Association of Social Isolation of Long-term Care Facilities in the United States With 30-Day Mortality

Key Points Question To what extent are US long-term care facilities located in socially isolated neighborhoods and what is the association with the risk of 30-day mortality? Findings This cross-sectional study found that long-term care facilities in the US were approximately 8 times more likely to be located in neighborhoods with the highest percentage of individuals aged 65 years or older living alone compared with neighborhoods with the lowest percentage. Long-term care facilities in socially isolated neighborhoods were associated with increased risk of 30-day all-cause mortality among residents. Meaning The findings suggest the need for special attention and strategies to keep long-term care residents connected to their family and friends.


Introduction
In mid-March of 2020, long-term care (LTC) facilities were required by the Centers for Medicare and Medicaid Services to ban all nonessential visitors and group activities in response to COVID-19 safety rules. Within a few months of the lockdown, LTC staff and clinicians began reporting increased levels of depression, anxiety, worsening dementia, and failure to thrive in residents. 1 These accounts have drawn attention to the role of social isolation in LTC facilities.
Social isolation-generally defined as having few social network ties or infrequent social contact-is an important public health concern that affects many older adults. 2 Living alone or living in neighborhoods with a high proportion of single person households have been found to be predisposing factors to social isolation. 3 In the community setting, 28% (14.7 million) of all older adults aged 65 years or older live alone (5.0 million men, 9.7 million women). 4 By age 85, 39% of older adults live alone. 5 Socially isolated adults experience high rates of negative health outcomes, including premature mortality and a 50% increase in the risk of dementia. 6,7 Living alone is a strong risk factor for LTC placement. 8,9 Growing evidence suggests differences in social isolation by race/ethnicity, 10 although an association between mortality risk and social isolation has been found in both non-Hispanic White individuals and African American individuals. 11 In the LTC setting, social isolation is less well defined and rarely studied. By definition, LTC residents live in a facility with others so social isolation often refers to the loss of personal connection to family and friends outside the facility. Concern about social isolation is important to LTC residents: proximity to prior residence is the strongest factor associated with the choice of LTC facility. 12 Socially isolated LTC residents may be at higher risk of negative outcomes. In a study of 323 LTC residents with advanced dementia, 88% received outside visitors for a least 1 hour a week, but 12% never received any visitors during an 18-month period. 13 Among the LTC residents who had no visitors, reports were higher of pain, pressure ulcers, and dyspnea compared with residents with regular weekly visitors. At least 1 study has found evidence of shorter survival time among LTC residents who were admitted to facilities located in socially isolated areas at the county level. 14 To our knowledge, there is no information on the extent that LTC facilities are located in socially isolated neighborhoods. The objectives of this study were to characterize the social isolation of LTC facilities in the US and to assess short-term (30-day) all-cause mortality risk in residents within the LTC facilities that experience the social isolation. We posit that the social isolation of LTC facilities may be important if it is associated with barriers to connections with friends and family and negative health risks to their residents that have been documented in the community setting.

Data Sources
We linked the following data sets: (1) 2011 Certification and Survey Provider Enhanced Reporting Data detailing the location of LTC facilities were extracted from the CASPER data set. CASPER is a repository of the validated, federally mandated, on-site surveys of all Medicare-and Medicaidcertified long-term care facilities in the US. The LTC surveys are conducted by state survey agencies every 15 months at minimum, or in the event that a complaint is filed. CASPER data include information about the facility operational characteristics and aggregate patient characteristics. The reliability and validity of the CASPER data have been demonstrated previously. [15][16][17] Data on the zip code Tabulation Area (ZCTA)-level percentage of older adult residents who were aged 65 and older and living alone came from the decennial 2010 US Census. 18 All other ZCTA-level aggregate data on demographic and socioeconomic characteristics were drawn from the 5-year combined 2009 to 2013 ACS, centered in 2011. 19 The ACS and US Census data were abstracted using American Fact Finder from the US Census Bureau. 20 The MDS is a federally mandated clinical assessment of residents living in LTC facilities that are certified by Medicare or Medicaid. The MDS data contain information about active diagnoses, psychosocial well-being, and physical functioning at admission, quarterly intervals, and when there are changes in health status. The reliability and validity of the MDS data have been demonstrated previously. 21,22 Medicare data came from the Medicare program and included enrollment information, including the date of death. The Medicare Parts A and B claims data included information about diagnoses and health care utilization.
Because all data used in this study were deidentified, this study was approved and classified as exempt by the Northeastern University institutional review board and informed consent was waived, in accordance with 45 CFR §46. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cross-sectional studies.

Study Population
Out of a total of 33 120 ZCTAs in the US, 8652 ZCTAs had at least 1 LTC facility. This cross-sectional study included 14 224 LTC facilities and 730 524 LTC residents from these 8652 ZCTAs in 2011.
Detailed information about the characteristics of LTC residents was reported in a previous study. 14

Measures Social Isolation of the ZCTAs
Our primary exposure variable of interest was the social isolation in ZCTA. The degree of social isolation is defined as the percentage of households in the ZCTA with individuals aged 65 years or older who lived alone. 23 The degree of social isolation was calculated by dividing the number of individuals aged 65 years or older living alone by the number of households with individuals aged 65 years or older in the ZCTA. These results were then categorized into quartiles: quartile 1, <30.77% (the lowest social isolation); quartile 2, 30.77% to 36.69%; quartile 3, 36.70% to 42.39%; and quartile 4, >42.39% (the highest social isolation) of households in the ZCTA with individuals 65 years or older who lived alone.

Presence of LTCs in ZCTAs
To estimate whether LTCs were located in areas with the highest levels of social isolation for older adults, the binary dependent variable was having at least 1 LTC facility in the ZCTA area at the ZCTA level. The zip codes of LTC facilities were obtained from the CASPER data and then linked with ZCTA information from US Census data.

30-Day All-Cause Mortality
To assess whether 30-day all-cause mortality was higher in areas of highest social isolation, the binary dependent variable was 30-day all-cause mortality after admission to an LTC facility at the individual level. We obtained 30-day all-cause mortality of both short-term (<100 days) and longterm (Ն100 days) LTC residents.

Covariates
We included 3 levels of covariates: individual, LTC facility, and ZCTA levels. ZCTA-level covariates were used to estimate whether LTC facilities were located in areas of highest social isolation.
Individual, LTC facility, and ZCTA covariates were used to assess whether 30-day mortality after admission to LTC facility was higher in areas of highest social isolation. We controlled for individual-level covariates that consisted of age groups (aged Յ64 years, 65-79 years, Ն80 years), sex (male, female), race/ethnicity (White, African American, Hispanic, other races), marital status (married, never married, widowed, separated, divorced), length of LTC stay (<100 days, Ն100 days), These variables were extracted using the MDS data. In addition, we included Medicaid enrollment and comorbid conditions using the validated Prescription Drug

JAMA Network Open | Geriatrics
Hierarchical Condition Category index.
We used the self-reported race/ethnicity measure contained in the Medicare administrative data and collected as part of federally mandated long-term care assessments. We measured race and ethnicity in this study to assess any disparities in social isolation of LTC facilities.
At the LTC facility level, we controlled for the percentage of LTC residents who were aged 65 years and older, the percentage of female LTC residents, the percentage of LTC residents by various races (eg, White, African American, Hispanic, other races), and the percentage of LTC residents by marital status (married, never married, widowed, separated, divorced). These variables were extracted using the MDS data. In addition, we included the percentage of LTC residents with Medicaid enrollment, and whether the LTC facility belonged to a chain and was a profit vs nonprofit institution.
At the ZCTA level, using 5-year average data from the ACS data, we included the following ZCTAlevel covariates: the percentage of adults who were aged 65 years and older, the percentage of female residents, the percentage of races (eg, White, African American, Asian, Hispanic, other races), the percentage of education levels, the percentage of married, the percentage of owner-occupied housing, the percentage of Medicaid enrollment, the urbanicity of area, and the 9 US Census divisions.

Statistical Analyses
First, we examined the unadjusted differences in area characteristics between the ZCTAs with and without LTC facilities using a t test for continuous covariates and a χ 2 test for categorical variables.
Second, we examined the generalized estimating equations model with logit link to estimate the association between the percentage of households with individuals aged 65 years or older who lived alone and the presence of nursing homes within that ZCTA. We estimated adjusted odds ratios (ORs) and 95% CIs of having LTC facilities by the quartile of socially isolated neighborhoods at the ZCTA level, controlling for covariates. We also conducted subgroup analyses in ZCTAs with a majority population of White, African American, or Hispanic populations as defined by US Census percentages above the national median. All model standard errors were adjusted by clustering within the ZCTA.
Third, using multilevel logistic regression models, we estimated adjusted ORs and 95% CIs of individual risk of 30-day morality from any cause by the quartile of socially isolated neighborhoods, controlling for covariates. We further performed subgroup analyses by racial and Hispanic ethnicity baseline demographic characteristics of these LTC residents was reported in a previous study. 14 The association between socially isolated neighborhoods and the location of LTC facilities is visualized in Figure 1. Figure 1A displays the presence of at least 1 LTC facility by the 4 levels of social isolation across the contiguous US map, while Figure 1B shows areas without LTC facilities. We see the first map was mostly red, which indicates that LTC facilities were mostly in areas with the highest levels of social isolation of older adults. In contrast, the second map was mostly yellow, which  indicates that areas of low social isolation of older adults do not usually have LTC facilities. Most of the socially isolated LTC facilities are in the Midwest section of the US (as seen in Figure 1A).   A, the first map was mostly red, which indicates that LTC facilities were mostly in areas with the highest levels of social isolation of older adults. B, the second map was mostly yellow, which indicates that areas of low social isolation of older adults do not usually have LTC facilities.  The odds of 30-day mortality were approximately 16% to 17% higher in neighborhoods with the 2 highest quartiles of older adults single-occupancy households (quartile 1: OR, 1.17; 95% CI, 1.10-1.25;

Discussion
To our knowledge, this is the first study to characterize the geographic locations of LTC facilities within neighborhoods according to the percentage of older adults who live alone. Nationwide, we found that most LTC facilities were located in areas with high social isolation as indicated by the high proportions of older adults living in single-person households in the surrounding ZCTA. This association held across areas with a majority population of African American and Hispanic residents, although the association was stronger in neighborhoods with a majority population of White residents.
Our work suggests that LTC residents may be at increased risk of social isolation that is exacerbated by the location of the facility. This may be important if it creates barriers to connections with friends and family. A review of barriers to family visitations to LTC residents found that travel time to the facility and access to transportation are substantial factors. 25 Socially isolated LTC facilities may also confer increased risk of negative health risks to their residents, which has been consistently documented in the community setting. Our prior research has found evidence of a shorter survival time based on deaths from all causes in LTC residents who were admitted to facilities in socially isolated neighborhoods at the county level. 14 In this study, our findings further support that there is an increased risk of short-term 30-day mortality after admission to a LTC facility in socially isolated neighborhoods at the areas level in the same LTC residents. This risk held across Our analysis of the social isolation of LTC facilities revealed an interesting geographic variation.
Most of the socially isolated LTC facilities are in the Midwest section of the US. We have no explanation for this finding, but believe it deserves further investigation.

Strengths and Limitations
This study had some strengths and limitations. Strengths of our study included its use of a large, nationally representative sample of LTC facilities; a multilevel study design with adjustment for multiple county-level factors; and exploration of differences in associations across race and ethnicity subpopulations.
Limitations of the study relate to the ZCTA-level analyses. Our analyses suggest that the external environments of LTC facilities may influence the social isolation of residents, although we do not test that assumption. Neighborhoods with many older adults living alone may not be a direct measure of social isolation. Similarly, an LTC facility located in an area with large numbers of older adults living alone may not translate to an experience within the facility of social isolation. We lacked individual-level measures of the visit experience of family and friends, which we hypothesize to be associated with the facility's location. Such measures would be useful in future studies. Additionally, our findings with the risk of mortality are associations that deserve more investigation before causality can be determined.

Conclusions
This study represents a novel area of inquiry given our growing understanding of the importance of social isolation in older adults who live in LTC facilities. Our work has found that LTCs are often located in socially isolated neighborhoods. This suggests that there may be a need for special attention and strategies to keep LTC residents connected to their family and friends. Such measures could eventually contribute to improved health trajectories in the US population that is increasingly aging and at growing risk of entering LTC facilities.