Associations of Skilled Nursing Facility Quality Ratings With 30-Day Rehospitalizations and Emergency Department Visits

Abstract

Skilled nursing facilities (SNFs) increasingly provide care to patients after hospitalization. The Centers for Medicare & Medicaid Services reports ratings for SNFs for overall quality, staffing, health inspections, and clinical quality measures. However, the relationship between these ratings and patient outcomes remains unclear. In this retrospective cohort study, we reviewed the electronic health records of 3,923 adult patients discharged from the hospital and admitted to 9 SNFs served by a health care delivery system. We used Cox proportional hazards models to examine associations between the overall quality and individual ratings and our primary outcomes of 30-day rehospitalizations and 30-day emergency department visits. Patients in higher-rated facilities had a 13% lower risk of 30-day rehospitalization than patients in lower-rated facilities (hazard ratio, 0.87; 95% CI, 0.76–0.99). The risk of emergency department visits was also lower for patients in facilities with a higher overall quality rating and a higher quality measures rating. Staffing and health inspection ratings were not associated with our primary outcomes. These findings may help inform providers and nursing home policy makers.

Introduction

Skilled nursing facilities (SNFs) provide care to many patients, usually older patients with complex conditions, and SNFs are increasingly used as a transitional care setting after hospitalization. A goal is to decrease time in the acute care setting in the hospital and reduce readmissions.¹ This population has high rates of acute and unplanned health care use, and approximately one-fifth of patients return to the hospital from SNFs.¹ In an analysis of Medicare data for SNFs across 5 states, 28.6% of patients were readmitted, and almost half the readmissions occurred within 7 days of hospital discharge. In addition, 3.8% of patients died, and only 60% ultimately returned home.² Because readmission is potentially avoidable for at least 40% of patients,^{1, 3–7} readmission rates and emergency department (ED) visits are increasingly used as markers of quality of care for this population.

In 2002, the Centers for Medicare & Medicaid Services (CMS) introduced Nursing Home Compare, a publicly available website that provides ratings of CMS-certified nursing homes across the United States. Nursing Home Compare details specific quality measures for residents receiving postacute or long-term care. This system was updated in 2008 to include the 5-Star Quality Rating System, which gives an overall quality rating and individual ratings for health inspections, staffing, and clinical quality measures (scale, 1–5 stars).⁸ Since its introduction, this system has influenced consumer demand for nursing homes, with a not-unexpected increase in demand observed for 5 star–rated facilities.⁹ However, a higher overall quality rating has not been associated with a better quality of life for long-term residents,¹⁰ nor do the ratings reflect consumer satisfaction ratings.¹¹

Few studies have evaluated clinical outcomes and the updated components of the 5-star rating system, and results were variable. Some studies showed that overall quality,^{12, 13} staffing,^13–15 and quality measures ratings^{12, 14} influenced risk of hospital readmission, whereas others showed that quality measures,¹⁶ staffing, and health inspection ratings did not.¹² These studies were usually limited to patients with specific medical conditions such as heart failure, did not control for provider care processes at the nursing home, or did not focus on ED visits. We conducted this study to determine the relationship between overall quality and individual component ratings of SNFs and the outcomes of rehospitalization and ED visit within 30 days for patients who were discharged to an SNF for postacute care. We also aimed to evaluate which components of the 5-star rating (ie, health inspection, staffing, and quality measures ratings) were significantly associated with these outcomes.

Methods

Study Design and Sample

This is a retrospective cohort study of adult (≥18 years) patients discharged from Mayo Clinic in Rochester, Minnesota to 9 local SNFs that are served by the Division of Employee and Community Health at Mayo Clinic from January 1, 2013, through March 31, 2016. We determined admission to the SNF by using the administrative database and confirmed admission with billing data. We included only patients whose hospital discharge date was the same as the SNF admission date to ensure that patients were directly discharged from the hospital to the SNF for postacute care. We excluded patients who declined consent for chart review for research purposes. All patients who are treated at our institution are asked to provide consent for chart review for medical research purposes. Patients can choose to opt out of this. A total of 246 patients selected to not have their health records reviewed for research purposes and so were not included in our study. Patient and SNF data were deidentified to maintain anonymity of patients and facilities. We obtained patient-level data from the administrative database and electronic health records (EHRs), including age, sex, length of stay, hospital stays or ED visits in the preceding 6 months, and comorbid conditions as categorized by the Charlson Comorbidity Index.¹⁷ Parts of the study data were collected and managed with electronic data capture tools (REDCap) hosted at Mayo Clinic.¹⁸ This study was approved by the Mayo Clinic Institutional Review Board.

SNF Data

We obtained ratings data for CMS-certified SNFs from the publicly available Nursing Home Compare website.¹⁹ We used ratings data from 9 of 13 SNFs within 25 miles of Rochester, Minnesota. We excluded 1 SNF affiliated with our institution because it used a different provider care model, 1 SNF because it was a new facility with no ratings available when this study was initiated, and 2 other SNFs because they were not covered by Mayo Clinic providers. The Division of Employee and Community Health provides similar postacute care at all 9 SNFs included in the current study; providers are from the Department of Family Medicine and the Division of Primary Care Internal Medicine at Mayo Clinic. On-site medical care at the SNFs was provided by a core group of geriatric medicine providers, which consisted of physicians, nurse practitioners, and physician assistants, from the Department of Community Health. After-hours coverage is provided via phone coverage and an on-call system with physicians from the Department of Community Health.

The data included overall quality ratings and individual ratings for health inspections, staffing, and clinical quality measures (Table 1). The health inspection rating, based on the results of the 3 most recent, annual independent inspections, forms the core of the overall quality rating.⁸ The staffing rating is based on registered nurse hours per resident per day and total staffing hours (registered nurses, licensed practical nurses, and nurse aides) per resident per day. The quality measures rating is based on a combined score generated from a subset of the quality measures recorded by CMS in the Minimum Data Set and Medicare claims data; these quality measures are specifically related to the clinical and functional status of the residents, and 9 measures pertain to residents receiving postacute care. The overall quality rating is a composite score; the health inspection rating forms the core of the overall quality rating and is adjusted higher or lower on the basis of the staffing and quality measures ratings.⁸ To control for variation among states, only a percentage of facilities in a given state can receive a 5-star health inspection rating, and the top 10% of facilities for health inspections rating are awarded a 5-star overall quality rating on the basis of their relative performance within the state.

Table 1.

Rating Categories and Components of CMS’s 5-Star Quality Rating System

Rating Categorya	Components
Overall quality	Composite score based on the health inspection, staffing, and quality measures ratings
Health inspection	Based on the results of the last 3 annual independent on-site inspections
Staffing	Based on RN hours per resident per day and total staffing (RN, LPN, and nurse aid) hours per resident per day
Clinical quality measures	Based on CMS quality measures related to a resident’s functional and clinical status

Abbreviations: CMS, Centers for Medicare & Medicaid Services; LPN, licensed practical nurse; RN, registered nurse.

^aEach category is rated on a scale from 1 to 5 stars.

Data from Centers for Medicare & Medicaid Services.⁸

Outcomes

Our outcomes were rehospitalizations and ED visits within 30 days of hospital discharge. The discharge date of the index hospitalization was the first day of the 30-day period. We used billing data to determine 30-day rehospitalization after admission to the SNF, and we confirmed rehospitalization by the presence of a discharge note in the EHR. ED visits were determined in a similar manner by using the billing codes in the EHR. We used International Classification of Diseases, Ninth Edition, codes through September 30, 2015, and International Classification of Diseases, Tenth Edition, codes afterward.

Statistical Analysis

Baseline demographic and clinical variables were summarized as number (percentage) for categorical variables or categorized and reported as number (percentage) for continuous variables. We used Cox proportional hazards models to examine unadjusted and adjusted relationships between overall quality ratings and 30-day rehospitalizations and ED visits. To deidentify SNFs and facilitate the analysis, we categorized SNFs as receiving 1- to 2-star ratings (“group A”) or 3- to 5-star ratings (“group B”). Analysis was conducted at the patient level and adjusted for age, sex, Charlson Comorbidity Index (<6 or ≥6), and number of ED visits and hospital stays in the previous 6 months (0, 1–3, or ≥4). For the analysis, patients were assigned to ratings groups A or B on the basis of their SNF’s rating on their hospital discharge day. To examine the association between the individual component ratings and outcomes, we created an adjusted model that included all 5 ratings and all covariates (age, sex, Charlson Comorbidity Index, prior 6-month hospital and ED visits). We also separately examined 5 univariate (unadjusted) models, each of which included 1 of the 5 ratings. The Charlson Comorbidity Index score was treated as a continuous covariate for analysis in the Cox proportional hazards model but was dichotomized (<6 or ≥6) only for data-presentation purposes. All statistical analysis was performed using JMP statistical software version 13 (SAS Institute Inc).

Results

Characteristics of the Patients

The study cohort included 3,923 adult patients who were discharged from Mayo Clinic to 9 SNFs in the greater Rochester area during the study period. Of these patients, 61.1% were women, 69.2% were 75 years of age or older, and 24.4% had a Charlson Comorbidity Index greater than 6 (Table 2).

Table 2.

Demographic and Clinical Characteristics of 3,923 Patients Admitted to Skilled Nursing Facilities, Stratified by 30-Day Hospital Readmission Status and Facility Rating

Characteristic	No. of Patients	Hospital Readmission Within 30 d (%)	Not Readmitted (%)
Age, y
<55	110	26 (23.6)	84 (76.4)
Group A	49	12 (24.5)	37 (75.5)
Group B	61	14 (23.0)	47 (77.0)
≥55 to <65	377	80 (21.2)	297 (78.8)
Group A	142	40 (28.2)	102 (71.8)
Group B	235	40 (17.0)	195 (83.0)
≥65 to <75	722	138 (19.1)	584 (80.9)
Group A	262	54 (20.6)	208 (79.4)
Group B	460	84 (18.3)	376 (81.7)
≥75 to <85	1,263	220 (17.4)	1,043 (82.6)
Group A	437	89 (20.4)	348 (79.6)
Group B	826	131 (15.9)	695 (84.1)
≥85	1,451	245 (16.9)	1,206 (83.1)
Group A	546	96 (17.6)	450 (82.4)
Group B	905	149 (16.5)	756 (83.5)
Sex
Female	2,396	391 (16.3)	2,005 (83.7)
Group A	850	158 (18.6)	692 (81.4)
Group B	1,546	233 (15.1)	1,313 (84.9)
Male	1,527	318 (20.8)	1,209 (79.1)
Group A	586	133 (22.7)	453 (77.3)
Group B	941	185 (19.7)	756 (80.3)
Charlson Comorbidity Index
>6	957	246 (25.7)	711 (74.3)
Group A	390	97 (24.9)	293 (75.1)
Group B	567	149 (26.3)	418 (73.7)
≤6	2,966	463 (15.6)	2,503 (84.4)
Group A	1,046	194 (18.5)	852 (81.5)
Group B	1,920	269 (14.0)	1,651 (86.0)
Hospital stays within previous 6 mo, No.
0	2,383	326 (13.7)	2,057 (86.3)
Group A	820	123 (15.0)	697 (85.0)
Group B	1,563	203 (13.0)	1,360 (87.0)
1–3	1,449	351 (24.2)	1,098 (75.8)
Group A	582	155 (26.6)	427 (73.4)
Group B	867	196 (22.6)	671 (77.4)
≥4	91	32 (35.2)	59 (64.8)
Group A	34	13 (38.2)	21 (61.8)
Group B	57	19 (33.3)	38 (67.7)
ED visits within previous 6 mo, No.
0	2,250	334 (14.8)	1,916 (85.2)
Group A	768	124 (16.1)	644 (83.9)
Group B	1,482	210 (14.2)	1,272 (85.8)
1–3	1,514	314 (20.7)	1,200 (79.3)
Group A	598	137 (22.9)	461 (77.1)
Group B	916	177 (19.3)	739 (80.7)
≥4	159	61 (38.4)	98 (61.6)
Group A	70	30 (42.9)	40 (57.1)
Group B	89	31 (34.8)	58 (65.2)

Abbreviation: ED, emergency department; group A, patients in lower-rated facilities; group B, patients in higher-rated facilities.

Outcomes

In total, 709 (18.1%) were readmitted within 30 days. Of these, 351 (49.5%) had from 1 to 3 hospital stays in the prior 6 months and 32 (4.5%) had 4 or more hospital stays. In terms of ED visits, 314 (44.3%) presented 1 to 3 times in the prior 6 months and 61 (8.6%) presented 4 or more times.

The rate of 30-day readmission was 16.8% for patients in higher-rated facilities for the overall quality rating compared with 20.3% for patients in lower-rated facilities (Table 3). In the adjusted analysis, patients in higher-rated facilities had a lower risk of 30-day rehospitalization than patients in lower-rated facilities (hazard ratio, 0.87; 95% CI, 0.76–0.99). Quality measures, health inspection, and staffing ratings were not significantly associated with rehospitalization.

Table 3.

Unadjusted and Adjusted Models for Risk of 30-day Rehospitalization

Rating	Rehospitalization Within 30 d, No. (%)	Not Admitted, No. (%)	Unadjusted HR (95% CI)	Adjusted HR (95% CI)a
Overall quality
Group A	291 (20.3)	1,145 (79.7)	1 [Reference]	1 [Reference]
Group B	418 (16.8)	2,069 (83.2)	0.81 (0.70–0.92)	0.87 (0.76–0.99)
Health inspection
Group A	542 (17.7)	2,516 (82.3)	1 [Reference]	1 [Reference]
Group B	167 (19.3)	698 (80.7)	1.06 (0.90–1.24)	1.02 (0.87–1.20)
Quality measures
Group A	85 (25)	255 (75)	1 [Reference]	1 [Reference]
Group B	624 (17.4)	2,959 (82.6)	0.67 (0.54–0.82)	0.83 (0.67–1.02)
Staffingb,c
Group A	18 (18.2)	81 (81.8)	1 [Reference]	1 [Reference]
Group B	687 (18.1)	3,117 (81.9)	1.03 (0.67–1.59)	1.00 (0.65–1.54)
Registered nurse staffingc
Group A	15 (25.9)	43 (74.1)	1 [Reference]	1 [Reference]
Group B	690 (17.9)	3,155 (82.1)	0.74 (0.450–1.21)	0.80 (0.49–1.32)

Abbreviation: group A, patients in lower-rated facilities; group B, patients in higher-rated facilities; HR, hazard ratio.

^aAdjusted for age, sex, Charlson Comorbidity Index, and hospital and emergency department visits within the previous 6 months.

^bRefers to total staffing hours (registered nurses, licensed practical nurses, and nurse aids).

^cData were missing for 20 patients.

ED Visits

In the adjusted analysis for overall quality rating, patients in higher-rated facilities had a 14% lower risk of ED visits than patients in lower-rated facilities (Table 4). Patients in higher-rated facilities for clinical quality measures had a 20% lower risk of ED visits than patients in lower-rated facilities. These findings remained significant after adjustment for patient factors. Health inspection and staffing ratings were not significantly associated with ED visits.

Table 4.

Unadjusted and Adjusted Models for Risk of ED Visit

Rating	ED Visit Within 30 d, No. (%)	Not Admitted, No. (%)	Unadjusted HR (95% CI)	Adjusted HR (95% CI)a
Overall quality
Group A	335 (23.3)	1,101 (76.7)	1 [Reference]	1 [Reference]
Group B	477 (19.2)	2,010 (80.8)	0.80 (0.70–0.91)	0.86 (0.75–0.98)
Health inspection
Group A	621 (20.3)	2,437 (79.7)	1 [Reference]	1 [Reference]
Group B	191 (22.1)	674 (77.9)	1.07 (0.92–1.25)	1.04 (0.89–1.21)
Quality measures
Group A	100 (29.4)	240 (70.6)	1 [Reference]	1 [Reference]
Group B	712 (19.9)	2,871 (80.1)	0.65 (0.54–0.79)	0.80 (0.66–0.98)
Staffingb,c
Group A	22 (22.2)	77 (77.8)	1 [Reference]	1 [Reference]
Group B	784 (20.6)	3,020 (79.4)	0.94 (0.63–1.40)	0.91 (0.61–1.36)
Registered nurse staffingc
Group A	16 (27.6)	42 (72.4)	1 [Reference]	1 [Reference]
Group B	790 (20.5)	3,055 (79.5)	0.79 (0.49–1.28)	0.85 (0.52–1.37)

Abbreviations: ED, emergency department; group A, patients in lower-rated facilities; group B, patients in higher-rated facilities; HR, hazard ratio.

^aAdjusted for age, sex, Charlson Comorbidity Index, and hospital and ED visits within the previous 6 months.

^bRefers to total staffing hours (registered nurses, licensed practical nurses, and nurse aids).

^cData were missing for 20 patients.

Discussion

This retrospective study of hospitalization, ED visits, and CMS ratings of SNFs showed that patients in SNFs with higher overall quality ratings (ie, 3–5 stars) had lower risk of 30-day rehospitalization and ED visit compared with patients in lower-rated facilities (1 or 2 stars), even when the models were adjusted for patient and clinical factors. Higher overall quality ratings were associated with a 13% lower risk of rehospitalization and a 14% lower risk of ED visits.

Previous studies of overall quality ratings focused on readmission risk. A study used unadjusted models and Medicare data from US patients with heart failure residing in SNFs to show that the risk of readmission was 8% higher for patients in overall 1 star–rated SNFs than patients in 5 star–rated SNFs, but this association was weaker in the fully adjusted model.¹² Notably, the investigators evaluated 90-day readmission and not 30-day readmission, as in the current and other studies. A recent national study of 30-day readmissions for nursing home patients with pneumonia, heart failure, and acute myocardial infarction (AMI) showed that the overall quality rating was significantly associated with readmission risk.¹⁵ An increase in the overall quality rating by 1 star was associated with a decrease in the standardized readmission ratio of 0.26% for patients with AMI, 0.73% for patients with heart failure, and 0.45% for patients with pneumonia.¹⁵ Similarly, in their study of an orthopedic cohort, Kimball et al¹³ used Medicare data to show that patients admitted to overall 5 star–rated SNFs after total knee arthroplasty and total hip arthroplasty had hazard ratios of 0.76 and 0.64, respectively, for readmission relative to patients admitted to overall 1 star–rated SNFs. Our study findings build on these results and, importantly, confirm lower risk of readmission in a broad clinical cohort that was not limited to specific clinical conditions. Overall, older patients composed a larger proportion of the readmitted cohort. Patients in the oldest age group who were readmitted accounted for 6% (245 of 3,923 patients) of the entire cohort, whereas patients in the youngest age group who were rehospitalized accounted for less than 1% (26 of 3,923 patients).

When we evaluated the individual components of these ratings, patients in higher-rated facilities for quality measures ratings also had a 20% lower risk of ED visits. In the adjusted models, quality measures rating was no longer significantly associated with 30-day rehospitalization. Staffing and health inspection ratings were not significantly associated with risk of rehospitalizations or ED visits. A possible explanation for these findings is that the effects of these individual component ratings on our outcomes may have been attenuated by assigning facilities to groups A and B, which was done to maintain anonymity, rather than analyzing outcomes according to individual star ratings. Previous studies report conflicting results about quality measures ratings. A prior study of patients with heart failure in 17 SNFs in western Massachusetts reported that quality measures ratings did not correlate with 30-day readmission rates (risk-adjusted readmission rate of 10.2% for 1 star–rated facilities vs 8.8% for ≥2 star–rated facilities, P=.54).¹⁶ In another cohort of heart failure patients, compared with patients in 5 star–rated facilities, patients in 1 star–rated facilities for quality measures had a 7% higher risk of readmission, although this risk was not associated with staffing or health inspection ratings.¹² In a national study of staffing that used readmission data for patients with pneumonia, heart failure, and AMI, a higher registered nurse staffing rating was associated with a 0.19% decline in hospital readmissions for patients with AMI, 0.40% for patients with heart failure, and 0.21% for patients with pneumonia.¹⁵ The association with staffing was also noted in the aforementioned orthopedic study: When nurse staffing ratings increased, 30-day readmissions decreased.¹³ In their large study of national Medicare data, Neuman et al¹⁴ used unadjusted models to determine that patients in SNFs with higher inspection, staffing, and certain quality measures ratings had lower risks of readmission and death at 30 days. These outcomes were attenuated after the models were adjusted for patient, hospital, and facility characteristics. Patients in SNFs with the highest health inspection ratings had a 23.7% risk of 30-day rehospitalization or death compared with 23% for patients in the lowest-rated SNFs (P<.001), but the association with the staffing rating was not significant after adjustment. Our study expands on this prior research and also examined the risk of ED visits, and unlike studies that used data from different health care systems and areas, our study mostly controlled for care processes at the clinical level because providers were from a single Department of Community Health and 1 health care system.

The associations of quality measures and overall quality ratings with ED visits are novel. In the United States, nursing home residents account for approximately 2 million ED visits annually.²⁰ To our knowledge, only 1 previous study recently evaluated the relationship between Nursing Home Compare star ratings and potentially preventable admissions and ED visits.²¹ Notably, that study evaluated a long-term care population, in contrast to our postacute population, and showed only a weak association between overall quality ratings and potentially preventable events.²¹ From our findings, one can hypothesize that day-to-day, facility-level factors, which are captured by quality metrics, have the most influence on patient outcomes such as ED visits. Until recently, nursing homes had no financial incentives (such as those offered to hospitals) to reduce their readmission rates²² or quality measures to evaluate outcomes, such as readmission or discharge to home or a long-term care facility.²³ However, with the introduction of the Skilled Nursing Facility Value-Based Purchasing Program, CMS now pays SNF providers on the basis of quality of care.²⁴ Hence, we need to quickly identify those residents most at risk for readmission and intervene accordingly. In addition, the SNF inspection process used to determine the individual ratings did not affect rehospitalization and ED visits in our cohort. This process is detailed, labor intensive, and costly as a team of health care professionals spends several days on site to perform an unannounced inspection,⁸ and our results suggest that inspection may not be so useful for the outcomes studied here.

Our study has limitations. This retrospective study was performed at a single health care delivery system that provides care to a population composed mostly of white patients, albeit at different SNFs with separate management structures, which may include differences in nursing and in facility or resource factors that cannot be controlled.²⁵ Also, patient-level factors, such as length of stay, timing of rehospitalization, and reason for index hospitalization, may have influenced the results and were not considered. This limits generalizability. However, use of a single health care system controlled as much as possible for variations in provider factors and health care processes that may have influenced the studied outcomes. Administrative data are also subject to inaccuracies in medical coding and nursing home reporting. Patients may have obtained care at another institution, and therefore data may be missing. Also, for analytic purposes, the ratings were obtained at a single time. In addition, because the health inspection rating reflects the results of the 3 most recent annual surveys, SNFs may have implemented changes to improve their practices that are not yet reflected in the rating. Because the health inspection rating substantially affects the overall quality rating, lower-rated facilities in our area may have performed better on the staffing and quality measures ratings, but this would not have improved their overall quality rating.

Our study has several strengths. It includes a large cohort served by a single health care delivery system, which may have controlled for variations in provider care processes that may have influenced the studied outcomes. This is a novel approach compared with previous studies, which evaluated data from several health care providers. Our institution’s EHR system allowed us to accurately capture clinical outcomes, including ED visits. All patients were transferred to a single center, which permitted us to capture data on all ED visits and hospital readmissions.

Conclusions

The risk of 30-day rehospitalization was lowest for patients in SNFs with higher overall quality ratings. The risk of ED visit was lower for patients in SNFs with higher overall quality and higher quality measures ratings, thereby indicating that patient-level factors are most influential. These findings will help inform acute care providers and nursing home administrators as we work towards fewer rehospitalizations and ED visits for patients. Our results need to be confirmed in other care settings and locations.

Abbreviations

AMI

acute myocardial infarction

CMS

Centers for Medicare & Medicaid Services

ED

emergency department

EHR

electronic health record

SNF

skilled nursing facility