Objective: To determine mortality and factors that might predict outcome in severe community-acquired pneumococcal pneumonia treated by a standard protocol.
Design: Prospective, non-concurrent study.
Setting: Respiratory intensive care unit (ICU) in a teaching hospital.
Patients: 63 patients who were diagnosed by positive blood culture or Gram stain and culture of sputum or tracheal aspirate were included.
Measurements and results: Clinical features, severity scores including Acute Physiology and Chronic Health Evaluation (APACHE) II, organ failure and lung injury scores, and the clinical course in the ICU were documented; 79% of patients required mechanical ventilation. Bacteraemia was present in 34 patients (54%); there were no distinguishing clinical features between bacteraemic and non-bacteraemic cases. The overall mortality was 21%, with only 5 deaths (15% mortality) in the bacteraemic group. Shock and a very low serum albumin (< 26 g/l) were the only clinical features that differentiated survivors from non-survivors; lung injury, APACHE II and multiple organ failure scores were all predictive of outcome. The positive predictive value and specificity in predicting death in individuals for the modified British Thoracic Society rule 1 were 26 and 64%; APACHE II > 2057 and 88%; > 2 organ failure 64 and 92%; and lung injury > 233 and 73%, respectively.
Conclusions: These results suggest that even in bacteraemic cases mortality should be below 25% with intensive care management and that conventional scoring systems, while predictive of group mortality, are unreliable in individuals.