Predicting overall death risk in patients with H7N9 infection using a nomogram (Preprint)

BACKGROUND To date, almost all of these studies have identified multiple risk factors but did not offer practical instruments for routine use in predicting death risk in human H7N9 infection cases. Such an instrument could be useful in identifying high-risk H7N9 patients who can benefit from reducing the risk of death. OBJECTIVE We aimed to create a clinical nomogram to predict the overall death (OD) risk of patients with H7N9 virus infection (VI). METHODS We reviewed specific factors and outcomes regarding patients with H7N9 VI to determine relationships and developed a nomogram to calculate individualized patient risk. This model was used to predict each individual patient’s probability of death based on results obtained from the multivariate binary logistic regression analysis. RESULTS We examined 227 patients with H7N9 VI enrolled in our study over a nearly 6-year period. Stepwise selection was applied to the data, which resulted in a final model with 7 independent predictors. The nomogram model was constructed for maximum predictive accuracy. The concordance index of this nomogram was 0.906 and 0.822 for the training and validation sets, respectively, which indicates adequate discriminatory power. The calibration curves for the OD showed optimal agreement between nomogram prediction and actual observation in the training and validation sets, respectively. A decision-curve analysis of the clinical benefit indicates that the prognostic model, including age ≥ 60 years , chronic disease, poor hand hygiene , time from illness onset to the first medical visit, incubation period of ≤ 5 days, peak C-reactive protein ≥120 mg/L, and increased initial neutrophil count factors, resulted in a higher net benefit across a wide range of decision threshold probabilities (i.e., an approximately 6 – 98% risk of death). With the cutoff threshold values of 30% and 20% predicted probabilities, nomogram models showed sensitivities of 85.7% and 80.9% and specificities of 78.3% and 73.1% when applied to the training and validation sets, respectively. CONCLUSIONS We established and validated a novel nomogram that can predict OD for patients with H7N9 VI. This practical prognostic model may help clinicians in decision making, clinical diagnosis, and treatment selection.