Postoperative mortality and pulmonary complication rankings: How well do they correlate at the hospital level?

BACKGROUND. Postoperative mortality rankings are used alone for quality assessment. OBJECTIVES. To determine the correlation between hospital rankings of postoperative respiratory failure, pneumonia, and mortality rates and to assess the influence of hospital volume, type of surgery, and time on these correlations. To compare hospital outlier detection with and without pulmonary complication rates. RESEARCH DESIGN. Prospective observational study. SUBJECTS. 103,176 noncardiac surgery patients from 123 VA hospitals enrolled between 1/1/94 and 8/31/95. Preoperative pneumonia, ventilator dependent, comatose, or do-not-resuscitate patients were excluded. MEASURES. Respiratory failure was defined as greater than 48 hours of ventilator assistance or postoperative reintubation. Pneumonia was defined as positive sputum cultures with antibiotic treatment or chest x-ray infiltrate diagnosed as pneumonia or pneumonitis. Mortality was defined as death within 30 days of surgery. Hospital rankings were assigned using risk-adjusted observed-to-expected ratios. RESULTS. There was significant, but weak correlation between mortality and pulmonary complication rankings (r = 0.21, P = 0.02 for pneumonia; r = 0.22, P = 0.01 for respiratory failure). Correlations with mortality rankings were highest for thoracic (r = 0.42, P < 0.001 for pneumonia; r = 0.38, P < 0.001 for respiratory failure) and vascular surgery (r = 0.26, P = 0.02 for pneumonia; r = 0.35, P < 0.001 for respiratory failure). Supplementing mortality with pulmonary complication outlier designations enhanced outlier detection for 47% of hospitals overall, and for 29% in the lowest caseload quartile. CONCLUSIONS. Pulmonary complication rankings correlate weakly with mortality overall, but have higher correlations in thoracic, vascular, and upper abdominal surgery. Examining pneumonia and respiratory failure outlier status with mortality outlier status enhances hospital outlier detection even in low-volume hospitals.