Background: Identifying COVID-19 patients at risk for long lengths of stay (LOS) may help hospitals anticipate resource limitations. There are no large representative studies from US hospitals assessing factors associated with LOS. This information is vital for resource allocation for the current and potential future pandemics. In this study, we examine factors associated with longer LOS using a national registry.
Methods: The American Heart Association COVID-19 Cardiovascular Disease registry is a retrospective registry including adult patients admitted with COVID-19 to any of over 130 participating hospitals across 30 states. We included patients with index admission for COVID-19 between March and November 2020. The distribution of patient, hospital, and geographic characteristics was assessed by LOS tertile. To assess the association of the above characteristics and LOS, we applied a multivariate Poisson regression model with identity link and reported ß coefficients for days change in LOS and 95% confidence intervals. Multiple imputations were performed for missing data.
Results: 20,793 patients were included in our analysis. Median LOS across the tertiles was 2, 5, and 15 days (Table 1). In the multivariate model, LOS decreased through the study period by 1.789 days. There was a decline in the percentage of patients in the third tertile of LOS over time (37.2% between March-May 2020, 31.5% between June-August 2020, and 24.8% between September-November 2020). Mortality rates increased across LOS tertiles (11.1%, 12.8%, and 20.8%, respectively). Mortality rates were the highest March-May 2020 at 17.4% but stayed at 9.7% between June-August 2020 and September-November 2020. Older age, tachypnea at admission, certain comorbidities, and transfer from another hospital were associated with longer LOS. Females and patients with obesity had significantly shorter LOS by 0.65 days and 0.08 days, respectively. Compared to patients with a commercial payor, patients with Medicare, Medicaid, or government insurance had longer LOS by more than half a day. New thrombotic event, new hemodialysis, or intubation during the hospitalization were associated with increased LOS by 5-10 days. Large hospitals had LOS > 1 day longer compared to small hospitals, and urban hospitals had LOS >0.5 days shorter than rural hospitals. Regionally, LOS was longest in hospitals in the West by > 1 day.
Conclusions: In this study, we identified several patient, in-hospital, and geographic characteristics associated with LOS. Unsurprisingly, elderly patients, certain comorbidities, or more severe illness had longer LOS. This was not true for obese patients and suggests not all factors associated with increased severity of COVID-19 correlate with increased LOS. Patients had a longer LOS in the earlier months of the pandemic with decrease over time, likely due to improvement in disease management and care and quarantine options outside the hospital, allowing for earlier discharge from hospitals. A similar decline in mortality rates over time may have been due to improvement in treatment protocols or inclusion of patients with less severe illness. Larger hospitals had longer LOS possibly due to patients with more comorbidities and more clinical severity. Larger hospitals are also more likely to have accepted transfers from other hospitals. Urban hospitals had shorter LOS perhaps to accommodate a higher volume of patients.