Proactive Healthcare Analytics: Early Detection of Diabetes with SDOH Insights and Machine Learning (Published)
This white paper presents a proactive healthcare analytics framework for early diabetes detection, combining Social Determinants of Health (SDOH) with machine learning. Traditional models only use clinical biomarkers, ignoring socioeconomic factors like income levels, food access and healthcare availability. By including SDOH data from CDC, County Health Rankings and USDA Food Access Atlas we improve predictive accuracy and get population level insights. Using optimized XGBoost our model has an R² of 0.88 and MAE of 0.63, beating baseline models. The study shows how healthcare analytics can move diabetes prevention from reactive to proactive and support personalized interventions and public health initiatives. We propose integration into healthcare systems via real-time APIs and predictive analytics dashboards. This research highlights the importance of SDOH aware models in addressing health disparities and informing data driven policy decisions.
Keywords: Diabetes, Healthcare Analytics, SDOH, XGB, machine learning
A Web-Based Clinical Decision Support System for the Management of Diabetes Neuropathy Using Naïve Bayes Algorithm (Published)
Diabetes Neuropathy is a chronic health problem with devastating, yet preventable consequences. Due to this shortage of specialists, there is a need for a Clinical Decision Support System that will diagnose and manage diabetes neuropathy. This work therefore aimed at designing a web-based Clinical Decision Support System for the management of early diabetes neuropathy. Four pattern classification algorithms (K-nearest neighbor, Decision Tree, Decision Stump and Rule Induction) were adopted in this work and were evaluated to determine the most suitable algorithm for the clinical decision support system. Datasets were gathered from reliable sources; two teaching hospitals in Nigeria, these were used for the evaluation Benchmarks such as performance, accuracy level, precision, confusion matrices and the models building’s speed were used in comparing the generated models. The study showed that Naïve Bayes outperformed all other classifiers with accuracy being 60.50%. k-nearest neighbor, Decision Tree, Decision Stump and Rule induction perform well with the lowest accuracy for x- cross validation being 36.50%. Decision Tree falls behind in accuracy, while k-nearest neighbour and Decision Stump maintain accuracy at equilibrium 41.00%. Therefore, Naïve Bayes is adopted as optimal algorithm in the domain of this study. The rules generated from the optimal algorithm (Naïve Bayes) forms the back-end engine of the Clinical Decision Support System. The web-based clinical decision support system was then designed The automatic diagnosis of diabetes neuropathy is an important real-world medical problem. Detection of diabetes neuropathy in its early stages is a key for controlling and managing patients early before the disabling effect present. This system can be used to assist medical programs especially in geographically remote areas where expert human diagnosis not possible with an advantage of minimal expenses and faster results. For further studies, researchers can improve on the proposed clinical decision support system by employing more than one efficient algorithm to develop a hybrid system.
Keywords: Accuracy, Algorithm, Classification, Diabetes, Neuropathy, precision