If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Collaborative for Healthcare Delivery Science, Medical College of Wisconsin, Milwaukee, WI, USAInstitute for Health and Equity, Medical College of Wisconsin, Milwaukee, WI, USA
Address correspondence to L. Silvia Munoz-Price, MD, PhD, Division of Infectious Diseases, Department of Medicine, Medical College of Wisconsin, ID Division–HUB A8167, 8701 Watertown Plank Road, Milwaukee, WI 53226, USA.
Evidence suggests that quality, location, and staffing levels may be associated with COVID-19 incidence in nursing homes. However, it is unknown if these relationships remain constant over time. We describe incidence rates of COVID-19 across Wisconsin nursing homes while examining factors associated with their trajectory during 5 months of the pandemic.
Design
Retrospective cohort study.
Setting/Participants
Wisconsin nursing homes.
Methods
Publicly available data from June 1, 2020, to October 31, 2020, were obtained. These included facility size, staffing, 5-star Medicare rating score, and components. Nursing home characteristics were compared using Pearson chi-square and Kruskal-Wallis tests. Multiple linear regressions were used to evaluate the effect of rurality on COVID-19.
Results
There were a total of 2459 COVID-19 cases across 246 Wisconsin nursing homes. Number of beds (P < .001), average count of residents per day (P < .001), and governmental ownership (P = .014) were associated with a higher number of COVID-19 cases. Temporal analysis showed that the highest incidence rates of COVID-19 were observed in October 2020 (30.33 cases per 10,000 nursing home occupied-bed days, respectively). Urban nursing homes experienced higher incidence rates until September 2020; then incidence rates among rural nursing homes surged. In the first half of the study period, nursing homes with lower-quality scores (1-3 stars) had higher COVID-19 incidence rates. However, since August 2020, incidence was highest among nursing homes with higher-quality scores (4 or 5 stars). Multivariate analysis indicated that over time rural location was associated with increased incidence of COVID-19 (β = 0.05, P = .03).
Conclusions and Implications
Higher COVID-19 incidence rates were first observed in large, urban nursing homes with low-quality rating. By October 2020, the disease had spread to rural and smaller nursing homes and those with higher-quality ratings, suggesting that community transmission of SARS-CoV-2 may have propelled its spread.
The novel coronavirus (SARS-CoV-2), causative agent of the coronavirus disease 2019 (COVID-19), has spread quickly across the globe, becoming a major public health emergency. As of May 2021, more than 32 million COVID-19 cases and 584,975 related deaths have been reported in the United States.
Nursing homes have been particularly affected by the pandemic, which is probably due to their congregate nature and the older age and higher prevalence of comorbidities among their residents. Nursing home residents have experienced more than 650,000 confirmed COVID-19 cases and 132,305 related deaths, constituting one-fourth of the total deaths in the United States.
COVID-19 outbreaks continue to occur in nursing homes despite the use of various infection control measures, including social distancing, cancellation of group activities, elimination of visitation, and daily symptom screening.
Some studies have found that quality ratings, nursing home location, resident demographics, and staffing levels may be associated with higher incidence of COVID-19; however, these studies did not examine if these relationships remained constant over time or were impacted by local or state COVID-19 incidence rates.
As part of their ongoing response to the COVID-19 pandemic, the Centers for Medicare & Medicaid Services (CMS) has published weekly COVID-19 reports about morbidity and mortality across US nursing homes.
Using this unique data set, we aimed to characterize the associations between nursing home–level characteristics and nursing home–level COVID-19 incidence rates. Specifically, we examined these associations cross-sectionally and longitudinally throughout the first year of the COVID-19 pandemic in Wisconsin.
Methods
Study Data and Participants
We identified all skilled nursing facilities in the State of Wisconsin using the LeadingAge, Wisconsin Healthcare Association, and Wisconsin Department of Health Services nursing home directories.
Location information was obtained from the US Department of Agriculture's (USDA) database, whereas resident characteristics were obtained from Brown School of Public Health's database, Long Term Care Facility focus (LTCFocus).
Two measures of relative socioeconomic status—the area deprivation index (ADI) and the social vulnerability index (SVI)—were collected from the Neighborhood Atlas and the Centers for Disease Control and Prevention (CDC), respectively, and merged into the analytical file.
The CMS 5-star quality rating is an aggregate of 3 nursing home domains: health inspections, quality measures, and staffing rating. The health inspection rating is a composite score based on points assigned to deficiencies identified in each provider's 3 most recent state health inspections. The quality measure rating is a composite score based on performance in 15 quality measures. The staffing rating is based on total nurse staffing hours per resident per day. In addition to their 5-star quality ratings, we also obtained information on nursing homes' characteristics, including ownership status, total beds, average daily census, adjusted registered nurse hours per patient per day, and adjusted total staff hours per patient per day.
Location, designated as either urban or rural, was determined for each nursing home using the Rural-Urban Continuum Codes (RUCC) 2010 primary codes and the Washington, Wyoming, Alaska, Montana, and Idaho (WWAMI) Rural Health Research Center's categorization C.
Finally, county-level data regarding COVID-19 testing was used to calculate percentage positive values for Wisconsin from June 1, 2020, to October 31, 2020.
Nursing homes were divided into those with no reported cases and those with 1 or more cases. Categorical variables were reported as frequencies and percentages, whereas continuous variables were described as mean (standard deviation) or median [interquartile range (IQR)] as per their distribution (Kolmogorov-Smirnov test). We then analyzed nursing home characteristics based on COVID-19 cases (0 vs ≥1) using Pearson chi-square tests for categorical variables and Mann-Whitney U test for continuous variables. Nursing home COVID-19 incidence rates were calculated by dividing the number of COVID-19 cases by monthly occupied-bed days, and then multiplying the result by 10,000. County-level percentage positive values were calculated by dividing positive tests performed by total tests and multiplying the result by 100. Variables of interest and those with a P value of <.3 were considered for the multivariate analysis. A multiple linear regression model was performed to address the association between COVID-19 cases and rurality while controlling for possible confounders. The dependent variable was the number of cases in nursing homes from June 1 to October 31, 2020, which was then transformed into the natural logarithm of the number of cases + 1 to reduce the effect of outliers. Data were analyzed using RStudio, version 1.3 (R Foundation for Statistical Computing, Vienna, Austria), and SPSS, version 24.0 (IBM Corp, Armonk, NY).
Ethical Aspects
This study was approved by the appropriate Institutional Review Board with a waiver of informed consent.
Results
All 363 nursing homes in the state of Wisconsin were included in the study. Of those, 352 reported data to CMS between June 1, 2020, and October 31, 2020. Two hundred forty-six (67.3%) nursing homes had 1 or more COVID-19 case, with a total of 2459 cases occurring during the study period. Among nursing homes with at least 1 COVID-19 case, the median number of confirmed cases was 4 (IQR 2-14).
The characteristics of nursing homes, stratified by COVID-19 cases (0 vs ≥1), are shown in Table 1. Higher incidence of COVID-19 cases was associated with urban location (P = .065), higher number of beds (P < .001), average count of residents per day (P < .001), government ownership (P = .014), nonprofit ownership (P = .155), and low SVI score (Q1, P = .175). Of note, neither the 5-star rating or its individual components were associated with having ≥1 COVID-19 cases.
Table 1Characteristics of the Nursing Homes Included in the Study by Number of COVID-19 Cases
Given our focus on the trajectory of COVID-19 across rural and urban nursing homes, we further stratified the analyses by location (rural vs urban; Table 2). Consistent with their smaller size [median number of beds = 60 (IQR 50-90) vs 81 (IQR 50-112); P < .001] and lower occupancy rates [median count of residents per day = 46.4 (IQR 34.34-63.89) vs 59.24 (IQR 41.42-82.42); P < .001], rural nursing homes had a lower number of COVID-19 cases [median 2 (IQR 0-8) vs 3 (IQR 1-13); P = .018]. Rural homes tended to be located in more disadvantaged areas, as measured by ADI [median 7 (IQR 6-8) vs 5 (IQR 3-7); P < .001] and SVI [median 0.67 (IQR 0.52-0.77) vs 0.52 (IQR 0.32-0.77); P < .001]. Regarding ownership status, there was a larger share of governmental ownership in rural areas [33 (18.9%) vs 17 (8.8%); P = .005] and nonprofit ownership in urban areas [64 (33%) vs 41 (24.3%); P = .067]. Rural nursing homes also had a higher percentage of Medicaid residents than urban nursing homes [median 61 (IQR 51-71) vs 52 (IQR 37-64); P < .001].
With respect to quality ratings, the overall 5-star rating was higher among rural nursing homes [median 3.9 (IQR 2.9-4.7) vs 3.75 (IQR 2-4.8); P = .232], driven by a better health inspection rating [median 3 (IQR 2-4) vs 2.3 (IQR 1.4-4); P = .021]. However, other dimensions of the 5-star rating including quality [median 4.1 (IQR 3.2-4.8) vs 4.3 (IQR 3.5-4.9); P = .110] and staffing [median 3.7 (IQR 3-4.3) vs 4 (IQR 3-4.74); P = .110] were lower in rural nursing homes compared to urban ones.
Subcategorization of rural and urban nursing homes into those with no cases of COVID-19 and those with ≥1 case showed that an increased number of beds, higher average count of residents per day, governmental ownership, SVI, lower staffing score, and lower registered nurse staffing scores were significantly associated with having at least 1 COVID-19 case in both rural and urban settings (Table 2). In contrast to rural nursing homes, low quality rating, higher share of residents with Medicaid, nonprofit status, and high ADI were significantly associated with COVID-19 cases in urban nursing homes.
We also performed a multivariate linear regression (Table 3), examining the impact of location on incidence of COVID-19 (continuous variable) while controlling for the month of the pandemic, 5-star rating, ADI, number of beds, and month by location (our interaction variable). Over time, the number of cases across nursing homes increased (β = 0.14, P < .01). Number of beds in a nursing home was also associated with an increased number of cases; however, at a reduced level (β = 0.001, P < .01). Finally, a statistically significant interaction was identified between the progression of the pandemic from June to October 2020 and increasing COVID-19 cases in rural locations (β = 0.05, P = .03).
Table 3Multivariate Linear Regression Using Number of COVID-19 Cases as the Dependent Variable
Temporal Dynamics of COVID-19 Cases in Wisconsin Nursing Homes
Table 3 presents the incidence of COVID-19 cases in Wisconsin nursing homes during the study period. The highest COVID-19 incidence rates were observed during October 2020 (30.33 cases per 10,000 occupied-bed days). Urban nursing homes (Figure 1A and B) and those with a lower 5-star rating (overall score between 1 and 3; Figure 2A and B) initially experienced a higher incidence of COVID-19 earlier in the pandemic; subsequently, this trend reversed with increased incidence rates among rural nursing homes and nursing homes with higher 5-star quality ratings (overall score = 4 or 5) (Figure 1C–F and Figure 2A and B, respectively). Larger nursing homes were greatly affected at the start of the pandemic, but differences in incidence of the disease by nursing home size disappeared by the end of the observation period (Figure 2C). COVID-19 positivity rates appeared to move in a similar fashion to the spread of COVID-19 from nursing homes in urban to rural areas (Figure 1A–F).
Fig. 1Temporal analysis of various nursing home characteristics vs COVID-19 incidence. Top panel: variables analyzed include (A) location (urban vs rural), (B) overall quality rating, and (C) number of beds. Bottom panel: Calculated COVID-19 incidence in Wisconsin nursing homes during the study period. Graphs created using Excel 2016 (Microsoft Corp) and Inkscape 2020 (Inkscape Project).
Fig. 2Map of COVID-19 incidence vs county level positivity in Wisconsin nursing homes from June to October 2020. Nursing home COVID-19 incidence is indicated using heatmaps within circles (lowest to highest using the colors green, yellow, and red). County-level spread was stratified into tertiles (shaded based on category values) and is indicated by background county colors blue, yellow, and red (lowest to highest). Panels A through E show incidence of COVID-19 from June to October 2020, respectively. Outlined regions represent the 6 most populous metropolitan areas in the state of Wisconsin as of 2019 (United States Census Bureau). Solid lines were used to indicated named cities. Map created using Excel 2016 3D Maps (Microsoft Corp) and Inkscape 2020 (Inkscape Project).
This study examined the statewide incidence of COVID-19 among nursing homes in Wisconsin during 5 months of the pandemic, with a special focus on time trends by rural or urban location. Our findings indicate that COVID-19 cases were associated with urban location, larger nursing home size, governmental ownership and nonprofit ownership, lower 5-star quality rating, lower staffing rating, and lower nurse staffing rating. High ADI and increased share of residents with Medicaid were only found to be significantly associated with 1 or more COVID-19 cases in urban nursing homes. Subsequent temporal evaluation of incidence rates showed that at the beginning of the pandemic, nursing homes located in urban settings, those with a 5-star quality rating of 1-3, and those with high bed counts experienced at least 1 case of COVID-19. In contrast, by the end of our observation period, rural nursing homes, nursing homes with a 5-star quality rating of 4 or 5, and smaller nursing homes had similar or higher likelihood of having at least 1 case of COVID-19. Multivariate analysis supported this finding, showing that over the course of our study, rural location became increasingly associated with increased incidence of COVID-19. These findings are novel as they indicate that the relationships between nursing home characteristics and COVID-19 incidence fluctuated over time.
Our overall results are generally consistent with those of earlier studies.
Although we did not note an association between increased COVID-19 incidence and overall 5-star quality rating, this finding was inconsistently reported across other studies.
Another difference between our findings and previous investigations was the association between the percentage of Medicaid residents and COVID-19 cases; some studies showed that a higher percentage of Medicaid residents was associated with an increase in COVID-19 cases, a relationship that we only observed among urban nursing homes.
Governmental nursing home ownership and to a lesser extent nonprofit status were found to be significantly associated with increased COVID-19 incidence during analysis, again a finding that has been inconsistently documented in prior studies.
In contrast to the COVID-19 experience among the general population in the state, our analysis based on 2 proxies for social disadvantage—ADI and SVI—was mixed, showing a significant relationship between incidence rates and 1 quartile of SVI in rural and urban homes whereas high ADI was associated with increased COVID-19 cases in urban areas.
One plausible explanation for these inconsistent findings may relate to nursing home staff, many of whom may reside in different zip codes than the ones where the nursing homes are located.
Perhaps most striking was the temporal change observed for the association between 5-star rating and incidence rates of COVID-19 within individual nursing homes. In contrast to results from California where the relationships between nursing home quality and COVID-19 cases were reported to hold constant over the pandemic, our data show a reversal in trend over the 5-month study period.
Further, our multivariate analysis showed that as the pandemic progressed, rural location became increasingly associated with higher incidence of COVID-19 cases in nursing homes. A recent report published by the Wisconsin Department of Public Health provides context to this observation, suggesting that a large spike in confirmed SARS-CoV-2 infections in college and university settings in the month of September was followed by a dramatic rise in COVID-19 incidence in long-term care facilities.
report similar findings, arguing that increased levels of community transmission were responsible for outbreaks of COVID-19 in nursing homes. Mapping of COVID-19 county positivity rates alongside nursing home incidence seems to support these conclusions. Although nursing homes have restricted visitation to prevent introduction of COVID-19, staff continued to circulate between the community and nursing homes.
this movement might explain the connection observed between community incidence and nursing home outbreaks.
This study has several limitations. First, we did not examine facility testing capabilities or patients’ demographic and clinical characteristics, which could have influenced COVID-19 rates. We were also unable to account for physical characteristics of individual nursing homes such as room occupancy, building age, and ventilation. Because we used data provided by CMS, bias might have been introduced into our data set. Although CMS reporting is federally mandated and checked for accuracy, nursing homes during the initial period of data collection were not familiar with the survey forms used by the CDC and CMS and could have incorrectly entered data; therefore, we removed data prior to June 1, 2020. Next, our measures of socioeconomic disadvantage (ADI and SVI) are ecological variables that may not adequately reflect the situation of residents in a long-term care facility. Furthermore, although these measures consider varied economic, environmental, and social factors that impact communities, there is some overlap between categories.
Because we only analyzed data from Wisconsin nursing homes, our conclusions may not be generalizable to other areas. Lastly, we were unable to incorporate community transmission of COVID-19 into our univariate or multivariate analysis because of difficulty obtaining COVID-19 prevalence (longitudinally) at the zip code level.
Conclusions and Implications
Our findings suggest that although nursing home characteristics can predispose a particular nursing home to greater levels of COVID-19 risk, the location of a particular nursing home (rural vs urban) also impacts this risk. Further, the impact of rurality changed over time as the pandemic progressed, probably suggesting a role of community transmission.
Further investigations should be focused on examination of the factors that link community transmission, the home zip codes of the nursing home workers, and COVID-19 cases within nursing homes.
References
Centers for Disease Control and Prevention
Coronavirus disease 2019 (COVID-19) cases in the U.S.
This article was supported by a grant from the Advancing a Healthier Wisconsin Foundation (AHW; FP00018465/BU00033280). AHW had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
00760-X&id=gr1.jpg){kind=link}
00760-X&id=gr2.jpg){kind=link}