OBJECTIVES: To determine the association between claims-based risk adjustment scores and duration of observation and to ascertain whether statistical control for any association is possible. METHODS: We analyzed 18 months of claims data (1/1/1996-6/30/1998) from a large American company. We restricted the study population to those individuals that were continuously eligible for benefit coverage for the entire 18 months (n=42,054). Drug-based risk adjustment scores (Chronic Disease Score) and medical-based risk adjustment scores (Elixhauser Index) were calculated at 3,6,9,12, and 18 months. RESULTS: We found a pronounced effect of time, with the population’s mean Chronic Disease Score increasing 23% (p<0.05) in 1 year and the population’s mean Elixhauser Index increasing by 19% (p<0.05). More importantly, the relationship between time and the mean risk adjustment scores was linear and demonstrated a strong correlation (r2>0.96). CONCLUSIONS: The high correlation between some risk adjustment scores and duration of observation precludes direct comparisons between individuals with different lengths of observation. However, statistical control, e.g., standardized scores derived from linear regression, for the bias due to length of observation is possible. Thus, restricting analyses to individuals with a common duration of observation may be unnecessary.