Abstract
The study aim to construct an effective model for predicting the survival period of COVID-19 patients. METHODS: Clinical data of 386 COVID-19 patients were collected from December 2022 to January 2023. The patients were randomly divided into training and validation cohorts in a 7:3 ratio. LASSO regression and multivariate Cox regression analyses were used to identify prognostic factors, and a nomogram was constructed. Nomogram was evaluated using decision curve analysis, receiver operating characteristic curve, consistency index (c-index), and calibration curve. RESULTS: 86 patients (22.3%) died. A new nomogram for predicting the survival was established based on age, resting oxygen saturation, Blood urea nitrogen (BUN), c-reactive protein-to-albumin ratio (CAR), and pneumonia visual score. The decision curve indicated high clinical applicability. The nomogram c-indexes in the training and validation cohorts were 0.846 and 0.81, respectively. The area under the curves (AUCs) for the 15-day and 30-day survival probabilities were 0.906 and 0.869 in the training cohort, and 0.851 and 0.843 in the validation cohort. The calibration curves demonstrated consistency between predicted and actual survival probabilities. CONCLUSIONS: Our nomogram has the capacity to assist clinical practitioners in estimating the survival rate of COVID-19 patients, thereby facilitating more optimal management strategies and therapeutic interventions with substantial clinical applicability.