3007 — How Valid is the AHRQ Patient Safety Indicator Postoperative Metabolic Derangements?
Borzecki AM (Bedford COE), Shin M
(Boston COE), Cevasco M
(Tufts Medical Center), Rosen AK
Postoperative acute kidney injury (AKI) requiring dialysis is associated with in-hospital mortality rates above 60% and is among the most costly of postoperative complications. Because of its seriousness, high associated resource use, and potential preventability, the Agency for Healthcare Research and Quality developed Postoperative Metabolic Derangements (PMD) as a Patient Safety Indicator. PMD screens for preventable AKI requiring dialysis following elective surgery, plus abnormalities of glucose control, using administrative data. Its accuracy in identifying true events is unknown. Thus, we examined PMD’s positive predictive value (PPV).
We applied the PSI software to 2003-2007 VA discharge data from 28 hospitals. Two nurse-abstractors reviewed medical records of 119 software-flagged PMD cases. (This included all available diabetes cases.) We calculated PPVs overall and separately for AKI and diabetes-related complications. We examined false positives (FPs) to determine reasons for incorrect identification, and true positives (TPs) to determine clinical consequences and potential PMD risk factors. Abstractor inter-rater reliability was 98%.
Overall, 77 cases were TPs (PPV 65%, 95% CI 56-73%); of 104 AKI cases, 75 were TPs (PPV 72%, 63-80%); of 15 diabetes cases, 2 were TPs (PPV 13%, 2-39%). Of all FPs, 40% represented non-elective admissions, and 48% either had the complication present on admission (POA) or had end-stage renal disease POA. Of AKI TPs, 45% died; 27% were discharged on dialysis. Of AKI-associated index procedures, the largest proportion were cardiac 29%; 16% involved the aorta; 56% of TPs had preoperative chronic kidney disease. AKI was most commonly due to perioperative renal hypoperfusion (81% of TPs), followed by nephrotoxins (19%) including contrast (12%).
Given its low frequency and PPV, we recommend dropping diabetes complications from the indicator and focusing on AKI. Due to its moderate PPV, PMD representing AKI events should continue to be used as a screen rather than as a definitive quality measure for patient safety events. Introducing admission status and POA codes would significantly improve this indicator’s PPV.
Many PMD-identified cases appeared to be at high risk of postoperative AKI based on patient and procedure-related factors. Whether such cases could have been prevented through improvements in care requires further assessment.