Dynamic and Individualized Prediction of Acute Kidney Injury via Longitudinal Peripheral Blood Cell Trajectories among COVID-19 Inpatients: A Multicenter Prospective Cohort Study

Background Acute kidney injury (AKI) is a serious complication of COVID-19 inpatients, requiring early identification of high-risk individuals by dynamic biomarkers. We evaluated the association between longitudinal peripheral blood count (PBC) trajectories and AKI in these patients by joint model analysis. Methods In a prospective cohort of 3,691 hospitalized adults with COVID-19 from six tertiary hospitals in China, longitudinal lymphocyte, neutrophil, and platelet counts and four derived ratios (NLR, PLR, NPR and NLPR) were analyzed. Joint model was performed to evaluate associations between PBC trajectories and in-hospital AKI. PBC thresholds on admission were identified by restricted cubic splines for further stratified joint model analyses. Competing-risk joint models then performed to evaluate the associations between PBC trajectories and renal recovery. Results Among 3,691 COVID-19 inpatients, 768 (20.8%) developed AKI during hospitalization. Joint model analysis revealed that trajectories of all seven PBC parameters were significantly associated with AKI risk (all P < 0.001). Increasing lymphocyte (HR=0.46, 95% CI 0.39–0.53) and platelet (HR=0.63, 95% CI 0.55–0.72) levels were associated with decreasing AKI risk, while elevated neutrophils (HR=3.33, 95% CI 2.70–4.13) and PBC ratios (NLR, NPR, PLR, NLPR) were associated with increased AKI risk. Stratified analyses showed that in patients with lower NLR and NPR baseline levels, larger risk magnitudes were observed. No significant associations were observed between biomarker trajectories and renal recovery. Conclusion By joint model analyses, dynamic PBC trajectories enabled individualized, real-time AKI risk evaluation in COVID-19 patients.