JMIR Publications

ABSTRACT

Background:

The novel 2019 coronavirus SARS-CoV-2 and its associated disease, COVID-19, have caused worldwide disruption, leading countries to take drastic measures. As the virus continues to spread, hospitals have struggled to allocate resources to the patients most at risk. In this context, it becomes important to develop models that can accurately predict the severity of infection for each hospitalized patient, helping to guide triage, planning, and resource allocation.

Objective:

The aim of this study is to develop accurate models to predict mortality among hospitalized COVID-19 patients, using basic demographics and easily obtainable laboratory data.

Methods:

A retrospective study of 375 hospitalized patients in Wuhan, China infected with COVID-19 was undertaken. The patients were randomly split into derivation and validation cohorts. Regularized logistic regression and support vector machine classifiers were trained on the derivation cohort and accuracy metrics (F1-scores) were computed on the validation cohort. Two types of models were developed: i) using laboratory findings from the entire length of stay at the hospital, and ii) using admission laboratory findings obtained no later than 12 hours after admission. The models were further validated on a multicenter external cohort of 542 patients.

Results:

Of the 375 patients, 174 (46.4%) succumbed to the infection. The study cohort was composed of 60% (224/375) males and 40% (151/375) females, with a mean age of 58.83 years old. Models developed using patient data from throughout the length of stay had an accuracy as high as 97%, whereas models with admission laboratory variables had accuracy of up to 93%. The latter models developed using admission patient data predicted patient outcomes an average of 11.5 days in advance. Key variables such as lactate dehydrogenase, high-sensitivity C-reactive Protein, and the percent of lymphocytes in the blood were indicated by the models. In line with previous studies, age was also found to be an important variable in predicting mortality. In particular, the mean age of patients that survived COVID-19 infection (50.23 years) was significantly smaller than the mean age of patients (68.75 years) that did not survive the infection (P<.001).

Conclusions:

Machine learning models can be successfully employed to accurately predict COVID-19 patient outcomes. Models achieve high accuracies and predict outcomes more than a week in advance, a promising result that can greatly aid hospitals in resource allocation.