# Construction and verification of a risk prediction model for diabetic retinopathy patients complicated with deep capillary plexus diabetic macular ischaemia

## TL;DR

This study identifies risk factors for a specific eye condition in diabetic retinopathy patients and builds a model to predict its occurrence.

## Contribution

A novel risk prediction model for deep capillary plexus diabetic macular ischaemia in diabetic retinopathy patients is developed and validated.

## Key findings

- The model's ROC area was 0.918, showing strong predictive accuracy.
- Internal and external validations confirmed the model's reliability.
- The model provides clinical net benefit across different thresholds.

## Abstract

To explore the risk factors for deep capillary plexus diabetic macular ischaemia (DCP-DMI) in patients with diabetic retinopathy and to establish and validate a risk prediction model.

Diabetic retinopathy patients who visited the ophthalmology department of Suining Central Hospital were selected as the study subjects and divided into a DCP-DMI group and a non-DCP-DMI group based on the presence or absence of DCP-DMI. Independent risk factors for DCP-DMI in DR patients were screened through univariate analysis and binary logistic regression analysis. Draw a nomogram to show the risk prediction model, and internal validation was performed using the Bootstrap resampling method, while external validation was conducted using the time period validation method. The clinical application value of the model was assessed using decision curve analysis (DCA).

Binary logistic regression analysis revealed that glucose excursion, duration, hypertension, proliferative diabetic retinopathy (PDR), deep capillary plexus vascular density (DCP-VD), deep capillary plexus geometric perfusion deficit (DCP-GPD), and choroidal vascular density (CVD) are independent risk factors for DCP-DMI in DR patients. The area under the ROC curve of the risk prediction model was 0.918 (95% CI: 0.893–0.943), and the Hosmer–Lemeshow test showed P = 0.573. The area under the ROC curve for internal validation was 0.915 (95% CI: 0.911–0.918), and the Hosmer–Lemeshow test showed P = 0.262. The area under the ROC curve for external validation was 0.793 (95% CI: 0.741–0.846), and the Hosmer–Lemeshow test showed P = 0.246. The DCA decision curves for the model group and validation group indicated that the red line representing the column diagram model was above the two special reference lines, and there was a certain net benefit at different thresholds.

Glucose excursion, duration, hypertension, PDR, DCP-VD, DCP-GPD, and CVD are closely related to the occurrence of DCP-DMI in DR patients. The column diagram model established on this basis has certain clinical reference significance.