There are few data on the practice of evidence based medicine in the developing world, nor on the actual sources of evidence that clinicians use in practice. To test the hypothesis that there was variation between and within developing countries in the proposed management of a patient with hospital acquired pneumonia, and that part of the variation can be explained by the sources of evidence used. Questionnaire responses to hypothetical case history. Investigators from 6 centres within the International Clinical Epidemiology Network (INCLEN) in China, Thailand, India, Egypt, and Kenya. Doctors chosen to represent primary and secondary hospital practice in the regions of the study centres. Investigations and initial treatments which would be ordered for a hypothetical 60-year-old woman who develops pneumonia 5 days after hospital admission, whether local data on antibiotic sensitivities are available and where information would be obtained to guide management. Chest x-ray and sputum gram stain/culture were consistently the most commonly ordered investigations, there being much greater variation in the initial treatment choices with either penicillin, a third-generation cephalosporin or aminoglycoside being the most popular choice. Textbooks were the commonest form of information source, and access to a library, textbooks and journals were statistically significantly associated with appropriate choice of investigations, but not treatment. Access to local antibiotic sensitivities was associated with appropriate initial treatment choice. Improving access to information in the literature and to local data may increase the practice of evidence-based medicine in the developing world. Copyright (C) 2000 Elsevier Science Inc.