Aim In many countries, do‐not‐resuscitate (DNR) orders were not legislated, partly because rationale for proposal of DNR orders have not been studied in elderly pneumonia patients with cognitive and physical disorders. The aim of the present study was to elucidate the factors influencing physicians' proposal for DNR orders and their validity as prognostic predictor, by comparing elderly pneumonia cases with and without DNR orders. Methods Medical records of community‐acquired pneumonia patients aged 65 years or older were retrospectively studied (n = 641). The patients were categorized into two groups; one with DNR orders within 72 h after admission and the other without it. Results DNR was decided in 183 patients (28.5%). The DNR group, containing more elderly patients with poorer performance status, showed higher rates of malnutrition, dementia, aspiration, very severe pneumonia, respiratory failure and mortality. The choice of antimicrobials was not affected by the presence of DNR orders. Mortality rate within 30 days was higher in the DNR group than in the non‐DNR group (33.9% vs 2.8%, P < 0.001), as well as total hospital mortality (56.8% vs 4.8%, P < 0.001). Multiple logistic analysis identified factors involved in the decision‐making of DNR orders; that is, aspiration, healthcare‐associated pneumonia, respiratory failure, intensive airspace consolidation, age 75 years and older, performance status 3 and 4, and serum albumin<2.5 g/dL were positive factors for DNR orders. Conclusion The present study showed factors involved in the physicians proposal of DNR orders, demonstrating that DNR was empirically chosen based on comprehensive judgment of several prognostic predictors and it, in itself, was a good prognostic predictor. Geriatr Gerontol Int 2013; ●●: ●●–●●.