Validation of the OAKS prognostic model for acute kidney injury after gastrointestinal surgery

Wur Ahmed, Shipra Bhatia, Kenneth Mclean, Rachel A Khaw, D. E. A. Baker, Sivesh Kathir Kamarajah, S Bell, D Nepogodiev, Ewen M Harrison, James C Glasbey

Research output: Contribution to journalArticlepeer-review

Abstract / Description of output

Background: Postoperative acute kidney injury (AKI) is a common complication of major gastrointestinal surgery with an impact on short- and long-term survival. No validated system for risk stratification exists for this patient group. Therefore, this study aimed to externally validate a prognostic model for AKI after major gastrointestinal surgery in two multi-centre cohort studies.
Methods: The Outcomes After Kidney injury in Surgery (OAKS) prognostic model was developed to predict risk of AKI in the 7-days after surgery using 6 routine datapoints (age, sex, ASA grade, preoperative estimated glomerular filtration rate, planned open surgery and preoperative use of either an angiotensin-converting enzyme inhibitor or an angiotensin receptor blocker). Validation was performed within two independent cohorts: (1) a prospective multi-centre, international study (“IMAGINE”) of patients undergoing elective colorectal surgery (2018); (2) a retrospective regional cohort study (“Tayside”) in major abdominal surgery (2011-2015). Multivariable logistic regression was used to predict risk of AKI, with multiple imputation used to account for data missing at random. Prognostic accuracy was assessed for patients at high‐risk (>20%) of postoperative AKI.
Results: In the validation cohorts, 12.9% (n=661/5106) of patients in IMAGINE and 14.7% (n=106/719) in Tayside developed 7-day postoperative AKI. Using the OAKS model, 558 (9.6%) patients were classified as high risk. Less than 10% of patients classified as low-risk developed AKI in either cohort (negative predictive value >0.9). Upon external validation, the OAKS model retained an AUC ranging 0.655-0.681 (respectively: Tayside 95% CI: 0.596-0.714; IMAGINE 95% CI: 0.659-0.703), sensitivity values ranging 0.323-0.352 (respectively: IMAGINE 95% CI: 0.281-0.368; Tayside 95% CI: 0.253-0.461), and specificity ranging 0.881-0.890 (respectively: Tayside 95% CI: 0.853-0.905; IMAGINE 95% CI: 0.881-0.899).
Conclusion: The OAKS prognostic model can identify patients who are not at high-risk of postoperative AKI after gastrointestinal surgery with high specificity.

Original languageEnglish
JournalBJS Open
Publication statusPublished - 6 Jan 2022


Dive into the research topics of 'Validation of the OAKS prognostic model for acute kidney injury after gastrointestinal surgery'. Together they form a unique fingerprint.

Cite this