Finding Indicator Diseases of Psychiatric Disorders in BigData using Clustered Association Rule Mining

Psychiatric disorders represent critical non-communicable diseases of the 21st century and are ranked as the leading cause of years lived with disabilities. Nevertheless, data that could be used to improve our understanding of psychiatric diseases remain underutilized. In this research, we apply clustered association rule mining to find comorbidities and indicator diseases for patients with psychiatric illnesses. The model was trained with health insurance billing data from 60,115 patients with a total of 904,821 ICD-10 coded diseases. Nine association rules were found without clustering, 40 with clustering of F diagnoses. The approach proves suitable for further use in the implementation of indicator-based digital decision support systems in psychiatry.