Evaluation of risk-adjusted home time after acute myocardial Infarction as a Novel hospital-level performance Metric for Medicare beneficiaries

Abstract

Background:

Utility of 30-day risk-standardized readmission rate (RSRR) as a hospital performance metric has been a matter of debate. Home time is a patient-centered outcome measure that accounts for rehospitalization, mortality, and post-discharge care. We aim to characterize risk-adjusted 30-day home time in patients with acute myocardial infarction (AMI) as a hospital-level performance metric and evaluate associations with 30-day RSRR, 30-day risk-standardized mortality rate (RSMR), and 1-year RSMR.

Methods:

The study included 984,612 patients with AMI hospitalization across 2,379 hospitals between 2009 to 2015 derived from 100% Medicare claims data. Home time was defined as the number of days alive and spent outside of a hospital, skilled nursing facility (SNF), or intermediate/long-term acute care facility 30-days after discharge. Correlation between hospital-level risk-adjusted 30-day home time and 30-day RSRR, RSMR, and 1-year RSMR were estimated using Pearson’s correlation. Reclassification in hospital performance using 30-day home time vs. 30-day RSRR & 30-day RSMR was also evaluated.

Results:

Median hospital-level risk-adjusted 30-day home time was 24.0 days (range: 15.3–29.0). Hospitals with higher home time were more commonly academic centers, had available cardiac surgery and rehabilitation services, and had higher AMI volume and PCI utilization during the AMI hospitalization. Of the mean 30-day home time days lost, 58% were to intermediate/long-term care or SNF stays (4.7 days), 30% to death (2.5 days), and 12% to readmission (1.0 days). Hospital-level risk adjusted 30-day home time was inversely correlated with 30-day RSMR (r = −0.22, p<0.0001) and 30-day RSRR (r = −0.25, p<0.0001). Patients admitted to hospitals with higher risk-adjusted 30-day home time had lower 30-day readmission (Q1 vs. Q4: 21% vs. 17%), 30-day mortality rate (5% vs. 3%), and 1-year mortality rate (18% vs. 12%). Furthermore, 30-day home time reclassified hospital performance status in approximately 30% of hospitals versus 30-day RSRR and 30-day RSMR.

Conclusion:

30-day home time for patients with AMI can be assessed as a hospital-level performance metric using Medicare claims data. It varies across hospitals, is associated with post-discharge readmission and mortality outcomes, and meaningfully reclassifies hospital performance compared with the 30-day RSRR and 30-day RSMR metric.

Introduction

Over the last decade, improving quality of healthcare has been a major focus for health professionals and policy makers. The Hospital Readmissions Reduction Program (HRRP) was implemented with the Affordable Care Act to incentivize reduction in readmission burden and improve care quality among hospitalized patients by penalizing hospitals with higher than expected 30-day risk-standardized readmission rates (RSRR) for common conditions, which included acute myocardial infarction (AMI) and heart failure (HF).^{1, 2} However, the utility of 30-day RSRR as a performance metric for hospital care quality has been a matter of debate.^3–6 Several studies have demonstrated consistent reductions in 30-day readmission rates and associated health care costs for targeted conditions since implementation of the HRRP.^{5, 7, 8} However, others have raised concerns over the potential unintended consequences of HRRP due to selective prioritizing of readmission over other meaningful patient-reported and clinical outcomes, and the modest to weak association of 30-day readmission rates with adherence to performance measures and long-term clinical outcomes.^{3, 9–11} Importantly, the competing nature of mortality with 30-day readmission make it difficult to disentangle whether a lower readmission rate is driven by better care quality or a higher rate of post-discharge mortality.^{11, 12} Furthermore, the 30-day readmission metric does not account completely for variation in post-discharge care observed with use of skilled nursing or intermediate/long-term care facilities. A hospital performance metric that accounts for readmission, mortality, and other key elements of post-discharge care could overcome some of the above limitations of using readmission rate alone for incentivizing quality improvement.

“Home time” is defined as the time a patient spends alive and out of a healthcare institution i.e., hospital, intermediate/long-term care or skilled nursing facility (SNF). It is a novel patient-centered outcome that accounts for the competing risk of mortality on the risk of readmission.¹³ In recent patient-level analyses, home time was associated with mortality among patients with HF and stroke.¹⁴ However, the performance of home-time as a hospital-level quality metric in patients with common cardiovascular conditions such as acute MI has not been evaluated previously.

Accordingly, we characterized hospital-level 30-day post-discharge home time among patients hospitalized with acute MI and evaluated its association with the current Centers for Medicare and Medicaid Services (CMS) metrics of hospital performance such as the risk-standardized 30-day readmission and mortality rates.

Methods

The authors will not make the data, methods used in the analysis, and materials used to conduct the research available to any researcher for purposes of reproducing the results or replicating the procedure. All data analysis for this study was conducted at the University of Iowa and the University of Iowa Institutional Review Board approved the study and the need for consent was waived because of the use of encrypted patient identifiers.

Data Source & Study Population

This study utilized the CMS 100% sample Medicare Provider Analysis and Review (MedPAR) Part A from January 2009 to September 2015. The MedPAR data contain information on acute-hospitalizations and stays at rehabilitation or skilled nursing facilities for all Medicare patients with Part A coverage. Specifically, the data include information on admission date, discharge date, discharge disposition, International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) principal and secondary diagnosis codes, and Medicare reimbursement. MedPAR files include patient-level demographic data such as age, race, and sex. Characteristics of study hospitals were obtained from the American Hospital Association data for 2016. Data were used to classify hospitals as academic training centers, private or public, rural or urban, and geographic region.

Cohort derivation for our study is shown in Supplemental Figure I. Briefly, we included all patients age 66 years or older that were hospitalized between January 1, 2009 to September 2015 with a primary discharge diagnosis of acute MI (ICD-9 code: 410.xx) who were enrolled in Medicare fee-for-service for at least 1-year prior to the index AMI. For patients with multiple episodes of AMI hospitalizations within a 30-day period, only the first hospitalization was included. Patients discharged alive within 1 day of hospitalization, discharged against medical advice, discharged to hospice, or those receiving palliative care within the past 30-days of index hospitalization were excluded. Consistent with CMS criteria for excluding centers with <25 MI cases over 3-year period (<~8/year) for estimation of 30-day RSRR, we excluded hospitals who treated less than 60 AMI cases over the study period (~7 years). Patients who died during the hospital stay (n=78,213) were excluded from the home time and readmission calculation but included for the mortality calculation. This is consistent with the CMS approach for calculation of 30-day RSRR and 30-day risk-standardized mortality rates (30-day RSMR).

Primary Exposure Variable of Interest: 30-day Home Time

The primary exposure variable for this analysis was risk-adjusted 30-day home time. Home time was defined as the number of days alive and spent outside of a hospital, SNF, or intermediate/long-term acute care facility 30-days after discharge from an acute care hospital. Figure 1 includes scenarios on the relative impact of readmission, facility stays, and post-discharge mortality on 30-day home time. MedPAR claims file and SNF claims file were used to identify time spent in a facility, and any part of a day spent in the hospital or a facility was counted against home time. Hospital days during the index MI hospitalization were not included in home time calculations. Home-time was modelled using generalized linear mixed models with a log link and Poisson distribution using maximum likelihood estimates. In the above models, hospital site was included as a random effect and patient variables were included as fixed effects.

Figure 1: Conceptualization of 30-day home time.

30-day home time calculation for different post-discharge scenarios in following index hospitalization for acute myocardial infarction

The covariates used in risk-adjustment models were derived similarly to CMS models for the 30-day RSRR and 30-day RSMR estimations and included patient demographics, chronic condition categories obtained from CMS software, and other co-morbid conditions. A detailed description of candidate selection and risk-adjustment model derivation is described in Supplemental Methods. All 189 Condition Categories (CCs) were reviewed (Supplemental Table I). Briefly, the variable selection strategy for development of a risk-adjustment model followed selection strategies utilized by CMS.¹⁵ First, the cohort was divided into two equal-sized halves, one for model development and the other for model validation. Next, 100 bootstrap samples from the model development cohort were generated and logistic regression models with stepwise variable selection were constructed for each bootstrap sample to identify candidate risk-adjustment variables. Models were generated separately for each outcome (30-day and 1-year mortality, 30-day readmission, and 30-day home-time). For each outcome, variables that were selected in more than 80% of the bootstrapped regression models (general linear model for home time, logistic models for other binary outcomes) were selected for final models. The risk-adjustment models was subsequently evaluated by examining model fit statistics (e.g., area-under-the-curve for mortality and readmissions, correlation between actual and predicted values for home stay) (Supplemental Table II) in the validation cohort and demonstrated adequate and comparable performance to the development cohort (for each metric) and the CMS risk adjustment models (for 30-day RSMR and 30-day RSRR).¹⁵

Because of the focus on hospital-level performance, we also validated the reliability of hospital predicted mean 30-day mortality, 30-day readmission, and 30-day home time by correlating the mean hospital-level predicted values in the development and validation samples. The correlations were very strong with coefficients of 0.91, 0.90, and 0.90 for 30-day home time, 30-day readmission, and 30-day mortality, respectively. Final cohort variables and parameter estimates are highlighted in Supplemental Tables III-VI. Consistent with the CMS approach, socioeconomic status and race were not included in the risk adjustment. For patients who were transferred from one facility to another, the estimated 30-day home time and readmission were attributed to the discharging hospital, while estimated 30-day mortality was attributed to the original admitting hospital, which is consistent with the CMS approach. Expected and predicted home time were estimated using these models with and without the linear unbiased prediction modeling for estimation of random effects, similar to the previously described approach used by the CMS.^{16, 17} The risk-adjusted home-time was determined as the ratio of the predicted to expected home time multiplied by the overall unadjusted home time. Hospitals were stratified into quartiles of risk-adjusted 30-day home time with the lowest quartile identifying the lowest-performing and highest quartile identifying the highest-performing hospital based on the home-time metric.

Outcomes of interest: risk-adjusted readmission and mortality rates

The outcomes of interest were hospital-level 30-day RSRR, 30-day RSMR, and 1-year RSMR. Planned hospitalization events, as identified by CMS, that occurred within 30-days of discharge from an index acute MI hospitalization were excluded from 30-day RSRR calculation.¹⁸ Risk-adjusted readmission and mortality rates were estimated as described previously using hierarchical logistic regression models with each outcome as the dependent variable, patient demographic characteristics and chronic condition categories from the CMS and other co-morbidities as covariates, and a random intercept for the hospital. The detailed list of the covariates that were included for risk adjustment and their parameter estimates are included in Supplemental Tables IV-VI. The risk-adjusted rate of each outcome was determined using the ratio of the predicted to expected rate of the respective outcome multiplied by the overall unadjusted rate. Hospital performance based on a specific readmission or mortality metric was determined according to the data-derived quartiles of 30-day RSRR and 30-day RSMR with Quartile 1 identifying the highest-performing and Quartile 4 identifying the lowest-performing hospitals for each metric.

Statistical Analysis

Distribution of hospital-level 30-day risk-adjusted home-time, 30-day RSRR, 30-day RSMR, and 1-year RSMR in the study cohort were assessed using histogram bar-plots. Baseline hospital- and patient-level characteristics were reported across the categories of risk-adjusted 30-day home-time as medians for continuous variables and percentages for categorical variables. The baseline characteristics were compared across the study groups using ANOVA for continuous variables and chi-square test for categorical variables. Proportion of perfect 30-day home time lost to stay at an intermediate/long term care or skilled nursing facility, an acute care facility following readmission, or death were estimated for the overall cohort as well as across the risk-adjusted 30-day home time categories. Days lost were calculated for each patient and then summed and divided by perfect home time (30 days) to calculate mean values of home time lost.

Significant patient-level factors associated with 30-day home time were determined using exponentiated parameter estimates of patient demographics and clinical covariates included in the risk adjustment models. Similarly, temporal change in home time during the study period was also assessed by including a time variable (after vs. before 2012) in the risk adjustment model.

The 30-day readmission, 30-day mortality, and 1-year mortality among patients across the 30-day home time-based hospital groups were compared using an ANOVA test. Correlations between hospital level risk-adjusted 30-day home-time and other performance metrics (30-day RSRR, 30-day RSMR, and 1-year RSMR) were estimated using Pearson’s correlation coefficient. Reclassification in hospital performance using 30-day home time metric vs. the 30-day RSRR and 30-day RSMR metric was also evaluated. We compared hospital categorization into quartiles based on the 30-day home time (Q1 to Q4: low to high performing) with their categorization on the 30-day RSRR (Q1 to Q4: high to low performing). A hospital was assigned a positive score (+1 to +3) if they were in a higher quartile of 30-day home time compared to 30-day RSRR and a negative score (−1 to −3) if they were in a lower quartile of 30-day home time compared to 30-day RSRR. Hospitals that ranked in the same quartile on both metrics were assigned a score of 0. Reclassification scores above +1 was considered as meaningful up-classification in performance status and a score below −1 was considered as meaningful down-classification.

Sensitivity Analysis

Consistent with the CMS time frame for hospital performance assessment (3-year period), sensitivity analyses were performed to evaluate risk-adjusted 30-day home time and its association with other performance measures over the last 3 years of the study period (2013–2015). Furthermore, performance reclassification based on 30-day home time vs. 30-day RSRR and 30-day RSRR were evaluated among hospital stratified by their median bed size. All analyses were conducted using SAS software, version 9.4 Level of significance was set at p<0.05.

Results

The study included 984,612 patients from 2,379 facilities admitted for acute MI between 2009 to 2015. There was substantial hospital-level variability in 30-day home time as shown in Figure 2. Median risk-adjusted home time for the study population was 24.0 days (range: 15.3 – 29.0). Risk-adjusted 30-day home time was higher among hospitals with higher AMI volume (Supplemental Table VII). Among patient-levels factors, older age, presence of cardiac comorbidities such as heart failure, non-cardiac comorbidities such as history of chronic lung disorders, urinary tract infections, COPD, ESRD, and dementia were each associated with lower home time (Supplemental Table VIII). In contrast, male sex, history of PCI, and hospitalization during the latter half of the study period (after 2012) were each associated with higher 30-day home time. At 30-day follow-up, the highest proportion of days lost from perfect home-time were due to intermediate/long-term care or skilled nursing facility stay (58%, mean = 4.7 days) followed by mortality (30%, mean = 2.5 days) and acute care readmission (12%, mean = 1 day) (Figure 3). Of all the patients who died within 30-days post-discharge (N = 47,259), 59% died without a preceding readmission.

Figure 2: Hospital-level variation in 30-day home time and other performance metrics in the study cohort.

Distribution of hospital level 30-day risk-adjusted home time (Panel A), 30-day risk-standardized readmission rates (Panel B), and 30-day risk-standardized mortality rate (Panel C) in the study cohort

Figure 3: Components of the days lost from perfect 30-day home time.

Proportional distribution of perfect home time days lost to different factors within 30-day post discharge

Baseline patient- and hospital-level characteristics across hospital groups stratified by risk-adjusted 30-day home time are shown in Table 1 & 2. Patients hospitalized at hospitals with higher 30-day home time were younger and more commonly men. Although differences in prevalence of cardiovascular risk factors such as diabetes, hypertension, chronic kidney disease, and other co-morbidities were statistically significant due to large sample size, they were not clinically meaningful. At hospitals in the highest quartile of 30-day home time, patients were more likely to undergo PCI during the index admission and were more likely to be discharged home compared as with patients in the lower quartiles of 30-day home time. Among hospital-level characteristics, hospitals with higher 30-day home time were larger in size, were more likely to be teaching hospitals, and more likely to be located in an urban area. Availability of cardiac surgery and cardiac rehabilitation was also higher at hospitals with higher 30-day home time. Similar trends in patient-level and hospital-level characteristics were noted across 30-day home time hospital categories in sensitivity analyses limited to the last 3 years of the study period (Supplemental Table IX).

Table 1:

Baseline patient-level characteristics across quartiles of risk-adjusted 30-day home time

Characteristics	Low performing High performing				p-value
	Quartile 1 (lowest) N = 594 Median 21.78 days	Quartile 2 N = 595 Median 23.49 days	Quartile 3 N = 595 Median 24.43 days	Quartile 4 (highest) N = 595 Median 25.42 days
Number of AMI patients	171,997	271,905	324,037	294,886
Age (mean, SD)	80.81 (8.42)	79.11 (8.15)	78.57 (7.99)	78.30 (7.94)	<0.0001
Female Sex	52.95	48.64	47.24	46.03	<0.0001
Race					<0.0001
White	88.59	89.05	88.26	86.91
Black	7.00	7.01	7.42	7.19
Others	4.41	3.94	4.32	5.90
Diabetes	34.77	34.58	34.41	34.11	<0.0001
Heart Failure	42.19	36.88	35.97	35.99	<0.0001
CKD	23.28	23.02	22.44	23.16	<0.0001
Hypertension	76.12	76.51	76.65	76.11	<0.0001
Charleston co-morbidity index (mean, SD)	3.06 (1.73)	3.00 (1.72)	2.97 (1.70)	3.00 (1.71)	<0.0001
PCI during hospitalization	25.11	40.33	43.58	44.06	<0.0001
Discharge to home	39.25	54.26	60.38	65.14	<0.0001
Discharge to SNF/LTAC	27.06	20.01	15.97	12.59	<0.0001
Breakdown of the 30-day post-discharge period
	30-day risk-adjusted home time Quartile 1 N = 594	30-day risk-adjusted home time Quartile 2 N = 595	30-day risk-adjusted home time Quartile 2 N = 595	30-day risk-adjusted home time Quartile 2 N = 595
Days spent Home (mean, SD)	18.68 (13.18)	21.39 (12.30)	22.62 (11.70)	23.55 (11.15)	<0.0001
Days spent in acute care (mean, SD)	1.17 (3.28)	0.99 (3.00)	0.93 (2.91)	0.87 (2.81)	<0.0001
Days spent in Intermediate or Long-Term Care (mean, SD)	0.99 (4.48)	0.83 (4.12)	0.68 (3.69)	0.59 (3.43)	<0.0001
Days spent in Skilled Nursing Facility (mean, SD)	6.34 (10.87)	4.32 (9.36)	3.44 (8.48)	2.67 (7.57)	<0.0001
Days lost to death (mean, SD)	2.82 (8.65)	2.49 (8.17)	2.34 (7.95)	2.33 (7.94)	<0.0001

Abbreviations: AMI- acute myocardial infarction; SD – standard deviation; CKD – chronic kidney disease; PCI – percutaneous coronary intervention; SNF – skilled nursing facility; LTAC – long-term acute care

Table 2:

Baseline hospital-level characteristics across quartiles of risk-adjusted 30-day home time

Characteristics	Low performing High performing				p-value
	Quartile 1 N = 594	Quartile 2 N = 595	Quartile 3 N = 595	Quartile 4 N = 595
30-day home time, days, median (range)	21.78 (15.30 – 22.81)	23.49 (22.82 – 24.00)	24.43 (24.01 – 24.87)	25.42 (24.88 – 30.00)
30-day RSRR median (range)	0.19 (0.15 – 0.26)	0.19 (0.14 – 0.25)	0.18 (0.14 – 0.24)	0.18 (0.12 – 0.23)	<0.0001
30-day RSMR median (range)	0.11 (0.07 – 0.17)	0.11 (0.08 – 0.16)	0.10 (0.06 – 0.15)	0.10 (0.07 – 0.15)	<0.0001
1-year RSMR median (range)	0.21 (0.15 – 0.26)	0.20 (0.16 – 0.26)	0.20 (0.15 – 0.26)	0.20 (0.15 – 0.25)	<0.0001
Median hospital AMI volume (range)	179 (61 – 2290)	356 (61 – 2237)	420 (62 – 3079)	397 (61 – 2738)	<0.0001
Hospital size (median bed no, range)	152 (28 – 2013)	212 (24 – 2829)	245 (16 – 2338)	244 (29 – 2700)	<0.0001
Rural hospital location (%)	25.75	18.56	13.45	14.53	<0.0001
Teaching hospital (%)	52.91	56.01	64.66	64.71	<0.0001
Available cardiac surgery (%)	25.22	54.33	66.34	64.93	<0.0001
Cardiac rehab available (%)	71.30	78.67	84.24	80.76	<0.0001

Abbreviations: RSRR – risk-standardized readmission rate; RSMR – risk-standardized mortality rate; AMI – acute myocardial infarction; no – number; SD – standard deviation

Clinical outcomes on follow-up across 30-day home time-based hospital categories

Across hospital groups with increasing risk-adjusted 30-day home time, a graded decline was observed in the 30-day readmission (Q1 to Q4: 21% to 17%) and 30-day mortality rates (Q1 to Q4: 5% to 3%), and 1year mortality rates (Q1 to Q4: 18% to 12%) of patients hospitalized with acute MI (Figure 4). The correlation between hospital-level risk-adjusted 30-day home time and other measures of hospital performance (30-day RSRR, 30-day RSMR, and 1-year RSMR) are shown in Figure 5. A significant inverse correlation was observed between 30-day home time and 30-day RSRR (r = −0.25, p<0.0001), 30-day home time and 30-day RSMR (r = −0.22, p<0.0001), and 30-day home time and 1-year RSMR (r = −0.28, p<0.0001). In sensitivity analysis limited to the last 3 years of the study period, similar patterns of association were noted between 30-day home time and other performance metrics (Supplemental Table X).

Figure 4: Readmission and mortality rates for patients across different hospital-level 30-day home time categories.

Comparison of 30-day readmission, 30-day mortality, and 1-year mortality rates among patients hospitalized across increasing hospital-level 30-day home time categories

Figure 5: Correlation between hospital-level 30-day home time and other hospital performance metrics.

Correlation of risk-adjusted 30-day home time with other hospital-level performance metrics including 30-day risk-standardized readmission rate (Panel A), 30-day risk-standardized mortality rate (Panel B), and 1-year risk-standardized mortality rate (Panel C)

Reclassification of hospital performance using 30-day home time vs. 30-day RSRR and 30-day RSMR

In analysis comparing hospital performance based on 30-day home time vs. 30-day RSRR metrics, 29.5% hospitals had a meaningful reclassification in their performance status based on 30-day home time as compared with 30-day RSRR with a similar proportion of hospitals up-classified (15.1%) and down-classified (14.5%) (Figure 6). Similarly, when compared against 30-day mortality, use of 30-day home time metric led to meaningful reclassification in the performance status of 28.7% hospitals. In subgroup analysis stratified by median hospital bed size, similar patterns of reclassification in performance status of hospitals was observed with use of 30-day home time vs. 30-day RSMR or 30-day RSRR (Supplemental Figure II).

Figure 6: Reclassification in hospital performance by risk adjusted 30-day home time vs. other CMS performance metrics.

Reclassification in hospital performance with use of risk-adjusted 30-day home time as compared to risk-standardized 30-day readmission rate and risk-standardized 30-day mortality rate.

Abbreviations: CMS – Center for Medicare & Medicaid Services; RSRR – risk-standardized readmission rate; RSMR – risk-standardized mortality rate

Discussion

In this study of Medicare beneficiaries hospitalized with acute MI, we evaluated the performance of risk-adjusted 30-day home time as a metric for hospital performance. We observed several key findings in our study. First, there was a substantial hospital-level variability in 30-day home time; hospitals with higher risk-adjusted home time were large, more commonly academic, had higher acute MI case-volumes and greater availability of PCI and cardiac surgery. Second, stays in intermediate/long-term care and skilled nursing home facilities contributed the most to loss of perfect home time followed by mortality and acute care readmission. Third, 30-day risk-adjusted home time significantly correlated with 30-day, and 1-year RSMR and 30-day RSRR. Finally, 30-day home time reclassified hospital performance status for ~ 30% of hospitals as compared with 30-day RSRR and 30-day RSMR, the currently utilized CMS metric for hospital performance-based incentives. Taken together, our study findings highlight the potential utility of 30-day home time as an objective, graded, patient-oriented metric of hospital performance that can be derived from administrative claims data and accounts for differences in post-discharge rehabilitation/skilled nursing facility utilization, mortality, and readmission across hospitals.

Previous studies have evaluated patient-level associations of home-time post-discharge for patients with measures of functional status and mortality after hospitalization for stroke and HF.^{13, 14, 19} Fonarow, et al. demonstrated that in a population of older patients with stroke, higher home time was associated with better measures of functional status at 90 days and 1-year post discharge.²⁰ Similarly, in a cohort of hospitalized patients with acute HF, Greene et al. demonstrated a significant association between home time and long-term mortality at 1- and 2-year follow up post index hospitalization.¹⁴ To our knowledge, this is the first study to evaluate the potential role of 30-day home time as a hospital-level performance metric for patients hospitalized with acute MI using a nationally representative cohort of 100% Medicare beneficiaries.

Intermediate/long-term care and skilled nursing home stays most contributed to loss of perfect home time, followed by 30-day mortality and 30-day readmission. These results are similar to findings from Greene et al who observed a similarly high percentage of home time lost due to skilled nursing facility stays among patients discharged following a HF hospitalization.¹⁴ Furthermore, consistent with Greene et al’s findings in patients with HF, we also observed that cardiac comorbidities such as heart failure, chronic lung disease, chronic kidney disease, and dementia were significant predictors of lower home time.¹⁴ Future studies are needed to determine if improved management of these co-morbidities may help to improve home time in the AMI population.

There are several advantages to risk-adjusted 30-day home time as a metric for hospital performance. Home time is a patient-centered outcome that has been previously correlated with self-reported health, functional status, burden of depression, and difficulty in self-care.^{21, 22} It may be more relatable and easily understood by patients and caregivers given its simplicity and direct implications for quality of life. Targeting improvements in 30-day home time may incentivize better care quality over the long-term including better transition of care from inpatient as well as long-term care facilities, and greater use of disease modifying therapies that may lower mortality risk. Moreover, from a healthcare cost standpoint, home time may be more reflective of the overall burden of the health system by accounting for the days spent in health care facilities outside of the inpatient care, recurrent hospitalization events, and observation stays instead of only accounting for readmissions.^{23, 24}

Hospitals with the higher home time were larger, more commonly academic and in urban location, had higher availability of cardiac surgery and cardiac rehabilitation, and had higher AMI case volumes with greater proportional use of PCI during the index hospitalization. This representation of high performing hospitals based on home time for acute MI is consistent with the prior literature for patients with acute stroke.²⁰ It also fits our intuitive understanding, as structural quality measures such as hospital volume, PCI availability, and PCI use in patients with acute MI have been associated with better outcomes.²⁵ In contrast, previous studies comparing high vs. low performing hospitals based on the 30-day RSRR metric for MI demonstrated no differences care quality metrics for AMI.⁹ Consistent with these observations, we observed a substantial degree of reclassification in the hospital performance status based on 30-day home time vs. 30-day RSRR.

Our study findings have important health policy implications. Reducing hospital readmission rates has been the central focus of CMS policies and efforts aimed at improving care quality and lowering costs associated with hospitalizations for common acute conditions.²⁶ Accordingly, hospital-level 30-day RSRR is the main driver of performance-based financial incentive based on the current CMS policies. In 2019, the CMS penalized 2,583 Medicare hospitals (83% of penalty-eligible hospitals), amounting to an estimated $563 million in withheld Medicare payments.²⁷ Since implementation of the readmission penalties in 2012, consistent declines in 30-day readmission rates have been noted across all targeted conditions.⁸ However, some concerns have been raised regarding the use of 30-day readmission rates as a metric of hospital performance.^{28, 29} Hospital performance based on 30-day RSRR for cardiovascular conditions such as acute MI and HF has not been associated with objective measures of in-hospital care quality and long-term clinical outcomes.^{9, 10, 30} Furthermore, the factors underlying the observed decline in hospital-level 30-day RSRR and the extent to which variation in hospital-level 30-day RSRR is associated with other meaningful clinical and patient-centered outcomes in not well-established.^{31, 32} Findings from our study suggest that risk-adjusted 30-day home time may be a potential performance metric that better captures the risk of mortality and morbidity including repeated hospitalizations, longer length of stay, need for inpatient rehabilitation or skilled nursing facility, and is associated with both 30-day and 1-year clinical outcomes.

Similar to 30-day home time assessed in our study, the Medicare Payment Advisory Commission (MedPAC) has evaluated “Healthy Days at Home” (HDAH) as a population-based health metric to better define how healthcare organizations care for patient populations.³³ MedPAC analyzed this metric across all Medicare beneficiaries and estimated 1-year HDAH across a population regardless of whether the patient had an acute hospitalization. There was relatively little variation in 1-year HDAH across beneficiaries and differences in HDAH were largely driven by risk of mortality. In contrast, we observed a greater hospital-level variation in 30-day home time after hospitalization for AMI which was largely driven by post-acute rehabilitation or skilled nursing facility stay, followed by mortality. Several factors may underlie these differences between 1-year HDAH assessed by MedPAC versus 30-day home time assessed in our study. A home time metric may have more variability for patient’s following an acute hospitalization, particularly over short-term follow-up whereby differences in home time may be more reflective of care differences across hospitals. In contrast, when estimated for all Medicare beneficiaries over a 1-year period, home time or healthy days of the population might be largely reflective of the healthier individuals and less driven by specific acute care episodes. Taken together, our study findings suggest that home time may be a more useful as a hospital performance metric when estimated over short-term periods following acute care episodes such as AMI hospitalization as compared with a population health metric.

We observed a modest but statistically significant increase in 30-day home time over time during the study period (after 2012). This is consistent with the decline in 30-day readmission and mortality rates for patients with AMI that has been reported in the since 2012 across several studies and highlights the potentially modifiable nature of the home time metric.^{8, 34} It is noteworthy that the largest contributor to loss of home time is the long-term care and skilled nursing facility stays. This is particularly relevant as skilled nursing and long-term care stays are important contributors to variation in post-discharge outcomes and Medicare spending in United States.^35–37 Future studies are needed to determine if implementation of home-time metric would lead to a more judicious use of in-patient rehabilitation/SNF or preferential use of home health care over these facilities. It is plausible that use of home time metric disincentivize hospitals to refer patients to rehabilitation or skilled nursing facilities who would normally require these services at discharge. However, that would be short-sighted if such patients get readmitted or experience mortality – which are also accounted for in the home time metric.

This study is not without limitation. First, our study findings are based on observational analysis of administrative claims data for Medicare beneficiaries and may not be generalizable to younger cohorts or private payor-based systems. Our observations may also be influenced by coding errors and missingness in co-morbidities/diagnosis captured only in the outpatient setting. We also do not have outpatient and carrier file data on the Medicare beneficiaries, which includes information on emergency department visits. Second, owing to the nature of the administrative claims data, we could not completely account for the severity of index MI event and other co-morbidities in the risk adjustment models for home time. Third, the effect of external factors such as availability of social support, home health services, geographic location, and caregiver assistance on the home-time metric could not be assessed. Patients may not have equal access to skilled nursing and rehabilitation facilities, and this may have impacted discharge disposition. Finally, our study could not address the patient-specific value of home-time as a meaningful measure of hospital care quality.

In conclusion, home time is a novel, patient-centered quality metric for hospital performance in acute MI that can be derived from administrative claims data. Contributors to loss of home time were predominantly due to rehabilitation and skilled nursing facility stay and death, with less contribution from readmission. 30-day home time correlated with 30-day RSMR and RSRR and meaningfully reclassified hospital performance status in 30% of hospitals as compared to 30-day RSRR and 30-day RSMR, which are current CMS metrics for hospital performance-based incentives.

Clinical perspective.

What is new?

30-day home time, a novel, patient-centered quality metric for hospital performance in acute myocardial infraction, can be derived from administrative claims data and is associated with short-term and long-term clinical outcomes, post-discharge care utilization, and meaningfully reclassifies hospital performance compared with the 30-day readmission and mortality metric.

What are the clinical implications?

Risk-adjusted 30-day home time may be a potential hospital-level performance metric that better captures the risk of mortality and morbidity in patients discharged following an acute myocardial infarction.

Abbreviations:

RSRR

risk-standardized readmission rate

RSMR

risk-standardized mortality rate

AMI

acute myocardial infarction

SNF

skilled nursing facility

CMS

Centers for Medicare and Medicaid Services

MedPAR

Medicare Provider Analysis and Review

CC

Condition Categories

COPD

chronic obstructive pulmonary disease

ESRD

end-stage renal disease

PCI

percutaneous coronary intervention

HF

heart failure

HDAH

healthy days at home

MedPAC

Medicare Payment Advisory Committee