Participants and methods:We compared HA-COVID-19 cases, defined as patients with a positive severe acute respiratory coronavirus virus 2 (SARS-CoV-2) test > 5 days after hospital admission, with hospitalized CA-COVID-19 cases, defined as those who tested positive within 5 days of admission. The composite primary outcome was patient transfer to an intensive care unit (ICU) or an intermediate care unit (IMCU) and/or all-cause in-hospital mortality. We used cause-specific Cox regression and Fine-Gray regression to model the time to the composite clinical outcome, adjusting for confounders and accounting for the competing event of discharge from hospital. We compared our results to those from a conventional approach using an adjusted logistic regression model where time-varying effects and competitive risk were ignored.
Results:Between February 19, 2020, and December 31, 2020, we included 1,337 HA-COVID-19 cases and 9,068 CA-COVID-19 cases. HA-COVID-19 patients were significantly older: median, 80 (interquartile range [IQR], 71–87) versus median 70 (IQR, 57–80) (P < .001). A greater proportion of HA-COVID-19 patients had a Charlson comorbidity index ≥ 5 (79% vs 55%; P < .001) than did CA-COVID-19 patients. In time-varying analyses, between day 0 and 8, HA-COVID-19 cases had a decreased risk of death or ICU or IMCU transfer compared to CA-COVID-19 cases (cause-specific hazard ratio [csHR], 0.43; 95% confidence interval [CI], 0.33–0.56). In contrast, from day 8 to 30, HA-COVID-19 cases had an increased risk of death or ICU or IMCU transfer (csHR, 1.49; 95% CI, 1.20–1.85), with no significant effect on the rate of discharge (csHR, 0.83; 95% CI, 0.61–1.14). In the conventional logistic regression model, HA-COVID-19 was protective against transfer to an ICU or IMCU and/or all-cause in-hospital mortality (adjusted odds ratio [aOR], 0.79, 95% CI, 0.67–0.93).