Key factors predicting the in-hospital mortality of patients with severe cutaneous adverse reactions in Thailand

BackgroundAt present, no predictive models are available to determine the probability of in-hospital mortality rates (HMRs) in all phenotypes of severe cutaneous adverse reactions (SCARs). ObjectivesOur study explored whether simple clinical and laboratory assessments could help predict the HMRs in any phenotypes of SCAR patients. MethodsFactors influencing HMRs in 195 adults diagnosed with different SCAR phenotypes were identified, and their optimal cut-offs were determined by Youden's index. Predictive equations for HMRs for all SCAR patients and Stevens-Johnson syndrome (SJS)/toxic epidermal necrolysis (TEN) patients were determined using the exact logistic regression models. ResultsAcute generalized exanthematous pustulosis (AGEP) patients were significantly older, with a short time from drug exposure to reaction, and higher neutrophil count compared to SJS/TEN and drug reaction with eosinophilia and systemic symptoms (DRESS, p < 0.001). Peripheral blood eosinophilia, atypical lymphocytosis and elevated liver transaminase enzymes were significantly higher in DRESS. SJS/TEN phenotype, age >= 71.5 years, neutrophil-to-lymphocyte ratio >= 4.08 (high NLR) and systemic infection were factors predicting in-hospital mortality in all SCAR subjects. The ALLSCAR model developed from these factors demonstrated high-diagnostic accuracy for predicting HMRs in all SCAR phenotypes (area under the receiver-operator curve (AUC) = 0.95). The risk of in-hospital death was significantly increased in SCAR patients with high NLR after adjusting for systemic infection. The model derived from high NLR, systemic infection and age yielded higher accuracy than SCORTEN (AUC = 0.77) for predicting the HMRs in SJS/TEN patients (AUC = 0.97). ConclusionsBeing older, having systemic infection, having a high NLR and SJS/TEN phenotype increases ALLSCAR scores, which in turn increases the risk of in-hospital mortality. These basic clinical and laboratory parameters can easily be obtained in any hospital setting. Despite its simple approach, further validation of the model is warranted.