Machine Learning-Based Hyperglycemia Prediction: Enhancing Risk Assessment in a Cohort of Undiagnosed Individuals

Abstract

BackgroundNoncommunicable diseases (NCDs) continue to pose a significant health challenge globally, with hyperglycemia serving as a prominent indicator of potential diabetes. This study employed machine learning algorithms to predict hyperglycemia in a cohort of asymptomatic individuals and unraveled crucial predictors contributing to early risk identification.MethodsThis dataset included an extensive array of clinical and demographic data obtained from 195 asymptomatic adults residing in a suburban community in Nigeria. The study conducted a thorough comparison of multiple machine learning algorithms to ascertain the most effective model for predicting hyperglycemia. Moreover, we explored feature importance to pinpoint correlates of high blood glucose levels within the cohort.ResultsElevated blood pressure and prehypertension were recorded in 8 (4%) and 18 (9%) individuals respectively. Forty-one (21%) individuals presented with hypertension (HTN), of which 34/41 (82.9%) were females. However, cohort-based gender adjustment showed that 34/118 (28.81%) females and 7/77 (9.02%) males were hypertensive. Age-based analysis revealed an inverse relationship between normotension and age (r = -0.88; P < 0.05). Conversely HTN increased with age (r = 0.53; P < 0.05), peaking between 50-59 years. Isolated systolic hypertension (ISH) and isolated diastolic hypertension (IDH) were recorded in 16/195 (8.21%) and 15/195 (7.69%) individuals respectively, with females recording higher prevalence of ISH 11/16 (68.75%) while males reported a higher prevalence of IDH 11/15 (73.33%). Following class rebalancing, random forest classifier gave the best performance (Accuracy Score = 0.894; receiver operating characteristic-area under the curve (ROC-AUC) score = 0.893; F1 Score = 0.894) of the 27 model classifiers. The feature selection model identified uric acid and age as pivotal variables associated with hyperglycemia.ConclusionsRandom Forest classifier identified significant clinical correlates associated with hyperglycemia, offering valuable insights for early detection of diabetes and informing the design and deployment of therapeutic interventions. However, to achieve a more comprehensive understanding of each feature’s contribution to blood glucose levels, modeling additional relevant clinical features in larger datasets could be beneficial.