Earlier confirmation of glaucoma progression with risk-weighted visual field and OCT scheduling
In the authors' words
Glaucoma is monitored with visual field (VF) tests of function and optical coherence tomography (OCT) of structure, but models that predict progression do not determine when each test should be repeated. Here we present GEAR (gated expected-delay allocation rule), a policy that pairs a longitudinal risk predictor with a decision layer and gives each eye separate VF and OCT intervals. In the decision layer, clinical gates define the admissible test intervals, and an explicit objective selects the intervals that minimize risk-weighted detection delay plus a penalty on expected testing, with penalties fixed to a routine-testing budget. In retrospective replay of held-out records (5,935 visits from 1,681 patients), GEAR with risk-based confirmation timing confirmed simulated progression 3.4 months earlier than annual VF and OCT testing (95% confidence interval, 2.7-4.1) at a comparable testing rate. The reduction persisted against a guideline-informed comparator and a reconstructed Kalman comparator, under an alternative onset assumption and with a VF-only adaptation in an external dataset.
Appeared: Thursday, November 19. medRxiv. Preprint, not yet peer-reviewed.