Abstract: SA-PO060

A Predictive Model for CKD Development after Severe AKI in Children: A Single Center Prospective Study

Session Information

Category: Acute Kidney Injury

  • 003 AKI: Clinical and Translational

Authors

  • Goldstein, Stuart, Cincinnati Children's Hospital, Cincinnati, Ohio, United States
  • Claes, Donna J., Cincinnati Children's Hospital, Cincinnati, Ohio, United States
  • Englert, Jacob R, Northern Kentucky University, Highland Heights, Kentucky, United States
  • Kain, Parker J, Northern Kentucky University, Highland Heights, Kentucky, United States
  • Driehaus, Rachel Marie, Northern Kentucky University, Highland Heights, Kentucky, United States
  • Kasturiratna, Dhanuja, Northern Kentucky University, Highland Heights, Kentucky, United States
Background

The association between hospital acquired AKI and CKD development in children has only been studied in either select cohorts comprised of convenience samples of available patients (pts) or from administrative data.

Methods

We conducted a prospective study of all non-BMT pts who developed severe AKI (sAKI, Pediatric Modified RIFLE-I or F) for at least 2 days to assess for CKD development after sAKI. Risk factors included standard of care lab data, AKI severity and duration, solid organ transplant status and nephrotoxic medication burden. The primary outcome was eGFR <90 ml/min/1.73m2 (CKiD Schwartz formula) at 6 months after sAKI. Significant predictors were identified based on a series of bivariate analyses (p<0.3), which were used to build several logistic regression models. The models were then compared and validated by Receiver Operator Characteristic (ROC) and Leave One Out Cross Validation (LOOCV).

Results

193 pts who developed Stage 2 AKI were enrolled in this study (99M, 94F, mean age 10.5+6.2yrs, mean weight 43.7+31.0 kg). The median [IQR] AKI duration was 12 days [7, 21]. 45 pts required RRT. 79 pts had an eGFR assessment at 6 months, 24 of whom had an eGFR <90. On bivariate analysis, age, AKI duration, number of nephrotoxic medications, history of transplantation, lower eGFR, and lower serum albumin at AKI diagnosis was associated with CKD development at 6 months. Interestingly, RRT provision was not associated with CKD development. Seven candidate models were developed with AUCs that ranged from 0.81 to 0.82 and LOOCV that ranged from 0.67 to 0.76. The best candidate model (see table) based on a combination of AUC, LOOCV and simplicity (5 variables vs. 7) had an AUC 0.81 [95%CI 70.2-92.6] and LOOCV value of 0.76.

Conclusion

A simple 5 variable model was developed to predict presence of CKD at 6 months in children who developed pRIFLE-I/F AKI from any cause. Further prospective work will need to be conducted to validate and calibrate this model.

Best Predictive Model for AKI to CKD Development
VariableCoefficientsp-valuesOdds Ratios(95% CI)
Age0.09040.121.09(0.98,1.24)
AKI Days0.03830.091.04(1.0,1.1)
RRT (yes)-1.69670.130.18(0.013,1.3)
eGFR (nadir)-0.02840.0040.97(0.95,0.98)
Transplant (yes)1.89160.036.63(1.27,43.4)