# Development and validation of early-stage and progression prediction models for chronic kidney disease: a retrospective study

## TL;DR

This study developed and validated models to predict early-stage chronic kidney disease and its progression, showing strong accuracy and potential clinical utility.

## Contribution

The novel contribution is the creation of high-performance nomograms for CKD risk assessment and progression prediction.

## Key findings

- The early-stage CKD prediction model achieved AUCs of 0.981 (training) and 0.969 (validation).
- The progression model showed AUCs of 0.984 (training) and 0.972 (validation).
- Decision curve analysis confirmed the models' clinical relevance and applicability.

## Abstract

Chronic kidney disease (CKD) poses a significant public health burden. This study aimed to evaluate the associations between clinical laboratory indices and CKD and to develop prediction and prognostic models for CKD risk assessment and disease progression.

Between January 2008 and June 2018, we enrolled 500 healthy controls, 445 patients with early-stage CKD (G1–G2), and 527 patients with CKD G5 at the First Hospital of China Medical University. Logistic regression analyses were performed to identify independent predictors for the presence of CKD and progression to advanced disease, which were subsequently incorporated into visual nomograms. Model performance was evaluated using area under the receiver operating characteristic curves (AUC) and calibration plots. Clinical utility was assessed using decision curve analysis (DCA) and clinical impact curves (CIC).

The early-stage CKD prediction nomogram achieved an AUC of 0.981 in the training set and 0.969 in the validation set. The progression nomogram demonstrated AUC values of 0.984 and 0.972 in the training and validation sets, respectively. DCA and CIC analyses further confirmed the clinical relevance and potential applicability of both models.

We developed and validated early-stage prediction and progression assessment for CKD, demonstrating high discriminative ability, good calibration, and significant clinical utility. These models may facilitate early detection and dynamic risk assessment in CKD management.