Objectives To assess nursing home staffing data reported in the Online Survey Certification and Reporting (OSCAR) system database for research and policy. Design Comparisons were made between OSCAR and a concurrent research survey of staffing data collected for the same facilities, using inter-rater agreement and correlation analyses. Setting Freestanding nursing homes from New York State (NYS) in 1997 (N = 327). Measurements Selected staffing variables were defined in comparable terms in both OSCAR and the NYS survey. Results The two data sources were in substantial agreement on the reported availability of a full-time physician (other than medical director) and of a physician assistant or nurse practitioner (Kappa >0.7), and they correlated well in the full-time equivalent (FTE) number of such staff (Spearman correlation >0.6). The correlation was 0.8 for FTE registered nurses (RNs), 0.7 for licensed practical nurses (LPNs), and 0.8 for certified nurse aides (CNAs). In terms of average nurse hours per patient day, separately for RNs, LPNs, CNAs, and all combined, the correlation was relatively weak (between 0.3 and 0.6). Overall staffing levels tended to be lower in OSCAR than in NYS. Conclusion The OSCAR data are useful for exploring relationships between staffing and various quality of care outcomes, but may not be accurate enough on a case-specific basis, or to determine policy regarding minimal staffing levels using average nurse hours per patient day measures. More systematic and timely efforts are needed to refine the OSCAR content and survey methodology to document nursing home staffing information. To assess nursing home staffing data reported in the Online Survey Certification and Reporting (OSCAR) system database for research and policy. Comparisons were made between OSCAR and a concurrent research survey of staffing data collected for the same facilities, using inter-rater agreement and correlation analyses. Freestanding nursing homes from New York State (NYS) in 1997 (N = 327). Selected staffing variables were defined in comparable terms in both OSCAR and the NYS survey. The two data sources were in substantial agreement on the reported availability of a full-time physician (other than medical director) and of a physician assistant or nurse practitioner (Kappa >0.7), and they correlated well in the full-time equivalent (FTE) number of such staff (Spearman correlation >0.6). The correlation was 0.8 for FTE registered nurses (RNs), 0.7 for licensed practical nurses (LPNs), and 0.8 for certified nurse aides (CNAs). In terms of average nurse hours per patient day, separately for RNs, LPNs, CNAs, and all combined, the correlation was relatively weak (between 0.3 and 0.6). Overall staffing levels tended to be lower in OSCAR than in NYS. The OSCAR data are useful for exploring relationships between staffing and various quality of care outcomes, but may not be accurate enough on a case-specific basis, or to determine policy regarding minimal staffing levels using average nurse hours per patient day measures. More systematic and timely efforts are needed to refine the OSCAR content and survey methodology to document nursing home staffing information.