OBJECTIVES:  As compared to non-diabetics, diabetic patients are more likely to be hospitalized with longer hospital stay. A variety of risk adjustment (RA) measures are available, which makes it difficult to select the best performing measure to predict outcomes. The aim of this study is to compare performance of risk adjustment measures to predict in-hospital mortality and length of stay (LOS) in diabetic patients.  METHODS: A retrospective cross-sectional study was conducted using the HCUP Nationwide inpatient Sample (NIS)-2009 data. All adults (age >18 years) diagnosed with Diabetes were included in the study. Charlson-Deyo Adaptation and Elixhauser Co-morbidity Index were constructed using the International Classification of Disease, 9th Revision, Clinical Modification (ICD-9-CM). Two proprietary measures (APR-DRG & Disease Staging) were compared with comorbidity measures for predicting in-hospital mortality and LOS. Logistic regression was used to predict in-hospital mortality and c-statistics were used to assess the comparative performance of different models. Adjusted R2from Linear regression models was compared to do the same for the continuous outcome, LOS. RESULTS:  The prevalence of diabetes was found to be 28.2% with mortality rate of 2.09% and median LOS of 2.77 ± 0.01 days. Hospital stays were predominantly by white females (<65 years). Models containing APR-DRG measure outperformed all other measures for both outcomes (in-hospital mortality, c-statistics=0.91-0.90 and; LOS, adjusted R2=0.172-0.163). The model containing all demographic variables along with APR-DRG and Elixhauser comorbidity index outperformed all other models for predicting in-hospital mortality (c-statistics=0.91) and LOS (Adjusted R2=0.172). CONCLUSIONS: The APR-DRG, being a clinical model, is superior to other comorbidity measures for risk-adjusting in-hospital mortality and LOS.  Addition of comorbidity measures to APR-DRG improves the model performance when predicting in-hospital mortality and LOS.