Travis C. Porco

BACKGROUND Public perceptions of LASIK are shaped in part by social media, where people describe concerns and experiences without the constraints of a clinical questionnaire. Whether the issues discussed in social media correspond to those emphasized in formal preoperative consent and postoperative patient-reported outcome assessment is unknown. OBJECTIVE To characterize LASIK social media discourse and compare post-surgery concerns with those represented in formal informed-consent and patient-reported outcome instruments. METHODS We collected 16,529 unique posts from blogs, forums, Twitter/X, YouTube, and Instagram (April 2023–March 2025). Six large language models independently proposed and classified themes; latent class analysis combined their ratings, and two blinded human raters validated classifications. Stance and time since surgery were assessed in additional analyses. We compared post-surgery themes with the Ophthalmic Mutual Insurance Company (OMIC) LASIK consent form and the Patient-Reported Outcomes with LASIK (PROWL) questionnaire. RESULTS Two-thirds of posts (66.8%) were classified as reflecting direct patient experience followed by secondary reports of others’ experiences (18%). Stance was divided overall (28.8% supportive, 26.6% opposed, 32.5% neutral) and varied by theme. Across all posts, the most frequent themes were regret (26.7%), psychological impact (23.2%), pain (22.2%), and fear (20.3%), with positive outcomes at 15.0%. Among post-surgery posts, pain (40.4%), regret (35.5%), and positive outcomes (35.3%) were most frequent. Recorrection concerns appeared at a median of 11 years after surgery, compared with 1.5 years for pain. OMIC addressed 7 of 17 themes; its item count showed weak association with social-media frequency (ρ=0.35), with no clear association between theme frequency and presentation order (ρ=0.25; n=7). The full 110-item PROWL battery addressed 9 themes, whereas the 33-item PROWL-SS subset addressed 2. Five themes were absent from both instruments, all decisional or economic. CONCLUSIONS Social media provides a complementary view of the concerns people choose to discuss without prompting. Differences from formal consent and outcome assessment were greatest for psychological, decisional, and economic concerns, identifying areas for attention in counseling and patient-reported assessment.

The aim of this study was to determine the antiamoebic efficacy of povidone-iodine when used as an adjunct to topical biguanide therapy for Acanthamoeba keratitis. Patients with smear- or culture-positive Acanthamoeba keratitis received topical chlorhexidine 0.04%, half-hourly for 2 days, then hourly for days 3–6, then eight times per day for days 7–28. Participants were randomized in a 1:1 ratio to receive either adjunctive topical povidone-iodine at the same frequency as chlorhexidine or no additional therapy (i.e., control group). Povidone-iodine was initially administered as a 2.5% solution but was reduced to a 1% solution after several reports of intolerance. The primary outcome was Acanthamoeba growth from the culture of corneal scrapings at 1, 2, and 4 weeks after randomization, as assessed by laboratory staff masked to treatment allocation. Of 49 patients enrolled, 25 were randomized to adjunctive povidone-iodine (mean age 48 [SD 16] years; 32% female) and 24 to control (mean age 43 [15] years; 33% female). Four (16%) participants discontinued 2.5% povidone-iodine prematurely due to intolerance. The proportion of participants with a positive culture during the initial 4 weeks of antiamoebic treatment was not significantly different in the povidone-iodine group (6/18 [33%] at 4 weeks) and control group (5/16 [31%] at 4 weeks) (RR 0.90, 95%CI: 0.35–2.29; P = 0.82). When used as an adjunctive to standard biguanide therapy, povidone-iodine did not markedly accelerate the clearance of Acanthamoeba organisms from the corneal surface over the initial month of therapy. This study is registered with ClinicalTrials.gov as NCT03484507 .

Background: Although healthcare-associated transmission of Clostridioides difficile is a recognized public health concern, community-onset infections represent an important component of the overall disease burden. This paradox likely reflects the underappreciated interplay between these settings. We aimed to quantify in-hospital transmission and the hospital’s contribution to community colonization by estimating the intrinsic reproduction number (R _i ) and introducing the colonization amplification index (A _i ), defined as the ratio of colonized patients at discharge to those at admission. Given the potential contribution of external cases, we also evaluated interventions targeting asymptomatic carriers at admission to reduce disease burden. Methods and findings: We developed a compartmental model informed by data from UCSF Medical Center to capture C. difficile transmission dynamics among symptomatic and asymptomatic patients. Across simulations, the median R _i was 0.61 (Q1–Q3: 0.53, 0.71), consistently indicating limited sustained in-hospital transmission (R _i < 1). In contrast, A _i was 1.9 (Q1–Q3: 1.7, 2.1), suggesting substantial amplification of colonization during hospital stay. Amplification of colonization persisted in sensitivity analyses, with R _i exceeding 1 under boundary values of low colonization prevalence at admission, a low proportion of colonized patients progressing to symptomatic disease, and prolonged incubation periods. Redefining healthcare-associated C. difficile infection (CDI) using thresholds from ≥1-day post-admission instead of the standard threshold of >3 days post-admission increased A _i and R _i by 11% and 21%, respectively. A threshold of ≥5 days post-admission reduced these metrics by 5% and 8%, respectively. Interventions targeting asymptomatic carriers through contact precautions and/or prophylactic treatment reduced both A _i and CDI incidence, with combined interventions yielding the greatest reductions, followed by contact precautions alone. Our main limitation stems from uncertainty in some parameters describing the disease’s natural history, particularly colonization prevalence at admission, progression to symptomatic disease, and the incubation period, which may affect the precision of our estimates despite sensitivity analyses. Conclusions: Our findings indicate that in most potential scenarios, in-hospital transmission of C. difficile is limited (R _i < 1) and likely sustained by continuous importation of cases from the community. Nevertheless, hospitalization amplifies colonization (A _i > 1), which potentially contributes to community transmission. These results underscore the importance of interventions addressing asymptomatic carriers, a currently overlooked source of spread. Our study highlights the need to broaden metrics beyond R _i to capture hospitals’ contribution to the C. difficile burden. Future infection control strategies should address colonization dynamics at admission and potentially at discharge to mitigate transmission and reduce the overall burden of C. difficile.

The spread of antibiotic resistance is a major concern in the setting of azithromycin mass drug distribution for childhood mortality. We performed long-read whole-genome sequencing and phenotypic resistance analysis on Streptococcus pneumoniae isolated from the nasopharynx of Nigerien children from communities treated with either six twice-yearly azithromycin distributions or placebo. Here, we showed that co-selection for macrolide, tetracycline, and trimethoprim-sulfamethoxazole genetic resistance was detected with repeated azithromycin mass drug distributions. CLINICAL TRIALS: This study is registered with ClinicalTrials.gov as NCT02047981 .

: The community-randomized, placebo-controlled Macrolides Oraux pour Réduire les Décès avec un Oeil sur la Résistance (MORDOR) trial revealed that biannual mass azithromycin distributions to preschool children reduced child mortality in sub-Saharan Africa. Other benefits may also be possible. Surveillance of local health post records was assessed from the Niger study site from 2014 to 2016. Community-level data on all clinic visits, diagnoses, and antibiotic prescriptions recorded in the health post logbooks were retrospectively extracted for the period spanning from 12 months before baseline to 12 months after baseline. During the 12-month period after the communities were randomized in the MORDOR trial, a total of 5,229 clinic visits were documented in the azithromycin group, compared with 7,647 in the placebo group (incidence rate ratio [IRR]: 0.82; 95% CI: 0.79–0.85); 2,011 malaria diagnoses were documented in the azithromycin group, compared with 3,366 in the placebo group (IRR: 0.71; 95% CI: 0.46–1.39); and 2,218 courses of antibiotics were documented in the azithromycin group, compared with 3,297 in the placebo group (IRR: 0.81; 95% CI: 0.77–0.85). The reduction in clinic visits and clinic-prescribed antibiotics is consistent with a reduction in infectious diseases in the azithromycin-treated communities.

Large-scale, placebo-controlled, cluster-randomized trials in high-mortality settings in several African countries demonstrated a 14-18% reduction in childhood mortality following twice-annual mass drug administration (MDA) of azithromycin among children aged 1–59 months [1–3]. Azithromycin MDA also selects for antimicrobial resistance (AMR), particularly macrolide resistance in treated populations [4–6]. It is unknown whether the genetic selection of AMR from azithromycin MDA could spill over to neighboring untreated populations. If present, such geographic spillover effects could lead the trials to underestimate the risks of AMR selection from azithromycin MDA. Here, we assessed between-village geographic spillover effects of genotypic resistance to macrolides and other antibiotic classes in rectal swabs collected from 1200 children in 30 monitoring villages in Niger after two years of MDA in 594 surrounding villages. We found no evidence of geographic spillover of macrolide resistance in untreated villages, as the genetic load of AMR remained at baseline levels in placebo-treated villages regardless of surrounding azithromycin treatment intensity. Sensitivity analyses confirmed the robustness of findings to the metric used to quantify the effect of proximal azithromycin MDAs on macrolide AMR, and no geographic spillover effects were detected for AMR to other antibiotic classes. Our results suggest that azithromycin MDA-induced selection of macrolide AMR is localized to treated villages without extending to children in neighboring, untreated villages, mitigating some concerns about geographic spillover of AMR to untreated populations. This analysis illustrates the value of randomized trial designs in assessing indirect effects of large-scale public health interventions.

Background: The use of web-based search and social media can help identify epidemics, potentially earlier than clinical methods or even potentially identifying unreported outbreaks. Monitoring for eye-related epidemics, such as conjunctivitis outbreaks, can facilitate early public health intervention to reduce transmission and ocular comorbidities. However, monitoring social media content for conjunctivitis outbreaks is costly and laborious. Large language models (LLMs) could overcome these barriers by assessing the likelihood that real-world outbreaks are being described. However, public health actions for likely outbreaks could benefit more by knowing additional epidemiological characteristics, such as outbreak type, size, and severity. Objective: We aimed to assess whether and how well LLMs can classify epidemiological features from social media posts beyond conjunctivitis outbreak probability, including outbreak type, size, severity, etiology, and community setting. We used a validation framework comparing LLM classifications to those of other LLMs and human experts. Methods: We wrote code to generate synthetic conjunctivitis outbreak social media posts, embedded with specific preclassified epidemiological features to simulate various infectious eye disease outbreak and control scenarios. We used these posts to develop effective LLM prompts and test the capabilities of multiple LLMs. For top-performing LLMs, we gauged their practical utility in real-world epidemiological surveillance by comparing their assessments of Twitter/X, forum, and YouTube conjunctivitis posts. Finally, human raters also classified the posts, and we compared their classifications to those of a leading LLM for validation. Comparisons entailed correlation or sensitivity and specificity statistics. Results: We assessed 7 LLMs for effectively classifying epidemiological data from 1152 synthetic posts, 370 Twitter/X posts, 290 forum posts, and 956 YouTube posts. Despite some discrepancies, the LLMs demonstrated a reliable capacity for nuanced epidemiological analysis across various data sources and compared to humans or between LLMs. Notably, GPT-4 and Mixtral 8x22b exhibited high performance, predicting conjunctivitis outbreak characteristics such as probability (GPT-4: correlation=0.73), size (Mixtral 8x22b: correlation=0.82), and type (infectious, allergic, or environmentally caused); however, there were notable exceptions. Assessing synthetic and real-world posts for etiological factors, infectious eye disease specialist validations revealed that GPT-4 had high specificity (0.83-1.00) but variable sensitivity (0.32-0.71). Interrater reliability analyses showed that LLM-expert agreement exceeded expert-expert agreement for severity assessment (intraclass correlation coefficient=0.69 vs 0.38), while agreement varied by condition type (κ=0.37-0.94). Conclusions: This investigation into the potential of LLMs for public health infoveillance suggests effectiveness in classifying key epidemiological characteristics from social media content about conjunctivitis outbreaks. Future studies should further explore LLMs’ potential to support public health monitoring through the automated assessment and classification of potential infectious eye disease or other outbreaks. Their optimal role may be to act as a first line of documentation, alerting public health organizations for the follow-up of LLM-detected and -classified small, early outbreaks, with a focus on the most severe ones.

Objective: Azithromycin has been shown to reduce all-cause child mortality. This subgroup analysis investigates azithromycin’s mortality impact by underweight status using Azithromycine pour la Vie des Enfants au Niger: Implementation et Recherche (AVENIR) trial data. Design: The AVENIR trial randomised communities into three arms: azithromycin for children aged 1–59 months, azithromycin for infants aged 1–11 months or placebo. Weight-for-age z-score was used to categorise children into subgroups of either moderate to severe underweight or not and severe underweight or not. Setting: 2880 communities with a population of less than 2500 people in the Dosso and Tahoua regions of Niger that participated in the AVENIR trial were included. Participants: 97 572 children aged 1–59 months who had weight captured during at least one census participated. Results: Underweight subgroups had higher overall mortality compared with non-underweight subgroups. IRDs of deaths in children aged 1–11 months comparing communities receiving azithromycin to children 1–59 months of age to placebo were −6.2 deaths per 1000 person-years (95% CI −9.3 to −2.6) overall, −8.0 (95% CI −15.9 to −0.4) in the moderate to severe subgroup and −11.2 (95% CI −26.0 to −2.1) in the severe subgroup. Similar trends were noted in the azithromycin 1–11 month comparison. Malnutrition was not a statistically significant effect modifier for either comparison. Conclusions: Although analyses suggest the potential for stronger effects in more severe underweight subgroups, we were unable to demonstrate underweight status as an effect modifier. In fact, azithromycin mass drug administration to children 1–59 months old reduced mortality in all subgroups, and, especially as the number of lives saved would be the highest by treating all subgroups, our results do not support restricting eligibility for this intervention. Trial registration number: clinicaltrials.gov NCT04224987 .

Children with acute malnutrition are at high risk of morality. Mass azithromycin distribution reduces all-cause mortality among children aged 1–59 months, and effects may be greater in underweight infants. Here, we evaluate the efficacy of azithromycin for reducing all-cause mortality in children aged 6–59 months with acute malnutrition (mid-upper arm circumference, MUAC, < 12.5 cm). Communities in Nouna District, Burkina Faso were 1:1 randomized to biannual mass distribution of single dose azithromycin or placebo to all children aged 1–59 months. Mortality was assessed during each census and treatment round. MUAC measurements were collected for all children. We evaluated the effect of azithromycin on mortality in subgroups of children aged 6–59 months defined by acute malnutrition (MUAC < 12.5 cm versus MUAC ≥ 12.5 cm). In children with MUAC < 12.5 cm, mortality rates were 51% lower among those living in azithromycin communities compared to placebo (incidence rate ratio 0.49, 95% confidence interval, CI, 0.25 to 0.99; incidence rate difference -18.1 deaths per 1,000 person-years, 95% CI -37.0 to -0.01), which was greater than the reduction in mortality among children with MUAC ≥ 12.5 cm (P-value for interaction on the relative scale = 0.09; P-value for interaction of the additive scale = 0.03). Children with acute malnutrition may benefit from single dose azithromycin above and beyond those without acute malnutrition. Trial registration: ClinicalTrials.gov NCT03676764; https://clinicaltrials.gov/study/NCT03676764

Background: Manually analyzing public health–related content from social media provides valuable insights into the beliefs, attitudes, and behaviors of individuals, shedding light on trends and patterns that can inform public understanding, policy decisions, targeted interventions, and communication strategies. Unfortunately, the time and effort needed from well-trained human subject matter experts makes extensive manual social media listening unfeasible. Generative large language models (LLMs) can potentially summarize and interpret large amounts of text, but it is unclear to what extent LLMs can glean subtle health-related meanings in large sets of social media posts and reasonably report health-related themes. Objective: We aimed to assess the feasibility of using LLMs for topic model selection or inductive thematic analysis of large contents of social media posts by attempting to answer the following question: Can LLMs conduct topic model selection and inductive thematic analysis as effectively as humans did in a prior manual study, or at least reasonably, as judged by subject matter experts? Methods: We asked the same research question and used the same set of social media content for both the LLM selection of relevant topics and the LLM analysis of themes as was conducted manually in a published study about vaccine rhetoric. We used the results from that study as background for this LLM experiment by comparing the results from the prior manual human analyses with the analyses from 3 LLMs: GPT4-32K, Claude-instant-100K, and Claude-2-100K. We also assessed if multiple LLMs had equivalent ability and assessed the consistency of repeated analysis from each LLM. Results: The LLMs generally gave high rankings to the topics chosen previously by humans as most relevant. We reject a null hypothesis (P<.001, overall comparison) and conclude that these LLMs are more likely to include the human-rated top 5 content areas in their top rankings than would occur by chance. Regarding theme identification, LLMs identified several themes similar to those identified by humans, with very low hallucination rates. Variability occurred between LLMs and between test runs of an individual LLM. Despite not consistently matching the human-generated themes, subject matter experts found themes generated by the LLMs were still reasonable and relevant. Conclusions: LLMs can effectively and efficiently process large social media–based health-related data sets. LLMs can extract themes from such data that human subject matter experts deem reasonable. However, we were unable to show that the LLMs we tested can replicate the depth of analysis from human subject matter experts by consistently extracting the same themes from the same data. There is vast potential, once better validated, for automated LLM-based real-time social listening for common and rare health conditions, informing public health understanding of the public’s interests and concerns and determining the public’s ideas to address them.