Review Proposes Integrated Ophthalmic Surgical Triage Framework

A mini-review published in Frontiers in Medicine proposes a framework for integrating quantitative ocular biometrics with automated electronic health record (EHR) risk stratification to support perioperative decision-making in ophthalmology. The authors, Jacob Craft and David Fleischman from the University of North Carolina School of Medicine, argue that current clinical systems fail to combine device-derived ocular measurements with medical, social, and utilization data. This separation limits the ability of automated tools to assess the full risk profile of a surgical patient, leaving a gap in real-time, patient-specific perioperative decision support.

The review examines how automated EHR-based risk stratification and quantitative ocular biometrics could be combined to address surgical readiness, technical complexity, and adverse outcomes. It highlights that ophthalmology is distinctive among surgical specialties due to the volume and precision of the quantitative data it generates. Routine care produces highly reproducible, procedure-specific ocular measurements, including biometry, keratometry, anterior chamber depth, and axial length. However, these measurements typically reside in separate device databases, while medical comorbidity and social information remain within the general EHR.

One critical application of this integrated approach is the prediction of surgical cancellations, which pose a significant burden in ophthalmic practice. The review notes that cancellation rates approaching 9% have been documented for glaucoma surgery, with mean delays of 63 days per cancelled case. In safety-net hospitals, baseline cancellation rates have exceeded 17%, with most occurring on the day of surgery. Among 123 glaucoma-practice cancellations analyzed, only 28% were judged preventable, with the leading preventable causes being lack of transportation, lack of timely clearance, and insurance-related issues. This indicates that any cancellation-prediction system operates against a ceiling of roughly a quarter of events, driven by logistical rather than ocular factors.

Across surgical specialties, cancellation models using demographic, comorbidity, socioeconomic, and appointment-related variables have achieved development areas under the curve (AUC) of 0.75 to 0.92, with validation ranges of 0.70 to 0.74. A systematic review of six such models encompassing 759,337 elective surgical procedures found that while discriminative performance was moderate, three of the six models had a high risk of bias. The review emphasizes that discriminative performance alone should not be taken as evidence of readiness for deployment, particularly given the heterogeneity and limited calibration reporting in existing studies.

In cataract surgery, the Epic Health Composite Score has been evaluated as an EHR-derived tool for predicting same-day cancellations. In a retrospective cohort of 1,204 cataract surgery patients, each 5-point increase in this score was independently associated with a 23% higher odds of same-day cancellation. However, the score provided only limited discrimination, with an AUC of 0.632, which is considered inadequate for individual-level triage. Crucially, the same study found that this composite score did not predict operative duration, anesthesia selection, surgical complications, or visual outcomes. These outcomes remained more closely associated with ophthalmic-specific factors, such as cataract complexity and prior ocular surgery, highlighting the limitation of generic health scores in capturing ocular-specific risk.

The review identifies several candidate predictors of technical complexity and surgical outcomes in cataract and glaucoma literature. For cataract surgery, high-risk ocular characteristics include small pupil size, pseudoexfoliation, zonular weakness, and shallow anterior chamber depth. The Triemli Cataract Score, a tool designed to classify cases as routine, complex, or highly complex, has shown potential in guiding resource allocation. Following its implementation at a Swiss hospital, intraoperative complication rates trended downward, although the score did not reach conventional statistical significance as an independent predictor of posterior capsule rupture in subsequent multivariable analysis. This suggests that while structured triage tools may influence outcomes, they require further validation before widespread clinical adoption.

Glaucoma surgery presents distinct challenges due to the heterogeneity of procedures and their specific postoperative demands. Quantitative anterior segment OCT has been used to inform preoperative planning, with studies finding that a significant portion of open-angle eyes had angle-opening distances too small to safely accommodate standard tube shunt devices. Additionally, conjunctival imaging parameters have been associated with filtration surgery outcomes, with lower conjunctival stromal thickness independently associated with surgical failure. These findings demonstrate the potential of quantitative biometrics to inform procedure-specific risks that are not captured by general health metrics.

Frailty and social vulnerability also play significant roles in glaucoma surgical outcomes and access to care. Studies have shown that geriatric assessment measures can improve the prediction of severe hemorrhagic complications after glaucoma surgery. Furthermore, patients residing farther from clinics or in areas of higher social vulnerability have greater odds of loss to follow-up and delayed surgery. These factors suggest that frailty may influence both perioperative risk and access to treatment, although it remains unclear whether lower surgical utilization reflects appropriate clinical selection or undertreatment.

The proposed framework envisions a modular architecture that maintains a shared EHR-based infrastructure while allowing the relative weighting of input domains to vary by procedure type. The readiness and cancellation model would run when surgery is requested, automatically populating patient-level EHR variables such as high-risk systemic diagnoses and social vulnerability indicators. In parallel, the complexity and outcome models would depend on the ocular stream, firing later in the pathway upon completion of biometry or finalization of the relevant examination. This approach ensures that outputs requiring unavailable inputs are suppressed and labeled as “not assessed,” improving adaptability in settings where ocular biometry is not always available.

The authors emphasize that prediction must be linked to an effective intervention to be clinically useful. The Wongtangman et al. care bundle, which combined automated screening with patient engagement tools, reduced same-day cancellation rates from 12.3% to 8.4% in an ambulatory otorhinolaryngology program. However, the bundle design prevented the isolation of the risk score’s independent contribution, with patient engagement via portal or text being the largest single driver of improvement. This underscores the need for prospective studies to assess whether risk-stratified decision support improves the appropriateness, safety, and accessibility of surgical care.

Several limitations warrant consideration in the proposed framework. The evidence reviewed is methodologically heterogeneous, including administrative datasets, single-center cohorts, and variable outcome definitions. The readiness and cancellation arm rests on larger multi-institution data, while the complexity and outcome arms rely on single-center, externally unvalidated work. The authors caution that these three components should not be read as equally ready for clinical use. Additionally, the proprietary nature of tools like the Epic Health Composite Score underscores the importance of transparency in future models, which should favor explainable logic with auditable components.

The next priority identified by the review is the prospective development and validation of a modular family of ophthalmology-specific models. These models should combine transparent EHR-derived variables with quantitative ocular biometrics and procedure-specific clinical findings. Evaluation should extend beyond discrimination to include calibration, external validity, workflow burden, and clinical utility. Explicit safeguards are required to ensure that risk markers, such as race, do not become access barriers, with the authors suggesting that race be reserved for outcome auditing to improve surgical equity. By combining automated patient-level risk assessment with quantitative, procedure-specific ocular data, ophthalmology can leverage its uniquely rich measurement environment to provide decision support that neither data stream could deliver alone.

Subject of Research: Integration of quantitative ocular biometrics and EHR-based risk stratification for ophthalmic surgical triage

Article Title: Integrating quantitative biometrics and automated risk stratification for ophthalmic surgical triage

Article References: Craft, J., & Fleischman, D. (2026). Integrating quantitative biometrics and automated risk stratification for ophthalmic surgical triage. Frontiers in Medicine, 13, Article 1975712. https://doi.org/10.3389/fmed.2026.1975712

Image Credits: AI Generated

DOI: 10.3389/fmed.2026.1975712

Keywords: ophthalmology, surgical triage, electronic health record, risk stratification, ocular biometrics, cataract surgery, glaucoma surgery, clinical decision support, surgical cancellation, perioperative care, Integrating, quantitative