General inpatient noncardiac, nonvascular, nonobstetric, nonurologic procedures. Most patients had abdominal surgery.
Overall assessment: Moderate quality.
Prediction accuracy: Good accuracy (c-statistic 0.80).
Risk to reproducibility: Low. The model was developed in a large sample of 57,075 patients of whom 561 experienced acute kidney injury and internally validated in a randomly-selected cohort of 18,872 patients of whom 201 developed acute kidney injury.
Risk of data reporting bias: Moderate. The model was developed and tested in retrospective data and patients with missing preoperative serum creatinine data were excluded. Statistical imputation did not appreciably alter the results.
Risk to transportability: Moderate. The model was developed and tested in retrospective data from 121 study centers in the United States. Prospective validation in needed.
Prediction accuracy: Good accuracy (c-statistic 0.80).
Risk to reproducibility: Low. The model was developed in a large sample of 57,075 patients of whom 561 experienced acute kidney injury and internally validated in a randomly-selected cohort of 18,872 patients of whom 201 developed acute kidney injury.
Risk of data reporting bias: Moderate. The model was developed and tested in retrospective data and patients with missing preoperative serum creatinine data were excluded. Statistical imputation did not appreciably alter the results.
Risk to transportability: Moderate. The model was developed and tested in retrospective data from 121 study centers in the United States. Prospective validation in needed.