We prospectively studied the incidence of hospital-acquired pneumonia in 1886 consecutive admissions to an 1800 bed hospital in Bombay; 991 of them to general medical wards and 895 to a 17-bed medical intensive care unit (ICU). The average bed occupancy in the general wards was 56 patients in a ward with 40 beds. Staffing in the general ward was two nurses for 56 patients, and in the ICU three nurses for 17 beds. One hundred and sixty-eight patients developed nosocomial pneumonia: 18 (1.8%) in general wards and 150 (16.7%) in the ICU. Common isolates included Pseudomonas spp (44%) and Klebsiella spp (34%). The most frequently used antibiotics were cefotaxime (34%), amikacin (25%), gentamicin (23%) and ofloxacin (13%). Crude mortality in general ward patients was 88.9 vs 14.6% in patients without pneumonia. The corresponding figures for ICU patients were 67.4 vs 37.1%; 40% of the crude mortality in ICU patients with pneumonia was attributable to the infection. Infected patients stayed an additional 5.8 days in the ICU and 6.7 days in the general ward. Costs of additional stay and antibiotics accounted for 18.6% of the ICU budget. The incidence of nosocomial pneumonia was lower than expected, despite occupancy exceeding bed capacity, low nurse:patient ratios, and extensive reuse of disposable respiratory therapy equipment. Nevertheless, nosocomial pneumonia imposes a significant financial burden on the already scarce resources available for intensive care in developing countries like India.