A simple risk score for chronic kidney disease using administrative data: A population-based cohort study

Abstract

AbstractBackgroundWe did this study to develop and validate a risk score for new chronic kidney disease (CKD), focusing on predictors that are typically available in Canadian administrative health datasets.MethodsThis was a retrospective population-based cohort study using data from the Alberta Kidney Disease Network database: 3,558,192 adult participants were followed from April 1, 2007 to March 31, 2019. We developed a simple score to predict reduced glomerular filtration rate using bootstrapping (100 iterations with replacement) and internally validated the score using the original dataset.FindingsThe final score had a maximum total of 9 points: age 50-70 years, moderate albuminuria, hypertension, diabetes and heart failure all received a single point, and age >70 years and severe albuminuria received three points. The C-statistic of the score for incident CKD was 0.9272 and the Brier score was 0.0053, indicating excellent discrimination. Graphical analysis demonstrated that predicted risk closely aligned with the observed risk of developing CKD, indicating a well-calibrated model.InterpretationWe have derived and internally validated a risk score for new CKD which is suitable for application to routinely collected Canadian administrative health data.FundingDavid Freeze chair in health services research