In pleural infection, medical treatment failure (chest-tube drainage and antibiotics) requires surgery and increases mortality. It would be helpful to predict which patients will fail this approach. We examined clinical predictors in 85 consecutive patients with pleural infection receiving chest drainage and intrapleural fibrinolytics, and recorded age, length of history, antibiotic delay and choice, time to drainage, blood/pleural fluid (PF) bacteriology, PF pH, lactate dehydrogenase (LDH), glucose and appearance, effusion size, pleural thickness on computed tomographic (CT) scan, and survival from time of drainage. Failures (surgery/death) were compared with successes. There were 13 (15%) medical failures. PF purulence was more frequent in medical failures (10 of 13 versus 29 of 72 successes, p < 0.02 chi-square). Absence of purulence was a useful predictor of success (positive predictive value [PPV] 93%). Purulence was not useful in predicting medical failure (PPV 26%). There was a trend for positive blood culture to predict failure (5 of 13 failures versus 11 of 72 successes, p = 0.05 chi-square), but no significant differences in other endpoints. Twelve (14%) patients died in follow-up, all with comorbidity within 400 d after drainage. Probability of survival at 4 yr was 86%. Of endpoints considered in this study, PF purulence was the only useful predictor of outcome with medical therapy in pleural infection. There is good long-term survival from pleural infection.