayments per stay and per day.” [20] A deductible is typically exceeded during the first day of a hospital stay, leaving no financial incentive for a patient to leave the hospital earlier. In contrast, a per diem structure retains an incentive for a patient to leave the hospital throughout his or her stay. Thus, changing a plan’s benefit structure from a deductible to a per diem could mean lower out-of-pocket spending for beneficiaries with shorter lengths of stay, but greater out-of-pocket costs for hospitalized beneficiaries with longer lengths of stay, and subsequently could lead to decreased utilization. This study highlights the tradeoffs of this benefit change (potentially lower inpatient utilization but perhaps much higher cost-sharing requirements for sicker enrollees). In this study, we examined the impact of per-diem copayment and increased levels of cost sharing on the use of hospital care among Medicare Advantage enrollees age 65 and older. We hypothesized that changing the inpatient benefit structure from a deductible at admission to a per diem will result in reduced hospital utilization at the plan level.

In the year before intervention plans replaced the inpatient deductible with a per-diem copayment, the average inpatient deductible in intervention plans was $376 (interquartile range [IQR], $250 to $500) and in control plans was $349 (IQR, $200 to $600). In the year after intervention plans changed their inpatient benefit, intervention plans replaced their inpatient deductible with a $165 (IQR, $110 to $225) average daily copayment and the average inpatient deductible in control plans remained unchanged, by design. (Table 1) 71.6% of all hospitalized enrollees had only 1 inpatient admission in a given year, 18.6% had 2 inpatient admissions and the remaining 9.8% had 3 or more inpatient admissions in any given year. (Hospitalization data not reported in Table 1, but used to generate Fig. 1) The demographic characteristics of enrollees (e.g., age, sex, race, etc.) in the intervention and control plans were similar. (Table 1) Skilled nursing facility cost sharing was unchanged in the intervention and control plans. For outpatient cost sharing, intervention plans exhibited a $4 increase in average specialist copayments and a $2 increase in average primary care copayments compared to a $1 decrease in average primary care copayments and no change in specialist copayments in the control plans. Emergency department copayments remained unchanged in the intervention and control plans. (Table 1) Average monthly premiums for intervention plans decreased slightly while control plan premiums remained stable. Intervention plans had a higher percentage of zero premium plans as compared to control plans. (Table 1).

Unadjusted inpatient admissions per 100 enrollees decreased in intervention plans by 0.4 admissions per 100 enrollees from the year before the benefit change to the year after the benefit change. In control plans, inpatient admissions per 100 enrollees increased by 1.1 admissions per 100 enrollees. Therefore, the difference-in-difference estimate of intervention plans relative to control plans was − 1.4 admissions per 100 enrollees (95% CI, − 1.8 to − 0.9). Adjusting for age, sex, race and Part D subsidy and including a year fixed effect resulted in a difference-in-difference estimate of − 1.3 admissions per 100 enrollees (95% CI, − 1.8 to − 0.9) for intervention plans relative to controls. (Table 2).

We observed similar results for inpatient days per 100 enrollees and for the proportion of enrollees hospitalized. Unadjusted inpatient days per 100 enrollees in intervention plans decreased by − 6.3 days per 100 enrollees relative to control plans (95% CI, − 9.3 to − 3.3) and the unadjusted proportion of enrollees hospitalized in intervention plans decreased by 1.0 percentage points relative to control plans (95% CI, − 1.3 to − 0.7). The adjusted difference-in-difference estimate of inpatient days per 100 enrollees in intervention plans relative to control plans was − 6.9 days (95% CI, − 10.1 to − 3.8). The adjusted difference-in-difference estimate of the proportion of enrollees hospitalized in intervention plans relative to control plans was − 0.7 percentage points (95% CI, − 1.0 to − 0.4). Neither unadjusted nor adjusted difference-in-difference estimates of mean length of stay reached conventional measures of significance at the 95% level. Results were similar when considering enrollees continuously enrolled for an entire 24-month period in the same plan and when examining beneficiaries enrolled for any amount of time. (Table 2).

We observed similar enrollee characteristics and inpatient utilization for beneficiaries staying enrolled in the same plan across intervention and control plans in the year prior to the benefit changes. We also see similar demographic characteristics among enrollees that exited plans in the year prior to exiting the plan and among enrollees that entered plans in the year prior to entering one of our study plans. A higher proportion of enrollees that exited intervention and control plans were hospitalized in the baseline year, 24%, as compared to those enrollees remaining in the same plan, 11%. There was no difference, though, when comparing intervention and control plans. Inpatient utilization for enrollees entering intervention and control plans in the year prior to entering was also similar. Beneficiaries disenrolled at a slightly higher rate from intervention plans, 16%, as compared to control plans, 15.5%. (Table 3).

For enrollees who were not hospitalized in their baseline year, the unadjusted difference-in-difference estimate of the proportion of enrollees hospitalized in intervention plans relative to control plans was − 0.4 percentage points (95% CI, − 0.6 to − 0.1). For those enrollees hospitalized 1 time in their baseline year, the unadjusted difference-in-difference estimate was − 0.9 percentage points (95% CI, − 1.8 to − 0.03). Finally, for enrollees hospitalized 2 or more times in the baseline year, the unadjusted difference-in-difference estimate of the proportion of enrollees hospitalized in intervention plans relative to control plans was − 2.2 percentage points (95% CI, − 3.8 to − 0.6). (Fig. 1) Unadjusted difference-in-difference estimates of admissions and days per 100 enrollees and average length of stay across strata of baseline hospitalizations did not reach conventional measures of significance at the 95% level.

We studied the effect of changing an inpatient insurance benefit from a deductible at admission to a per diem (charge per day) in a large sample of Medicare Advantage enrollees age 65 and older. We found that the change in benefit structure was associated with significant declines in inpatient admissions and days per 100 enrollees. We also found a 0.7 adjusted percentage point reduction in the proportion of enrollees hospitalized in intervention plans relative to control plans. Enrollees with greater use of hospital care in the year prior to the cost-sharing change experienced greater declines in hospital utilization, perhaps indicating that patients, once hospitalized, were more acutely aware of the cost-sharing burden and took steps to avoid future admissions, or, perhaps, the plan itself more actively managed the patient to avoid additional inpatient expenses. Once enrollees were hospitalized, though, we did not find any significant difference in the adjusted length of stay between intervention and control plans.

Our findings are consistent with research on inpatient utilization and its association with supplemental Medigap insurance. Medigap coverage, or decreased cost sharing, has been found to be associated with increases in inpatient utilization, and we find a directionally similar result, with increased cost sharing associated with decreases in inpatient utilization, though it should be noted that the increased cost sharing only applies, on average, to patients hospitalized 3 days or more (70.7% of hospitalized patients in our sample). Our findings are also broadly consistent with two studies of hospital cost sharing among non-elderly populations. The United Mine Workers Study demonstrated that the imposition of a $250 hospital deductible in 1977 led to a “45% decline in the probability of a hospital admission” from a baseline of 6.8%, but resulted in an increase in length of stay among those hospitalized [5]. Our results show about a 7% relative reduction in the probability of a hospitalization from a baseline of about 20%. The RAND experiment also found increases in cost sharing were associated with reductions in the likelihood of seeking care, but not the intensity of care once the patient was hospitalized [2, 4]. We also observe a reduction in the proportion hospitalized and reduction in inpatient admissions and days per 100 among enrollees exposed to the cost-sharing change. However, like the RAND study, we did not observe a corresponding increase in length of stay, perhaps because this outcome is more strongly influenced by the decisions of hospital physicians with relatively little impact by patients [24]. A central but often unappreciated finding from the RAND study is that cost sharing was not associated with reductions in the use of health services after patients initiate contact with the health care system.

Inpatient utilization remains the most expensive component of Medicare spending [25]. Therefore, if the purpose of the change in benefit structure was to reduce overall spending on hospital care, the intervention plans in our study likely achieved this objective by reducing aggregate inpatient utilization. However, there were substantial increases in out-of-pocket spending among hospitalized patients, particularly among patients with longer lengths of stay. For instance, a person with a median length of stay would experience expected out-of-pocket costs in the baseline year of $376. In the year after the benefit change, the out-of-pocket costs for an enrollee in an intervention plan with a median length of stay of 4.4 days would increase by 93% to $726. However, persons with longer lengths of stays would have substantially greater increases in out-of-pocket costs. For instance persons at the 75th percentile of length of stay (5.5 days) could expect to pay $908 dollars for an admission after the cost-sharing changes went into effect, an increase of 141%. This highlights the importance for Medicare Advantage enrollees, and for those helping consumers navigate the market, to understand the full package of benefits as it is quite possible some enrollees hospitalized after the benefit change were unaware of the higher copayment until after the hospitalization.

Strengths of our study include the use of a large sample of over 400,000 beneficiaries in 59 MA plans across the country. By observing the entire structure of each plans’ benefit, we were able to identify plans that only changed their inpatient cost sharing, without making significant changes to skilled nursing facility, physician office or emergency department cost sharing. We matched plans by geographic region and confirmed that hospitalization utilization trends in intervention and control plans were similar prior to the change in cost sharing. To our knowledge, this is the first study to quantify the impact of changes in inpatient cost sharing among Medicare Advantage enrollees. We do, though, find greater cost sharing is associated with reduced inpatient utilization, similar to the studies done in the Medigap program.

Our study has limitations. First, we are unable to observe MA plans’ strategies that, apart from changes in benefits, may have impacted hospital utilization. For instance, intervention plans may have implemented more stringent utilization management practices, or made changes in their network of preferred hospitals. However, this would assume that these strategies were implemented in intervention plans and not in control plans and occurred at the same time that inpatient cost sharing was changed. Second, hospitalization and length of stay decisions are complex and include many factors and decision makers, including physicians, care managers, other providers, and patients; our study did not directly observe these processes and exclusively relied on administrative data to quantify changes in inpatient utilization. Third, since traditional Medicare applies an inpatient deductible with no per diem, it would be difficult to generalize the findings to traditional Medicare beneficiaries. However, current policy debates related to cost sharing in the traditional Medicare program could benefit from these findings and, more broadly, the findings may also have implications for benefit design for other kinds of insurance, both public and private. Fourth, it is possible that enrollees selectively disenrolled from intervention plans, anticipating hospitalizations in the following year. However, we ran our regressions on enrollees that exited plans in the year they were included in our study and in the year after exiting one of our study plans and found no significant difference in utilization in our adjusted difference-in-difference models and in all but one (proportion hospitalized) of our unadjusted models. Fifth, we are limited in the covariates we have access to with our data, so there may have been substantial unobserved differences in key variables such as comorbidity and clinical complexity among enrollees in intervention and control plans that could have influenced our findings. However, we did not observe large baseline differences in hospital utilization between intervention and control plans, nor did we observe differences in utilization among those enrollees entering or exiting the intervention and control plans. Finally, we are limited to a small set of plans that met our matching criteria. There are differences between our study population and the overall Medicare Advantage population, therefore there are limits to the generalizability of our study. However, given the paucity of studies on this topic, this provides a foundation for future studies to better understand the effects of inpatient cost sharing on utilization.

In conclusion, we found that changing from a deductible to a per diem copayment structure was associated with reductions in utilization of inpatient care among Medicare Advantage enrollees, particularly among those with greater use of hospital care prior to the copayment change. These reductions appear to be driven by decreases in the number of admissions and the probability of hospitalization, without significant changes in length of stay. Although the use of hospital care, but not the duration of admissions, may be sensitive to daily out-of-pocket costs, the financial burden of changing from a deductible to a per-diem falls heavily on seniors with longer hospital stays.