 |
International Society of Science and Applied Technologies |
| Integrating Socioeconomic Determinants with Medical Risk Factors for Diabetes | ||||
| Author | James McCormack
|
|||
| Co-Author(s) | Naeem Seliya
|
|||
| Abstract | Diabetes mellitus is a metabolic disease of inadequate control of elevated blood glucose levels. Among the subtypes, Type 2 has a higher prevalence in our population. The likelihood of a Type 2 diagnosis is affected by medical risk factors such as age, body mass index, hyperglycemia, and heart health. The variation of people with diabetes is significant across many levels of social, economic, geographic, cultural, and other non-medical aspects. In the literature, medical risk factors are the predominant features in machine learning based studies of diabetes. However, in comparison, we observe a significant gap in examining non-medical data as potential, important predictive attributes. In this work, an objective is to study the effects of incorporating social and economic determinants with medical risk factors, to form the features in a dataset for developing learners for a binary (yes/no) prediction of diabetes. The case study data, part of a large health-related survey, includes social and economic determinants as well as medical risk factors as features. A total of six machine learners were developed with the integrated dataset. A feature-importance study with SHapley Additive exPlanations provided insight into the social and economic determinants as features. The main observations include: most learners yielded an F-1 of 82%, AUC of 80%, and AUPRC of 81%. The feature-importance study indicated Blood Pressure, Cholesterol, and BMI as important medical risk factors, as well as various social and economic determinants were among the top 10 features. The relative importance of the latter features lends to including them in predictive diabetes analysis.
|
|||
| Keywords | Diabetes, Socioeconomic Attributes, Medical Risk Factors, Machine Learning, Feature Importance | |||
| Article #: RQD2025-172 | ||||
Proceedings of 30th ISSAT International Conference on Reliability & Quality in Design |