Development of a risk prediction model for vancomycin-associated acute kidney injury: a multicentre retrospective study

Objectives To clarify the relationship between acute kidney injury (AKI) and vancomycin use in patients with renal impairment and to establish a risk score for AKI. Methods In this retrospective, multicentre, observational cohort study, trough and peak blood samples were collected from patients who initiated vancomycin therapy. The cumulative incidence of AKI within 14 days was compared among three groups classified according to renal function (estimated glomerular filtration rate ≥ 60, 30–59 and <30 mL/min/1.73 m2). The risk score and predicted probability of AKI incidence were calculated. AKI was defined in accordance with the Kidney Disease Improving Global Outcomes criteria. Results The incidence of AKI was 11.7% (99/847). No statistically significant difference was detected in the cumulative incidence of AKI among the three groups (P = 0.103). Cox proportional hazard analysis showed that the use of tazobactam/piperacillin [HR: 3.3, 95% CI (2.18–4.99), 2 points], vasopressors/inotropes [HR: 3.0, 95% CI (2.02–4.51), 2 points] and area under the concentration-time curve (AUC) on Day 2 [500–600 µg·h/mL: HR, 2.4, 95% CI (1.50–3.89), 1 point; >600 µg·h/mL: HR, 4.4, 95% CI (2.62–7.37), 3 points] were significantly related to the development of AKI. The predicted probabilities of AKI incidence were <5% (low-risk), 5% to <20% (moderate-risk), 20% to <40% (high-risk) and 40% to 67% (very high-risk), with total points of 0, 1–2, 3–4 and ≥5, respectively. Conclusions A risk prediction model was developed for AKI based on AUC exposure and concomitant medications, and no difference in AKI risk was observed across renal function categories.