Predicting illness trajectory and hospital resource utilization of COVID-19 hospitalized patients - a nationwide study

Roimi, M.; Gutman, R.; Somer, J.; Ben Arie, A.; Calman, I.; Bar-Lavie, Y.; Ziv, A.; Eytan, D.; Gorfine, M.; Shalit, U.

Importance: The spread of COVID-19 has led to a severe strain on hospital capacity in many countries. There is a need for a model to help planners assess expected COVID-19 hospital resource utilization. Objective: Provide publicly available tools for predicting future hospital-bed utilization given a succinct characterization of the status of currently hospitalized patients and scenarios for future incoming patients. Design: Retrospective cohort study following the day-by-day clinical status of all hospitalized COVID-19 patients in Israel from March 1st to May 2nd, 2020. Patient clinical course was modelled with a machine learning approach based on a set of multistate Cox regression-based models with adjustments for right censoring, recurrent events, competing events, left truncation, and time-dependent covariates. The model predicts the patient's entire disease course in terms of clinical states, from which we derive the patient's hospital length-of-stay, length-of-stay in critical state, risk of in-hospital mortality, and overall hospital-bed utilization. Accuracy assessed over 8 cross-validation cohorts of size 330, using per-day Mean Absolute Error (MAE) of predicted hospital utilization over time; and area under the receiver operating characteristics (AUROC) for individual risk of critical illness and in-hospital mortality, assessed on the first day of hospitalization. We present predicted hospital utilization under hypothetical incoming patient scenarios. Setting: 27 Israeli hospitals. Participants: During the study period, 2,703 confirmed COVID-19 patients were hospitalized in Israel for 1 day or more; 28 were excluded due to missing age or sex; the remaining 2,675 patients were included in the analysis. Main Outcomes and Measures: Primary outcome: per-day estimate of total number of hospitalized patients and number of patients in critical state; secondary outcome: risk of a single patient experiencing critical illness or in-hospital mortality. Results: For random validation samples of 330 patients, the per-day MAEs for total hospital-bed utilization and critical-bed utilization, averaged over 64 days, were 4.72 {+/-} 1.07 and $1.68 {+/-} 0.40 respectively; the AUROCs for prediction of the probabilities of critical illness and in-hospital mortality were 0.88 {+/-} 0.04 and 0.96 {+/-} 0.04, respectively. We further present the impact of several scenarios of patient influx on healthcare system utilization, demonstrating the ability to accurately plan ahead how to allocate healthcare resources. Conclusions and Relevance: We developed a model that, given basic easily obtained data as input, accurately predicts total and critical care hospital utilization. The model enables evaluating the impact of various patient influx scenarios on hospital utilization. Accurate predictions are also given for individual patients' probability of in-hospital mortality and critical illness. We further provide an R software package and a web-application for the model.