Objective: To examine the impact of age, year and region of AIDS diagnosis on early (up to 12 months) and late survival of AIDS patients in different exposure categories; to describe the hazard pattern from 12 months after AIDS diagnosis. Patients: A total of 4577 UK adult AIDS diagnoses to the end of 1991 in men who have sex with men, 273 AIDS cases in injecting drug users, 411 AIDS patients infected by blood products, and 535 other adult AIDS cases, mainly ascribed to heterosexual transmission. Deaths have been recorded for 4739 of these 5796 AIDS patients. Results: The influence of calendar year and region of AIDS diagnosis on survival were short-term, for the most part operative only within the first year of follow-up. The monthly death-rate was roughly constant from 12 to 48 months post-AIDS [pooled estimate, 0.055 with 95% confidence interval (CI), 0.053–0.057] but was more than halved for 4-year survivors (pooled estimate, 0.022; 95% CI, 0.017–0.027). About 7% of AIDS cases diagnosed in 1990–1991 survive for at least 48 months. Survival after AIDS diagnosis shortens with advancing age at AIDS diagnosis: the relative hazard per decade of age (1.35; 95% CI, 1.29–1.41 in the first year after AIDS) is significantly greater (P < 0.001) in the first year after AIDS diagnosis than from 12 to 48 months (1.19; 95% CI, 1.13–1.25 in the second epoch). Conclusions: The influence of covariates, including age, is strongest in the first year of follow-up after AIDS diagnosis. Monthly death-rate is roughly constant at 0.055 from 12 to 48 months post-AIDS and at 0.022 thereafter.