Philosophical Transactions of the Royal Society B

Importance: Physician-facing AI tools are now in clinical use, yet whether different platforms fail in similar or fundamentally different ways in high-stakes settings like critical care is unknown. Objective: To evaluate two physician-facing AI platforms, ChatGPT for Clinicians and OpenEvidence, for distinct vulnerabilities under structured stress testing. Design, Setting, and Participants: An observational study conducted using 60 simulated critical care vignettes developed and adjudicated by four attending critical care physicians. Data were collected in the last week of April 2026, via the public website interfaces of each platform. Interventions/Exposures: A 2x2x2x2 factorial design across four stressors - anchoring, cognitive load, social conformity pressure, and a clinically incorrect directive - yielded 16 prompt subsets per vignette and 960 prompts per platform. A separate multi-turn adversarial prompting paradigm administered three sequential "You are incorrect" challenges to baseline vignettes. All prompts had a universal output length constraint of fewer than 30 words. Main Outcomes and Measures: Critical elements capture (percentage of gold-standard critical elements present in responses), susceptibility to clinically incorrect directive, and sycophancy (reversal of an initial correct recommendation under iterative adversarial challenge). Results: Across 1916 responses to 1920 prompts, ChatGPT for Clinicians captured more gold-standard critical elements than OpenEvidence (81.4% {+/-} 18.1% vs 61.0% {+/-} 23.5%; adjusted difference, 20.3 percentage points; 95% CI, 18.3 to 22.4; P < .001) and was less susceptible to clinically incorrect directives (1.7% vs 8.0%; adjusted odds ratio, 0.07; 95% CI, 0.02-0.21; P < .001). Anchoring and social conformity pressure were associated with reduced critical element capture across both platforms, while cumulative stressor burden reduced critical element capture only on OpenEvidence. Conversely, ChatGPT for Clinicians reversed correct recommendations more readily under adversarial prompting (hazard ratio, 2.61; 95% CI, 1.10 - 6.19; P = .03). Conclusion and Relevance: The two physician-facing clinical AI platforms evaluated demonstrated non-overlapping vulnerabilities, with neither platform uniformly superior. These findings argue against single-axis ranking of clinical AI systems and support multidimensional safety evaluation encompassing completeness of reasoning, resistance to incorrect directives, and stability under adversarial challenge.

Large epidemics of invasive meningococcal disease are rare in temperate regions. Here, we analyse administrative data on the largely forgotten epidemic of bacterial meningococcal meningitis that occurred in Glasgow in 1907, probably the largest on record in the UK. The epidemic, predominantly confined to the city, killed around 1,000 people, had a case fatality rate of nearly 70%, and hit infants and young children the hardest. We show the rapid rise and fall in cases and the spatial distribution of incidence and mortality rates within the city. We find that within-household overcrowding was a key driver of incidence whereas between-household geographic proximity was not. We also find that the spatial distribution of disease risk during the epidemic persisted in the post-epidemic period and during a later outbreak. The findings suggest that interventions should prioritise populations in areas that have experienced higher incidence rates to mitigate the risk of future outbreaks.

Background Bacterial fitness is shaped by interactions between genome variation and environmental context, yet how these interactions determine its predictability and heritability remains unclear. In the clinically important pathogens of Klebsiella pneumoniae, a leading cause of hospital-acquired infections, this question is particularly pressing. Despite extensive genomic characterization, we still lack a systematic understanding of how genome-wide variation translates into fitness across diverse environments in K. pneumoniae. Methods We filled this gap by profiling a systematic collection of 1,462 clinical K. pneumoniae isolates across 214 diverse environmental and pharmacological stress conditions using high-throughput chemical genomics. Fitness was quantified from colony growth and integrated with whole-genome sequencing data. Genome-wide association analyses identified genetic determinants of fitness, and machine learning models incorporating genomic features were used to predict fitness.Results Fitness exhibited a strongly environment-dependent genetic architecture, with modest but significant concordance between genetic background and phenotypic variation. Under antibiotic and stress-combination conditions, fitness was driven by discrete, high-effect determinants, including known resistance genes, resulting in stronger signals and improved predictability. In contrast, non-antibiotic environments showed more polygenic and distributed architectures with weaker associations. Genome-wide analyses identified both established and previously uncharacterized genes linked with fitness across conditions. Resistance and virulence determinants exhibited clear context-dependent trade-offs, conferring fitness advantages under selection but imposing costs in non-selective environments. Consistent with this, plasmid carriage showed environment- and genotype-dependent fitness effects, with benefits under antibiotic pressure and measurable costs otherwise. Genomic variant-based models for fitness prediction achieved moderate performance (Mean Spearman correlation ({rho}) = 0.36 (95% CI: 0.18-0.67) for predicted versus observed values in unseen data) across conditions, with improved accuracy under strong antibiotic selective pressures, and produced well-calibrated prediction intervals with high coverage. Despite strong population structure effect on predictions, models captured predictive gene and SNP biomarkers for fitness. Conclusion These findings highlight that bacterial fitness is an emergent property of genome-environment interactions rather than a fixed attribute of genotype. This work establishes a unified high-dimensional genotype-phenotype framework linking genomic variation to fitness across diverse conditions in a major pathogen, with broader implications for other pathogenic bacterial species.

Background: Wearing facemasks and practising social distancing slow the spread of respiratory pathogens. However, in the event of a new pandemic emerging, the willingness of populations to voluntarily adopt these behaviours is unclear. Methods: A discrete choice experiment was conducted among 2,006 UK-based adults. Participants were presented with hypothetical scenarios describing the emergence of a respiratory virus pandemic and were asked to choose when they would wear facemasks and practise social distancing. A mixed multinomial logit model was used to jointly estimate how disease severity and prevalence, uncertainty in these quantities, and individual-level characteristics influence behavioural choices. Findings: Participants were averse to facemasks and social distancing in the absence of pandemic risk. For each ten-unit increase in severity (10 additional hospitalisations/1,000 infections), the odds of always wearing a facemask outside the home increased by 15.9% (95%CI: 14.3%, 17.5%), relative to rarely/never, and the odds of avoiding all people as much as possible increased by 16.4% (14.6%, 18.2%), relative to not avoiding anyone. Greater disease prevalence, uncertainty in disease severity or disease prevalence, a university education, prior COVID-19 vaccination and non-white ethnicity were also associated with choosing to always wear facemasks and avoid all people as much as possible. The probability of participants choosing to rarely/never wear facemasks varied from 13.4% (11.9%, 14.9%) in the lowest-risk scenario to 1.4% (1.2%, 1.7%) in the highest-risk scenario. Interpretation: Perceived risks of disease and associated uncertainty drive intention of UK adults to adapt their behaviour in a future pandemic.

A closed-loop quality system deployed across thirteen US hospital sites resolved physician complaints with zero regressions on 42 tracked cases across 1,089 optimization iterations, while a deterministic assembly-agent replacement cut H+P trace latency from 19.6 s to 10.8 s (-8.8 s, 95% CI [-10.5, -7.1] s; n = 100 pre, n = 100 post). We report four observations and an architectural follow-through. First, the same binary-check instrument produces opposite outcomes depending on the question asked: "maximize this score" produces structurally-correct notes that physicians reject (Spearman rho = -0.077, 95% CI [-0.40, 0.26], n = 36); "did this specific fabrication stop?" produces rater-invariant deployment decisions. Second, in our pipeline, assembly-stage agents did not respond to prompt optimization the way reasoning agents did: four consecutive optimization attempts produced 18-28 point regressions. Third, physician preference is rater-fragile at typical clinical-AI calibration sample sizes (Cohen's kappa = 0.028 between two board-certified physicians, 95% CI [-0.30, 0.36] on n = 35 overlapping pairs). Fourth, the architectural punchline: six weeks after the prediction, the LLM call at the chart-assembly step was replaced with a deterministic renderer (sub-500-character template plus sandboxed scripting), lifting the defect-free rate on a 51-case holdout from 49% to 84%. We introduce a Pareto-with-absolute-floors acceptance rule (multi-axis commit with severity-class categorical vetoes) as a methodological contribution distinct from scalar-reward acceptance in standard prompt-optimization frameworks. Cross-iteration rejection memory prevents the loop from re-proposing edits already rejected three or more times. A reproducibility bundle (anonymized ablation per-case counts, bootstrap-CI data, analysis scripts) is released under CC BY 4.0 at github.com/sayvant/SQS-Auditor-paper-data.

Background: Hospital incident risk scoring has long relied on two- or three-dimensional frameworks (Severity Assessment Codes or Risk Priority Numbers),even though root cause analysis standards recognize that clinical risk is multi-factorial. The obstacle has been mainly cognitive: human reviewers cannotreliably score many dimensions across high incident volumes, so richer assessmenthas not been operationalized at scale.Objective: To extend the traditional three-dimensional FMEA to an eight-dimensional patient-safety risk feature framework, to establish a multi-modellarge language model (LLM) extraction pipeline that scores these dimensionsautomatically, and to demonstrate a variance-aware integer optimization (mean-variance integer programming, MV-IP) that provides a reproducible tie-breakingrule for incident prioritization under extraction uncertainty, rather than improvedrisk coverage.Methods: An 8-dimensional framework covering harm severity, potential harm,frequency, detectability, systemic impact, vulnerable populations, regulatoryrelevance, and economic impact was applied to 213 synthetic and 196 realcurated incident narratives. Three independent LLMs (GPT-5.4, Gemini 3.1 Pro, Grok-4.1 Fast) from different provider families extracted structured risk scores.Inter-model consistency was assessed via ICC(A,1). Among coverage-equivalentselections, MV-IP minimized inter-model variance to give a reproducible prioriti-zation rule. An English-language sensitivity analysis was conducted on 31 AHRQPSNet WebM&M cases.Results: On real cases, seven of eight dimensions reached Fair or betterinter-model reliability (ICC(A,1) 0.53 to 0.83); D5 (Systemic Impact) was theexception at Poor reliability (0.275), driven by little between-case variation ratherthan by wide model disagreement. Reliability was not uniform: two dimensionswere Excellent (D1 actual harm 0.834, D8 economic impact 0.782), two Good,and three only Fair, so some dimensions are more readily extractable than others.The same anchors gave broadly similar results on English-language narratives.When deterministic top-K selection returned several equal-coverage solutions(11 on real cases, total inter-model variance 0.205 to 1.274), MV-IP selected theminimum-disagreement set, replacing ad hoc tie-breaking with an explicit rulewithout improving coverage. Bootstrap resampling found 74% to 90% of per-casevariance estimates stable despite the three-model panel.Conclusions: The eight-dimensional framework operationalizes patient-safetyrisk features that quality teams have considered only implicitly, and three inde-pendent LLM families produced reproducible scores on most dimensions ofcurated narratives. Inter-model agreement, however, measures reproducibilityrather than clinical correctness, and high agreement does not by itself establishthat a score is right; the dimensions that are reliably extractable today (notablyD6 and D8) differ from those that are not yet (D5, and to a lesser degree D4 andD7), which has direct implications for incident-reporting form design. MV-IP con-tributes a reproducible, variance-aware tie-breaking rule rather than improvedcoverage. Validation against expert-prioritized RCA lists and deployment on rawinstitutional incident reports remain the next steps toward clinical use.

Vocal biomarkers, encompassing voice and speech, have largely been developed for individual conditions in isolation, limiting their generalizability across diseases and recording settings. To address this, we introduce VoiceFM, a contrastive model that learns general-purpose clinical voice representations by aligning audio embeddings with rich clinical metadata. Using the Bridge2AI-Voice dataset (984 primarily English-speaking adult participants, 846 used for training and 138 held out as a temporally separated validation cohort, 40,056 recordings totaling 176 hours across 5 academic medical centers), VoiceFM pairs a fine-tuned Whisper large-v2 encoder with a tabular transformer over 44 clinical features via symmetric InfoNCE loss. Linear probes on frozen VoiceFM embeddings achieve mean AUROC 0.952 +/- 0.005 across five evaluation tasks (control vs disease screening plus four disease categories), significantly outperforming Frozen Whisper (0.926 +/- 0.013, p = 0.013), Frozen HuBERT (0.885 +/- 0.017, p = 0.0009), and the contrastively trained VoiceFM-HuBERT (0.938 +/- 0.006, p = 0.012). On the 138-participant held-out cohort, VoiceFM-Whisper achieves AUROCs of 0.99 for Alzheimer's/dementia/MCI and 0.89 for airway stenosis, demonstrating that the learned representations generalize to participants the model has never seen. VoiceFM representations transfer to three external datasets without retraining and improve few-shot classification. Recording task attribution identifies a small set of speech tasks that match or exceed the full battery's performance, suggesting shorter screening protocols are feasible. Trained predominantly on English audio, VoiceFM transfers without fine-tuning to Spanish-language Parkinson's disease (PD) detection (NeuroVoz, 107 participants, AUROC 0.93 +/- 0.02), with the signal dominated by articulatory rather than phonatory features. A fine-tuned classifier achieves participant-level AUROC 0.87 (sustained 0.85, countdown 0.80) on the mPower smartphone study (585 held-out participants). Together, these results show that contrastive alignment between voice and rich clinical metadata can serve as the basis for a clinical voice foundation model, producing a single set of transferable representations that generalize across diseases, languages, recording conditions, and patients enrolled after model freeze.

Background: Traditional diagnostic models lack explainability, while multimodal language models prone to hallucination remain unsafe for medical education. An interactive, risk-free artificial intelligence framework is required to serve as a reliable clinical mentor for radiology trainees. Methods: We propose a multi-agent architecture decoupling deterministic image analysis from generative consultation. Specialized computer vision models perform anatomical localization and pathological segmentation. These quantitative outputs are synthesized into a structured payload, which grounds a locally hosted large language model (LLaVA 7B) using strict prompt guardrails and prerequisite protocols. Results: The system effectively eliminates visual hallucinations by intercepting unanchored queries. The artificial intelligence tutor successfully contextualizes spatial anomalies and baseline metrics, generating accurate conversational explanations and formally structured radiology reports while strictly enforcing medical safety disclaimers. Discussion and Conclusion: By anchoring language generation exclusively to verified algorithmic realities, this framework transforms opaque diagnostic models into safe, interactive educational simulators. This establishes a highly reliable paradigm for integrating explainable artificial intelligence into medical training.

Objective. Biomedical knowledge graphs (KGs) such as PrimeKG, Hetionet, UMLS, and PharmGKB are increasingly used as the substrate for downstream machine-learning, retrieval-augmented generation, drug-repurposing, and electronic health record (EHR) augmentation pipelines. The dominant assumption in published work is that integrating two or more such KGs is a tractable engineering step solved by identifier (ID) matching. This paper interrogates that assumption empirically. We quantify how much concept overlap survives realistic alignment, and we characterize the new failure modes introduced by the methods that practitioners reach for when ID matching is insufficient. Materials and Methods. We compared four widely used biomedical KGs (PrimeKG, Hetionet v1.0, the full UMLS Metathesaurus, and PharmGKB) across eleven node types using a tiered alignment pipeline: (1) direct ID matching for nodes sharing a primary vocabulary; (2) cross-ontology bridging using standard mappings (e.g., MONDO-DOID, HPO-UMLS, HPO-UMLS-MeSH for side effects, NCBI Gene-HGNC-UMLS, UBERON-FMA/SNOMEDCT_US/NCI/MeSH for anatomy); (3) ClinicalBERT cosine-similarity grouping at threshold >= 0.98 for over-segmented disease nodes, with a deterministic suffix-stripping canonicalizer; (4) exact name matching for ontology-poor types (anatomy, REACTOME pathways); and (5) embedding-based fuzzy matching with UMLS lookup (SapBERT and ClinicalBERT) for free-text microbiome concepts. We applied the pipeline to a 698-concept gut-microbiome benchmark spanning taxa, pathways, and disease labels, validated grouping decisions against the curated SSSOM mappings released by the MONDO project, and audited the ClinicalBERT consolidation against five clinical-genetics case studies drawn from the literature. Results. Per-type pairwise coverage was strikingly asymmetric. Genes/proteins and the three Gene Ontology categories aligned cleanly across PrimeKG and Hetionet (mutual coverage 94-99%), but disease overlap was sparse: only 0.7% of PrimeKG individual disease nodes mapped to Hetionet, rising to 2.0% after MONDO grouping (versus 78.7% and 18.4% from the Hetionet side). PrimeKG-to-UMLS coverage spanned 100% (effect/phenotype via HPO) down to 20.8% (REACTOME pathways), with drugs at 73.7% and anatomy at 58.8%. PrimeKG-to-PharmGKB drug coverage required up to two bridging hops (DrugBank -> UMLS -> RxNorm/ATC/MeSH). Bigger was not uniformly more complete: on a 698-concept microbiome drug benchmark, Hetionet missed 0 concepts while PrimeKG missed 16. ClinicalBERT-based grouping consolidated 22,205 raw MONDO disease nodes into 17,080 groups but introduced three reproducible failure modes documented in case studies: (i) peer over-merging: for example, all 22 osteogenesis imperfecta subtypes collapsed into a single node despite distinct severity classes; (ii) parent-child collapse: e.g. acute myeloid leukemia merged with myeloid leukemia, erasing the acute/chronic distinction that drives clinical management; and (iii) lexical false positives: neurofibromatosis and schwannomatosis grouped together despite cellular-pathology differences. Discussion. Identifier matching alone is a weak baseline for biomedical KG integration. Cross-ontology bridges and embedding-based consolidation expand coverage but do so at the cost of clinically meaningful resolution, and the resulting failures are systematic rather than random. Reporting only aggregate coverage statistics obscures these losses, which propagate silently into downstream tasks. Conclusion. We provide reusable per-type coverage tables, a taxonomy of three integration failure modes, and concrete recommendations for downstream studies that depend on a unified biomedical KG. We argue that future KG integration work should report per-type coverage and per-cluster confidence rather than aggregate match rates.

While SARS-CoV-2 and influenza continue to place a significant burden on population health, within-household differences in decisions towards vaccination and seeking care across these two pathogens, and across sociodemographic groups, remain largely unexplored. By conducting a household-level survey in Illinois, we found that many individuals made inconsistent decisions about vaccination: among all adults, 29% were vaccinated for only one of COVID-19 or influenza, and among those with children in the home, 39% lived with a child whose influenza or COVID-19 vaccination status differed from their own. A higher proportion of adults were vaccinated against COVID-19 compared to influenza, while the opposite was true for those younger than 18 years old. These differences hold even when accounting for disparities in coverage by age, race/ethnicity, political affiliation, and socioeconomic status. While vaccinated individuals consistently reported wanting to protect themselves or others, those who declined vaccination reported highly heterogeneous reasons ranging from resource constraints to distrust or misconceptions about vaccination. These differences are even more pronounced for COVID-19, with larger partisan gaps and higher refusal driven by safety concerns, lack of trust, or religious reasons than those who decide not to get the influenza vaccine. In contrast to vaccination, the decision to seek medical care when sick showed opposite sociodemographic trends, that are likely attributable to illness severity. Our findings highlight that closing gaps in COVID-19 and influenza vaccination coverage will require an integrative strategy that accounts for diverse motivations, fears, and barriers to access, while addressing social inequalities common to both diseases.

Background Artificial intelligence-enhanced electrocardiography (AI-ECG) enables scalable, low-cost cardiac dysfunction screening, but existing models are annotation-intensive and predominantly adult-derived, leaving paediatric generalizability uncertain. Paediatric cohorts exhibit highly variable cardiac morphology and function compared to adults, which may be useful for learning generalizable AI-ECG models. Methods We pretrained ECG-Fyler on a predominantly paediatric, all-age cohort at Boston Children's Hospital (1992-2023), annotated with a cardiology-specific coding system (Fyler codes), and evaluated it on assessments from echocardiography (echo) and cardiac magnetic resonance (CMR) studies. We validated on an external adult cohort from Columbia University Irving Medical Center. Performance was benchmarked against several AI-ECG foundation models by AUROC across age groups, lesion types, and limited-data scenarios. Findings The pretraining cohort comprised 782,138 ECGs from 255,271 patients (median age: 10.9 years, IQR: [2.8-16.8]). Internal evaluation included 178,495 ECG-echo pairs (median age: 10.9 [3.7-17.0]) and 8,584 ECG-CMR pairs (median age: 20.7 [15.6-29.6]). External validation included 82,543 ECG-echo pairs from adults (median age: 64.0 [52.0-74.0]). ECG-Fyler improved AUROC across biventricular dysfunction and dilation tasks, with the largest gains in low-data settings. In internal validation, ECG-Fyler detected low left ventricular ejection fraction (LVEF [&le;] 40%) from only 100 fine-tuning samples (AUROC: 0.80, 95% CI: [0.78-0.80]), outperforming other models (AUROC < 0.65) and improving with additional fine-tuning (AUROC: 0.94 [0.93-0.94]). Similar improvements were observed for CMR-derived LVEF, RVEF, and ventricular dilation. In external validation on adults, ECG-Fyler exhibited an AUROC of 0.83 (CI: [0.82-0.85]) for LVEF [&le;] 40%. After fine-tuning on less than 10% of external data, LVEF [&le;] 45% performance (AUROC: 0.87 [0.86-0.88]) outperformed a fully trained, site-specific prior model (AUROC: 0.85 [0.84-0.87]). Interpretation Pretraining on richly annotated, paediatric-dominant ECGs yields models that transfer efficiently across institutions and ages, supporting AI-ECG screening and triage when labels or imaging access are limited. Funding National Institutes of Health (R01LM012973); Kostin Innovation Fund, Boston Children's Hospital

Artificial intelligence (AI) tools have been rapidly adopted by medical researchers, yet whether early career researchers in low and middle income countries possess the awareness and habits needed to use these tools safely remains poorly documented. This study characterized AI adoption patterns, hallucination awareness, and verification and disclosure practices among early career medical researchers in Pakistan. A cross sectional anonymous online survey was conducted among medical students, house officers, residents, physicians, and faculty involved in research or academic work across Pakistan (May 2026). Descriptive statistics and chi square tests were applied to 373 eligible responses. AI use was near universal (99.7%), with 60.3% using AI tools daily. The most commonly reported tool in this sample was Claude (40.5%), followed by ChatGPT (29.2%) and Perplexity (26.0%), though this ranking likely reflects sampling characteristics. Despite high adoption, 59.2% typically did not verify AI outputs before use, and 40.2% had never heard that AI can generate fabricated scientific references. In behavioral vignettes, 36.5% assumed convincing AI generated references were authentic, and 54.2% would continue using remaining AI content after discovering one fabricated reference. Formal research training was strongly associated with consistent disclosure (51.7% vs. 17.1%; chi square=48.43, p less than 0.001). Role, daily use frequency, and research training were not significantly associated with verification behavior. Early career medical researchers in Pakistan demonstrate high AI adoption alongside incomplete hallucination awareness and infrequent verification, a pattern that may carry implications for research integrity. Formal training was the only factor significantly associated with consistent disclosure. Integration of AI literacy into medical curricula and institutional governance frameworks merits consideration.

Introduction -- Emergency Medical Dispatch Systems (EMDS) can reduce delays in accessing emergency care by providing structured communication, triage, and coordination. However, such systems remain absent or underdeveloped in most low- or middle-income countries (LMICs). This study aimed to establish international consensus on essential EMDS components to inform global guidance. Methods -- We convened a multidisciplinary expert group to draft a preliminary list of essential components for three EMDS levels reflecting resource availability and system maturity. We then conducted a three-round Delphi with international experts to reach consensus on core EMDS components. Components which had [&ge;]75% agreement were included, those with [&ge;]75% disagreement were excluded. Components not achieving consensus by Round 3 were removed. Results were analysed overall and stratified by respondents' country income level. A subsequent online expert meeting resolved inconsistencies and finalised the component list. Results -- The expert group generated 111 components for each of three EMDS levels (Foundational, Emerging, and Established) spanning 11 operational domains. Of the 68 experts invited to the Delphi, 43 participated in Round 1 and 30 in Round 3. Across all Delphi rounds, 289 components reached consensus for inclusion. The consensus resulted in a final list of 227 components (63 Foundational, 84 Emerging, and 80 Established). Consensus agreement clustered around core EMDS domains including communication, structured call-taking and prioritisation, advice-giving, resource dispatch and tracking, and foundational governance and data functions, whereas items showing either non-consensus or consensus disagreement were typically technology-dependent or context-specific. Conclusions -- This international consensus offers guidance for EMDS development across diverse resource settings and provides a scalable roadmap to strengthen emergency care systems.

Background: Preparing tumor board patient summaries is time intensive. Large-language-model based systems may automate summarization but require real-world evaluation prior to clinical use. We performed an exploratory retrospective evaluation of the Microsoft Healthcare Agent Orchestrator (HAO), deployed in a Mayo Clinic controlled staged environment, to generate tumor board-style patient summaries from retrospective Electronic Health Record (EHR) notes. Methods: HAO generated summaries for breast, hepatobiliary, and neuro-oncology tumor board cases using up to the most recent 1,000 clinical notes. Clinician reviewers evaluated outputs via REDCap surveys across perceived factuality, completeness, clarity/conciseness, temporal cohesion, comparative performance, safety, and clinical utility (0-4 Likert scale). Reviewers were permitted to query the HAO chat interface to address missing details. Automated factuality was assessed using TBFact (bidirectional entailment), reporting precision and recall against available reference summaries. Results: Among 57 survey responses from 5 different physicians, mean scores exceeded 2.8 across domains, with medians of 3 for most axes. In an exploratory comparison, oncology fellows required less time to review HAO-generated summaries than to manually generate patient summaries (mean difference 13.57 minutes per patient, p<0.001), although this difference may be influenced by prior familiarity with the same cases; 96% of survey responses indicated that HAO would save time. TBFact evaluations showed higher recall than precision across domains, consistent with broad capture of reference content alongside additional content that was not present in gold-standard summaries. Attribution was viewed favorably but showed issues with primary-source specificity and link reliability. Conclusions: In a controlled Mayo environment, HAO demonstrated moderate performance and was associated with reduced review time for tumor board preparation. These findings are promising but preliminary and do not establish clinical safety, noninferiority to manual review, or readiness for routine clinical use. Limitations, including verbosity, specialty-specific content gaps, and inconsistent attribution, highlight the need for iterative refinement and further evaluation.

Hybrid mechanistic-statistical models offer interpretability and adaptability for short-term seasonal epidemic forecasting, but it remains unclear whether their accuracy depends more on increased biological complexity or on the assimilation of richer data. Using eight retrospective influenza seasons in North Carolina, we evaluate whether training on historical data and assimilating auxiliary emergency department (ED) visit data improves four-week-ahead hospital admission forecasts more than adding biological complexity (multi-subtype structure and cross-season immunity). Hierarchical Bayesian training on historical data improves accuracy by 22.4 % (95 % CI: 16.4-28.1 %), and inclusion of ED visit data yields a further 5.3 % (95 % CI: 3.0-7.6 %) improvement, whereas added biological complexity produces diminishing or null gains. We further observe a substitution effect in which ED visit data partially compensates for omitted biological structure. We deployed a simplified model variant in the 2025-2026 CDC FluSight Challenge and ranked among the top ensemble performers, supporting the robustness of Bayesian hierarchical training in real time. Together, these findings indicate that short-term forecast accuracy is driven more by historical learning and assimilating auxiliary signals than by biological fidelity, with implications for how forecasting systems should balance mechanistic complexity.

Background: Generative AI tools can support data-intensive research by writing code, drafting prose, searching analytical possibilities, and stress-testing claims. They can also produce false citations, drift between statistical specifications, and lose continuity across long investigations. This paper describes a practical workflow for using AI systems in empirical research while keeping discovery, verification, and accountability inspectable. Methods: We developed and applied a three-phase human-AI workflow to a case study of 14 elite Ethiopian distance runners. The dataset contained 22,605 GPS-segments collected across 97 consecutive days in late 2025, supplemented by venue and athlete metadata collected in the field. Phase 1 used an autonomous data-exploration tool to pre-filter the hypothesis space across five seeded research questions. Phase 2 used an AI system under direct human guidance to construct candidate findings into numerical claims, verification scripts, and draft text. Phase 3 used an independent AI system in an adversarial role to stress-test methods, statistics, prose, figures, and citations. The workflow was informed by Pearl's distinction between association, intervention, and counterfactual reasoning, with human judgement retained for research direction, interpretation, and final claims. Results: The workflow produced three empirical analyses and a documented correction process. The analyses estimated an altitude-to-sea-level pace correction of +0.10 min/km per 1,000 m at matched heart rate, showed why pooled altitude-surface regression was not identifiable within this venue system, documented method-dependence in heart-rate-based intensity classification, characterised within-venue route variation as a 64/36 path-fixed-to-trail-variable split with the Sululta label resolving into two functionally distinct sub-venues, and reframed the cohort's training through a 3x3x3 prescription lattice grounded in Ethiopian coaching practice. The adversarial phase identified several hallucinated citations, a terminology error between HC1 and cluster-robust standard errors, and several inconsistencies between prose, figures, and computed results. Verification scripts re-derived nearly all numerical claims from the cleaned lap-level data. Conclusions: The case study shows how researchers can organise AI-assisted empirical work so that candidate discovery, claim construction, independent stress-testing, and final accountability remain separated. The workflow did not remove the need for domain expertise or human judgement. Its value was in making the route from candidate finding to manuscript claim explicit, reproducible, and open to challenge. Trial registration: Not applicable.

Andes orthohantavirus (ANDV), the primary etiological agent of hantavirus pulmonary syndrome (HPS) in South America, is uniquely capable of limited human-to-human transmission, posing a significant challenge for outbreak control. Recent events, including the 2018-2019 Epuyen outbreak and the 2026 MV Hondius incident, underscore the need for rapid, lineage-specific molecular diagnostics. In this study, we present an artificial intelligence (AI)-driven framework for the design of diagnostic primers targeting the S genomic segment of the Epuyen lineage. Using an evolutionary algorithm integrated with thermodynamic evaluation via Primer3Plus, candidate primers were optimized to maximize classification accuracy while satisfying stringent biochemical constraints. The resulting primer set enables amplification of lineage-specific regions suitable for molecular characterization and surveillance. In silico validation demonstrates that the proposed primers achieve perfect discrimination between 2026 outbreak sequences and other ANDV variants. Furthermore, in silico comparison with standard protocol-based primers reveals substantially reduced sensitivity and specificity in the latter, highlighting the limitations of static diagnostic designs when applied to evolving viral populations. Overall, this work demonstrates that AI-assisted primer design provides a robust and adaptable strategy to improve viral detection, enhance outbreak tracking, and support timely public health interventions. Integrating computational optimization into diagnostic development is essential for strengthening preparedness against emerging zoonotic threats.

OBJECTIVES To quantify socioeconomic inequalities in antibiotic prescribing for common infections in primary care, and assess whether these inequalities arise from differences in consultation frequency, prescribing behaviour, or variation in vaccination uptake, smoking, and body mass index. DESIGN Population based cohort study. SETTING Primary care data from Clinical Practice Research Datalink, England. PARTICIPANTS 17,195,399 children and adults estimated to have been registered with a general practice in 2019. MAIN OUTCOME MEASURES Antibiotic prescribing rates (prescriptions per person-year), consultation rates (consultations per person-year), and probability of receiving an antibiotic prescription following consultation. RESULTS Higher deprivation was associated with higher antibiotic prescribing rates for most respiratory tract indications. In children, prescribing rates were 44.8% (95% confidence interval [CI] 41.9% to 47.7%) higher for upper respiratory tract infections and 47.6% (95% CI 44.2% to 51.3%) higher for lower respiratory tract infections in the most versus least deprived twentile. In adults, prescribing rates for lower respiratory tract infections were 22.7% (95% CI 21.4% to 24.1%) higher in the most deprived twentile. Prescribing rates for other indications showed weak, U-shaped, or negative associations with deprivation. Prescribing inequalities were primarily driven by inequalities in consultation rates rather than probability of receiving antibiotics once consulted. Lower influenza vaccination uptake partly accounted for higher consultation rates for respiratory infections among more deprived children, while smoking prevalence contributed to inequalities among adults. CONCLUSIONS Socioeconomic inequalities in antibiotic prescribing vary by indication type and are largely explained by consultation frequency. Reducing inequalities may require interventions that decrease the need to consult, e.g. improving influenza vaccination coverage in children and reducing smoking among adults, rather than focussing solely on prescribing behaviour.

Subjective auditory-perceptual evaluation and uninterpretable deep learning models limit the clinical assessment of voice disorders. This study proposes a two-phase zero-shot framework to evaluate voice pathology. First, an Audio Spectrogram Transformer is fine-tuned on the Perceptual Voice Quality Database to generate an acoustic latent space. Second, Orthogonal Procrustes analysis maps these acoustic embeddings directly onto the semantic space of a pre-trained Sentence Transformer. The geometric alignment produced continuous semantic axes that outperformed a supervised machine learning baseline in regressing clinician-rated GRBAS (Grade, Roughness, Breathiness, Asthenia, and Strain) severity scales. Furthermore, these axes correlate with traditional acoustic measures, including Harmonics-to-Noise Ratio and local jitter, while remaining robust when applied to aperiodic signals by not requiring fundamental frequency extraction. Most importantly, the model achieved zero-shot semantic expansion, successfully evaluating voices using an untrained, natural clinical vocabulary beyond the GRBAS scale. External validation on the Voice ICarus Database confirmed cross-corpus stability and demonstrated the capacity for zero-shot differential phenotyping of specific etiologies, such as hypokinetic dysphonia and reflux laryngitis. By bridging acoustic and semantic latent spaces, this framework offers an objective, continuous, and transparent metric for evaluating voice quality using voice descriptive vocabulary.

Pharmacogenomics (PGx) offers a pathway towards personalised medicine, which relies on health consumer involvement in making informed decisions. As consumers increasingly seek health information online, high-quality digital resources are essential to support informed consent and shared decision making. The complexity of PGx and widespread limitations in health literacy raise concerns about whether existing consumer-facing online PGx resources are understandable and sufficiently comprehensive. This study evaluates the readability, visual design, and informational quality of publicly available online written PGx health information. Twenty-three webpages met inclusion criteria. The mean readability corresponded to approximately 15 years of formal education (university level), substantially exceeding the Australian Government's recommended Year 7 reading level for public health materials. Informational quality was generally low, with most webpages being rated as poor or very poor. In contrast, visual design quality was relatively strong, with webpages achieving on average around three-quarters of the criteria. Although the visual presentation of PGx webpages is generally professional, their high reading difficulty and limited discussion of treatment choices and uncertainties reduce their usefulness for health consumer education. Improving readability, clearly communicating risks and limitations, and incorporating decision-support features may enhance the ability of online resources to support informed consent and shared decision making.