Objective To explore the difference of influencing factors on the death between young and middle-aged tuberculosis patients and elderly tuberculosis patients and construct an age-stratified nomogram prediction model for the death risk of tuberculosis patients.

Methods A total of 100 811 confirmed cases of tuberculosis treated and closed in 14 designated tuberculosis hospitals in Hangzhou from January 2005 to December 2021 were included as the training set, and 7 669 confirmed cases treated and closed from January 2022 to June 2023 were included as the validation set. A multivariate Cox regression model was used to evaluate the independent risk predictors of death of young and middle-aged and elderly tuberculosis patients, and an age-stratified nomogram model was constructed. The predictive performance of the model was evaluated based on the index of consistency (C-index), area under the curve (AUC), and calibration curve.

Results Age (HR = 1.056, 95%CI: 1.047–1.066), gender (HR = 1.353, 95%CI: 1.060–1.727), current residence (HR = 1.792, 95%CI: 1.434–2.240), pathogenic diagnosis results (HR = 2.338, 95%CI: 1.711–3.193), drug resistance (HR = 1.724, 95%CI: 1.098–2.707), human immunodeficiency virus (HIV) infection (HR = 4.342, 95%CI: 1.391–13.552), history of tuberculosis (HR = 1.627, 95%CI: 1.225–2.161), treatment plan (HR = 1.345, 95%CI: 1.054–1.716), use of fixed-dose combination (FDC) in the whole process or not (HR = 2.342, 95%CI: 1.310–4.188) were independent influencing factors for the death of young and middle-aged patients. Age (HR = 1.086, 95%CI: 1.080–1.092), gender (HR = 1.548, 95%CI: 1.390–1.724), pathogenic diagnosis results (HR = 2.144, 95%CI: 1.888–2.435), treatment management methods (HR = 1.182, 95%CI: 1.044–1.337), history of tuberculosis (HR = 1.217, 95%CI: 1.077–1.374), and use of FDC in the whole process or not (HR = 1.200, 95%CI: 1.010–1.426) were independent influencing factors for the death of elderly patients.

Conclusion The factors influencing the death of young and middle-aged tuberculosis patients were different from those of the elderly. The age-stratified nomogram model constructed demonstrated good performance in predicting the death of tuberculosis patients.