Likelihood of hospital readmission in Medicare Advantage and Fee‐For‐Service within same hospital

Abstract

Objective

To assess the extent to which all‐cause 30‐day readmission rate varies by Medicare program within the same hospitals.

Study Design

We used conditional logistic regression clustered by hospital and generalized estimating equations to compare the odds of unplanned all‐cause 30‐day readmission between Medicare Fee‐for‐Service (FFS) and Medicare Advantage (MA).

Data Collection

Wisconsin Health Information Organization collects claims data from various payers including private insurance, Medicare, and Medicaid, twice a year.

Principal Findings

For 62 of 66 hospitals, hospital‐level readmission rates for MA were lower than those for Medicare FFS. The odds of 30‐day readmission in MA were 0.92 times lower than Medicare FFS within the same hospital (odds ratio, 0.93; 95 percent confidence interval, 0.89‐0.98). The adjusted overall readmission rates of Medicare FFS and MA were 14.9 percent and 11.9 percent, respectively.

Conclusion

These findings provide additional evidence of potential variations in readmission risk by payer and support the need for improved monitoring systems in hospitals that incorporate payer‐specific data. Further research is needed to delineate specific care delivery factors that contribute to differential readmission risk by payer source.

What This Study Adds.

• Previous studies have shown that all‐cause 30‐day readmission rate in Medicare Fee‐For‐Service (FFS) is higher than in Medicare Advantage (MA).

• Beneficiaries in MA are more often hospitalized at teaching, larger, and/or not‐for‐profit hospitals than those in Medicare FFS.

• It is unclear whether variations in readmission rate can be ascribed to payer type beyond intrinsic between‐hospital variation.

• Patients in Medicare FFS were more likely to be readmitted within 30‐days after discharge than those in MA when they used the same hospital.

1. INTRODUCTION

Reducing hospital readmissions remains an urgent priority for national health policy in light of the detrimental influence on patient outcomes and care costs. ¹ Readmissions among Medicare beneficiaries account for 56 percent of 30‐day readmissions in the United States and cost $26 billion. ¹ To address this, the Centers for Medicare and Medicaid Services (CMS) has undertaken several initiatives to reduce readmissions among Medicare beneficiaries including the Hospital Readmissions Reduction Program (HRRP) in 2012. ² Following implementation of these initiatives, the 30‐day readmission rate in Medicare Fee‐for‐service (FFS) declined from 19‐19.5 percent during 2007‐2011 to 17.5 percent in 2013, nationally. ² However, considering that a significant proportion of 30‐day readmissions in Medicare is still considered avoidable, more efforts to achieve further readmission reduction is necessary. ³

One possible way to identify opportunities to further reduce avoidable readmissions is by comparing Medicare FFS and Medicare Advantage (MA), the two‐main programs for Medicare coverage, that could develop benchmarks for readmission rates and produce helpful information to identify opportunities to reduce readmissions. Specifically, comparing the risk of readmission between Medicare FFS and MA within the same hospital could produce important evidence on whether all patients are receiving the same quality of care during index hospitalizations regardless of payer type.

Substantial differences between MA and Medicare FFS could affect readmission rates. While several studies have documented that MA beneficiaries receive higher quality of care than those in Medicare FFS, Medicare beneficiaries generally rated MA lower on health care access than Medicare FFS. ⁴ , ⁵ , ⁶ Also, MA beneficiaries are more likely to be younger and healthier. ⁶ , ⁷ Under capitation payment systems, MA plans have a strong motivation to attract younger and healthier beneficiaries. Conversely, beneficiaries' individual health risk factors can influence selection into Medicare FFS over MA due to the less restrictive provider networks. ⁸ MA plans can steer beneficiaries to a limited network of hospitals and providers. ⁹ Beneficiaries in MA are more often hospitalized at teaching, larger, and/or not‐for‐profit hospitals than those in Medicare FFS. ¹⁰ However, it is unclear whether variations in readmission rate by payer type are attributable to the differences in hospitals and providers where patients receive care, or the benefits from the limited network such as better coordination care and/or the services within hospitals.

Previous studies have found that the all‐cause 30‐day readmission rate in Medicare FFS ranges 15‐30 percent higher than in MA. ¹¹ , ¹² However, these estimates reflect a combination of differences between the hospitals where patients receive care, and the care and services provided to patients within the same hospital. While recent studies have found that adverse patient safety events within the same hospital differ by source of coverage, there is limited evidence regarding whether risk of readmission within the same hospital differs between Medicare FFS and MA. ¹³ , ¹⁴ Identifying whether the 30‐day readmission risk between Medicare FFS and MA differs within the same hospital is important for ascribing any difference in outcome to payer type, rather than to intrinsic between‐hospital variation and related care quality that could affect readmission rates.

Therefore, this study examines variations in the risk of 30‐day readmissions among Medicare FFS and MA beneficiaries within hospitals using Wisconsin Health Information Organization (WHIO) all‐payer claims data (APCD), focusing on hospitalizations from October 2013 to September 2014.

2. METHODS

2.1. Data sources

This retrospective cohort study used the WHIO APCD version 13 of the data mart (DMV13) which includes medical claims data from October 2012 to December 2014 and covers 75 percent of the Wisconsin population. WHIO is a state‐wide collaboration of insurance companies, healthcare providers, large employers, and public agencies. WHIO developed the state‐wide health insurance claims database in 2005 to provide data useful for examining healthcare issues related to quality, efficiency, and safety in Wisconsin. The WHIO data contain medical, dental, and pharmaceutical claims data with patient demographics from various payers including private insurance, Medicare, and Medicaid. In DMV13, Medicare claims account for 20 percent of the total claims, 38 percent are commercial claims, 20 percent are Medicaid fee‐for‐service claims, and 22 percent are Medicaid Health Maintenance Organization claims.

WHIO data were linked to the 2013 American Hospital Association's (AHA) annual survey data using the hospital identification number to generate hospital‐level variables. The AHA annual Survey of Hospitals is an annual survey of more than 6400 hospitals in the US that collects data on a variety of topics including hospital organizational structure, facilities and services, utilization data, physician arrangements, staffing, and community orientation.

2.2. Study sample

This study focused on individuals aged 66 years or older at the time of index hospitalization, who were alive upon discharge, and hospitalized in an acute care hospital in Wisconsin. We included individuals continuously enrolled in the same Medicare program for the 12 months prior to the index admission, and continuously enrolled 30 days after discharge. Individuals with potentially incomplete data due to railroad benefits, those transferred to another acute care hospital on discharge or discharged against medical advice, were excluded. ¹⁵ Our study sample included 102 556 discharges from 66 hospitals.

2.3. Variables

2.3.1. Readmission

We used the 30‐day all‐cause hospital readmission measure developed for CMS by the Yale School of Medicine Center for Outcomes Research & Evaluation as the outcome in this study. ¹⁶ Medicare beneficiaries age 66 or older who were hospitalized at a short‐stay acute‐care hospital and experienced an unplanned readmission for any cause to an acute care hospital within 30 days of discharge were defined as readmitted. The measure uses a 30‐day time frame because older adult patients are more vulnerable to adverse health outcomes during this time and also to be consistent with the readmission measures approved by the National Quality Forum and publicly reported by CMS. If there is more than one unplanned admission within 30 days of discharge from the index hospitalization, only the first is considered as a readmission.

2.3.2. Medicare program

We focused on individuals enrolled in Medicare FFS and MA plans who are continuously enrolled in the same Medicare program for 12 months prior to the index hospitalization and 30 days after discharge from the index hospitalization. We define individuals continuously enrolled in Medicare Parts A as enrolled in Medicare FFS.

2.3.3. Covariates

We selected individual characteristics based on the Andersen's Model of Health Care Utilization including, ¹⁷ predisposing factors (age, sex, and patient residence in a rural area) and need characteristics (principal diagnosis for index hospitalization based on major diagnostic categories, patient‐level number of chronic conditions, number of hospitalizations and average days of hospitalization in 12 months prior to index hospitalization).

We included hospital characteristics (teaching status, size, urban/rural location, type, Medicaid inpatient, and staff level) to examine the overall association between Medicare programs and readmissions. Appendix S1 provides a list of covariate categories.

2.4. Statistical analysis

We compared baseline characteristics of hospitalizations stratified by Medicare program. Additionally, the percentage of total readmissions by the number of days after discharge and by Medicare program were examined to observe how the distribution of readmissions within the 30‐day timeframe varies by Medicare program. Next, we used conditional logistic regression within hospitals to examine the association between Medicare programs and 30‐day readmissions within hospitals. Conditional regression models have advantages such as distributional assumptions are not required for random subject effects. ¹⁸ Generalized estimating equations (GEE) with independent working correlation were used to perform multivariate logistic regression estimation to examine the marginal association between Medicare programs and 30‐day readmissions. Statistical significance was set at P < .05. Using the multivariate logistic regression model, we calculated the mean predicted probabilities of readmission by Medicare program and hospital. Analyses were conducted in R Studio 3.3.2 using the “geem” and “clogit” package.

We also performed sensitivity tests to examine the robustness of our results. Specifically, we examined the models with all‐cause 3‐ and 7‐day readmissions, which may more directly reflect associations with inpatient care. Also, as CMS does not reimburse readmissions within 24 hours after discharge for the same condition, ¹⁹ we ran our analyses based on readmissions between 2 through 30 days from date of discharge. Moreover, according to the Agency of Healthcare Research and Quality, hospitals with fewer than 30 discharges are not recommended to use in comparative analyses. ¹⁴ As we are comparing Medicare FFS and MA within the same hospital, we applied the sample exclusion criteria for each Medicare programs in hospitals for a reliable comparison between Medicare FFS and MA. Lastly, because the use of post‐acute care facilities varies by individual health conditions and Medicare program, ²⁰ we examined the variation in discharge locations by Medicare program, and the readmission rates by Medicare program and discharge location.

3. RESULTS

The baseline characteristics of 102 556 discharges by Medicare program are shown in Table 1. Between October 1, 2013 and September 30, 2014, nearly three out of four discharges were from patients in Medicare FFS and the other one out of four discharges were from patients in MA. The Medicare FFS sample was more likely to be older (85 years or older—FFS: 26.9 percent, MA: 21.7 percent), has a longer stay for their index hospitalizations (4 or more days—FFS: 47.1 percent; MA: 42.7 percent), and has a greater number of chronic conditions (6 or more—FFS: 31.2 percent; MA: 20.9 percent). While patients in Medicare FFS were more likely to be discharged to a skilled nursing facility (FFS: 27.9 percent; MA: 19.6 percent), patients in MA were more likely to discharged to home (FFS: 69.8 percent; MA: 76.1 percent).

TABLE 1.

Characteristics of discharges by Medicare Program, Wisconsin Health Information Organization All‐Payer Claims Data, 2013‐2014 (N = 102 556)

	Overall	FFS	MA	P‐value
N	102 556	77 237	25 319
Age (%)
66‐69	16.4	16.3	16.8	<.001
70‐74	19.6	19.0	21.3
75‐79	19.6	19.1	21.0
80‐84	18.8	18.6	19.2
85‐89	13.3	13.5	12.4
90+	12.4	13.4	9.3
Gender = Female (%)	57.3	58.4	53.8	<.001
Residence = Rural (%)	13.0	12.5	14.3	<.001
Number of hospitalizations a (mean [SD])	0.9 (1.4)	0.9 (1.4)	0.8 (1.2)	<.001
Avg days of hospitalizations b (mean [SD])	2.3 (4.4)	2.4 (4.1)	2.2 (5.1)	<.001
Number of chronic conditions (%)
0‐1	18.4	16.8	23.4	<.001
2‐3	29.2	27.9	32.9
4‐5	23.7	24.0	22.8
6‐7	15.8	16.8	12.8
8‐9	8.4	9.3	5.6
10 or more	4.5	5.1	2.5
Principle diagnosis for index hospitalization (%)
Circulatory system	23.0	22.7	24	<.001
Digestive system	11.1	10.9	11.7
Musculoskeletal system & connective tissue	18.3	17.6	20.5
Respiratory system	12.3	12.5	11.8
Kidney & urinary tract	7.1	7.3	6.3
Nervous system	6.9	6.8	7.4
Infectious & parasitic diseases	6.8	6.9	6.6
Factors influencing health status	2.8	2.8	2.5
Endocrine, nutritional, & metabolic	3.0	3.8	0.6
Hepatobiliary system & pancreas	2.1	2.0	2.3
Skin, subcutaneous tissue & breast	2.1	2.2	1.8
Blood and immunology	1.2	1.3	1.2
Injuries, poisonings, & drug toxicity	0.9	0.9	0.9
Ear, nose, mouth, & throat	0.7	0.7	0.6
Female reproductive system	0.3	0.3	0.3
Male reproductive system	0.3	0.3	0.3
Eye	0.1	0.1	0.1
Multiple significant trauma	0.1	0.1	0.1
Unknown	0.8	0.8	0.9

Abbreviations: Avg, average; FFS, Medicare Fee‐for‐Service; MA, Medicare Advantage.

^aNumber of hospitalizations in 12 mo prior to index hospitalization.

^bAverage days of hospitalization in 12 mo prior to index hospitalization.

Figure 1 displays the distribution of the readmissions by number of days after discharge and by Medicare program. Generally, the percentage of readmissions decreases as the number of days after discharge increases starting from the second day. The percentage of readmissions for the first day after discharge in Medicare FFS is nearly half than that of MA. In both Medicare programs, more than 30 percent of the readmissions occur within the first 7 days after discharge.

FIGURE 1.

Percentage of total 30‐d hospital readmissions by number of days after discharge by Medicare Program. FFS, Medicare Fee‐for‐Service; MA, Medicare Advantage

Table 2 reports unadjusted and adjusted odds ratios (OR) from our logistic regression models. The results from the conditional logistic regression models show the unadjusted odds for readmission (Unadjusted) in MA were 0.84 times lower than Medicare FFS (OR, 0.84; 95 percent CI, 0.80‐0.88; P < .001). While controlling for characteristics such as age, gender, and rural/urban (model 1) shows similar result to the unadjusted model (OR, 0.83; 95 percent CI, 0.79‐0.87; P < .001), adding need characteristics (model 2) including the number of chronic conditions narrows the difference in odds for readmission between MA and FFS (OR, 0.93; 95 percent CI, 0.89‐0.98; P < .001).

TABLE 2.

Adjusted association between Medicare Programs and 30‐d hospital readmissions, Wisconsin Health Information Organization All‐Payer Claims Data, 2013‐2014 (Reference group: Medicare Fee‐for‐Service; N = 102 556)

	Conditional logistic regression models a		GEE models
	Odds ratio (95% CI)	P‐value	Odds ratio (95% CI)	P‐value
Unadjusted	0.84 (0.80‐0.88)	<.001	0.81 (0.77‐0.86)	<.001
Model 1 b	0.83 (0.79‐0.87)	<.001	0.81 (0.77‐0.86)	<.001
Model 2 c	0.93 (0.89‐0.98)	<.001	0.93 (0.87‐0.98)	.012
Model 3 d	–	–	0.93 (0.87‐0.99)	.014

Abbreviations: CI, confidence interval; GEE, generalized estimating equations.

^aConditional logistic regression models are clustered by hospital.

^bCovariates include predisposing factors (age, sex, and patient residence in a rural area).

^cCovariates include predisposing factors and need characteristics (principle diagnosis for index hospitalization based on major diagnostic categories, patient‐level number of chronic conditions, number of hospitalizations, and average days of hospitalization in 12 mo prior to index hospitalization).

^dCovariates include predisposing factors, need characteristics, and hospital characteristics (teaching status, size, urban/rural location, type, Medicaid inpatient, and staff level).

The results from the logistic regression models with GEE show similar results to the conditional logistic regression models. The odds for readmission in MA were 0.81 times lower than FFS in the unadjusted model (OR, 0.81; 95 percent CI, 0.77‐0.86; P < .001) and adjusting for need characteristics (model 2) narrows the difference in odds for readmission between MA and FFS within the same hospital (OR, 0.93; 95 percent CI, 0.87‐0.98; P = .012). The final model (model 3), adjusted for individual and hospital characteristics, reports a similar odds ratio to model 2 (OR, 0.93; 95 percent CI, 0.88‐0.99; P = .014).

Figure 2 shows the adjusted overall readmission rates and adjusted hospital‐level readmission rates by Medicare program. The adjusted overall readmission rate of Medicare FFS and MA was 14.9 percent and 11.9 percent, respectively. For 62 of 66 hospitals, we find that hospitals' readmission rates for MA were lower than those for Medicare FFS. Hospital‐level adjusted readmission rates range from 3.0 to 24.2 percent in MA and 4.2 to 35.2 percent in Medicare FFS. Among hospitals where the readmission rate was lower in MA than in Medicare FFS, the hospital‐level readmission rates for Medicare FFS were 1.01 to 2.78 times higher than those for MA.

FIGURE 2.

Hospital‐level risk adjusted 30‐d hospital readmission rates by medicare program. FFS, Medicare Fee‐for‐Service; MA, Medicare Advantage

Results of sensitivity analyses were qualitatively consistent to the main analysis (Appendices S2 and S3), demonstrating that MA was associated with lower odds of readmission than Medicare FFS with both the 7‐day readmission measure (OR, 0.91; 95 percent CI, 0.83‐0.99) and the 3‐day readmission measure (OR, 0.85; 95 percent CI, 0.74‐0.97). Also, our sensitivity analysis (Appendix S6‐b) examining the readmission rates by Medicare program and discharge location shows that the readmission rate in Medicare FFS was higher than MA across discharge locations.

4. DISCUSSION

Findings from this study contribute new and relevant information to the literature, as previous literature has not adequately controlled for difference in access and use of hospitals between Medicare FFS and MA in the context of all‐cause 30‐day readmissions. The present study demonstrated that patients in Medicare FFS were more likely to be readmitted within 30‐days after discharge than those in MA when they used the same hospital. Also, the study results show hospital‐level adjusted readmission rates were higher in Medicare FFS than MA in 62 of the 66 hospitals. These results suggest that older adults in Medicare FFS face higher risk of readmission compared with those in MA, despite using the same hospital.

The results from this study are in line with previous studies suggesting that the readmission rate in Medicare FFS is higher than in MA. Lemieux et al have reported that 30‐day readmission rates for patients in MA were nearly 13‐20 percent lower than those in Medicare FFS. ¹² A report from America's Health Insurance Plans, 30‐day, 90‐day, and 1‐year readmission rates were about 27‐29 percent lower in MA than in Medicare FFS. ¹¹ Our study extends this literature by examining whether the readmission rate between Medicare FFS and MA differs even when patients use the same hospital. Assessing whether readmission rates are different within the same hospital by payer type is important to improve our understanding around the risk of readmission and whether patients are receiving the same quality of care. Furthermore, it produces information on whether current monitoring system for quality of care in Medicare could be improved to provide more helpful information to Medicare beneficiaries.

Considering MA patients are more likely to be hospitalized at teaching, larger, and/or not‐for‐profit hospitals than Medicare FFS patients, ¹⁰ there are possibilities that differences in the hospital characteristics where patients obtain care contributes to the difference in readmission rate by Medicare program. The difference in quality of care and health outcomes by hospital characteristics are well documented in that hospital characteristics such as teaching and not‐for‐profit hospital, high staffing hospitals, and/or larger and urban hospitals are known to be associated with higher quality of care. ²¹ , ²² , ²³ The narrowed difference in the odds of readmission between Medicare FFS and MA in the unadjusted conditional logistic regression model compared with the unadjusted GEE model implies the potential possibility that the readmission rate difference by Medicare program could be partially explained by the difference in hospitals where patients receive care.

Also, the quality of post‐discharge care, which also involves transitional care, could explain the difference in the risk of readmission between Medicare FFS and MA even if patients receive same quality of care during the index hospitalization. Hospitals are adopting interventions such as promoting follow‐up visits and post‐discharge follow‐up phone calls to improve post‐discharge care. ²⁴ , ²⁵ , ²⁶ Along with such hospital efforts, compared to FFS beneficiaries, MA beneficiaries are anticipated to receive higher quality of post‐discharge care including better coordination and continuity of care with approaches such as gatekeeping system, robust information system, and network contracting. ²⁷ , ²⁸ , ²⁹ , ³⁰ Such difference therefore could be affecting quality of post‐discharge care, explaining the difference in the risk of readmission by Medicare program.

The difference in the risk of readmission between Medicare FFS and MA within the same hospital could be explained by differences in quality of care during index hospitalizations, including but not limited to discharge processes and transitional care. It is known that high‐quality inpatient and transitional care are important components to reduce readmissions. Friedman et al have found that hospitalized patients who experienced adverse safety events had higher risk of readmission. ³¹ Also, interventions to improve the discharge process and transitional care including Project Re‐Engineered Discharge and Project Better Outcomes for Older adults through Safe Transitions are known to be associated with lower risk of readmission. ²⁴ , ²⁵ , ²⁶ Furthermore, recent models which encourage providers to manage care coordination and other factors affecting risk of readmission, such as accountable care organizations, are effective in reducing readmissions. ³² , ³³

To our knowledge, there is no study which has directly examined whether the quality of inpatient care differs between Medicare FFS and MA, but there are several considerations that support this. MA beneficiaries are known to receive higher quality of care than those in Medicare FFS. Ayanian et al have reported that MA beneficiaries were more likely to receive appropriate ambulatory care than Medicare FFS beneficiaries based on breast cancer screening, diabetes care, and cholesterol testing for cardiovascular disease measures. ⁵ Moreover, a recent study by Timbie et al have found that MA outperformed Medicare FFS on clinical quality care measures including diabetes care, cancer screening, and immunization measures. ⁶ Recent studies by Spencer et al suggest that within‐hospital differences in quality exist across payer types by identifying patients with public insurance including Medicare and Medicaid were more likely to experience adverse patient safety events within the same hospital compared to those enrolled in private insurance. ¹³ , ¹⁴ While MA is not the same as private insurance, considering MA plans are offered and operated by private insurance companies, we might expect similar results for MA beneficiaries as for private insurers. Moreover, previous studies have found variations in available resources and characteristics of physicians by payor type. ³⁴ , ³⁵ These studies support possible differences in the quality of inpatient care by Medicare program, while patients use the same hospital.

There are also external factors after discharge that could affect the risk of readmission despite patients in different Medicare programs receiving the same quality of care within the same hospital. Previous studies have found that factors such as lack of social support, living situation, and community and neighborhood factors are associated with the risk of readmission. These could be considered as non‐medical factors outside of hospitals' control. ³⁶ , ³⁷ , ³⁸ However, our further analyses based on all‐cause 3‐ and 7‐day readmissions, which may more directly reflect associations with inpatient care, supports the possibility that the difference in quality of care by Medicare program within the same hospital contributes to the difference in readmission risks between different payers. Considering the similar and bigger differences in the risk of 3‐ and 7‐day readmissions between Medicare FFS and MA (Appendix S2) than for 30‐day readmissions, and that more than 30 percent of readmissions occur in the 7‐day time frame after discharge, it seems there are substantial opportunities to achieve further readmissions reduction by improving the quality of inpatient and transitional care. Additional studies to directly examine whether there is difference in quality of inpatient and transitional care, between Medicare FFS and MA and whether those differences explain the difference in readmission rates will be needed to develop methods to reduce readmissions rate.

Also, differences in unmeasured health characteristics between Medicare FFS and MA could explain some of the difference in the risk of readmission by Medicare program within the same hospital. Findings from this study suggest that adding adjustment for need characteristics narrows the difference in odds for readmission between Medicare FFS and MA. Considering this study used administrative data, there could be unmeasured health characteristics such as functional limitations that are associated with readmission risk but not systematically identified. ³⁹ Such characteristics could vary between patients in Medicare FFS and MA. Further studies adjusting for more health characteristics such as self‐reported functional limitations will be helpful to better understand the difference in risk of readmission between Medicare FFS and MA.

This study shows that hospital‐level readmission rates for MA were lower than those for Medicare FFS in all but four of the 66 hospitals in Wisconsin. However, there is variance in the readmission rate difference between Medicare FFS and MA across hospitals. The readmission rates of hospitals were 1.01‐2.78 times higher in Medicare FFS compared with MA. Future studies to examine factors and characteristics related to the variance in readmission rate difference by Medicare program across hospitals will be helpful to identify opportunities to not only reduce readmissions but also reduce the gap by Medicare program.

This study has the following limitations. First, the primary limitation of this analysis is the use of administrative data. Because administrative data have limited clinical information, there could be unmeasured factors that influence readmissions. We tried to overcome this limitation by including principal diagnosis for index hospitalization, patient‐level number of chronic conditions, number of hospitalizations in the baseline period, and days hospitalized in 12 months prior to index hospitalization. Due to lack of information on race/ethnicity in the Wisconsin data sources, we are unable to differentiate differences in sub‐populations by racial/ethnic background; findings also may not be generalizable to other states. However, considering Medicare is a federal program and majority of MA plans operate in multiple states, ⁴⁰ the study findings provide useful information to Medicare and MA health plans. Third, this analysis focused on data in the early years of several initiatives that CMS undertaken such as the Medicare Hospital Readmission Reduction Program, therefore, may not reflect the full effect of these initiatives. Lastly, we could not remove all planned readmissions with perfect accuracy using these data. However, by using the Yale formulation of all‐cause readmission, a method widely accepted by providers and payers, we were able to eliminate many types of planned readmissions.

In conclusion, this study presents evidence that risk of readmission is higher among Medicare FFS beneficiaries than among those compared to those in MA within hospitals in Wisconsin. CMS has undertaken several initiatives to reduce readmissions in Medicare such as Hospital Readmissions Reduction Program in 2012. While these efforts contribute to readmission reduction, this study indicates that further efforts in improving the quality of care and reducing gap in the care by payer type could lead to further reductions in readmissions. Additional studies to identify factors and healthcare services that cause differences in quality of care related to readmissions by Medicare program will be helpful for policy makers to develop strategies and incentives that influence hospitals, providers, and patients. Our study findings also support the need to improve current monitoring systems for hospitals by including payer‐specific data that could increase the transparency of hospital care and help consumers choose care venues.

Supporting information

Appendix S1‐S6

Jung DH, DuGoff E, Smith M, Palta M, Gilmore‐Bykovskyi A, Mullahy J. Likelihood of hospital readmission in Medicare Advantage and Fee‐For‐Service within same hospital. Health Serv Res. 2020;55:587–595. 10.1111/1475-6773.13315