Objective: To test the statistical significance of the existing World Health Organization (WHO) clinical case definition for the diagnosis of acquired immunodeficiency syndrome (AIDS) in areas where diagnostic resources are limited in India, and frame a prediction rule that is sensitive and specific for the clinical diagnosis of AIDS. Methodology and Results: In a case-load study, we evaluated the WHO clinical case definition of AIDS in 162 patients, the training set (58 seropositives) at the All India Institute of Medical Sciences, New Delhi. This definition had a sensitivity of about 29%, specificity of about 29%, positive predictive value of about 19% and negative predictive value of about 42%. We analyzed the clinical data of these 162 patients using stepwise logistic regression analysis. Generalized lymphadenopathy, oropharyngeal candidiasis, presumptive esophageal candidiasis, herpes zoster in the last 5 years and a total lymphocyte count of less than 1,000 × 109/l emerged as important predictors. A prediction rule for the diagnosis of AIDS in the Indian context was then developed empirically using the coefficients of these predictors. We further tested the rule prospectively in 50 patients, the test set (11 seropositives) and found it to have a sensitivity of 81.8%, specificity of 97.4%, positive predictive value of 90%, and a negative predictive value of 95%. Conclusion: The WHO definition was not significant statistically for clinical use in India. The new prediction rule may help in AIDS surveillance in India.