Derivation and validation of new prehospital phenotypes for adults with COVID-19

Objective. To characterize phenotypes of prehospital patients with COVID-19 to facilitate early identification of at-risk groups. Methods. Multicenter observational noninterventional study of a retrospective cohort of 3789 patients, analyzing 52 prehospital variables. The main outcomes were 4 clusters of prehospital variables describing the phenotypes. Secondary outcomes were hospitalization, mechanical ventilation, admission to an intensive care unit, and cumulative mortality inside or outside the hospital on days 1, 2, 3, 7, 14, 21, and 28 after hospitalization and after start of prehospital care. Results. We used a principal components multiple correspondence analysis (factor analysis) followed by decomposition into 4 clusters as follows: cluster 1, 1090 patients (28.7%); cluster 2, 1420 (37.4%); cluster 3, 250 (6.6%), and cluster 4, 1029 (27.1%). Cluster 4 was comprised of the oldest patients and had the highest frequencies of residence in group facilities and low arterial oxygen saturation. This group also had the highest mortality (44.8% at 28 days). Cluster 1 was comprised of the youngest patients and had the highest frequencies of smoking, fever, and requirement for mechanical ventilation. This group had the most favorable prognosis and the lowest mortality. Conclusions. Patients with COVID-19 evaluated by emergency medical responders and transferred to hospital emergency departments can be classified into 4 phenotypes with different clinical, therapeutic, and prognostic characteristics. The phenotypes can help health care professionals to quickly assess a patient's future risk, thus informing clinical decisions.