Development and Validation of a Risk Prediction Model for In-Hospital Mortality of Adult Acute Kidney Injury Patients Requiring Intermittent Hemodialysis, Gondar, Northwest Ethiopia
Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Introduction
Acute kidney injury is a common complication in hospitalized patients and is strongly associated with increased morbidity, mortality, and healthcare costs. In spite of advancements in dialysis techniques and critical care, mortality rates in acute kidney injury patients requiring intermittent hemodialysis remain high, depending on the severity of illness and underlying comorbidities. This Study aimed to develop and validate a risk prediction model of In-hospital mortality for adult acute kidney injury patients requiring intermittent hemodialysis.
Method
We retrospectively recruited 829 patients admitted to University of Gondar and Felege Hiwot Comprehensive Specialized Hospitals from 2 October 2016 to 30 March 2025.Demographic, clinical and laboratory candidate variables were selected based on a literature review and expert opinion. A stepwise backward elimination technique was used and the role of each predictor in the multivariable analysis was assessed by the Akaike Information Criterion . Presence of multicollinearity was checked using variance inflation factor .Bootstrapping technique was used for internal validation cohorts. We assessed model discrimination using the area under the receiver operator characteristic curve. Clinical utility and applicability of the model was assessed with decision curve analysis and nomogram respectively.
Result
A total of 829 adult patients were included in the study. Of these 247 (29.8%) (95%CI: 24.53-32.53) patients died during hospitalization. The predictive model based on seven predictors (age, vasopressor, mechanical ventilation, immunosuppression, cardiovascular disease, platelet count, and serum creatinine) exhibited a predictive performance, as assessed by AUC of 0.85(95%CI: .81-0.89) in development dataset, and 0.8(95%CI: .77-.83) in internal validation for the discrimination of the model.
Conclusion
This study successfully developed and internally validated a risk prediction model for in-hospital mortality in adult patients with acute kidney injury requiring intermittent hemodialysis. The model demonstrates good potential to support clinical decision-making by enabling individualized mortality risk estimation at the time of initiating renal replacement therapy. With external validation in independent cohorts, it could become a valuable tool to guide prognosis and facilitate shared decision-making in clinical practice.