Risk Factors for Nondiagnostic Imaging in a Real-World Deployment of Artificial Intelligence Diabetic Retinal Examinations in an Integrated Healthcare System: Maximizing Workflow Efficiency Through Predictive Dilation
Authors' conclusion
We used factors associated with nondiagnostic results to design a novel, predictive dilation workflow, where patients most likely to benefit from pharmacologic dilation are dilated a priori. This new workflow has the potential to be more efficient than reflexive dilation, thus maximizing the number of at-risk patients receiving their diabetic retinal examinations.
Study context
- Population
- Patients assessed with autonomous artificial intelligence (AI) diabetic retinal examination at Johns Hopkins Medicine (8/2020 to 5/2021), before reflexive dilation; 59% female, median age 59
- Participants
- 241 patients
- Exposure
- Patients with vs without nondiagnostic autonomous AI diabetic retinal examination results
- Outcome
- Factors associated with nondiagnostic results
Methodological detailssingle health system, multivariable logistic regression
- Single health system
- Multivariable logistic regression
Main finding
As reported in the abstract
Nondiagnostic results
123
51%
T1D, aOR
5.82 (1.45-23.40), P = .01 (95% CI)
Smoking, aOR
2.86 (1.36-5.99), P = .005 (95% CI)
Age, every 10-year increase, aOR
2.12 (1.62-2.77), P < .001 (95% CI)
Observational data: associations, not effects.
Further findings
Predictive model (T1D, smoking, age, race, sex, hypertension)
- AUC in five-fold cross-validation
- 0.76
- After feature elimination
Original publication
Composed by AI from the published abstract and checked against it. How Briefs are made →