Abstract: TH-PO070
Development and Validation of a Risk Score for AKI After Cardiac Surgery
Session Information
- AKI: Epidemiology, Risk Factors, Prevention - I
November 07, 2019 | Location: Exhibit Hall, Walter E. Washington Convention Center
Abstract Time: 10:00 AM - 12:00 PM
Category: Acute Kidney Injury
- 101 AKI: Epidemiology, Risk Factors, and Prevention
Authors
- Chander, Subhash, Brigham and Women's Hospital, Boston, Massachusetts, United States
- Sharma, Shreyak, Brigham and Women's Hospital, Boston, Massachusetts, United States
- Short, Samuel, Brigham and Women's Hospital, Boston, Massachusetts, United States
- Rawn, James, Brigham and Women's Hospital, Boston, Massachusetts, United States
- Leaf, David E., Brigham and Women's Hospital, Boston, Massachusetts, United States
Background
AKI is a frequent and important complication of cardiac surgery. However, existing prediction models for cardiac surgery-associated AKI are limited by reliance on diagnostic and billing codes, lack of external validation, and inclusion of variables that can only be determined postoperatively (e.g., cardiopulmonary bypass time). Most importantly, existing models are mainly limited to prediction of severe AKI, whereas accurate risk stratification of more mild forms of AKI is crucial for enrichment of patients in early phase clinical trials.
Methods
We collected data from adults who underwent cardiac surgery between 2008 and 2018 at two major academic medical centers in Boston, MA. Data were obtained by querying the Society for Thoracic Surgeons National Database, along with electronic medical records. The primary exposures included clinically plausible demographics, comorbidities, laboratory values, and surgical characteristics, each of which was available preoperatively. The primary endpoint was AKI, defined according to KDIGO criteria as follows: an increase in serum creatinine ≥ 0.3 mg/dl within 48 hours, ≥50% in 7 days, or dialysis. We used forward selection, with a p value cutoff of 0.05, to develop the model, using a development (n=10,387) and external validation (n=8132) approach. The final model included 12 variables.
Results
The AKI event rate in the development and validation cohorts was 11% and 19%, respectively. A higher score monotonically predicted a higher risk of AKI in both cohorts (AUC 0.75 in the development cohort; AUC 0.71 in the validation cohort). The positive predictive value of a score >8 was 37% and 48% in the development and validation cohorts, respectively. The model was similarly predictive of more severe stages of AKI.
Conclusion
We present a 12-variable risk prediction score for AKI following cardiac surgery. We propose that this model could be used for both clinical prognostication and for research purposes.