SEVERE COMMUNITY-ACQUIRED PNEUMONIA IN ICUS - PROSPECTIVE VALIDATION OF A PROGNOSTIC SCORE

Objective: To determine predictors of intensive care unit (ICU) mortal ity in patients with community-acquired pneumonia (CAP), to develop a pneumonia-specific prognostic index, and to evaluate this index prospe ctively. Design: Combined retrospective and prospective clinical study over two periods: January 1987-December 1992 and January 1993-Decembe r 1994. Setting: Four medical ICUs in the north of France. Patients: D erivation cohort: 335 patients admitted to one ICU were retrospectivel y studied to determine prognosis factors and to develop a pneumonia-sp ecific prognostic index. Validation cohort: 125 consecutive patients, admitted to four ICUs, were prospectively enrolled to evaluate this in dex. Results: In the derivation cohort, 16 predictors of mortality wer e identified and assigned a value directly proportional to their magni tude in the mortality model: aspiration pneumonia (-0.37), grading of sepsis greater than or equal to 11 (-0.2), antimicrobial combination ( -0.01), Glasgow score >12+mechanical ventilation (MV) (+0.09), serum c reatinine greater than or equal to 15 mg/l (+0.22), chest involvement shown by X-ray greater than or equal to 3 lobes (+0.28), shock (+0.29) , bacteremia (+0.29), initial MV (+0.29), underlying ultimately or rap idly fatal illness (+0.31), Simplified Acute Physiology Score greater than or equal to 12 (+0.49), neutrophil count less than or equal to 35 00/ mm(3) (+0.52), acute organ system failure score greater than or eq ual to 2 (+0.64), delayed MV (+0.67), immunosuppression (+1.38), and i neffective initial antimicrobial therapy (+1.5). An index was obtained by adding each patient's points. According to a receiver operating ch aracteristic curve, the cut-off value of this index was 2.5. In the va lidation cohort, an index of greater than or equal to 2.5 could predic t death with a positive predictive value of 0.92, sensitivity 0.61, an d specificity 0.98. Conclusion: This index, which performs well in cla ssifying patients at high-risk of death, may help physicians in initia l patient care (appropriateness of the initial antimicrobial therapy) and guide future clinical research (analysis and design of therapeutic trials).