The Karnofsky Scale of Performance Status (KPS) is a measure of health status that is widely used for HIV-infected persons, although few studies have documented its reliability and validity for HIV. We developed a modified KPS in an attempt to make it more appropriate for use in HIV-infected persons and evaluated its inter-rater reliability in a multicenter study. Patients (N = 657) were recruited from five hospitals in three Italian cities; responses were scored using the modified scale by three different raters in each hospital: one experienced physician, one young physician, and one nurse. Kendall's correlation test showed a strong inter-rater agreement for the total of observations: 0.82 between the two physicians, 0.77 between the experienced physician and the nurse, and 0.76 between the young physician and the nurse. The analysis of variance showed a strong effect of the stage of the disease on the KPS score. The specific hospital had little effect, whereas the effect of the type of rater disappeared when the nurse was excluded from the analysis. This modified scale demonstrated a good reliability, although training of raters could help to eliminate differences among raters and among hospitals.