Objectives: To derive a clinically relevant age-independent physiologic failure scoring system and to use this system to examine aspects of the association of physiologic failure, age, and comorbidity with inpatient mortality.
Design: Retrospective, secondary analysis of a derivation and validation cohort selected from the Cleveland Health Quality Choice Coalition data set.
Setting: Thirty hospitals in greater Cleveland.
Participants: Thirty-one thousand nine hundred seventy-six inpatients aged 50 and older discharged in 1993 with a diagnosis of congestive heart failure, pneumonia, or stroke.
Measurements: The Inpatient Physiologic Failure Score (IPFS) was developed and used to calculate physiologic failure. Forty-four candidate variables were examined for their association with inpatient mortality, and 12 were selected. A point value (2, 3, 4, or 6) based on adjusted odds ratio was assigned for an abnormal result for each of the 12 common physiologic variables. Each patient's abnormal physiology points were summed to produce a physiologic failure score (range 0-39). Comorbidity was quantified using the Patient Management Category Severity Scale. The association between mortality and increasing physiologic failure, increasing age and comorbidity, and distribution of physiologic failure with increasing age and comorbidity were examined. A threshold age was sought. Models for predicting inpatient mortality were developed.
Results: Twelve physiologic variables constitute the IPFS. Increasing physiologic failure, age, and comorbidity were associated with increasing mortality. Increasing physiologic failure was not associated with increasing age or comorbidity. We did not find a threshold age. The area under the receiver operating characteristic (ROC) curve for predicting inpatient mortality for IPFS was 0.730, and for comorbidity was 0.741 (not significant). The area under the ROC curve for a mortality prediction model based on age was significantly less (0.603). Accounting for patient age did not significantly improve the predictive ability of the IPFS model (area = 0.752, P <.05). The complete model best predicted mortality (0.829).
Conclusions: The IPFS represents a clinically relevant method for scoring physiologic failure. Physiologic failure, age, and comorbidity are independently and differently associated with inpatient mortality. Physiology fails independent of age and comorbidity.