OBJECTIVES:  Using Machine-Learning, this research identifies key diagnostic reading levels that lead to the prediction of diabetes among readmitted patients, and provides analysis as to why this is the case. METHODS:  Gathered by researchers at the University of California Irvine, the dataset encompasses 10 years (1999-2008) of clinical care at 130 US hospitals and integrated delivery networks. It includes over 50 features representing patient and hospital outcomes. The dataset includes information concerning demographics, diagnostics, prescriptions, the existence of diabetes and readmission. I used the Complex Tree method for analysis, using readmission as the independent variable, with the thresholds of diagnostic readings from A1C test selected for their sensitivity and specificity. RESULTS:  Roc curve with Area Under Curve 67.8% and False positive class 64% True positive Rate (TPR) of the current classifier and positive class 83.3%. A positive class was determined by levels 7 and 8 of the A1C test, while and normal readings of the A1C diagnostic test were classified as negative class. CONCLUSIONS: A higher threshold on the A1C test indicates allows for an increased likelihood of high sensitivity and true predictive power.