Large national claims databases are sources of vital information concerning health care resource utilization. However, the comparability of data obtained from such databases has not yet been ascertained. OBJECTIVE: To compare prevalence rates of low back disorders obtained from two large national inpatient claims databases and to study variations in length of stay and corresponding costs. METHOD: Data were obtained from two independent databases with inpatient claims information including ICD-9 codes for specific diagnoses, demographics, length of stay (LOS), and payments or charges made. One of the databases is a 20% national inpatient sample of all community hospitals in the U.S. (HCUP). The other national database consists of data gathered for privately insured population (MarketScan). Claims for specific diagnoses of low back disorders (ICD-9 codes: 720.0-724.9) for 1994 were obtained. Using age, gender, and diagnosis adjusted rates (direct method), the overall low back disorder rates were compared. Also age, gender, and diagnosis specific low back disorder rates were compared between two databases. RESULTS: The overall adjusted prevalence rates of low back disorders were 1.49 and 1.88 per 100 admissions for HCUP and MarketScan, respectively. Significant difference was observed in the age, gender adjusted rates for diagnosis of displacement of lumbar intervertebral disc without myelopathy, with MarketScan showing a higher rate as compared to HCUP (1.06 vs 0.78/100 admissions). The adjusted average LOS and age, gender, and diagnosis specific LOS, were higher for HCUP than MarketScan. The specific and adjusted payments (based on MarketScan) were however, higher than the charges reported in HCUP. CONCLUSIONS: The prevalence rate of low back disorders is higher in the MarketScan database than HCUP. The differences in the length of stay and associated costs might be attributable to other variables such as geographical variations.