|
Abstract
|
BACKGROUND: Patients with laryngeal cancer have more than five times the incidence of suicide compared with the general population. In this study, we aimed to develop an online risk stratification system, named Larysuicide, to identify patients at high risk of suicide after the laryngeal cancer diagnosis.
METHODS: Forty-two thousand and sixty-six American patients from the SEER-18 database and 4207 Chinese patients from our center were included in this study. We randomly assigned American patients into the training set and validation set at a ratio of 7:3, and all Chinese patients remained as an independent external testing set. LASSO regression model was applied for data dimension reduction, feature selection, and Larysuicide building. The performance of model was evaluated and validated by C-index, AUC, calibration curves, decision curve analysis (DCA), and univariate regression analysis.
RESULTS: The Larysuicide developed with seven selected features-age, race, cancer site, pathological subtype, grade, stage at presentation, and radiation. The model showed good discrimination, with a C-index of 0.745 (95% CI 0.723-0.767) in training set, 0.759 (95% CI 0.722-0.800) in validation set, and 0.749 (95% CI 0.730-0.769) in testing set. The AUC was 0.745 in training set, 0.759 in validation set, and 0.749 in testing set. The calibration curves showed good calibration. Decision curve analysis demonstrated that Larysuicide was clinically useful. The univariate regression analysis presented patients in the high-risk group identified by Larysuicide suffered a significantly higher risk of committing suicide after cancer diagnosis.
CONCLUSION: We constructed an online risk stratification system which could help health-care professionals efficiently identify patients at high risk of suicide after the laryngeal cancer diagnosis. Larysuicide could be a useful tool for health-care professionals to implement an early and appropriate psychological intervention in context of precision medicine.
Language: en |