A novel model for predicting mortality in the management of deep neck infections

Abstract

Objectives: Deep neck infections (DNIs) are a common and intractable disease encountered in ENT clinics that impose a significant medical and financial burden on affected individuals and their families. However, insufficient data are currently available for predicting outcomes in cases of DNI. The present study thus sought to develop a novel model capable of predicting treatment outcomes of DNI patients just using indicators at the visit. Methods: Patients with DNIs treated from 2010 to 2022 were included in the present study. Patient data were retrospectively collected from medical records. Risk factors associated with mortality were identified using logistic regression models. A predictive model was constructed based on odds ratios for factors calculated using a multivariate regression model. Results: In total, 153 patients were enrolled in the present study. Risk factors associated with mortality included age >50 years, residence in a rural area, dyspnea at visit, the involvement of multiple infected sites, serum albumin<34 g/L, renal insufficiency, mediastinitis, pulmonary infection, and septic shock. A multivariate regression model revealed that mediastinitis (OR: 7.308, P < 0.001), serum creatinine>95 μmol/L (OR: 23.363, P < 0.05), and serum albumin<34 g/L (OR: 13.837, P < 0.05) were independent predictors of mortality in deep neck infection patients, with serum creatinine>95 μmol/L being particularly critical to the outcomes. Diabetes was not the predictor of mortality but was associated with long-term hospitalization (P < 0.001). Conclusions: In summary, the model constructed in the present study was capable of estimating the potential for poor outcomes in DNI patients before the initiation of treatment. These findings may help improve doctor–patient communication, especially for those struggling financially.