Development and Validation of a Diabetic Retinopathy Risk Stratification Algorithm
Authors' conclusion
Relatively simple logistic regression models using nine readily available clinical variables can be used to rank order patients for onset of diabetic eye disease and thereby more efficiently prioritize and target screening for at risk patients.
Study context
- Population
- Patients from the Kaiser Permanente Northern California Diabetes Registry with no evidence of DR at a baseline DR screening, 2008–2020; 51.9% male, 42.1% White, mean age 60.0 (±13.1) years
- Participants
- 276,794 patients
- Exposure
- XGBoost machine learning algorithm vs nine-covariate logistic regression model
- Outcome
- Onset of proliferative DR, diabetic macular edema and referable retinopathy detected through routine screening; model performance by AUC
Methodological detailsprediction model development and internal validation
- Prediction model development and internal validation
Main finding
As reported in the abstract
AUC for onset of proliferative DR, diabetic macular edema and referable retinopathy
| Group | AUC proliferative DR | AUC diabetic macular edema | AUC referable retinopathy |
|---|---|---|---|
| XGBoost | 0.86 (95% CI, 0.86-0.87) | 0.76 (95% CI, 0.75-0.77) | 0.78 (95% CI, 0.78-0.79) |
| Logistic regression | 0.82 (95% CI, 0.80-0.83) | 0.73 (95% CI, 0.72-0.74) | 0.75 (95% CI, 0.75-0.76) |
AUC proliferative DR
XGBoost0.86 (95% CI, 0.86-0.87)
Logistic regression0.82 (95% CI, 0.80-0.83)
AUC diabetic macular edema
XGBoost0.76 (95% CI, 0.75-0.77)
Logistic regression0.73 (95% CI, 0.72-0.74)
AUC referable retinopathy
XGBoost0.78 (95% CI, 0.78-0.79)
Logistic regression0.75 (95% CI, 0.75-0.76)
Observational data: associations, not effects.
Original publication
Development and Validation of a Diabetic Retinopathy Risk Stratification Algorithm. Diabetes Care. Published 17 March 2023. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →