We prospectively identified cases of pneumococcal pneumonia and used stringent criteria to stratify them into bacteremic and nonbacteremic cases. Although patients were distributed among racial groups in proportion to all patients seen at this medical center, the proportion of African-Americans with bacteremic disease was significantly increased. All patients had at least 1 underlying condition predisposing to pneumococcal infection, and most had several. Although the mean number of predisposing factors was greater among bacteremic patients than nonbacteremic patients, only alcohol ingestion was significantly more common. Nearly one-third of patients had substantial anemia (hemoglobin < or = 10 g/dL) on admission, which may have predisposed to infection. In the case of other laboratory abnormalities, such as albumin, creatinine, and bilirubin, it was difficult to determine which abnormality might have predisposed to pneumococcal infection and which might have resulted from it. The radiologic appearance was varied. Airspace consolidation and air bronchogram on chest X-ray were highly associated with bacteremic disease, as was the presence of pleural effusion. Although the Pneumonia Patient Outcomes Research Team (PORT) risk score was a predictor of mortality, it did not help to predict the presence of bacteremia in an individual case. Most patients who died in the first week in hospital were bacteremic, and a high PORT risk score with bacteremia reliably predicted a high likelihood of a fatal outcome. Eleven patients had extrapulmonary disease with meningitis, empyema, and septic arthritis predominating; all of these patients were bacteremic. The antibiotic susceptibility of our strains correlated well with those that have been reported in the United States during the years of this study. The use of numerous antibiotics of different classes in many patients, especially those who were the most ill, precluded analysis of outcome based on antibiotic therapy. Only 17 patients had been vaccinated. Since nearly all patients had conditions for which pneumococcal vaccine is recommended and more than one-third had been hospitalized in the preceding 6 months, the low rate of vaccination can be regarded as a missed opportunity to administer a potentially beneficial vaccine.