Development of a Cloud-Based Clinical Decision Support System for Ophthalmology Triage Using Decision Tree Artificial...

Area of Science:

• Ophthalmology
• Medical Informatics
• Telemedicine

Background:

• Clinical decision support systems (CDSS) are increasingly vital in teleophthalmology for optimizing triage and referral processes.
• Traditional methods for on-call ophthalmology consults can be inefficient and lack standardization.

Purpose of the Study:

• To develop and implement a novel cloud-based decision tree CDSS for on-call ophthalmology consults.
• To standardize the triage and referral process, aiming for more accurate provisional diagnoses and urgency assessments.

Main Methods:

• A web-based decision tree algorithm was created using current guidelines and expert consensus.
• The CDSS collected specific patient ophthalmic information to generate a provisional diagnosis and urgency level.
• Statistical analyses, including Spearman-rho correlations and Cohen's kappa, were used to compare CDSS outputs with referring provider and ophthalmologist assessments.

Main Results:

• The CDSS demonstrated comparable performance to referring providers in determining diagnostic category (κ = 0.5898).
• The CDSS showed superior performance over referring providers in determining provisional diagnosis (agreement = 53.1%) and urgency (ρ = 0.5014).
• The CDSS identified a lower urgency level in 22.9% of cases compared to referring providers (6.3%).

Conclusions:

• This represents the first cloud-based CDSS in ophthalmology designed to enhance triage and referral.
• The CDSS improves diagnostic accuracy and urgency assessment, standardizes information collection, and moves beyond paper-based consults.
• Future work may involve machine learning models (e.g., random forest, CNN) to further refine speed and accuracy, particularly sensitivity.