Risk Patterns for Suicide Attempts among People with Diabetes Using Association Rules Mining

Prevalence of suicide attempts among people with diabetes was approximately 10%, higher than the general population (<5%). This is the first study using association rules mining to discover the risk patterns for suicide attempts in people with diabetes using large multi-centered EHR data. We retrieved data on 3,265,041 diabetes patients from 2010-2020, and 0.2% had a record of suicide attempts. Figure 1 shows the visualization of the top ten association rules. The first rule, {benzodiazepines, no. antidepressants ≥2} ≤ {suicide=Yes}, Support=0.19, Confidence=0.87, means “19% of patients had a record of benzodiazepines, the number of antidepressants ≥2 and suicide attempts together.” In addition, “87% of patients with benzodiazepines and the number of antidepressants ≥2 also had a record of suicide attempts.” We found new findings related to suicide attempts in people with diabetes, including the use of benzodiazepines, atypical antipsychotic agents, two and more antidepressants, none or one diabetes medication, and white in the race. Further studies are needed to validate the results and investigate the causality of suicide attempts and diabetes. Our results can be explainable and practically used by providers during outpatient clinic visits. Figure 1. Visualization of the top ten association rules represented by circles. Darker color denotes higher confidence, and a larger circle means greater support. Disclosure P.Narindrarangkura: None. P.Alafaireet: None. U.Z.Khan: None. M.Kim: None.