Predicting hospital readmissions in patients with diabetes

The objective of this study was to determine whether 11 independent variables or combinations of variables help to predict a diabetes-related hospital readmission for patients with diabetes within 60 days from discharge. The variables were categorized into four main groups: (a) patient characteristics, (b) lifestyle, (c) biomarkers, and (d) disease management aspects. A convenience sample of 389 historical medical records of patients who were admitted to a rural hospital in northeastern North Carolina with a diagnosis of, or relating to, diabetes was studied. After comparing predictive discriminant analysis (PDA) and logistic regression (LR), PDA performed better and was chosen to analyze a convenience sample of patients admitted to the hospital for a diabetes-related diagnosis from January, 2004 to December, 2006. The best overall subset accurately classified 27 cases with six predictors that included (a) systolic blood pressure, (b) smoking status, (c) blood glucose range, (d) ethnicity, (e) diabetes education, and (f) diastolic blood pressure. In an effort to simplify the prediction process, the subsets of two predictors were examined. The results of the analysis returned four subsets of 2-predictor variable combinations that correctly classified cases for readmission. Each of the four subsets has two predictors that are statistically and practically significant for predicting readmissions for a diabetes-related problem within fewer than 60 days. These combinations are the predictor subsets of (a) smoking status and being treated by a specialist or non-specialist physician, (b) a religious affiliation or a lack thereof and smoking status, (c) gender and smoking status, and (d) smoking status and ethnicity.