Abstract
Background:
Skilled nursing facility (SNF) care is frequently used after cardiac surgery, but the patterns and determinants of use have not been well understood. The objective of this study was to evaluate determinants and outcomes associated with SNF use after isolated coronary artery bypass grafting (CABG).
Methods:
A retrospective analysis of Medicare fee-for-service claims linked to The Society of Thoracic Surgeons clinical data was conducted on isolated CABG patients without prior SNF use in Michigan between 2011 and 2019. Descriptive analysis evaluated the frequency, trends, and variation in SNF use across 33 Michigan hospitals. Multivariable mixed effects regression was used to evaluate patient-level demographic and clinical determinants of SNF use and its effect on short and long-term outcomes.
Results:
In our sample of 8614 patients, the average age was 73.3 years, 70.5% were male, and 7.7% were listed as Non-White race. A SNF was utilized by 1920 (22.3%) patients within 90 days of discharge and varied from 3.2% to 58.3% across the 33 hospitals. Patients using SNFs were more likely to be female, older, non-white, with more comorbidities, worse cardiovascular function, a perioperative morbidity, and longer hospital lengths of stay. Outcomes were significantly worse for SNF users, including more frequent 90-day readmissions and emergency department visits, and less use of home health and rehabilitation services. SNF users had higher risk-adjusted hazard of mortality (HR=1.41, 95% CI: 1.26-1.57, p<0.001) compared to non-SNF users, and had 2.7-percentage point higher five-year mortality rate in a propensity matched cohort of patients (18.1% vs. 15.4%, p<0.001).
Conclusions:
The use of SNF care after isolated CABG was frequent and variable across Michigan hospitals and associated with worse risk-adjusted outcomes. Standardization of criteria for SNF use may reduce variability among hospitals and ensure appropriateness of use.
Keywords: Postacute care, coronary artery bypass grafting, outcomes
INTRODUCTION
Skilled nursing facility (SNF) care is an expensive and intensive form of transitional post-acute care that is used by almost 20% of cardiac surgical patients.1–3 Prior studies have demonstrated wide hospital-level variation in SNF use after isolated coronary artery bypass grafting (CABG). 1,4,5 As a component of overall episode of care spending, SNF use accounts for 5% of overall expenditures and a third of post-acute care-specific health care expenditures. Consequently, payment reform efforts have pressured providers to reduce unnecessary post-acute care spending, including SNF use, while still maintaining high-quality care.6,7
Despite the increasing emphasis on reducing SNF care in cardiac surgery, few have identified factors that contribute to SNF use and associated outcomes. While prior multicenter studies have predominantly leveraged administrative claims data to identify predictors of SNF use, these data lack important clinical granularity and risk prediction.5,8 Other investigators have leveraged clinical registry data to evaluate outcomes for patients discharged to SNF and other post-acute care facilities, but have had limited generalizability (e.g., often single-center studies).2,9 Moreover, while studies have already demonstrated hospital-level variation in SNF use, it is unclear if clinical factors account for variation in SNF use across hospitals.1,5 Finally, few studies have examined the association between SNF use and mortality beyond thirty days and on other forms of health care utilization such as readmissions and emergency department (ED) visits.2 Given the clinical, economic, and policy importance of SNF use after cardiac surgery, it is critical to develop and understand empirical evidence on its use.
The current study evaluated determinants and outcomes associated with SNF use after isolated CABG. To accomplish this aim, we leveraged a dataset of patient-level clinical data from The Society of Thoracic Surgeons (STS-ACSD) Adult Cardiac Surgery Database collected by the Michigan Society of Thoracic and Cardiovascular Surgeons Quality Collaborative (MSTCVS-QC) that was linked with Medicare fee-for-service claims available through the Michigan Value Collaborative (MVC) registry.
METHODS
This study was deemed exempt from human subject protections by the University of Michigan Institutional Review Board (HUM00175541). The underlying data in this study cannot be made available due to existing data use agreements.
Data Sources and Study Sample
The study sample initially included 19,783 beneficiaries undergoing isolated CABG procedures from 2011 to 2019 who had no prior SNF use within 6 months of index admission within Medicare administrative claims data from the MVC registry. All beneficiaries had continuous enrollment in Medicare fee-for-service Part A and B during six months prior to the index admission and 90-days following discharge. The MVC registry includes comprehensive administrative claims for Medicare beneficiaries, including SNF use and other health care utilization.10 Beneficiaries were excluded if they were admitted to non-Michigan hospitals or outside the study range of January 1, 2011 to December 31, 2018 for a sample of 15,886 beneficiaries. As previously described, a deterministic matching algorithm (e.g., demographic and clinical factors) was used to link beneficiary-level claims data with STS-ACSD data on 37,042 CABG patients housed at the MSTCVS-QC.11 The MSTCVS-QC data warehouse includes patient demographic and clinical factors, laboratory and testing, clinical care and in-hospital or 30-day outcomes. Beneficiaries were excluded if they were unable to be matched to STS-ACSD data, had missing data on STS-ACSD covariates used in the study, died in-hospital, were classified as a reoperation, or were unable to be linked to Medicare SNF files. The final analytic sample included 8,614 isolated CABG patients (Figure 1).
Figure 1.
Flow diagram of data sources and final analytic sample.
Skilled Nursing Facility Use
Medicare defines a SNF as a facility providing supervised nursing and therapy care, and Medicare SNF files were used to identify SNF claims.12 A binary indicator of SNF use was created for each patient that reflected whether the beneficiary had a SNF claim within 90-days of discharge (i.e., yes versus no SNF use). The number of days in SNF within 90-days of discharge were estimated based on admission and discharge dates from SNF claims.
Patient Demographic and Clinical Factors
Patient factors were drawn from the STS-ACSD data collected by the MSTCVS-QC. To be considered a candidate variable for the analysis, the variables must have been available across STS-ACSD data collection forms 2.73, 2.81, and 2.90 throughout the study period and were determined to be clinically meaningful. Demographic factors included patient age in years, self-reported gender (male vs. female), self-reported race category based on STS definitions (Caucasian, Black, Asian, Native American, Native Hawaiian/Pacific Islander, other, multiple races), which was categorized as White versus non-White due to small sample size of individual categories. Clinical factors included body mass index (BMI), history of chronic diseases (diabetes, chronic lung disease, peripheral vascular disease, cerebrovascular disease, or hypertension), lab values (creatinine hematocrit, white blood cell count), left ventricular ejection fraction (<30%, 30-39%, 40-49%, or +50%), previous myocardial infarction, admission status (elective, urgent, emergent/salvage), number of diseased vessels (1-2 vs. 3 or more), hospital length of stay and intensive care unit length of stay. A binary indicator of major perioperative morbidity was created based on the presence of any STS defined stroke, renal failure, prolonged ventilation, deep sternal wound infection, or reoperation.13 Patient-level STS predicted risk of operative mortality (PROM) and mortality or major morbidity (PRMM) scores were also included.
Outcomes
Mortality was evaluated for each patient using the Medicare Beneficiary Summary File, which contained dates of death for each beneficiary. Dates of death were calculated for each patient: (1) for patients with a date of death present - the number of days from date of discharge to date of death and (2) for patients without a date of death - the number of days from date of discharge to the end of the follow up period (December 31, 2019). Binary indicator of death at 90 days and five-years post-discharge were also created for each patient as well. If no date of death is present, they are assumed to be alive through the end of the follow up period. Additional measures of health care utilization within 90 days of discharge were also obtained from the MVC registry, and included any hospital readmission, ED visits, inpatient rehabilitation stay, home health use, or cardiac rehabilitation (CR) use.
Statistical Analysis
Univariate analysis evaluated the frequency of SNF use within the sample, including the proportion of patients with any SNF use and number of days in SNF. Bivariate analyses compared SNF use across patient factors and year. Chi-square and Kruskal Wallis tests were used to compare categorical and continuous variables across SNF use categories. Cochrane-Armitage tests of trend evaluated annual trends in SNF use. Box plots were created to visualize the distributions of STS PROM scores for SNF users and non-users.
Multivariable mixed effects logistic regression models were developed to identify patient-level factors predictive of SNF use and included a hospital random effect. Significant patient factors identified during bivariate analyses were selected as model covariates, and a stepwise backwards selection process was conducted by removing model covariates that were not statistically significant at an alpha level of 0.05. Penalized criterion, including pseudo AIC and BIC, were also used to evaluate model fit with lower values preferred for better model fit. Variables removed from the backward selection process were individually added back into the reduced model and evaluated for statistical significance and pseudo-AIC/BIC. The final model fit was assessed using a c-statistic and Pearson goodness-of-fit test. Predicted probabilities of SNF use were derived from the final model using only the patient-level fixed effects (i.e., not including the hospital random effect), which would subsequently be used to estimate hospital-level expected rates of SNF use conditional on the patient case-mix for a given hospital. Rates of observed SNF use were estimated as the number of SNF users divided by all patients, while expected rates were estimated as the sum of the predicted probability of SNF use for all patients divided by the number of patients. The strength and significance of the correlation between observed and expected SNF rates were evaluated using a Pearson correlation coefficient. The observed to expected SNF rate ratio, or O/E ratio, was also calculated for each hospital along with a 95% confidence interval to evaluate significant hospital-level deviations in SNF use from the overall state average.
Crude rates of 90-day mortality and healthcare utilization were estimated for the overall cohort and stratified by SNF use. Multivariable mixed effects logistic regression models were used to estimate the adjusted relative odds of 90-day mortality and healthcare utilization in SNF users compared to non-SNF users. Kaplan-Meier survival curves and a log-rank test were used to test for significant differences survival between SNF users and non-users. Cox-proportional hazards models were used to estimate the unadjusted and adjusted hazard ratio of mortality by SNF use, adjusting for significant patient factors. We also used logistic regression models to compare five-year mortality among SNF users and non-users unadjusted for patient factors, adjusted for patient factors, and within a propensity matched sample of patients using a nearest-neighbor approach with caliper=0.15.
All analyses were performed using SAS 9.4 (SAS Institute, Cary, NC) and Stata 17, and statistical tests were deemed significant at alpha<0.05 (two-sided)
RESULTS
A total of 1920 patients in our sample (22.3%) used a SNF within 90 days of discharge, with a median SNF admission stay of 15 days (25th percentile: 8 days, 75th percentile 23 days) (Figure S1 in the Supplement). Patients with SNF use were more likely to be female, older in age, non-white, more comorbid, have worse left ventricular ejection fraction, have had a major perioperative morbidity, and longer ICU and hospital lengths of stay (Table 1). Cochrane-Armitage test for trend found no significant trend in SNF use over time (z=−0.2927, p=0.770) (Figure S2 in the Supplement). A box and whisker plot showing the distribution of STS PROM for non-SNF users and SNF users can be found in Figure S3 in the Supplement. Mean STS PROM was higher for SNF users (3.8%, SD=4.2%) compared to non-SNF users (2.3%, SD=3.2%) (p<0.001).
Table 1.
Overall study sample patient demographic and clinical factors and stratified by skilled nursing facility use.
| Patient factors | Overall (n=8614) | Skilled Nursing Facility Use | p-value | |
|---|---|---|---|---|
| No (n=6694) | Yes (n=1920) | |||
| Age, mean (SD) | 73.3 (5.7) | 72.6 (5.4) | 75.5 (6.2) | <0.001 |
| 65-70 | 2727 (31.7) | 2334 (34.9) | 393 (20.5) | <0.001 |
| 70-75 | 2569 (29.8) | 2075 (31.0) | 494 (25.7) | |
| 75-80 | 1932 (22.4) | 1444 (21.6) | 488 (25.4) | |
| 80-85 | 1057 (12.3) | 670 (10.0) | 387 (20.2) | |
| 85+ | 329 (3.8) | 171 (2.6) | 158 (8.2) | |
| Male, n (%) | 6074 (70.5) | 4961 (74.1) | 1113 (58.0) | <0.001 |
| Non-White Race, n (%) | 667 (7.7) | 473 (7.1) | 194 (10.1) | <0.001 |
| Body Mass Index, mean (SD) | 30.1 (9.8) | 30.0 (10.6) | 30.5 (6.3) | <0.001 |
| Diabetes, n (%) | 4011 (46.6) | 2976 (44.5) | 1035 (53.9) | <0.001 |
| Chronic Lung Disease, n (%) | 5995 (69.6) | 4783 (71.5) | 1212 (63.1) | <0.001 |
| Peripheral Vascular Disease, n (%) | 1584 (18.4) | 1145 (17.1) | 439 (22.9) | <0.001 |
| Cerebrovascular Disease, n (%) | 2270 (26.4) | 1614 (24.1) | 656 (34.2) | <0.001 |
| Hypertension, n (%) | 7943 (92.2) | 6131 (91.6) | 1812 (94.4) | <0.001 |
| Creatinine mg/dL, mean (SD) | 1.14 (0.82) | 1.12 (0.82) | 1.19 (0.82) | <0.001 |
| White Blood Count x109/L, mean (SD) | 7.8 (3.0) | 7.7 (2.8) | 8.2 (3.6) | <0.001 |
| Hematocrit, mean (SD) | 38.5 (5.2) | 39.0 (5.0) | 36.9 (5.5) | <0.001 |
| Left Ventricular Ejection Fraction, n (%) | ||||
| 50% | 6280 (72.9) | 4994 (74.6) | 1286 (67.0) | <0.001 |
| 40-49% | 1266 (14.7) | 949 (14.2) | 317 (16.5) | |
| 30-39% | 675 (7.8) | 488 (7.3) | 187 (9.7) | |
| <30% | 393 (4.6) | 263 (3.9) | 130 (6.8) | |
| Status, n (%) | ||||
| Elective | 3216 (37.3) | 2600 (38.8) | 616 (32.1) | <0.001 |
| Urgent | 5194 (60.3) | 3958 (59.1) | 1236 (64.4) | |
| Emergent | 204 (2.4) | 136 (2.0) | 68 (3.5) | |
| Previous MI, n (%) | 4113 (47.8) | 3046 (45.5) | 1067 (55.6) | <0.001 |
| 3 or More Diseased Vessels, n (%) | 6669 (77.4) | 5138 (76.8) | 1531 (79.7) | 0.006 |
| Presence of Major Morbidity, n (%) | 903 (10.5) | 543 (8.1) | 360 (18.8) | <0.001 |
| Hours in the ICU, mean (SD) | 69.5 (96.7) | 62.9 (93.6) | 92.4 (102.6) | <0.001 |
| Length of Stay (in days), mean (SD) | 11.3 (6.0) | 10.5 (5.1) | 14.1 (7.7) | <0.001 |
Abbreviations: SNF=skilled nursing facility, SD=standard deviation, MI=myocardial infarction, ICU=intensive care unit
Independent patient factors predictive of SNF use from the multivariable mixed effects logistic regression model can be found in Table 2. Significantly greater relative odds of SNF use were found for patients who were female, older, of non-white race, greater body mass index, diabetic, had history of chronic lung disease, had history of cerebrovascular disease, lower left ventricular ejection fraction category, higher white blood count, lower hematocrit, had a previous myocardial infarction, were elective versus urgent status, had 3+ diseased vessels, had a major morbidity, and longer hospital length of stay. The c-statistic for the final model was 0.75 with a non-significant Pearson goodness-of-fit test (chi-square test p=0.99).
Table 2.
Patient factors independently associated with skilled nursing facility use after isolated CABG surgery.
| Patient factors | Crude OR (95% CI) | p-value | Adjusted OR (95% CI) | p-value |
|---|---|---|---|---|
| Age, per year | 1.09 (1.08-1.10) | <0.001 | 1.10 (1.09-1.11) | <0.001 |
| Female (vs. Male) | 2.08 (1.87-2.31) | <0.001 | 1.91 (1.69-2.15) | <0.001 |
| Non-White Race (vs. White Race) | 1.48 (1.24-1.76) | <0.001 | 1.33 (1.08-1.64) | 0.007 |
| Body mass index, per 5-unit increase | 1.02 (1.00-1.05) | 0.050 | 1.04 (1.01-1.07) | 0.016 |
| Diabetes | 1.46 (1.32-1.62) | <0.001 | 1.43 (1.27-1.61) | <0.001 |
| Chronic Lung Disease | 1.46 (1.31-1.63) | <0.001 | 1.29 (1.14-1.46) | <0.001 |
| Cerebrovascular Disease | 1.63 (1.46-1.82) | <0.001 | 1.38 (1.22-1.56) | <0.001 |
| White Blood Count x109/L, per 1 unit increase | 1.04 (1.03-1.06) | <0.001 | 1.02 (1.00-1.04) | 0.031 |
| Hematocrit, per 1 unit increase | 0.93 (0.92-0.94) | <0.001 | 0.97 (0.96-0.99) | <0.001 |
| Left Ventricular Ejection Fraction | ||||
| 50% | Referent | - | Referent | - |
| 40-49% | 1.30 (1.13-1.49) | <0.001 | 1.18 (1.00-1.38) | 0.048 |
| 30-39% | 1.49 (1.24-1.78) | <0.001 | 1.20 (0.98-1.47) | 0.083 |
| <30% | 1.92 (1.54-2.39) | <0.001 | 1.41 (1.10-1.82) | 0.007 |
| Previous myocardial infarction | 1.50 (1.35-1.66) | <0.001 | 1.17 (1.03-1.32) | 0.014 |
| 3 or More Diseased Vessels | 1.18 (1.04-1.35) | 0.010 | 1.13 (0.98-1.30) | 0.084 |
| Status | ||||
| Elective | Referent | - | Referent | - |
| Urgent | 1.32 (1.18-1.47) | <0.001 | 0.84 (0.73-0.96) | 0.011 |
| Emergent | 2.11 (1.56-2.86) | <0.001 | 1.34 (0.96-1.97) | 0.085 |
| Presence of Major Morbidity | 2.61 (2.26-3.02) | <0.001 | 1.37 (1.13-1.65) | 0.001 |
| Length of Stay, per day increase | 1.10 (1.09-1.11) | <0.001 | 1.07 (1.05-1.08) | <0.001 |
Abbreviation: SNF=skilled nursing facility, CABG = coronary artery bypass grafting, OR = odds ratio, CI = confidence interval)
Hospital-level rates of observed and expected SNF use rates and distribution of select patient characteristics can be found in Table S1 in the Supplement. Observed rates of hospital SNF use varied from 3.2% to 58.3%, while expected rates of SNF use based on the final logistic regression model varied from 17.4% to 35.1% (Figure 2). There was a non-significant, weak correlation between hospital observed and expected SNF rates (Pearson r=0.194, p=0.279). The hospital O/E ratios varied from 0.16 to 2.87 across hospitals, with 16 hospitals having O/E ratios that had confidence intervals not including 1.0 suggesting they were outliers in SNF use.
Figure 2.
Observed and expected rates of skilled nursing facility use across Michigan hospitals and observed to expected skilled nursing facility rate ratio with confidence intervals comparing hospitals to the state average (n=33 hospitals, ordered from lowest to highest observed use rate).
Crude 90-day mortality and health care utilization rates can be found in Table 3 along with risk-adjusted associations. There was no significant difference in 90-day mortality rates between SNF users and non-users (2.9% vs. 1.6%, adjusted odds ratio (aOR)=0.75, 95% confidence interval (CI): 0.52-1.09, p=127). However, SNF users were significantly more likely to be readmitted (aOR=1.72, 95% CI: 1.51-1.95, p<0.001) and have an ED visit within 90-days of discharge (aOR=1.26, 95% CI: 1.11-1.43, p<0.001). SNF users were less likely to use inpatient rehabilitation (aOR=0.38, 95% CI: 0.31-0.48, p<0.001), home health care (aOR=0.39, 95% CI: 0.34-0.44, p<0.001), and cardiac rehabilitation (aOR=0.46, 95% CI: 0.40-0.52, p<0.001).
Table 3.
Crude rates of 90-day mortality and health care utilization measures, and adjusted odds ratios comparing skilled nursing facility users and non-users.
| 90-Day Measure | Overall, n (%) | SNF Use, n (%) | Adjusted OR (95% CI) | p-value | |
|---|---|---|---|---|---|
| No | Yes | ||||
| Mortality | 162 (1.9) | 106 (1.6) | 56 (2.9) | 0.75 (0.51, 1.09) | 0.127 |
| Readmissions | 1888 (21.9) | 1207 (18.0) | 681 (35.5) | 1.72 (1.51, 1.95) | <0.001 |
| ED Visit | 2137 (24.8) | 1578 (23.6) | 559 (29.1) | 1.26 (1.11, 1.43) | <0.001 |
| Inpatient Rehab Use | 822 (9.5) | 675 (10.1) | 147 (7.7) | 0.38 (0.31, 0.48) | <0.001 |
| Home Health Use | 6843 (79.4) | 5551 (82.9) | 1292 (67.3) | 0.39 (0.34, 0.44) | <0.001 |
| Cardiac Rehab Use | 4519 (52.5) | 3913 (58.5) | 606 (31.6) | 0.46 (0.40, 0.52) | <0.001 |
All odds ratios were adjusted for patient gender, age, non-white race, BMI, diabetes, chronic lung disease, cerebrovascular disease, white blood count, hematocrit, left ventricular ejection fraction, previous MI, surgical status, 3+ vs 1-2 diseased vessels, presence of major morbidity, hospital length of stay, and hospital random effect.
Abbreviations: SNF = skilled nursing facility, ED = emergency department, OR = odds ratio
Survival was significantly lower for SNF users compared with non-users (p<0.001), Figure 3. The crude hazard for mortality in SNF users was 2.40 times higher compared to non-SNF users (hazard ratio (HR)=2.40, 95% CI: 2.17-2.65, p<0.001). After adjustment, the hazard of mortality was 1.40 times higher in SNF users compared to non-SNF users (HR=1.40, 95% CI: 1.26-1.57, p<0.001). Mortality at five years post-discharge was significantly higher in SNF users compared to non-users (27.8% vs. 13.7%, p<0.001) (Table 4). Adjusting for patient factors, SNF use was associated with a 4.1 percentage point increase in predicted five-year mortality (95% CI: 2.3% to 5.9%, p<0.001). In a propensity matched cohort of 3,842 patients, SNF use was again associated with a significant increase in five-year mortality of 2.7 percentage points (1.6% to 3.8%, p<0.001).
Figure 3.
Kaplan-Meier survival curves for skilled nursing users and non-users.
Table 4.
Crude, adjusted, and propensity matched comparisons of five-year mortality among skilled nursing facility users and non-users.
| Model | Five-year mortality | Absolute Difference in Mortality (95% CI) | p-value | |
|---|---|---|---|---|
| No | Yes | |||
| Crude | 13.7% | 27.8% | 14.1% (11.9% to 16.3%) | <0.001 |
| Multivariable adjusted | 15.6% | 19.8% | 4.1% (2.3% to 5.9%) | <0.001 |
| Propensity matched | 15.4% | 18.1% | 2.7% (1.6% to 3.8%) | <0.001 |
All models were adjusted for patient gender, age, non-white race, BMI, diabetes, chronic lung disease, cerebrovascular disease, white blood count, hematocrit, left ventricular ejection fraction, previous MI, surgical status, 3+ vs 1-2 diseased vessels, presence of major morbidity, hospital length of stay, and hospital random effect.
Abbreviations: SNF = skilled nursing facility
DISCUSSION
This large, statewide experience of Medicare beneficiaries undergoing isolated CABG surgery evaluated determinants and outcomes of post-discharge SNF use. After excluding patients admitted to a SNF prior to their CABG surgery, almost a quarter of patients used SNF within 90-days of discharge (median SNF length of stay of 15 days). Despite identifying significant demographic and clinical risk factors for SNF use in multivariable modeling of STS-ACSD linked Medicare claims, appreciable unexplained interhospital variation in rates of SNF use persisted. This unexplained variation has important implications given that SNF users had significantly worse outcomes, long-term mortality, and less frequent use of cardiac rehabilitation. Additionally, SNF use was associated with further increased resource utilization including higher rates of readmission and ED visit within 90-days.
Predictors of and Variation in SNF Use
Findings in this study contribute to the assessment of 90-day episode outcomes. To date, much of the literature has focused on evaluating extended care facility use, which does not clearly separate out SNF care. The use of linked Medicare data provides contemporary descriptive data on SNF use after CABG across patients and hospitals. Prior studies of STS-ACSD have identified patient factors predictive of discharge to extended care facilities after cardiac surgery, such as older age, female gender, the presence of comorbid conditions, and higher preoperative risk of mortality.2,9 Findings from this study identify similar patient demographic and preoperative factors predictive of SNF use after CABG, while also highlighting the importance of perioperative morbidity and index hospital length of stay in predicting SNF use. There is also existing evidence showing that facility-based post-acute care use is frequent and variable across hospitals after cardiac surgery.1,5 The present study extends these data by documenting wide variation in SNF use across hospitals, and that the observed variation in SNF use across hospitals is not well explained by pre and perioperative factors.
Findings from this study support the need to reduce interhospital variability in post-acute care use after cardiac surgery, including through bundled payment initiatives.14,15 Prior work within other surgical settings has demonstrated that SNF use may be influenced by non-clinical factors (e.g., local practice patterns, per capita local availability of SNFs.16–18 Findings from this study suggest that the use of SNF after CABG may be a product of a decision around the appropriate type of post-acute care for each patient, as SNF users were significantly less likely to use inpatient rehabilitation or home health services within 90 days of discharge. Reductions in potentially avoidable SNF use may be realized by identifying important non-clinical determinants, which could be leveraged in efforts to reduce discretionary care through bundled payments. However, more robust quantitative and qualitative studies would need to be conducted to inform appropriateness criteria for SNF use that may be used to guide decision-making processes around post-discharge acute care use.
SNF Use and Patient Outcomes
Efforts to reduce post-acute care face challenges; however, as our study highlighted significantly worse outcomes among patients with SNF use. Data on the short and long-term outcomes associated with SNF use after cardiac surgery have been limited up to this point, with single-center studies documenting that patients discharged to SNF have worse outcomes.2,9 This study similarly showed worse long-term mortality among SNF users, with an estimated 5-year mortality rate of 35% among SNF users compared to 16% among non-SNF users, even after risk-adjustment for demographic and clinical factors. Identifying patients at elevated risk for discharge to SNF and subsequently poor outcomes prior to surgery may allow for more thoughtful selection of patients for surgical or non-surgical intervention. The shared decision-making process should include discussions around possible SNF use, optimization prior to surgery, and care coordination after hospital and SNF discharge. In addition, the advantage of linked Medicare and STS-ACSD data available in Michigan revealed SNF users had significantly higher rates of hospital readmission and ED visits within 90-days of discharge, potentially attributed to patient (e.g., frailty, social support, longitudinal clinical status) risk factors.19,20 Lower participation in cardiac rehabilitation in SNF users may also contribute to worse long-term outcomes post-CABG, which has been shown to extend life and reduce hospitalizations.21 More research is needed to ensure that SNF care is used appropriately and not withheld from patients that may need it.
There are limitations that should be acknowledged when interpreting this study’s findings. First, data used in this study are limited to Michigan residents insured by Medicare fee-for-service insurance and may not apply to other states and insurance providers. Second, while the data are reflect all 33 non-federal hospitals performing cardiac surgery in the state of Michigan, findings from this study may lack generalizability outside of this state. Third, study data are limited to isolated CABG patients, and SNF use may be different among other cardiac surgical patients, particularly in valve-related procedures that are more often performed among patients undergoing concomitant CABG procedures. Fourth, despite leveraging a merged clinical-administrative data, the STS-ACSD database is limited in its tracking of some potentially important determinants of SNF use (e.g., socioeconomic indicators), which may confound the relationship between SNF use and clinical outcomes. Further, proxies of patient frailty (e.g., five-meter walk test) have high missingness in the STS-ACSD.22
CONCLUSION
The 90-day use of SNF care after isolated CABG was 22% but demonstrated high variability across Michigan hospitals with a significant amount of this variation unexplained after risk-adjustment. Further, SNF use was associated with worse short and long-term outcomes. Efforts to reduce SNF use following CABG should be cautious to avoid serious unintended effects on outcomes in this at-risk population. However, efforts at preoperative optimization and standardized postoperative SNF use selection could reduce resource utilization. Further efforts at care coordination may be needed after discharge from SNF to reduce long-term mortality.
Supplementary Material
What is known?
Skilled nursing facility (SNF) care is frequently used after cardiac surgery, but the patterns and determinants of use have not been well understood.
Payment reform efforts have pressured providers to reduce unnecessary post-acute care spending, including SNF use, while still maintaining high-quality care.
What the study adds?
Almost a quarter of patients undergoing coronary artery bypass grafting used SNF within 90-days of discharge.
Even after accounting for demographic and clinical factors, there was appreciable unexplained interhospital variation in SNF use.
Patients with SNF use after discharge had worse short term outcomes and higher long-term mortality.
Acknowledgements:
The Michigan Value Collaborative (MVC) is a Blue Cross Blue Shield of Michigan funded collaborative quality initiative that includes 103 acute care hospitals and 40 physician organizations across the State of Michigan. MVC provides Michigan hospitals and POs with payment and utilization data for an episode of care from paid, adjudicated claims. For this analysis, MVC provided us with insurance claims data from the following payer sources: Medicare fee-for-service. Support for the Michigan Value Collaborative and the Michigan Society of Thoracic and Cardiovascular Surgeons Quality Collaborative is provided by Blue Cross Blue Shield of Michigan as part of the BCBSM Value Partnerships program; however, the opinions, beliefs and viewpoints expressed by the author do not necessarily reflect those of BCBSM or any of its employees.
Disclosures:
Dr. Thompson, Dr. Nathan, Dr. Pagani, P. Theurer, Dr. Prager, and Dr. Likosky, received support from a contract from Blue Cross Blue Shield of Michigan. However, the opinions, beliefs, and viewpoints expressed by the authors do not necessarily reflect those of Blue Cross Blue Shield of Michigan or any of its employees. Dr. Thompson receives extramural support from the Agency for Healthcare Research and Quality (1K01-HS027830). Dr. Likosky receives research funding from the Agency for Healthcare Research and Quality (R01HS026003 AHRQ) and the National Institutes of Health (R01HL146619) and serves as a consultant for the American Society of Extracorporeal Technology. Dr. Pagani receives research funding from the Agency for Healthcare Research and Quality (R01HS026003 AHRQ) and the National Institutes of Health (R01HL146619) and serves as a non-compensated scientific advisor for Medtronic, Abbott, FineHeart, and CH Biomedical. The other authors report no conflicts.
Funding Statement:
Funding for this study was provided by the Frankel Cardiovascular Center McKay Grant fund and the Anderson Heart of a Champion Prize. The project was also supported by the Agency for Healthcare Research and Quality career development award (MPT, K01HS027830). The Michigan Value Collaborative and Michigan Society of Thoracic and Cardiovascular Surgeons Quality Collaborative are both supported by Blue Cross Blue Shield of Michigan through its Value Partnerships Initiative.
Abbreviations
- CABG
Coronary artery bypass grafting
- LVEF
Left ventricular ejection fraction
- MSTCVS-QC
Michigan Society of Thoracic and Cardiovascular Surgeons Quality Collaborative
- MVC
Michigan Value Collaborative
- STS PROM
Society of Thoracic Surgeons Predicted Risk of Mortality
- STS PRMM
Society of Thoracic Surgeons Predicted Risk of Mortality and Morbidity
- SNF
Skilled Nursing Facility
- STS-ACSD
Society of Thoracic Surgeons Adult Cardiac Surgery Database
REFERENCES
- 1.Thompson MP, Yost ML, Syrjamaki JD, Norton EC, Nathan H, Theurer P, Prager RL, Pagani FD, Likosky DS. Sources of Hospital Variation in Postacute Care Spending After Cardiac Surgery. Circ. Cardiovasc. Qual. Outcomes 2020;13:e006449. [DOI] [PubMed] [Google Scholar]
- 2.Edgerton JR, Herbert MA, Mahoney C, Armstrong D, Dewey TM, Holper E, Roper K, Mack MJ. Long-Term Fate of Patients Discharged to Extended Care Facilities After Cardiovascular Surgery. Ann. Thorac. Surg 2013;96:871–877. [DOI] [PubMed] [Google Scholar]
- 3.Sacks GD, Lawson EH, Dawes AJ, Weiss RE, Russell MM, Brook RH, Zingmond DS, Ko CY. Variation in Hospital Use of Postacute Care After Surgery and the Association With Care Quality. Med. Care 2016;54:172–179. [DOI] [PubMed] [Google Scholar]
- 4.Guduguntla V, Syrjamaki JD, Ellimoottil C, Miller DC, Prager RL, Norton EC, Theurer P, Likosky DS, Dupree JM. Drivers of Payment Variation in 90-Day Coronary Artery Bypass Grafting Episodes. JAMA Surg. 2018;153:14–19. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Chen LM, Norton EC, Banerjee M, Regenbogen S, Cain-Nielsen A, Birkmeyer JD. Surgical Post-acute Care Spending Driven by Choice of Post-Acute Care Setting Rather than Intensity of Services. Health Aff. . 2017;36:83–90. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Centers for Medicare & Medicaid Services,. Bundled Payments for Care Improvement Initiative: General Information [Internet]. 2017. [cited 2017];Available from: https://innovation.cms.gov/initiatives/bundled-payments/
- 7.Barnett ML, Mehrotra A, Grabowski DC. Postacute Care - The Piggy Bank for Savings in Alternative Payment Models? N. Engl. J. Med 2019;381:302–303. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Chen LM, Acharya Y, Norton EC, Banerjee M, Birkmeyer JD. Readmission Rates and Skilled Nursing Facility Utilization After Major Inpatient Surgery. Med. Care 2018;56:679–685. [DOI] [PubMed] [Google Scholar]
- 9.Edgerton J, Filardo G, Ryan WH, Brinkman WT, Smith RL, Hebeler RF Jr, Hamman B, Sass DM, Harbor JP, Mack MJ. Risk of not being discharged home after isolated coronary artery bypass graft operations. Ann. Thorac. Surg 2013;96:1287–1292. [DOI] [PubMed] [Google Scholar]
- 10.Ellimoottil C, Syrjamaki JD, Voit B, Guduguntla V, Miller DC, Dupree JM. Validation of a claims-based algorithm to characterize episodes of care. Am. J. Manag. Care 2017;23:e382–e386. [PubMed] [Google Scholar]
- 11.Thompson MP, Yaser JM, Fliegner MA, Syrjamaki JD, Nathan H, Sukul D, Theurer PF, Clark MJ, Likosky DS, Prager RL, et al. High Socioeconomic Deprivation and Coronary Artery Bypass Grafting Outcomes: Insights From Michigan. Ann. Thorac. Surg 2022;113:1962–1970. [DOI] [PubMed] [Google Scholar]
- 12.SNF Care Coverage [Internet]. [cited 2023 Mar 29];Available from: https://www.medicare.gov/coverage/skilled-nursing-facility-snf-care
- 13.Shroyer ALW, Coombs LP, Peterson ED, Eiken MC, DeLong ER, Chen A, Ferguson TB Jr, Grover FL, Edwards FH, Society of Thoracic Surgeons. The Society of Thoracic Surgeons: 30-day operative mortality and morbidity risk models. Ann. Thorac. Surg 2003;75:1856–64; discussion 1864-5. [DOI] [PubMed] [Google Scholar]
- 14.Liao JM, Gupta A, Zhao Y, Zhu J, Martinez J, Cousins DS, Navathe AS. Association Between Hospital Voluntary Participation, Mandatory Participation, or Nonparticipation in Bundled Payments and Medicare Episodic Spending for Hip and Knee Replacements. JAMA. 2021;326:438–440. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Centers for Medicare & Medicaid Services,. BPCI Advanced [Internet]. 2018. [cited 2019 Dec 2];Available from: https://innovation.cms.gov/initiatives/bpci-advanced
- 16.Konetzka RT, Stuart EA, Werner RM. The effect of integration of hospitals and post-acute care providers on Medicare payment and patient outcomes. J. Health Econ 2018;61:244–258. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Buntin MB, Garten AD, Paddock S, Saliba D, Totten M, Escarce JJ. How Much Is Postacute Care Use Affected by Its Availability? Health Serv. Res 2005;40:413–434. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Makam AN, Nguyen O, Xuan L, Miller ME, Goodwin JS, Halm EA. Factors associated with variation in long-term acute care hospital vs skilled nursing facility use among hospitalized older adults. JAMA Intern. Med 2018;178:399–405. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Shinall MC Jr, Arya S, Youk A, Varley P, Shah R, Massarweh NN, Shireman PK, Johanning JM, Brown AJ, Christie NA, et al. Association of Preoperative Patient Frailty and Operative Stress With Postoperative Mortality. JAMA Surg. 2020;155:e194620. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Shah SJ, Fang MC, Wannier SR, Steinman MA, Covinsky KE. Association of Social Support With Functional Outcomes in Older Adults Who Live Alone. JAMA Intern. Med 2022;182:26–32. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Heran BS, Chen JM, Ebrahim S, Moxham T, Oldridge N, Rees K, Thompson DR, Taylor RS. Exercise-based cardiac rehabilitation for coronary heart disease. Cochrane Database Syst. Rev 2011;CD001800. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Bergquist CS, Jackson EA, Thompson MP, Cabrera L, Paone G, DeLucia A 3rd, He C, Prager RL, Likosky DS, Michigan Society of Thoracic and Cardiovascular Surgeons Quality Collaborative. Understanding the Association Between Frailty and Cardiac Surgical Outcomes. Ann. Thorac. Surg 2018;106:1326–1332. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.