Modelling of selected socio-economic and demographic predictors of diabetic kidney disease among diabetic patients: comparative analysis of cox regression and support vector machine models

Abstract

Diabetic kidney disease (DKD) accounts for one in three adults with diabetes and is a significant trigger for mortality among diabetic patients globally. Traditional predictive models of DKD in diabetic patients have mainly been based on patients’ clinical health histories but have overlooked socio-economic factors that may also be integral in DKD prevalence. This study aimed to develop an improved predictive model that considers the effects of socio-economic, demographic, and behavioural factors on the survival rates of diabetic patients prior to developing DKD. A retrospective survey design was conducted among 756 diabetic patients at Meru Teaching and Referral Hospital and Kerugoya Level 5 Hospital in Kenya. Patients’ records and semi-structured questionnaires were utilised for data collection. The extracted data were entered into Excel and analysed using R software. The Cox regression and Support Vector Machine (SVM) were employed to identify the predictors of DKD and to gauge someone’s risk of developing DKD over time based on their socio-economic characteristics. The research data were randomly split into a training set (70%) and a test set (30%) for developing the two models and identifying predictors of DKD. Age at diagnosis, history of cardiovascular disease, alcohol use, financial hardships, employment status, level of education, and gender were identified as significant predictors associated with DKD. The study found that the SVM model had a slightly higher C-index (0.7753) in comparison with the Cox model (0.770), indicating that SVM model was marginally more accurate in predicting DKD than the Cox model. Therefore, prompt policy changes and effective strategies in public health or clinical practice should be designed based on the identified socio-economic predictors and the developed models in an effort to prevent the development of DKD in diabetic patients.