OBJECTIVES: To evaluate performances of various risk adjustment scales in predicting risk of hospitalization in patients diagnosed with dementia. METHODS: This cross-sectional study was conducted using the household and medical provider component files of Medical Expenditure Panel Survey (MEPS) data from 2000 to 2003 (panel 5, 6, and 7). Dementia patients were identified using cccodex and International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9CM) codes, and all cause hospitalizations were recorded from the inpatient files. The risk adjustment scales evaluated in this study were- diagnosis based scales (identified using ICD-9CM codes): D’Hoore’s adaptation of Charlson comorbidity index (CCI), and Elixhauser comorbidity algorithm; and prescription based scales (identified using national drug codes): chronic disease scores (CDS-1 and CDS-2). Logistic regression models were constructed with all-cause hospitalization as binary outcome, adjusting for demographic, socio-economic (insurance, income, geographic region), and perceived health status covariates, and risk adjustment algorithms. Performance of the models in predicting hospitalization was measured by c-index. RESULTS: Total 392 dementia patients were identified from the household component files during the study period. Most of the patients were male (68.58%), white (54.11%), elderly (mean age 73 years), and from a low income family (59.85%); and almost half of them had an inpatient visit (46.70%). Baseline logistic model with the covariates only predicted hospitalization risk reasonably well (c= 0.602). Performance of risk adjustment models after adjusting for covariates were as follows- D’Hoore: c= 0.688, Elixhauser: c= 0.691, CDS-1: c= 0.666, CDS-2: c= 0.733, CDS-1+D’Hoore: c= 0.687, CDS-1+Elixhauser: c= 0.709, CDS-2+D’Hoore: c= 0.752, and CDS-2+Elixhauser: c= 0.757. CONCLUSIONS: Diagnosis based scales performed better than CDS-1 scale in predicting hospitalization. CDS-2 which is a comorbidity algorithm modified from CDS-1 to efficiently model healthcare utilization and cost performed superiorly. Among the combinations of diagnosis and prescription based scales CDS-2+Elixhauser predicted hospitalization risk most efficiently.