Key result
The CatBoost machine learning model predicted lifetime and past-year suicidal behaviors with 83.97% and 88.46% accuracy, respectively, outperforming traditional logistic regression.
Why the study?
Individuals with diabetes face increased risks of suicidal behaviors, but machine learning techniques have not been widely used alongside traditional statistical methods to investigate prevalence and risk factors.
Do machine learning models improve the prediction of suicidal behaviors in diabetic patients compared to traditional statistical methods?
Cross-Sectional (n=390)
Do machine learning models improve the prediction of suicidal behaviors in diabetic patients compared to traditional statistical methods?
Machine learning models like CatBoost offer improved predictive power for identifying suicidal behaviors in patients with diabetes compared to traditional statistical methods.
May improve suicidal behavior prediction in diabetes; hypothesis-generating and requires prospective validation before clinical use.
Background: Diabetes management poses significant challenges worldwide, with individuals often facing increased risks of mental health issues, including suicidal behaviors. While traditional statistical methods have been used to assess risk factors, this study introduces machine learning techniques to predict suicidal behaviors among diabetic patients. Therefore, this study aims to investigate the prevalence and risk factors of suicidal behaviors among individuals with diabetes in Bangladesh using both traditional statistical methods and machine learning models. Methods: A cross‐sectional study was carried out with 390 diabetic patients, assessing demographic characteristics, diabetes‐related variables, mental health indicators, and suicidal behaviors. Statistical analyses including chi‐square tests and logistic regression, and various machine learning models, were employed to examine associations and predict risk factors. Results: This study found that 13.8% of the individuals with diabetes reported lifetime suicidal behavior and 9.0% reported past‐year suicidal behavior. Significant risk factors associated with higher rates of suicidal behaviors included younger age, marital status, lower monthly income, higher educational attainment, being insulin users, not having a family history of diabetes, and suffering from anxiety. Among the machine learning models tested, the CatBoost model outperformed others, obtaining low log loss values of 0.45 and 0.32 for lifetime and past‐year suicidal behaviors, respectively. CatBoost demonstrated excellent accuracy rates of 83.97% and 88.46%, significantly higher than those of traditional logistic regression models. The most significant predictors identified by machine learning were treatment type, followed by occupation, anxiety, age group, and having chronic diseases. Conclusion: The findings suggest the urgent need for comprehensive interventions and support systems for individuals managing diabetes, particularly those at higher risk of suicidal behaviors. Machine learning offers improved predictive power and the ability to identify complex patterns in risk factors. Collaborative efforts between healthcare providers, mental health professionals, and public health initiatives are essential for prioritizing mental well‐being in diabetes management.
No takes yet. Share an insight, caveat, or question.
Mamun et al. (2024) conducted a cross-sectional in Diabetes (n=390). CatBoost machine learning model vs. Traditional logistic regression models was evaluated on Prediction of lifetime and past-year suicidal behaviors. The CatBoost machine learning model predicted lifetime and past-year suicidal behaviors with 83.97% and 88.46% accuracy, respectively, outperforming traditional logistic regression.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: