PURPOSE: The 2001 Sepsis Consensus Conference introduced the Predisposition, Insult/Infection, Response, and Organ Dysfunction (PIRO) concept as a hypothesis generating model for future research. There are few publications addressing the role of PIRO generally in sepsis, and specifically in pneumonia. The PIRO concept has also been tested in a substudy of SAPS3 database. We performed this retrospective study to test if the PIRO model derived from APACHE III database can be applied to predict the mortality rate of septic patients admitted to ICU.
METHODS: The study included patients with ICU admission diagnosis of infection and 2 or more of the systemic inflammatory response (SIRS) criteria. Hospital mortality was the primary outcome variable. We used patient age, gender, ethnicity, Do-not-Resuscitate (DNR) status, co-morbidity, and admission source to evaluate for predisposition (P); infection source, surgical status, and hospital days before ICU admission for infection (I); the number of SIRS criteria for response (R); and Acute Physiology Score (APS), mechanical ventilation, and vasopressors for organ dysfunction (O). The area under the receiver operating curve (AUC) and the Hosmer-Lemeshow statistic were used to assess the performance of the model.
RESULTS: During the 16-year ending on 30 September 2009, and after excluding repeat admissions and those who did not authorize research, 4,684 patients met the inclusion criteria. A total of 1135 patients (24.2%) died in the hospital. The AUC (95% CI) of the APACHE III probability of death in predicting hospital mortality was 0.817 (0.803-0.831) and the Hosmer-Lemeshow statistic was 58.7 with a p value < 0.001. The APACHE-PIRO model had AUC of 0.829 (0.816-0.842) and Hosmer-Lemeshow statistic 11.8 with a p value of 0.159.
CONCLUSION: The APACHE-PIRO model had better calibration in predicting the mortality rate of critically ill patients with sepsis.
CLINICAL IMPLICATIONS: Incorporation of infection source, DNR status, and SIRS criteria improves the performance of the APACHE model in predicting mortality.
DISCLOSURE: Oetiz-Diaz Enrique, No Financial Disclosure Information; No Product/Research Disclosure Information