Abstract: TH-PO002
Real-Time, Model-Driven Diagnostic and Therapeutic Evaluation of Patients at Risk of AKI: A Pilot and Feasibility Study
Session Information
- AKI: Epidemiology, Risk Factors, Prevention
October 25, 2018 | Location: Exhibit Hall, San Diego Convention Center
Abstract Time: 10:00 AM - 12:00 PM
Category: Acute Kidney Injury
- 101 AKI: Epidemiology, Risk Factors, and Prevention
Authors
- Wilson, Francis Perry, Yale School of Medicine, New Haven, Connecticut, United States
- Ugwuowo, Ugochukwu Caleb, Yale University , New Haven, Connecticut, United States
- Martin, Melissa, Yale School of Medicine, New Haven, Connecticut, United States
- Biswas, Aditya, Yale University , New Haven, Connecticut, United States
- Yamamoto, Yu, Yale University , New Haven, Connecticut, United States
Background
Acute kidney injury is typically defined by a rise in serum creatinine, but this is a late marker of the syndrome. Real-time, updated modeling of AKI risk using electronic health record (EHR) parameters may allow for targeting of diagnostic and therapeutic interventions earlier in the course of AKI.
Methods
We deployed a time-updated AKI prognostic model into the EHR of a large, tertiary care hospital. The model alerted study team members when any hospitalized adult had a greater than 30% risk of developing AKI within the next 48 hours: the “pre-AKI Alert”. Study personnel took urine and blood samples, examined the medical record, and followed patients for the development of creatinine-defined AKI. Our primary goal was to determine the feasibility of providing tailored pre-AKI recommendations in this clinical setting.
Results
Of 98 patients who met eligibility criteria, 68 were unable or refused to consent to study participation. Of the 30 who were enrolled, 5 developed AKI within 24 hours, 1 additional patient within 48 hours, and 1 later during the subsequent hospital admission. Lower-than-expected AKI rates were seen due to difficulty obtaining consent from the highest-risk patients. At the time of pre-AKI alert, 3 patients were receiving a potentially nephrotoxic agent, 24 had at least one electrolyte abnormality that could be addressed, and 3 had mean arterial pressure < 65mmHg (Table). Oxygen saturation was significantly lower in those who went on to develop AKI compared to those who did not with median (IQR) SPO2 94 (92 - 94) compated to 97 (95-98), p=0.03.
Conclusion
A real-time updated model successfully identified a group of patients at high risk for AKI, many of whom had nephrotoxin, electrolyte, blood pressure, or oxygenation abnormalities that could be proactively addressed. Model-based pre-AKI interventions are feasible and their efficacy should be explored in a larger study.
Characteristics at Pre-AKI Alert
| AKI within 48h (N=5) | No AKI Within 48h (N=25) | Total (N=30) | |
| Potassium > 5.4 meq/L | 1 | 2 | 3 |
| Bicarbonate < 22 meq/L | 3 | 18 | 21 |
| MAP < 65 mmHg | 1 | 2 | 3 |
| Receiving Nephrotoxin | 0 | 3 | 3 |
| Scheduled for Contrast Study | 0 | 2 | 2 |
Funding
- NIDDK Support