Author + information
- Liu Xing and
- Hong Yuan
A model for predicting surgery related AKI in hypertension patients is limited. This study was conducted to develop and validate a risk model of surgery related AKI using preoperative risk factors in a large cohort.
In total, 24357 hypertension patients aged ≥18 years who underwent general surgery were enrolled. Subjects were randomly classified into the train (n=17020) and validation (n=7337) cohorts. The AKI events were defined based on the KDIGO (Kidney Disease: Improving Global Outcomes) AKI classification system. Models were developed by stepwise multivariate logistic regression. To compare the performance of the models to other published models, the models were built using the published equations. A comprehensive search of Medline Database was performed. Studies were included if a model was developed to predict adult (≥18 years old) AKI based on KDIGO AKI classification system after any kinds of surgery and reported in the form of a scoring system or algorithm.
During hospitalization, surgery related AKI developed in 1982 patients (8.1%). In the train cohort, our model, named Xiang-ya Model, consisted of blood urea nitrogen (BUN), eGFR, NLR, pulmonary infection, age, prothrombin time, uric acid, and serum albumin. The C-index for the validation set (0.88, 95% CI 0.86-0.89) was almost equal to the performance achieved in the training set (C-index =0.88, 95% CI 0.86-0.89). When this model was tested in other published model, only one model by Kate et al 2014 fulfilled the inclusion criteria. Our model outperformed the model by Kate et al with an NRI of 82.60% (95%CI 69.1%-94.1%) and an IDI of 13% (95%CI 10.46%-15.70%) in those underwent cardiac surgery patients (n=1726), respectively.
An AKI risk model based on preoperative risk factors and biomarkers was developed and showed a feasible model performance for predicting events in a large cohort of hypertensive patients underwent general surgery.