Development of a nomogram to predict the incidence of acute kidney injury among ischemic stroke individuals during ICU hospitalization

Abstract

Background: Limited clinical prediction models exist to assess the likelihood of acute kidney injury (AKI) occurrence in ischemic stroke individuals. In this retrospective study, our aim was to construct a nomogram that utilizes commonly available clinical features to predict the occurrence of AKI during intensive care unit hospitalization among this patient population. Methods: In this study, the MIMIC-IV database was utilized to investigate potential risk factors associated with the incidence of AKI among ischemic stroke individuals. A predictive nomogram was developed based on these identified risk factors. The discriminative performance of the constructed nomogram was assessed. Calibration analysis was utilized to evaluate the calibration performance of the constructed model, assessing the agreement between predicted probabilities and actual outcomes. Furthermore, decision curve analysis (DCA) was employed to assess the clinical net benefit, taking into account the potential risks and benefits associated with different decision thresholds. Results: A total of 2089 ischemic stroke individuals were included and randomly allocated into developing (n = 1452) and verification cohorts (n = 637). Risk factors for AKI incidence in ischemic stroke individuals, determined through LASSO and logistic regression. The constructed nomogram had good performance in predicting the occurrence of AKI among ischemic stroke patients and provided significant improvement compared to existing scoring systems. DCA demonstrated satisfactory clinical net benefit of the constructed nomogram in both the validation and development cohorts. Conclusions: The developed nomogram exhibits robust predictive performance in forecasting AKI occurrence in ischemic stroke individuals.

Keywords

• Ischemic stroke
• Acute kidney injury
• MIMIC-IV
• Nomogram
• Prediction model