AKI Etiology Impression by Nephrologists: Evolution and Associated Outcomes

Background: The etiology of AKI in hospitalized patients often evolves between initial consultation and final evaluation. We investigated differences in characteristics and clinical outcomes of patients by nephrologist-determined AKI etiology. Methods: We included a cohort of patients from the KAT-AKI trial enrolled at 5 Yale New Haven Health affiliated hospitals (12/2021-2/2024) and who had a nephrology consult within 48h of AKI detection. EHRs were reviewed to capture consultant characteristics and the clinical impression of AKI etiology for the first 3 days and final day of assessment. We examined etiologies of AKI, their evolution, and their association with clinical outcomes (AKI progression, dialysis, mortality within 14d). Results: Table 1 shows baseline characteristics of n=363 patients included. 73 unique nephrologists were involved. By final assessment, 48% had a change in clinical impression of their AKI etiology. The most common etiologies across follow up were multifactorial, ATN, and unclear (Figure 1a). Figure 1b shows the change in etiology during follow-up, most commonly from multifactorial to ATN or CIN, and from ATN to multifactorial. In patients whose diagnosis remained multifactorial, the average count of presumed etiologies within that group decreased significantly over time (p=0.016). While a change in presumed AKI etiology was not significantly associated with death, dialysis, or AKI progression at 14 days (p>0.05), initial AKI etiology was predictive (p<0.05). Patients with an initial diagnosis of ATN had worse outcomes (aOR [95%CI]): mortality: 2.7 [1.20,5.85], dialysis: 4.2, [1.6,10.2], and AKI progression: 3.7, [1.8,8.0]. Excluding “multifactorial,” “unclear,” and “other” from both diagnostic fields showed high agreement (κ = 0.85, n=99). Excluding “multifactorial” alone from final diagnosis, agreement was moderate (κ = 0.53, n=148). Conclusion: AKI etiology impressions commonly evolve. While a change in the clinical impression was not associated with clinical outcomes, the initially presumed etiology was predictive of worse outcomes. Funding: Other U.S. Government Support - Characteristics-----------------------------------------Demographics Age, mean (IQR), years Female, n (%) Comorbidities, n (%) CKD HTN DM CHF Cirrhosis Malignancy Baseline Kidney Function sCr, median (IQR), mg/dL At Randomization mSOFA, median (IQR) sCr at AKI, median (IQR), mg/dL Hospital service, n (%) General medical floor Specialist medical floor Surgical floor ICU/SDU units Consult Cohort (n=363) ----------------------------------------- 72.5 [62.0, 81.3] 167 (46.0) 198 (54.5) 301 (82.9) 198 (54.5) 193 (53.2) 39 (10.7) 22 (6.1) 1.8 [1.20, 2.40] 3.0 [2.0, 4.0] 2.2 [1.6, 2.9] 203 (55.9) 55 (15.2) 26 (7.2) 79 (21.8) KAT-AKI Full Cohort (n=4003) ----------------------------------------- 71.9 [60.6, 81.3] 1874 (46.8) 1665 (41.6) 3065 (76.6) 1884 (47.1) 1785 (44.6) 270 (6.7) 230 (5.7) 1.18 [0.90, 1.57] 2.0 [1.0, 3.0] 1.5 [1.2, 2.0] 1996 (49.9) 553 (13.8) 667 (16.7) 786 (19.6)