Scoping Review on 30-day Readmission Prediction in Diabetes Management Using Artificial Intelligence

Now affecting over 500 million people around the world, diabetes poses an enormous burden on healthcare systems and necessitates constant control and self-management by the patient to prevent complications. This state-of-the-art review focuses on synthesizing existing prediction models of 30-day readmissions in diabetes care to identify successful strategies and gaps in current research. However, despite the optimistic results of machine learning algorithms applied at different success levels for readmission prediction, some critical areas still need to be explored. These include the impacts of demographic factors, including education, gender, and income, besides the unique challenges among underrepresented regions of Eastern India, and particularly those in its economically disadvantaged populations. Literature indicates that interventions at many levels, such as medication management, patient education programs, and lifestyle modifications, can result in better clinical outcomes and reduced secondary care utilization. Scant research focuses on the socioeconomic variables and geographical context that could influence readmission rates. Most existing models are based on data from more affluent and developed regions and are not generalizable or applicable in diverse healthcare settings. Future research should address these gaps using more inclusive datasets, fostering interpretability, robustness, and seamless integration of predictive models within health systems. This approach will enable health providers to develop better, more accurate, fair interventions across different patient groups and have a basis for using the least resources in different socioeconomic strata and geographical locations.