Use of Large Language Models to Assess the Likelihood of Epidemics From the Content of Tweets: Infodemiology Study.

BACKGROUND: Previous work suggests that Google searches could be useful in identifying conjunctivitis epidemics. Content-based assessment of social media content may provide additional value in serving as early indicators of conjunctivitis and other systemic infectious diseases. OBJECTIVE: We investigated whether large language models, specifically GPT-3.5 and GPT-4 (OpenAI), can provide probabilistic assessments of whether social media posts about conjunctivitis could indicate a regional outbreak. METHODS: A total of 12,194 conjunctivitis-related tweets were obtained using a targeted Boolean search in multiple languages from India, Guam (United States), Martinique (France), the Philippines, American Samoa (United States), Fiji, Costa Rica, Haiti, and the Bahamas, covering the time frame from January 1, 2012, to March 13, 2023. By providing these tweets via prompts to GPT-3.5 and GPT-4, we obtained probabilistic assessments that were validated by 2 human raters. We then calculated Pearson correlations of these time series with tweet volume and the occurrence of known outbreaks in these 9 locations, with time series bootstrap used to compute CIs. RESULTS: Probabilistic assessments derived from GPT-3.5 showed correlations of 0.60 (95% CI 0.47-0.70) and 0.53 (95% CI 0.40-0.65) with the 2 human raters, with higher results for GPT-4. The weekly averages of GPT-3.5 probabilities showed substantial correlations with weekly tweet volume for 44% (4/9) of the countries, with correlations ranging from 0.10 (95% CI 0.0-0.29) to 0.53 (95% CI 0.39-0.89), with larger correlations for GPT-4. More modest correlations were found for correlation with known epidemics, with substantial correlation only in American Samoa (0.40, 95% CI 0.16-0.81). CONCLUSIONS: These findings suggest that GPT prompting can efficiently assess the content of social media posts and indicate possible disease outbreaks to a degree of accuracy comparable to that of humans. Furthermore, we found that automated content analysis of tweets is related to tweet volume for conjunctivitis-related posts in some locations and to the occurrence of actual epidemics. Future work may improve the sensitivity and specificity of these methods for disease outbreak detection.

Shifts in ophthalmic care utilization during the COVID-19 pandemic in the US.

BACKGROUND: Healthcare restrictions during the COVID-19 pandemic, particularly in ophthalmology, led to a differential underutilization of care. An analytic approach is needed to characterize pandemic health services usage across many conditions. METHODS: A common analytical framework identified pandemic care utilization patterns across 261 ophthalmic diagnoses. Using a United States eye care registry, predictions of utilization expected without the pandemic were established for each diagnosis via models trained on pre-pandemic data. Pandemic effects on utilization were estimated by calculating deviations between observed and expected patient volumes from January 2020 to December 2021, with two sub-periods of focus: the hiatus (March-May 2020) and post-hiatus (June 2020-December 2021). Deviation patterns were analyzed using cluster analyses, data visualizations, and hypothesis testing. RESULTS: Records from 44.62 million patients and 2455 practices show lasting reductions in ophthalmic care utilization, including visits for leading causes of visual impairment (age-related macular degeneration, diabetic retinopathy, cataract, glaucoma). Mean deviations among all diagnoses are 67% below expectation during the hiatus peak, and 13% post-hiatus. Less severe conditions experience greater utilization reductions, with heterogeneities across diagnosis categories and pandemic phases. Intense post-hiatus reductions occur among non-vision-threatening conditions or asymptomatic precursors of vision-threatening diseases. Many conditions with above-average post-hiatus utilization pose a risk for irreversible morbidity, such as emergent pediatric, retinal, or uveitic diseases. CONCLUSIONS: We derive high-resolution insights on pandemic care utilization in the US from high-dimensional data using an analytical framework that can be applied to study healthcare disruptions in other settings and inform efforts to pinpoint unmet clinical needs.

The COVID-19 pandemic disrupted healthcare services globally, including eye care in the United States. Using a US eye disease database, we measured how the pandemic impacted patient visits for 261 eye diagnoses by comparing actual visit volumes for each diagnosis with what would have been expected without the pandemic. We identified groups of conditions with similar changes in visit levels and examined whether these shifts were related to characteristics of the diagnoses studied. We found extended decreases in patient presentations for most eye conditions, with greater reductions for less severe diagnoses, and with anomalies and differences in this trend across diagnosis categories and pandemic sub-periods. This highlights areas of potentially unmet need in vision care arising from the pandemic.

A Google Trends Approach to Identify Distinct Diurnal and Day-of-Week Web-Based Search Patterns Related to Conjunctivitis and Other Common Eye Conditions: Infodemiology Study.

BACKGROUND: Studies suggest diurnal patterns of occurrence of some eye conditions. Leveraging new information sources such as web-based search data to learn more about such patterns could improve the understanding of patients' eye-related conditions and well-being, better inform timing of clinical and remote eye care, and improve precision when targeting web-based public health campaigns toward underserved populations. OBJECTIVE: To investigate our hypothesis that the public is likely to consistently search about different ophthalmologic conditions at different hours of the day or days of week, we conducted an observational study using search data for terms related to ophthalmologic conditions such as conjunctivitis. We assessed whether search volumes reflected diurnal or day-of-week patterns and if those patterns were distinct from each other. METHODS: We designed a study to analyze and compare hourly search data for eye-related and control search terms, using time series regression models with trend and periodicity terms to remove outliers and then estimate diurnal effects. We planned a Google Trends setting, extracting data from 10 US states for the entire year of 2018. The exposure was internet search, and the participants were populations who searched through Google's search engine using our chosen study terms. Our main outcome measures included cyclical hourly and day-of-week web-based search patterns. For statistical analyses, we considered P<.001 to be statistically significant. RESULTS: Distinct diurnal (P<.001 for all search terms) and day-of-week search patterns for eye-related terms were observed but with differing peak time periods and cyclic strengths. Some diurnal patterns represented those reported from prior clinical studies. Of the eye-related terms, "pink eye" showed the largest diurnal amplitude-to-mean ratios. Stronger signal was restricted to and peaked in mornings, and amplitude was higher on weekdays. By contrast, "dry eyes" had a higher amplitude diurnal pattern on weekends, with stronger signal occurring over a broader evening-to-morning period and peaking in early morning. CONCLUSIONS: The frequency of web-based searches for various eye conditions can show cyclic patterns according to time of the day or week. Further studies to understand the reasons for these variations may help supplement the current clinical understanding of ophthalmologic symptom presentation and improve the timeliness of patient messaging and care interventions.

Sustained Reductions in Online Search Interest for Communicable Eye and Other Conditions During the COVID-19 Pandemic: Infodemiology Study.

Background: In a prior study at the start of the pandemic, we reported reduced numbers of Google searches for the term "conjunctivitis" in the United States in March and April 2020 compared with prior years. As one explanation, we conjectured that reduced information-seeking may have resulted from social distancing reducing contagious conjunctivitis cases. Here, after 1 year of continued implementation of social distancing, we asked if there have been persistent reductions in searches for "conjunctivitis," and similarly for other communicable disease terms, compared to control terms. Objective: The aim of this study was to determine if reduction in searches in the United States for terms related to conjunctivitis and other common communicable diseases occurred in the spring-winter season of the COVID-19 pandemic, and to compare this outcome to searches for terms representing noncommunicable conditions, COVID-19, and to seasonality. Methods: Weekly relative search frequency volume data from Google Trends for 68 search terms in English for the United States were obtained for the weeks of March 2011 through February 2021. Terms were classified a priori as 16 terms related to COVID-19, 29 terms representing communicable conditions, and 23 terms representing control noncommunicable conditions. To reduce bias, all analyses were performed while masked to term names, classifications, and locations. To test for the significance of changes during the pandemic, we detrended and compared postpandemic values to those expected based on prepandemic trends, per season, computing one- and two-sided P values. We then compared these P values between term groups using Wilcoxon rank-sum and Fisher exact tests to assess if non-COVID-19 terms representing communicable diseases were more likely to show significant reductions in searches in 2020-2021 than terms not representing such diseases. We also assessed any relationship between a term's seasonality and a reduced search trend for the term in 2020-2021 seasons. P values were subjected to false discovery rate correction prior to reporting. Data were then unmasked. Results: Terms representing conjunctivitis and other communicable conditions showed a sustained reduced search trend in the first 4 seasons of the 2020-2021 COVID-19 pandemic compared to prior years. In comparison, the search for noncommunicable condition terms was significantly less reduced (Wilcoxon and Fisher exact tests, P<.001; summer, autumn, winter). A significant correlation was also found between reduced search for a term in 2020-2021 and seasonality of that term (Theil-Sen, P<.001; summer, autumn, winter). Searches for COVID-19-related conditions were significantly elevated compared to those in prior years, and searches for influenza-related terms were significantly lower than those for prior years in winter 2020-2021 (P<.001). Conclusions: We demonstrate the low-cost and unbiased use of online search data to study how a wide range of conditions may be affected by large-scale interventions or events such as social distancing during the COVID-19 pandemic. Our findings support emerging clinical evidence implicating social distancing and the COVID-19 pandemic in the reduction of communicable disease and on ocular conditions.

Very late onset LASIK flap Acremonium fungal keratitis confirmed by metagenomic deep sequencing.

PURPOSE: To describe a unique case of LASIK flap fungal keratitis confirmed by next generation sequencing. OBSERVATIONS: A 56-year-old female presented with refractory keratitis involving her LASIK flap 21 years after surgery. Confocal was positive for filamentous structures. The patient underwent immediate flap amputation followed by topical antifungal treatment. Corneal culture was positive for Acremonium sp. Metagenomic deep sequencing confirmed Acremonium as the primary source of infection and also identified Fusarium as a likely contributor of a mixed fungal infection. Sequencing also identified hay as the likely source of the infection. Treatment resulted in eradication of the infection. The patient's final best corrected visual acuity was 20/30 with rigid contact lens overrefraction. CONCLUSIONS: Metagenomic deep sequencing is a novel diagnostic tool that is increasingly being utilized for diagnosis of refractory keratitis. This case demonstrates the diagnostic potential of deep sequencing for identifying post-LASIK keratitis and reinforces the utility of LASIK flap amputation in the setting of tectonic flap instability due to keratolysis. IMPORTANCE: This case highlights several important clinical points for treating LASIK flap keratitis and highlights the emerging role metagenomic sequencing has in the diagnosis of infectious keratitis. This is first known case using next generation sequencing to diagnose a post-LASIK infectious keratitis.

Keeping an eye on pink eye: a global conjunctivitis outbreak expert survey.

BACKGROUND: Recurrent conjunctivitis epidemics are prevalent worldwide. Aetiologies are often undetermined. METHODS: We surveyed conjunctivitis researchers about perceived trends in prevalence, incidence and aetiologies of conjunctivitis epidemics. RESULTS: Of the 155 participants, 7% endorsed globally variable and dynamic microbial aetiologies of conjunctivitis epidemics. Increased incidence of conjunctivitis epidemics over the last decade were reported by 21% of respondents. Peak seasons differed between the northern and southern hemispheres. CONCLUSIONS: There is regional equipoise regarding the increasing incidence and emerging underlying aetiologies of epidemic conjunctivitis. Further investigation of global surveillance and microbial characterization of conjunctivitis outbreaks could improve prevention and outcomes.

Preoperative Medical Testing and Falls in Medicare Beneficiaries Awaiting Cataract Surgery.

PURPOSE: Delaying cataract surgery is associated with an increased risk of falls, but whether routine preoperative testing delays cataract surgery long enough to cause clinical harm is unknown. We sought to determine whether the use of routine preoperative testing leads to harm in the form of delayed surgery and falls in Medicare beneficiaries awaiting cataract surgery. DESIGN: Retrospective, observational cohort study using 2006-2014 Medicare claims. PARTICIPANTS: Medicare beneficiaries 66+ years of age with a Current Procedural Terminology claim for ocular biometry. METHODS: We measured the mean and median number of days between biometry and cataract surgery, calculated the proportion of patients waiting ≥ 30 days or ≥ 90 days for surgery, and determined the odds of sustaining a fall within 90 days of biometry among patients of high-testing physicians (testing performed in ≥ 75% of their patients) compared with patients of low-testing physicians. We also estimated the number of days of delay attributable to high-testing physicians. MAIN OUTCOME MEASURES: Incidence of falls occurring between biometry and surgery, odds of falling within 90 days of biometry, and estimated delay associated with physician testing behavior. RESULTS: Of 248 345 beneficiaries, 16.4% were patients of high-testing physicians. More patients of high-testing physicians waited ≥ 30 days and ≥ 90 days to undergo surgery (31.4% and 8.2% vs. 25.0% and 5.5%, respectively; P < 0.0001 for both). Falls before surgery in patients of high-testing physicians increased by 43% within the 90 days after ocular biometry (1.0% vs. 0.7%; P < 0.0001). The adjusted odds ratio of falling within 90 days of biometry in patients of high-testing physicians versus low-testing physicians was 1.10 (95% confidence interval [CI], 1.03-1.19; P = 0.008). After adjusting for surgical wait time, the odds ratio decreased to 1.07 (95% CI, 1.00-1.15; P = 0.06). The delay associated with having a high-testing physician was approximately 8 days (estimate, 7.97 days; 95% CI, 6.40-9.55 days; P < 0.0001). Other factors associated with delayed surgery included patient race (non-White), Northeast region, ophthalmologist ≤ 40 years of age, and low surgical volume. CONCLUSIONS: Overuse of routine preoperative medical testing by high-testing physicians is associated with delayed surgery and increased falls in cataract patients awaiting surgery.

The Effect of Tube Location on Corneal Endothelial Cells in Patients with Ahmed Glaucoma Valve.

PURPOSE: To compare the effects of the Ahmed glaucoma valve (AGV; New World Medical, Rancho Cucamonga, CA) with sulcus versus anterior chamber (AC) tube placement on the corneal endothelial density and morphology over time. DESIGN: Nonrandomized, interventional study. PARTICIPANTS: This study included 106 eyes from 101 pseudophakic patients who had the AGV tube placed in the AC (acAGV) and 105 eyes from 94 pseudophakic patients who had the AGV tube placed in the ciliary sulcus (sAGV). METHODS: All patients underwent preoperative specular microscopy, which was repeated postoperatively in 2019. The patients' demographic information, glaucoma diagnoses, and basic ocular information were obtained on chart review. Anterior segment OCT was conducted for patients who underwent sAGV to evaluate the sulcus tube position. Gonioscopy was performed to document peripheral anterior synechiae (PAS). Linear mixed-effects models were used to compare the different ocular and endothelial measurements between the 2 groups and to identify risk factors for endothelial cell density (ECD) loss over time. MAIN OUTCOME MEASURES: Monthly change in corneal endothelial measurements, including ECD and coefficient of variation (CV), calculated as the difference between preoperative and postoperative measurements divided by the number of months from the time of surgery to postoperative specular microscopy. RESULTS: The acAGV and sAGV groups were comparable in all baseline characteristics except that the acAGV group had longer follow-up (37.6 vs. 20.1 months, respectively, P < 0.001). Mean monthly loss in central ECD was significantly more in the acAGV group (mean ± standard deviation: 29.3±29.7 cells/mm2) than in the sAGV group (15.3±20.7 cells/mm2, P < 0.0001). Mean monthly change in CV was similar between the 2 groups (P = 0.28). Multivariate analyses revealed that younger age and tube location in the AC were associated with faster central ECD loss (P = 0.02, P < 0.0001, respectively). For patients with sAGV, while PAS was associated with faster central ECD loss (P = 0.002), a more forward tube position tenting the iris was not (P > 0.05). CONCLUSIONS: Compared with anterior segment placement, ciliary sulcus tube implantation may be a preferred surgery approach to reduce endothelial cell loss in pseudophakic patients.

Single-piece S-plasty for small-incision IOL removal.

A foldable acrylic intraocular lens (IOL) with a 6.0 mm optic was removed through a 2.2 mm corneal incision as a single piece with minimal wound trauma. By viscodissection, the IOL was freed from its position in the capsular bag and moved into the anterior chamber. Vannas scissors were introduced through a 2.2 mm corneal incision. A 4.5 mm straight slit was cut at each optic-haptic junction on opposite sides, perpendicular to each haptic shaft. This pair of parallel cuts generated 3 IOL segments in an S configuration that remained connected. The IOL was grasped by 1 haptic and removed as a single piece.

Association of Postfungal Keratitis Corneal Scar Features With Visual Acuity.

Importance: Corneal opacity is a leading cause of visual impairment worldwide; however, the specific features of corneal scars, which decrease visual acuity, have not been well characterized. Objective: To investigate which features of a postfungal keratitis corneal scar contribute to decreased visual acuity after an episode of infectious keratitis and evaluate whether any corneal features may be used as outcomes for clinical trials. Design, Setting, and Participants: In this ancillary, prospective cross-sectional study, a subset of study participants treated for fungal keratitis (n = 71) as part of the Mycotic Ulcer Treatment Trial I (MUTT I) underwent best spectacle-corrected visual acuity (BSCVA) and best contact lens-corrected visual acuity examination, Scheimpflug imaging, and anterior segment optical coherence tomography at a referral hospital in India approximately 2 years after enrollment. Data were collected from December 3, 2012, to December 19, 2012, and analyses were performed from December 2, 2013, to October 2, 2019. Main Outcomes and Measures: Linear regression models were used to evaluate the importance of various corneal features for BSCVA and to assess whether these features could be used to differentiate the 2 treatment arms of the MUTT I trial. Results: Seventy-one patients (42 men [59.1%]; median age, 48 [range, 39-60] years) were examined at a median (IQR) time of 1.8 (1.4-2.2) years after enrollment. The mean (SD) logMAR BSCVA was 0.17 (0.19) (Snellen equivalent, 20/32). In multivariable linear regression models, BSCVA was most associated with irregular astigmatism (1.0 line of worse BSCVA per 1-line difference between BSCVA and contact lens visual acuity; 95% CI, 0.6-1.4) and corneal scar density (1.5 lines of worse vision per 10-unit increase in the mean central corneal density; 95% CI, 0.8-2.3). The thinnest point of the cornea was the metric that best discriminated between the natamycin- and voriconazole-treated ulcers in MUTT I, with 29.3 µm (95% CI, 7.1-51.6 µm) less thinning in natamycin-treated eyes. Conclusions and Relevance: Both irregular astigmatism and corneal scar density may be important risk factors for BSCVA in a population with relatively mild, healed fungal corneal ulcers. The thinnest point of the corneal scar may be a cornea-specific outcome that could be used to evaluate treatments for corneal ulcers.