Predictive accuracy of the TRISS survival statistic is improved by a modification that includes admission pH
F. H. Milham, M. Malone, J. Blansfield, W. W. LaMorte and E. F. Hirsch
Department of Surgery, Boston (Mass) University Medical Center.
OBJECTIVE: To determine if pH measured at the time of hospital admission
and corrected for PCO2 was an independent predictor of trauma survival.
DESIGN: Phase 1 was a retrospective case-control analysis of 1708 patients,
followed by multivariate multiple logistic regression analysis of a subset
of 919 patients for whom the Revised Trauma Score (RTS), Injury Severity
Score (ISS), and pH were available. Phase 2 was a prospective comparison of
a mathematical model of survival derived in phase 1 (pH-TRISS) with the
TRISS method in 508 of 1325 subsequently admitted trauma patients. SETTING:
Urban level 1 trauma center. PATIENTS: All patients admitted with blunt or
penetrating trauma during the study period. MAIN OUTCOME MEASURES: Survival
vs mortality. RESULTS: In phase 1, factors significantly associated with
mortality by t test and chi 2 analysis included the RTS, ISS< Glasgow
Coma Scale, corrected pH (CpH), and sum of the head, chest, and abdominal
components of the Abbreviated Injury Scale-85 (AIS85) (HCAISS) (for all, P
< .0001). The TRISS statistic was also a significant predictor of
survival (P < .004). Age, sex, and the extremity and soft tissue
components of the AIS85 were not associated with mortality. In a
multivariate analysis of the RTS, HCAISS, and CpH, all were significant
predictors of mortality. Even when controlling for RTS and HCAISS, CpH
remained a significant predictor of mortality (P < .008). In phase 2,
when pH-TRISS was tested prospectively against TRISS in a new group of
patients, the new statistic appeared to provide a more accurate prediction
of survival. CONCLUSIONS: The arterial pH measurement obtained on hospital
arrival and corrected for PCO2 is a significant independent predictor of
survival and adds to the predictive accuracy of the TRISS survival
statistic. Age, sex, and the extremity and soft tissue components of the
AIS85 did not contribute to the accuracy of the TRISS statistic in this
patient population.
