Neural Network-Based Risk Score Prediction of Severe AKI: Clinical Outcomes in a High-Risk Cohort

Background: Timely identification of patients at risk of severe acute Kidney Injury (AKI) may allow for clinical interventions and reduce adverse outcomes and the burden of AKI. Artificial intelligence-based risk scores may help identify patients earlier and improve outcomes. Methods: We identified inpatients at the University of Chicago who were deemed high risk for Stage 2 or higher AKI (severe AKI) based on having an ESTOP (recurrent neural network-based) AKI risk score > 0.091, within the last 8 hours. Demographics, history, AKI rates, mortality, and 90-day outcomes were collected. We recorded the incidence of KDIGO AKI for all patients at 2 and 7 days. We excluded patients who already had stage 2 or higher AKI during their admission as well as those with ESRD and kidney transplants. Results: We enrolled 95 patients, through day 7 43(54%) patients had severe AKI (stage2/3) while 37(46%) had non-severe AKI (stage 1) and 15(16%) never developed any AKI. In the severe AKI cohort, 30(70%) patients had a maximum of stage 2 AKI while 13(30%) developed stage 3 AKI; 3(7%) of whom received dialysis. Severe AKI developed within the first 48 hours in 31(33%) patients. The baseline serum creatinine (6-month prior median value) was 1.16 for the entire cohort, with no significant difference between those with and without severe AKI. For patients developing severe AKI across 7 days (n=43);12(28%) had serum creatinine(scr) based AKI, 22(51%) had UOP based and 9(21%) had severe AKI by both criteria. Outcomes data in the table shows no association between length of ICU or inpatient stay, 90 days repeat hospitalization across those with and without severe AKI. The inpatient mortality rate was high in those with severe AKI (p=0.01). Conclusion: Clinical implementation of ESTOP-AKI risk score identifies patients who are at risk of developing severe AKI. Funding: NIDDK Support Characteristics of Severe vs Non-severe AKI cohort - Characteristics Non-severe AKI(n=37) Severe AKI (n=43) p-value Bscr 1.12 (0.88 -1.3) 1.06 (0.86-1.36) 0.70 Enrollment Scr 1.4 (1.19-1.82) 1.6 (1.1-2.1) 0.35 ICU (n) 28 (76%) 31 (71%) 0.72 Total LOS (days) 9 (5-23) 12 (7-22) 0.90 LOS, ICU (days) 5.5 (3-17.5) 7 (2-22) 0.95 Mortality (n) 5 (13.7%) 16 (37%) 0.02 90 day re-admission (n) 13 (35%) 10 (23%) 0.31

Authors

Institutions

Publication Details

Journal
Journal of the American Society of Nephrology
Published
2025-10-01
DOI
https://doi.org/10.1681/asn.2025tmfksc0y
Primary Topic
Renal and Vascular Pathologies
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
article

Neural Network-Based Risk Score Prediction of Severe AKI: Clinical Outcomes in a High-Risk Cohort

Jay L. Koyner, Aiman Fatima, Oluwafunmibi Anjorin
Journal of the American Society of Nephrology
Renal and Vascular Pathologies
article

Neural Network-Based Risk Score Prediction of Severe AKI: Clinical Outcomes in a High-Risk Cohort

Jay L. Koyner, Aiman Fatima, Oluwafunmibi Anjorin
article en

Abstract

Background: Timely identification of patients at risk of severe acute Kidney Injury (AKI) may allow for clinical interventions and reduce adverse outcomes and the burden of AKI. Artificial intelligence-based risk scores may help identify patients earlier and improve outcomes. Methods: We identified inpatients at the University of Chicago who were deemed high risk for Stage 2 or higher AKI (severe AKI) based on having an ESTOP (recurrent neural network-based) AKI risk score > 0.091, within the last 8 hours. Demographics, history, AKI rates, mortality, and 90-day outcomes were collected. We recorded the incidence of KDIGO AKI for all patients at 2 and 7 days. We excluded patients who already had stage 2 or higher AKI during their admission as well as those with ESRD and kidney transplants. Results: We enrolled 95 patients, through day 7 43(54%) patients had severe AKI (stage2/3) while 37(46%) had non-severe AKI (stage 1) and 15(16%) never developed any AKI. In the severe AKI cohort, 30(70%) patients had a maximum of stage 2 AKI while 13(30%) developed stage 3 AKI; 3(7%) of whom received dialysis. Severe AKI developed within the first 48 hours in 31(33%) patients. The baseline serum creatinine (6-month prior median value) was 1.16 for the entire cohort, with no significant difference between those with and without severe AKI. For patients developing severe AKI across 7 days (n=43);12(28%) had serum creatinine(scr) based AKI, 22(51%) had UOP based and 9(21%) had severe AKI by both criteria. Outcomes data in the table shows no association between length of ICU or inpatient stay, 90 days repeat hospitalization across those with and without severe AKI. The inpatient mortality rate was high in those with severe AKI (p=0.01). Conclusion: Clinical implementation of ESTOP-AKI risk score identifies patients who are at risk of developing severe AKI. Funding: NIDDK Support Characteristics of Severe vs Non-severe AKI cohort - Characteristics Non-severe AKI(n=37) Severe AKI (n=43) p-value Bscr 1.12 (0.88 -1.3) 1.06 (0.86-1.36) 0.70 Enrollment Scr 1.4 (1.19-1.82) 1.6 (1.1-2.1) 0.35 ICU (n) 28 (76%) 31 (71%) 0.72 Total LOS (days) 9 (5-23) 12 (7-22) 0.90 LOS, ICU (days) 5.5 (3-17.5) 7 (2-22) 0.95 Mortality (n) 5 (13.7%) 16 (37%) 0.02 90 day re-admission (n) 13 (35%) 10 (23%) 0.31

Journal of the American Society of NephrologyVol. 36(10S)
University of Chicago (US)
Openalex Percentile: Top 57%
Renal and Vascular Pathologies
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.