Systematic review of risk prediction models for diabetic retinopathy in type 2 diabetes mellitus
10.3760/cma.j.cn115682-20221113-05488
- VernacularTitle:2型糖尿病视网膜病变风险预测模型的系统评价
- Author:
Xiyang LIU
1
;
Wei ZHANG
;
Ruobing ZHAO
;
Yimeng FAN
;
Chen CHEN
Author Information
1. 江苏大学医学院，镇江 212000
- Keywords:
Diabetes mellitus;
Diabetic retinopathy;
Prediction model;
Systematic review
- From:
Chinese Journal of Modern Nursing
2023;29(32):4429-4436
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To systematically review the risk prediction models for diabetic retinopathy in patients with type 2 diabetes mellitus (T2DM) .Methods:Literature regarding the risk prediction models for diabetic retinopathy in T2DM were searched from CINAHL, PubMed, Web of Science, Cochrane Library, Embase, Wanfang, CNKI, VIP, and China Biology Medicine disc, with the search timeframe extending to October 1, 2022. Two researchers independently selected literature and extracted data, and the bias risk and applicability of the included literature were analyzed using the PROBAST tool for assessing risk of bias in prediction model studies.Results:A total of 14 articles were included. Age, duration of diabetes, glycated hemoglobin, proteinuria, blood pressure, and serum creatinine levels were the main predictive factors for diabetic retinopathy in T2DM; the Area Under the Receiver Operating Characteristic (ROC) curve of the predictive models ranged from 0.683 to 0.984, with calibration conducted in 6 models; 4 studies used external validation, while the rest used internal validation. All studies demonstrated good applicability, but all presented bias risk.Conclusions:The risk prediction models for diabetic retinopathy in T2DM demonstrate good predictive performance. However, high bias risk due to methodological flaws (such as improper handling of missing data, inappropriate methods for variable selection, lack of blinding, etc.) indicates they cannot be directly applied to clinical practice. Future work should either conduct extensive validation on existing models or undertake large-scale, diverse prospective studies to develop predictive models with superior performance and user-friendly application.
