Predictors of mortality among long-term care residents with SARS-CoV-2 infection

J Am Geriatr Soc. 2021 Dec;69(12):3377-3388. doi: 10.1111/jgs.17425. Epub 2021 Sep 20.

Abstract

Background: While individuals living in long-term care (LTC) homes have experienced adverse outcomes of SARS-CoV-2 infection, few studies have examined a broad range of predictors of 30-day mortality in this population.

Methods: We studied residents living in LTC homes in Ontario, Canada, who underwent PCR testing for SARS-CoV-2 infection from January 1 to August 31, 2020, and examined predictors of all-cause death within 30 days after a positive test for SARS-CoV-2. We examined a broad range of risk factor categories including demographics, comorbidities, functional status, laboratory tests, and characteristics of the LTC facility and surrounding community were examined. In total, 304 potential predictors were evaluated for their association with mortality using machine learning (Random Forest).

Results: A total of 64,733 residents of LTC, median age 86 (78, 91) years (31.8% men), underwent SARS-CoV-2 testing, of whom 5029 (7.8%) tested positive. Thirty-day mortality rates were 28.7% (1442 deaths) after a positive test. Of 59,702 residents who tested negative, 2652 (4.4%) died within 30 days of testing. Predictors of mortality after SARS-CoV-2 infection included age, functional status (e.g., activity of daily living score and pressure ulcer risk), male sex, undernutrition, dehydration risk, prior hospital contacts for respiratory illness, and duration of comorbidities (e.g., heart failure, COPD). Lower GFR, hemoglobin concentration, lymphocyte count, and serum albumin were associated with higher mortality. After combining all covariates to generate a risk index, mortality rate in the highest risk quartile was 48.3% compared with 7% in the first quartile (odds ratio 12.42, 95%CI: 6.67, 22.80, p < 0.001). Deaths continued to increase rapidly for 15 days after the positive test.

Conclusions: LTC residents, particularly those with reduced functional status, comorbidities, and abnormalities on routine laboratory tests, are at high risk for mortality after SARS-CoV-2 infection. Recognizing high-risk residents in LTC may enhance institution of appropriate preventative measures.

Keywords: artificial intelligence; covid-19; machine learning; mortality; prognosis.

Publication types

  • Research Support, Non-U.S. Gov't

MeSH terms

  • Aged
  • Aged, 80 and over
  • Artificial Intelligence
  • COVID-19 / diagnosis*
  • COVID-19 / mortality*
  • COVID-19 / prevention & control
  • COVID-19 / transmission
  • COVID-19 Nucleic Acid Testing
  • Cause of Death
  • Comorbidity
  • Female
  • Humans
  • Long-Term Care / statistics & numerical data*
  • Machine Learning
  • Male
  • Nursing Homes
  • Ontario / epidemiology
  • Pandemics / prevention & control
  • Predictive Value of Tests
  • Risk Factors
  • SARS-CoV-2 / genetics
  • SARS-CoV-2 / isolation & purification*
  • Severity of Illness Index