Background. Despite a substantial disease burden. there is little descriptive epidemiology of acute pneumonia in sub-Saharan Africa. We did this study to define the aetiology of acute pneumonia, to estimate mortality at convalescence, and to analyse mortality risk-factors. Methods. We studied 281 Kenyan adults who presented to two public hospitals (one urban and one rural) with acute radiologically confirmed pneumonia during 1994-96. We did blood and lung-aspirate cultures, mycobacterial cultures, serotype-specific pneumococcal antigen detection, and serology for viral and atypical agents. Findings. Aetiology was defined in 182 (65%) patients. Streptococcus pneumoniae was the most common causative agent, being found in 129 (46%) cases; Mycobacterium tuberculosis was found in 26 (9%). Of 255 patients followed up for at least 3 weeks, 25 (10%) died at a median age of 33 years. In multivariate analyses, risk or protective factors for mortality were age (odds ratio 1.51 per decade [95% CI 1.04-2.19]), unemployment (4.42 [1.21-16.11), visiting a traditional healer (5.26 [1.67-16.51), visiting a pharmacy (0.30 [0.10-0.91]), heart rate (1.64 per 10 beats [1.24-2.16]), and herpes labialis (15.4 [2.22-107)). HIV-1 seropositivity, found in 52%, was not associated with mortality. Death or failure to recover after 3 weeks was more common in patients with pneumococci of intermediate resistance to benzylpenicillin, which comprised 28% of pneumococcal isolates, than in those infected with susceptible pneumococci (5.60 [1.33-23.6]). Interpretation. We suggest that tuberculosis is a sufficiently common cause of acute pneumonia in Kenyan adults to justify routine sputum culture, and that treatment with benzylpenicillin remains appropriate for clinical failure due to M tuberculosis, intermediate-resistant pneumococci, and other bacterial pathogens. However, interventions restricted to hospital management will fail associated with socioeconomic, behavioural factors.
