Objective To investigate the risk factors for acute kidney injury (AKI) following the use of amphotericin B deoxycholate and to develop a predictive model to guide clinical monitoring and intervention.

Methods A retrospective analysis was conducted on hospitalized patients who received amphotericin B deoxycholate between January 2014 and September 2024. Patients were divided into a training set and a validation set. Demographic data, laboratory findings, and medication orders were collected. Based on the occurrence of AKI during treatment and within 7 days after discontinuation, patients were classified into an AKI group and a non-AKI group. Univariate analysis was used to screen for potential risk factors, multivariate logistic regression was employed to construct a predictive model, and model performance was evaluated using the area under the receiver operating characteristic curve (AUC) and the Hosmer-Lemeshow test.

Results The training set included 473 patients, comprising 255 males (53.91%) and 218 females (46.09%), with a median age of 52(35, 62) years. The AKI group consisted of 191 cases (40.38%), and the non-AKI group consisted of 282 cases (59.62%). The validation set included 114 patients, comprising 80 males (70.18%) and 34 females (29.82%), with a median age of 43.5 (31.0, 58.5) years. The AKI group consisted of 42 cases (36.84%), and the non-AKI group consisted of 72 cases (63.16%). Univariate analysis revealed statistically significant differences between the two groups in 23 factors (all P < 0.05), including demographic data (age, admission to intensive care unit (ICU), death, length of hospital stay), laboratory findings (serum creatinine above normal, urea above normal, neutrophil percentage below normal, neutrophil percentage above normal, lymphocyte percentage below normal, serum sodium above normal), concomitant medications (nonsteroidal anti-inflammatory drugs (NSAIDs), antifungal agents, neurological drugs, adrenergic drugs, sodium bicarbonate tablets/injection, diphenhydramine/promethazine injection, NSAIDs⁺ diphenhydramine/promethazine, NSAIDs⁺ corticosteroids, diphenhydramine/promethazine+corticosteroids, NSAIDs⁺ diphenhydramine/prome-thazine+corticosteroids), and comorbidities (diabetes mellitus, kidney disease, cardiac insufficiency)(all P < 0.05). Multivariate analysis identified admission to the ICU (OR=2.128, 95% CI: 1.415-3.201), elevated serum creatinine at admission (OR=1.920, 95% CI: 1.235-2.985), and comorbid cardiac insufficiency (OR=3.394, 95% CI: 1.369-8.417) as risk factors, while prophylactic use of diphenhydramine/promethazine injection (OR=0.182, 95% CI: 0.083-0.399) or sodium bicarbonate tablets/injection (OR=0.512, 95% CI: 0.339-0.773) were protective factors.The prediction model equation is as follows. logit(P)=lnP/(1-P)= -0.479+0.755X₁+0.652X₂+1.222X₃-1.702X₄-0.67X₅ (X₁ admitted to the intensive care unit, X₂ admitted to the hospital with higher serum creatinine than the normal value, X₃ complicated with cardiac insufficiency, X₄ combined with diphenhydramine/promethazine injection, X₅ was combined with sodium bicarbonate tablets/injection, X was taken as "1" or "0" according to whether the conditions were met). The AUC of the model was 0.735 (95% CI: 0.691-0.780) in the training set and 0.699 (95% CI: 0.604-0.795) in the validation set. The Hosmer-Lemeshow test yielded a χ² value of 4.048 (P=0.774), indicating good model calibration.

Conclusions Admission to the ICU, elevated serum creatinine at admission, and comorbid cardiac insufficiency as potential risk factors for AKI, while prophylactic use of diphenhydramine/promethazine or sodium bicarbonate showed a protective association. A predictive model with good discrimina-tion and calibration was developed, which may provide a basis for early identification of high-risk patients and timely adjustment of treatment strategies in clinical practice.