Skip to main content

Table 4 Comparison of prediction models and model discrimination

From: Acute kidney injury as an independent risk factor for unplanned 90-day hospital readmissions

  Full model Best stepwise model Administrative data only model Biochemistry + age model Age alone model AKI alone
Characteristics
 Age * * * * *  
 Age term quadratic term *    * *  
 Male sex *      
 Residential care * * *    
 Deprived (highest quintile) *      
 Rural (settlement <3000) * * *    
Admission context
 Admissions in prior 1 year (per admission) * * *    
 Length of stay (per week) *      
 Emergency admission * * *    
 Medical ward admission *   *    
 Intensive care admission *      
Renal function
 AKI stages 0–3 * *   *   *
 Prior AKI count (per episode) *    *   
 Baseline eGFR (linear and quadratic) * *   *   
 Discharge creatinine 20% > baseline *      
Comorbidity
 Cancer * * *    
 Cardiac failure * * *    
 Cerebrovascular disease *      
 Dementia *      
 Diabetes * * *    
 Hemiplegia *      
 Liver disease *      
 Myocardial infarction *      
 Peptic ulcer disease *      
 Peripheral vascular disease *      
 Pulmonary * * *    
 Rheumatic disease *      
Model C statistic 0.699 0.698 0.685 0.655 0.594 0.587
95% confidence interval (0.688–0.709) (0.688–0.709) (0.675–0.696) (0.644–0.666) (0.582–0.605) (0.578–0.596)
P-value for AUC comparison with the next most complex model - 0.536 <0.001 <0.001 <0.001 0.344
  1. Abbreviations: AKI acute kidney injury, eGFR estimated glomerular filtration rate