Construction and evaluation of leukemia suicide risk predictive model based on SEER database.

Background: A marked increase in suicide rate has been detected among individuals diagnosed with leukemia. Our research aimed to develop a predictive model intended for assessing the suicide risk in leukemia patients. This novel tool aims to optimize the process of pinpointing individuals at high risk within clinical environments, thereby guaranteeing the timely provision of targeted intervention strategies. Methods: Between 2000 and 2020, our study involved a cohort of 194584 leukemia patients, extracted from the Surveillance, Epidemiology, and End Results (SEER) database. These patients were randomly stratified into distinct training and validation cohorts. We utilized the Cox proportional hazards model to screen for influential variables and construct a predictive nomogram within the training set. The concordance index (C-index) and receiver operating characteristic (ROC) curves were employed to evaluate model's discrimination, and calibration curves was used to assess the calibration ability. Furthermore, the validation set was utilized to conduct an internal validation process to ensure the robustness of nomogram. Results: Age, gender, race, residence, marital status, and histologic type were selected to construct the nomogram for predicting suicide risk of leukemia patients. In the training and validation sets, the C-indexes were 0.798 and 0.776, respectively. The calibration plots demonstrated a significant agreement between the predicted and actual outcomes. Ultimately, leukemia patients were divided into two groups, and Kaplan-Meier curves showed significant differences in the high- and low-risk groups, as confirmed in the validation set. Conclusions: We have successfully developed an intuitive and robust predictive model for assessing the suicide risk among leukemia patients. This model holds the potential to contribute to the reduction of preventable deaths.