# Development and validation of a prediction model for cognitive impairment in elderly patients with type 2 diabetes

## TL;DR

This study developed a model to predict cognitive impairment in elderly patients with type 2 diabetes, showing good accuracy and potential for early intervention.

## Contribution

The novel contribution is a validated prediction model for cognitive impairment in elderly T2DM patients using clinical and psychological factors.

## Key findings

- Age, HADS-D score, WMD, and HbA1c were identified as significant predictors of cognitive impairment in T2DM patients.
- The model achieved an AUC of 0.812 with good calibration and minimal overfitting after bootstrap validation.
- Decision curve analysis confirmed the model's clinical utility across a wide range of threshold probabilities.

## Abstract

China is experiencing rapid population aging, accompanied by a rising prevalence of type 2 diabetes mellitus (T2DM) and its complex complications. Cognitive impairment is one of the major complications of T2DM and currently lacks effective treatment. These two conditions can interact and aggravate each other, forming a vicious cycle.

This study aimed to identify reliable early predictors of cognitive impairment among elderly individuals with T2DM, in order to facilitate early intervention and delay disease progression.

A total of 202 elderly patients with T2DM hospitalized at Tianyou Hospital, affiliated with Wuhan University of Science and Technology, between May and September 2025 were enrolled. Cognitive function was assessed using the Montreal Cognitive Assessment (MoCA) with a cutoff score of 26. Seventy-two participants scoring ≥26 were assigned to the normal cognition group, and 130 participants scoring ≤25 were assigned to the cognitive impairment group. Demographic information, hematological and imaging parameters, and scale scores related to sleep quality, anxiety–depression status, and activities of daily living were collected. Statistical analyses were conducted using R version 4.5.

Least absolute shrinkage and selection operator regression selected 14 predictors. After analyzing the data, four factors remained independently associated with T2DM related cognitive impairment: age (OR = 1.100, 95% CI: 1.038–1.164, P = 0.001), HADS-D score (OR = 1.242, 95% CI: 1.081–1.426, P = 0.002), WMD (OR = 2.444, 95% CI: 1.137–5.254, P = 0.022), and HbA1c (OR = 1.264, 95% CI: 1.007–1.585, P = 0.043). The model demonstrated an AUC of 0.812 (95% CI: 0.778–0.891) and was well-calibrated (Hosmer-Lemeshow P = 0.661). After bootstrap validation, the optimism-corrected AUC was 0.751, indicating minimal overfitting. At the optimal cut-off of 0.685, the model achieved a sensitivity of 69.2% and a specificity of 81.9%, with a positive predictive value of 87.4% and a negative predictive value of 59.6%. DCA demonstrated a positive net benefit across threshold probabilities from 0.02 to 0.86, supporting the model’s clinical value.

This study developed a prediction model for T2DM related cognitive impairment in elderly Chinese patients. The model showed good discrimination, calibration, and clinical value, supporting its potential role for identifying high-risk populations. However, before using this model, more research is needed to confirm it’s performance in different people.