|
Abstract
|
BACKGROUND: Patients diagnosed with lung cancer have a higher suicide rate than the general population and other cancer patients. The aim of this study was to develop and validate a prediction model for the individual risk for suicide after the diagnosis of lung cancer.
METHODS: Patients diagnosed with lung cancer between 2007 and 2016 were selected from the Surveillance, Epidemiology, and End Results (SEER) database and randomly divided into training and validation cohorts. Cox proportional hazard models were used to identify relevant predictors and construct prediction models. Additionally, graphic visualization methods were used to predict the risk for suicide within 5 years after the diagnosis of lung cancer. We used bootstrapping for the internal validation, Harrell's C-index for the discrimination, and a calibration plot for the calibration of the proposed model.
RESULTS: We obtained complete information on 112372 patients diagnosed with lung cancer from the SEER cohort. Multivariate Cox regression identified sex, race, marital status, tumour grade, surgery, radiation, and chemotherapy as significant predictors for suicide. A nomogram and a risk matrix were developed to visualize the risk for suicide within 5 years after lung cancer diagnosis. The bootstrapped and validated C-indices of the nomogram were 0.77 and 0.78, respectively. The calibration plot indicated good agreement between the prediction and actual observation.
CONCLUSIONS: The proposed model demonstrated good discrimination and calibration performance for predicting the risk for suicide within 5 years after lung cancer diagnosis. Reliable and feasible risk assessment tools can be promising for preventing unnecessary suicides among lung cancer patients.
Language: en |