# Development and validation of a model that predicts the risk of diabetic kidney disease in type 2 diabetes mellitus patients: a retrospective study

## TL;DR

This study developed a model to predict the risk of diabetic kidney disease in type 2 diabetes patients using easily accessible clinical indicators.

## Contribution

A validated risk prediction model for diabetic kidney disease using primary care-obtainable variables is proposed.

## Key findings

- Duration of diabetes, BMI, Scr, WBC, TyG-BMI, hypertension, and HDL-C were identified as independent risk factors for DKD.
- The model showed good discrimination (AUC 0.725 in training, 0.698 in validation) and calibration.
- The model is practical for use in primary healthcare settings due to the availability of its predictors.

## Abstract

This study aimed to identify independent risk factors for DKD in T2DM patients and develop a risk prediction model with internal validation.

We retrospectively collected data from 1,049 T2DM patients undergoing community health checks in Longhua District (2024). Patients were divided into DKD and non-DKD groups, then randomly divided into training (n=735) and validation (n=314) sets in 7:3 ratio.

The results of the binary logistic regression analysis showed that the duration of diabetes (OR 1.037, 95% CI: 1.005-1.07, P = 0.024), BMI (OR 0.869, 95% CI: 0.762-0.992, P = 0.037), Scr (OR 1.019, 95% CI: 1.010-1.028, P = 0.000), WBC (OR 1.141, 95% CI: 1.019-1.279, P = 0.023), and TyG-BMI (OR 1.019, 95% CI: 1.1007-1.030, P = 0.002) were independent risk factors for the occurrence of DKD in T2DM. Seven predictors including duration of diabetes, BMI, Scr, WBC, TyG-BMI, hypertension, and HDL-C, which were identified via binary logistic analysis. We visualized the predictive model in the form of a nomogram and evaluated its predictive performance. The model demonstrated good discrimination (AUC: training 0.725, validation 0.698) and calibration (H-L test P>0.05 for both groups). Decision curve analysis confirmed its clinical utility by showing higher net benefit than extreme scenarios.

All seven indicators in this model are readily obtainable in primary healthcare settings, providing a practical tool for primary care physicians to conduct DKD risk prediction in general practice.