Kidney Week

Abstract: TH-PO071

The Scoring Model for Prediction of Acute Renal Replacement Therapy in Intensive Care Unit Patients with AKI in a Limited-Resource Country

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

• Pongsittisak, Wanjak, Navamindradhiraj University, Bangkok, Thailand

Wanjak Pongsittisak,

• Phonsawang, Kashane, Faculty of medicine, Vajira hospital, Bangkok, Thailand

Kashane Phonsawang,

Background

Acute kidney injury (AKI) is a common problem with high mortality in clinical practice, especially in the intensive care unit (ICU) patients. Making-a-decision of Acute Renal replacement therapy (ARRT) initiation has still been state-of-the-arts. This study aimed to generate and validate a scoring model for prediction of ARRT in ICU patients with AKI (ARRT score).

Methods

We performed a retrospective cohort study of ICU patients with AKI in a university hospital, Thailand. Risk factors of 7-day AKI requiring ARRT (7d-ARRT) were derived from the medical records between January 2013 and June 2015 (derivation cohort; der-cohort). We generate an ARRT score by the significant risk factors from the multivariate logistic analysis. To find the best model, we applied the area under the receiver operating characteristic curve (AUROC) analysis and Akaike information criterion (AIC). The ARRT score was validated by the data between June 2015 and December 2015 (validation cohort; val-cohort).

Results

The study included 292 patients in a der-cohort and 57 patients in a val-cohort. We found the best model to predict 7d-ARRT was oliguria (<0.5 ml/kg/hr after resuscitation), advanced CKD (eGFR < 45 ml/min/1.73m²) and severity of AKI at ICU admission [der-cohort: AUROC = 0.768, AIC = 201.02, val-cohort: AUROC = 0.845, AIC = 57.04]. These risk factors were used for generation of ARRT score by weighting their score from coefficients value of each risk factors (figure 1). At 4 points of the ARRT score, specificity was 84.2%, 81.6% and sensitivity was 55.8%, 73.7% for der-cohort and val-cohort respectively.

Conclusion

We strongly recommend that ARRT score ≥ 4 points could predict 7d-ARRT.
We suggest further large prospective cohort study to validate our ARRT scoring.

Funding

• Government Support - Non-U.S.