A weekly roundup of ophthalmic news from around the web.
Children take to eye patch wearing relatively well, as do their parents, say researchers who compared outcomes in 47 children aged 54–55 months with unilateral congenital cataract enrolled in the Infant Aphakia Treatment Study. All children had visual acuity of 20/200 or worse, and were divided into mild (0 to <15 minutes), moderate (15 to <120 minutes), or extensive (≥120 minutes) daily eye patching groups. There were no significant differences between the daily patching groups in terms of child behavioral issues, fine or gross motor skill problems, or mean parenting stress levels. The authors conclude that “health care practitioners may be reassured that continuing to prescribe patching of these young children is unlikely to cause significant harm to families’ level of stress and/or children’s development, even in those children with poor vision.” JAMA Ophthalmology
Age, gender, and other factors may predict glaucoma medication adherence. An observational study of 30,100 Danish adults with glaucoma found that 55% were adherent to their glaucoma medication during the first year from index date. This number dropped slightly in the second year from index date, to 52%. High medication adherence was associated with age >60 years, female sex, and lower Charlson Comorbidity Index score, while poor adherence was associated with greater health care costs and having adnexa, ischemic heart disease, and/or musculoskeletal diseases. Additional studies on the association between poor medication adherence and severity and progression of glaucoma disease may provide further illumination on this topic. BMJ Open Ophthalmology
Adding visual processing tests to standard dementia screening could improve future dementia prediction. Data from the population-based European Prospective Investigation into Cancer Norfolk (EPIC-Norfolk) cohort study were used to evaluate the ability of the standard simple and complex Visual Sensitivity Test (VST) tools to detect dementia. The VST and 2 dementia screening tools, the Hopkins Verbal Learning Test (HVLT) and the Short Form Extended Mental State Exam (SF-EMSE), were administered to 8623 EPIC-Norfolk participants aged 48–92 years. Low VST scores, indicating low visual processing, were associated with increased risk of future dementia (hazard ratios 1.39–1.56). While the VST had lower specificity and sensitivity than the HVLT and the SF-EMSE, the authors noted that the latter tools were not highly specific or sensitive, indicating that incorporating VST in cognitive assessment could improve overall screening accuracy. Scientific Reports
State-of-the-art large language model scores high on an ophthalmology exam. Questions from the UK Fellowship of the Royal College of Ophthalmology (FRCOphth) Part 2 examination were put into 4 large-language models (LLMs)—OpenAI’s ChatGPT-3.5 and -4, Meta AI’s LLaMA-2, and Google’s PaLM-2—and the answers were compared with those of trainee and fully qualified ophthalmologists. The FRCOphth questions were taken from a textbook not available online. GPT-4 performed better on a mock examination than the other 3 LLMs, and the mean scores seen with GPT-4 were statistically similar to the scores from the human ophthalmologists. Performance was consistent across question subjects and types. There needs to be more research to investigate and define the role of LLM in clinical practice. PLOS Digital Learning