Objective: The Agency for Healthcare Research and Quality (AHRQ) developed 20 patient safety indicators (PSIs) to identify potentially preventable complications of acute inpatient care based on administrative data. The objective of this patient safety performance study was to assess the impact of cases flagged by each PSI for diagnoses that were actually present on admission on unadjusted PSI rates.
Methods: The latest AHRQ PSI software, which allows users to produce 14 of the 20 PSIs for adult inpatients both without and with a "present on admission" (PoA) variable, was applied to administrative data for adult patients discharged from the University of Michigan Health System (UMHS) in 2006. The impact of the PoA values on unadjusted PSI rates was evaluated. Because of concerns about the accuracy of PoA values, results were compared with those of a prior analysis at UMHS that was similar but based on a review of medical records.
Findings: Thirteen PSIs had at least 1 case in the numerator. Rates for all but 1 of the 13 were lower using the PoA values and the reduction was statistically significant for 5 PSIs: decubitus ulcer (P < 0.001), foreign body left in (P = 0.033), selected infections due to medical care (P < 0.001), postoperative physiologic and metabolic derangement (P = 0.039), and postoperative pulmonary embolism or deep vein thrombosis (P < 0.001). Results were consistent with those of the analysis of medical records.
Conclusions: Unadjusted PSI rates at UMHS are substantially overstated, because the PSIs do not differentiate preexisting conditions from complications and therefore include false positive cases. Because of these findings and the lack of a broader study of the validity of the indicators, PSIs should not be used to profile hospital performance.