Table 2.

Prediction of severe disease.

	OR (95% CI)
PMH: DM	1·51 (0·51 - 4·45)
Smoker	3·13 (1·08 - 9·38)
ED: MAP low	2·59 (0·92 - 7·47)
ED: SpO2 low	5·36 (2·03 - 15·07)
ALC low	3·12 (1·00 - 9·8)
Lactate high	3·90 (0·63 - 22·91)
Glucose high	2·58 (0·92 - 7·30)
RNAaemic	6·72 (2·45 - 19·79)
N	191
AIC	134·74
AUROC	0·82

Potential predictors of severe (WHO 5–8) disease included: demographic features (age 60+ or 80+, sex), past medical history features (lung disease, cancer, diabetes, immunosuppression, heart disease, hypertension, angiotensin converting enzyme inhibitor or angiotensin receptor blocker use, stroke, dementia, deep venous thrombosis or pulmonary embolus, chronic kidney disease, tobacco smoking, obesity), binary indicators of abnormal ED vital signs (low or high mean arterial pressure, low or high heart rate, low or high respiratory rate, low oxygen saturation, low or high temperature), pneumonia on chest X-ray or CT, patient-reported symptoms (fever, chills, cough, sore throat, congestion, shortness of breath, chest pain, myalgias, nausea/vomiting/diarrhea, loss of taste, loss of smell, confusion, headache), and binary indicators of abnormal lab values (high or low leukocyte count, low absolute lymphocyte count, low haemoglobin, low or high platelet count, high D-dimer level, high fibrinogen level, low fibrinogen level, high prothrombin time, high partial thromboplastin time, high C-related peptide level, high lactate dehydrogenase level, high ferritin level, high troponin level, high lactate level, high or low sodium level, high or low potassium level, high or low chloride level, high or low bicarbonate level, high blood urea nitrogen level, high creatinine level, high or low calcium level, high or low magnesium level, high or low glucose level, high bilirubin level, high aspartate aminotransferase level, high alanine aminotransferase level, high alkaline phosphatase level).

To prevent over-fitting, predictors were selected via elastic net regression of severe disease on these features with 10-fold cross-validation, selecting the regularisation parameter λ minimizing mean cross-validated error, and yielding the features in the table above. In a logistic model regressing severe disease on these features, significant predictors of severe disease included: tobacco smoking, low oxygen saturation (SpO2), and RNAaemia. RNAaemia was associated with 6·7 times the odds of severe disease, adjusting for other features selected by elastic net penalised regression, an association comparable in magnitude to the association of hypoxia on initial presentation with eventual severe disease. Mean cross-validated area under the receiver-operating characteristic curve (AUROC) of the model in predicting severe disease was 0·82.