Risk Factors for Acute Kidney Injury in Patients Undergoing Total Joint Arthroplasty

TL;DR

This study identifies risk factors for acute kidney injury after joint replacement surgery, finding that age, health conditions, and certain medications increase the risk.

Contribution

The study introduces a machine learning approach to predict AKI risk factors in joint arthroplasty patients, achieving 85.2% accuracy with logistic regression.

Findings

AKI occurred in 14.7% of patients, with KDIGO1 and KDIGO2 classifications observed.

Logistic regression achieved the highest predictive accuracy (85.2%) among machine learning models.

Independent risk factors included advanced age, high ASA/CCI scores, diabetes, hypertension, NSAIDs, vancomycin, contrast material, and preoperative anemia.

Abstract

Objective: The present study investigates the incidence of postoperative acute kidney injury (AKI) and related risk factors in patients undergoing total joint arthroplasty. Methods: Included in the study were patients undergoing joint arthroplasty in 2015–2020. The patients with acute or chronic renal failure were excluded. The participants’ demographical data, American Society of Anesthesiologist (ASA) score, Charlson Comorbidity Index (CCI), type of operation, duration of surgery, presence of comorbidities, preoperative anemia, preoperative albumin levels, use of nephrotoxic agents, number of transfusions during perioperative period, presence of AKI according to Kidney Disease Improving Global Outcome (KDIGO) scores, and length of hospital and intensive care unit stay were evaluated. Results: The study was initiated with 1780 patients: 113 patients were excluded due to chronic kidney failure, 108 patients were excluded due to acute kidney failure, 648 patients were excluded because their data could not be reached, and finally, 911 patients were included in the study. AKI was detected in 134 patients (14.7%), and the number of patients in the KDIGO1 and KDIGO2 groups were 120 and 14, respectively. When evaluated according to the variable significance test result and clinical significance, the model consists of variables such as ASA, CCI, hypertension, nonsteroidal anti-inflammatory drugs (NSAIDs), vancomycin, beta lactam, contrast material and preoperative anemia, operation type, and anesthesia management. Machine learning analyses were performed using 11 variables (10 independent and 1 dependent variable). Logistic regression, naive Bayes, multilayer perceptron, bagging, and random forrest approaches were used for evaluation of the predictive performance. In an evaluation of the true classification ratio, the best result was obtained with the logistic regression method at 85.2%. Conclusions: The study revealed advanced age, high ASA and CCI, presence of diabetes and hypertension, NSAID, vancomycin and contrast material, and the presence of preoperative anemia to be independent risk factors for AKI.

Figures