Algorithmic Identification of Treatment-Emergent Adverse Events From Clinical Notes Using Large Language Models: A Pilot Study in Inflammatory Bowel Disease.

Outpatient clinical notes are a rich source of information regarding drug safety. However, data in these notes are currently underutilized for pharmacovigilance due to methodological limitations in text mining. Large language models (LLMs) like Bidirectional Encoder Representations from Transformers (BERT) have shown progress in a range of natural language processing tasks but have not yet been evaluated on adverse event (AE) detection. We adapted a new clinical LLM, University of California - San Francisco (UCSF)-BERT, to identify serious AEs (SAEs) occurring after treatment with a non-steroid immunosuppressant for inflammatory bowel disease (IBD). We compared this model to other language models that have previously been applied to AE detection. We annotated 928 outpatient IBD notes corresponding to 928 individual patients with IBD for all SAE-associated hospitalizations occurring after treatment with a non-steroid immunosuppressant. These notes contained 703 SAEs in total, the most common of which was failure of intended efficacy. Out of eight candidate models, UCSF-BERT achieved the highest numerical performance on identifying drug-SAE pairs from this corpus (accuracy 88-92%, macro F1 61-68%), with 5-10% greater accuracy than previously published models. UCSF-BERT was significantly superior at identifying hospitalization events emergent to medication use (P < 0.01). LLMs like UCSF-BERT achieve numerically superior accuracy on the challenging task of SAE detection from clinical notes compared with prior methods. Future work is needed to adapt this methodology to improve model performance and evaluation using multicenter data and newer architectures like Generative pre-trained transformer (GPT). Our findings support the potential value of using large language models to enhance pharmacovigilance.

Secondary (additional) findings from the 100,000 Genomes Project: Disease manifestation, health care outcomes, and costs of disclosure.

PURPOSE: The UK 100,000 Genomes Project offered participants screening for additional findings (AFs) in genes associated with familial hypercholesterolemia (FH) or hereditary cancer syndromes including breast/ovarian cancer (HBOC), Lynch, familial adenomatous polyposis, MYH-associated polyposis, multiple endocrine neoplasia (MEN), and von Hippel-Lindau. Here, we report disclosure processes, manifestation of AF-related disease, outcomes, and costs. METHODS: An observational study in an area representing one-fifth of England. RESULTS: Data were collected from 89 adult AF recipients. At disclosure, among 57 recipients of a cancer-predisposition-associated AF and 32 recipients of an FH-associated AF, 35% and 88%, respectively, had personal and/or family history evidence of AF-related disease. During post-disclosure investigations, 4 cancer-AF recipients had evidence of disease, including 1 medullary thyroid cancer. Six women with an HBOC AF, 3 women with a Lynch syndrome AF, and 2 individuals with a MEN AF elected for risk-reducing surgery. New hyperlipidemia diagnoses were made in 6 FH-AF recipients and treatment (re-)initiated for 7 with prior hyperlipidemia. Generating and disclosing AFs in this region cost £1.4m; £8680 per clinically significant AF. CONCLUSION: Generation and disclosure of AFs identifies individuals with and without personal or familial evidence of disease and prompts appropriate clinical interventions. Results can inform policy toward secondary findings.

QALYs and rare diseases: exploring the responsiveness of SF-6D, EQ-5D-5L and AQoL-8D following genomic testing for childhood and adult-onset rare genetic conditions in Australia.

BACKGROUND: Genomic testing transforms the diagnosis and management of rare conditions. However, uncertainty exists on how to best measure genomic outcomes for informing healthcare priorities. Using the HTA-preferred method should be the starting point to improve the evidence-base. This study explores the responsiveness of SF-6D, EQ-5D-5L and AQoL-8D following genomic testing across childhood and adult-onset genetic conditions. METHOD: Self-reported patient-reported outcomes (PRO) were obtained from: primary caregivers of children with suspected neurodevelopmental disorders (NDs) or genetic kidney diseases (GKDs) (carers' own PRO), adults with suspected GKDs using SF-12v2; adults with suspected complex neurological disorders (CNDs) using EQ-5D-5L; and adults with dilated cardiomyopathy (DCM) using AQol-8D. Responsiveness was assessed using the standardised response mean effect-size based on diagnostic (having a confirmed genomic diagnosis), personal (usefulness of genomic information to individuals or families), and clinical (clinical usefulness of genomic information) utility anchors. RESULTS: In total, 254 people completed PRO measures before genomic testing and after receiving results. For diagnostic utility, a nearly moderate positive effect size was identified by the AQoL-8D in adult DCM patients. Declines in physical health domains masked any improvements in mental or psychosocial domains in parents of children affected by NDs and adult CNDs and DCM patients with confirmed diagnosis. However, the magnitude of the changes was small and we did not find statistically significant evidence of these changes. No other responsiveness evidence related to diagnostic, clinical, and personal utility of genomic testing was identified. CONCLUSION: Generic PRO measures may lack responsiveness to the diagnostic, clinical and personal outcomes of genomics, but further research is needed to establish their measurement properties and relevant evaluative space in the context of rare conditions. Expected declines in the physical health of people experiencing rare conditions may further challenge the conventional application of quality of life assessments.

Carbon Monoxide Levels Produced by Propane/Isobutane Canister Stoves inside a Tent.

INTRODUCTION: Improper use of camp stoves in enclosed spaces has resulted in fatalities from carbon monoxide (CO) poisoning. Prior research has focused on the CO output of stoves burning white gas, unleaded gas, or kerosene. Stoves burning an isobutane/propane fuel have not been investigated and are the focus of this study. METHODS: Three stoves utilizing isobutane/propane fuel were used to heat a pot of water inside a 3-season tent under controlled settings. Multiple runs with each stove were performed, and CO measurements, in parts per million (ppm), were recorded at 1-min intervals for a total of 15 min using a RAE Systems gas monitor. Data are reported as mean with SD. Repeated measures analysis of variance was utilized to examine changes over time. Statistical significance was set at P<0.05. RESULTS: There was a statistically significant main effect of time and CO level, F (14, 168)=7.6, P<0.001. There was a statistically significant difference between-subjects effect of stove group F (2, 12)=8.6, P=0.005, indicating that CO levels were different depending on the stove. Tukey's post-hoc analyses revealed that stove A had the highest CO levels. The average level of stove A was statistically significantly higher than that of stove B and stove C, with a mean CO level difference of 79 ppm (95% CI, 3-156), P=0.043 and 117 ppm (95% CI, 40-194), P=0.004, respectively. CONCLUSIONS: Stoves utilizing isobutane/propane fuel can produce unsafe CO levels and should not be used in enclosed spaces.

Algorithmic identification of treatment-emergent adverse events from clinical notes using large language models: a pilot study in inflammatory bowel disease.

Background and Aims: Outpatient clinical notes are a rich source of information regarding drug safety. However, data in these notes are currently underutilized for pharmacovigilance due to methodological limitations in text mining. Large language models (LLM) like BERT have shown progress in a range of natural language processing tasks but have not yet been evaluated on adverse event detection. Methods: We adapted a new clinical LLM, UCSF BERT, to identify serious adverse events (SAEs) occurring after treatment with a non-steroid immunosuppressant for inflammatory bowel disease (IBD). We compared this model to other language models that have previously been applied to AE detection. Results: We annotated 928 outpatient IBD notes corresponding to 928 individual IBD patients for all SAE-associated hospitalizations occurring after treatment with a non-steroid immunosuppressant. These notes contained 703 SAEs in total, the most common of which was failure of intended efficacy. Out of 8 candidate models, UCSF BERT achieved the highest numerical performance on identifying drug-SAE pairs from this corpus (accuracy 88-92%, macro F1 61-68%), with 5-10% greater accuracy than previously published models. UCSF BERT was significantly superior at identifying hospitalization events emergent to medication use (p < 0.01). Conclusions: LLMs like UCSF BERT achieve numerically superior accuracy on the challenging task of SAE detection from clinical notes compared to prior methods. Future work is needed to adapt this methodology to improve model performance and evaluation using multi-center data and newer architectures like GPT. Our findings support the potential value of using large language models to enhance pharmacovigilance.

Important Considerations for Signal Detection and Evaluation.

Safety clinicians have a wealth of resources describing how to perform signal detection. Nevertheless, there are some nuances concerning approaches taken by regulatory authorities and statistical considerations that should be appreciated. New approaches, such as the FDA Medical Queries, illustrate the value of considering medical concepts over individual adverse events. One area which would benefit from further clarity is how safety signals may be evaluated for evidence of a causal relationship to the drug of interest. Just as such safety signals can take many forms, the types of tools and methods required to interrogate these signals are equally as diverse. An understanding of the complexity of this process can aid the safety reviewer in successfully characterizing the emerging safety profile of a drug during the pre-marketing phase of development.

Retrospective file review shows limited genetic services fails most patients - an argument for the implementation of exome sequencing as a first-tier test in resource-constraint settings.

BACKGROUND: Exome sequencing is recommended as a first-line investigation for patients with a developmental delay or intellectual disability. This approach has not been implemented in most resource-constraint settings, including Africa, due to the high cost of implementation. Instead, patients have limited access to services and testing options. Here, we evaluate the effectiveness of a limited genetic testing strategy and contrast the findings to a conceivable outcome if exome sequencing were available instead. RESULTS: A retrospective audit of 934 patient files presenting to a medical genetics clinic in South Africa showed that 83% of patients presented with developmental delay as a clinical feature. Patients could be divided into three groups, representing distinct diagnostic pathways. Patient Group A (18%; mean test cost $131) were confirmed with aneuploidies, following a simple, inexpensive test. Patient Group B (25%; mean test cost $140) presented with clinically recognizable conditions but only 39% received a genetic diagnostic confirmation due to limited testing options. Patient Group C - the largest group (57%; mean test cost $337) - presented with heterogenous conditions and DD, and 92% remained undiagnosed after limited available testing was performed. CONCLUSIONS: Patients with DD are the largest group of patients seen in medical genetics clinics in South Africa. When clinical features are not distinct, limited testing options drastically restricts diagnostic yield. A cost- and time analysis shows most patients would benefit from first-line exome sequencing, reducing their individual diagnostic odysseys.

Assessing adverse events in clinical trials during the era of the COVID-19 pandemic.

Interpretation of safety data for clinical trials that were ongoing at the onset of the COVID-19 pandemic or were started subsequent to the beginning of the pandemic may be affected in a variety of ways. Pandemic-related issues can influence the extent of study participation and introduce data collection gaps. A SARS-CoV-2 infection among study subjects as a post-randomization event may introduce a number of confounding factors that can alter the frequency of adverse events, in some cases appearing as an increase in the frequency of an adverse event associated with a study drug relative to a comparator. The authors discuss clinical challenges and statistical concerns, specifically the estimand framework, including examples for consideration, to address these challenges in safety evaluation wrought by the COVID-19 pandemic. Our aim is to shed light on the importance of starting an early dialogue among the drug development team on the evaluation of safety, critical for benefit-risk evaluation throughout the drug development process.

Graphene Oxide and Metal-Organic Framework-Based Breathable Barrier Membranes for Toxic Vapors.

Garments protective against chemical warfare agents (CWAs) or accidently released toxic chemicals must block the transport of toxic gases/vapors for a substantial time and allow moisture transport for breathability. These demands are challenging: either the barriers block CWAs effectively but have poor breathability or barriers have excellent breathability but cannot block CWAs well. Existing protective garments employ large amounts of active carbon, making them quite heavy. Metal-organic framework (MOF)-based adsorbents are being investigated as sorbents for CWAs. Breathable laminate of graphene oxide (GO) flakes supported on a porous membrane reduces permeation rates of CWA simulants substantially. We developed a multilayered membrane-based flexible barrier: GO laminate-based membrane over a MOF nanocrystal-filled expanded polytetrafluorethylene (ePTFE) membrane having submicrometer pores. The GO laminate-based layer developed a steady breakthrough concentration level almost 2 orders of magnitude below the usual breakthrough level. This highly reduced level of CWA was blocked by the MOF nanocrystal-filled membrane substrate layer over a highly extended period. We demonstrated the blocking of CWAs, mustard (HD), soman (GD), a sarin simulant [dimethyl methyl phosphonate (DMMP)], and ammonia for an extended period while the moisture transmission rate was substantial. The times for complete blockage of ammonia, HD, GD, and DMMP were 2750 min, 1075 min, 176 min, and 7 days, respectively. This remarkable performance resulted from a very low steady-state penetrant permeation through GO-laminate membrane and substantial penetrant sorption by MOF nanocrystals; furthermore, both layers show high moisture vapor transmission.

Financing and Reimbursement Models for Personalised Medicine: A Systematic Review to Identify Current Models and Future Options.

BACKGROUND: The number of healthcare interventions described as 'personalised medicine' (PM) is increasing rapidly. As healthcare systems struggle to decide whether to fund PM innovations, it is unclear what models for financing and reimbursement are appropriate to apply in this context. OBJECTIVE: To review financing and reimbursement models for PM, summarise their key characteristics, and describe whether they can influence the development and uptake of PM. METHODS: A literature review was conducted in Medline, Embase, Web of Science, and Econlit to identify studies published in English between 2009 and 2021, and reviews published before 2009. Grey literature was identified through Google Scholar, Google and subject-specific webpages. Articles that described financing and reimbursement of PM, and financing of non-PM were included. Data were extracted and synthesised narratively to report on the models, as well as facilitators, incentives, barriers and disincentives that could influence PM development and uptake. RESULTS: One hundred and fifty-three papers were included. Research and development of PM was financed through both public and private sources and reimbursed largely through traditional models such as single fees, Diagnosis-Related Groups, and bundled payments. Financial-based reimbursement, including rebates and price-volume agreements, was mainly applied to targeted therapies. Performance-based reimbursement was identified mainly for gene and targeted therapies, and some companion diagnostics. Gene therapy manufacturers offered outcome-based rebates for treatment failure for interventions including Luxturna®, Kymriah®, Yescarta®, Zynteglo®, Zolgensma® and Strimvelis®, and coverage with evidence development for Kymriah® and Yescarta®. Targeted testing with OncotypeDX® was granted value-based reimbursement through initial coverage with evidence development. The main barriers and disincentives to PM financing and reimbursement were the lack of strong links between stakeholders and the lack of demonstrable benefit and value of PM. CONCLUSIONS: Public-private financing agreements and performance-based reimbursement models could help facilitate the development and uptake of PM interventions with proven clinical benefit.

Factors that impact on women's decision-making around prenatal genomic tests: An international discrete choice survey.

OBJECTIVE: We conducted a survey-based discrete-choice experiment (DCE) to understand the test features that drive women's preferences for prenatal genomic testing, and explore variation across countries. METHODS: Five test attributes were identified as being important for decision-making through a literature review, qualitative interviews and quantitative scoring exercise. Twelve scenarios were constructed in which respondents choose between two invasive tests or no test. Women from eight countries who delivered a baby in the previous 24 months completed a DCE presenting these scenarios. Choices were modeled using conditional logit regression analysis. RESULTS: Surveys from 1239 women (Australia: n = 178; China: n = 179; Denmark: n = 88; Netherlands: n = 177; Singapore: n = 90; Sweden: n = 178; UK: n = 174; USA: n = 175) were analyzed. The key attribute affecting preferences was a test with the highest diagnostic yield (p < 0.01). Women preferred tests with short turnaround times (p < 0.01), and tests reporting variants of uncertain significance (VUS; p < 0.01) and secondary findings (SFs; p < 0.01). Several country-specific differences were identified, including time to get a result, who explains the result, and the return of VUS and SFs. CONCLUSION: Most women want maximum information from prenatal genomic tests, but our findings highlight country-based differences. Global consensus on how to return uncertain results is not necessarily realistic or desirable.

Eliciting risk preferences that predict risky health behavior: A comparison of two approaches.

Information on attitudes to risk could increase understanding of and explain risky health behaviors. We investigate two approaches to eliciting risk preferences in the health domain, a novel "indirect" lottery elicitation approach with health states as outcomes and a "direct" approach where respondents are asked directly about their willingness to take risks. We compare the ability of the two approaches to predict health-related risky behaviors in a general adult population. We also investigate a potential framing effect in the indirect lottery elicitation approach. We find that risk preferences elicited using the direct approach can better predict health-related risky behavior than those elicited using the indirect approach. Moreover, a seemingly innocuous change to the framing of the lottery question results in significantly different risk preference estimates, and conflicting conclusions about the ability of the indicators to predict risky health behaviors.

Defining a Core Data Set for the Economic Evaluation of Precision Oncology.

OBJECTIVES: Precision oncology is generating vast amounts of multiomic data to improve human health and accelerate research. Existing clinical study designs and attendant data are unable to provide comparative evidence for economic evaluations. This lack of evidence can cause inconsistent and inappropriate reimbursement. Our study defines a core data set to facilitate economic evaluations of precision oncology. METHODS: We conducted a literature review of economic evaluations of next-generation sequencing technologies, a common application of precision oncology, published between 2005 and 2018 and indexed in PubMed (MEDLINE). Based on this review, we developed a preliminary core data set for informal expert feedback. We then used a modified-Delphi approach with individuals involved in implementation and evaluation of precision medicine, including 2 survey rounds followed by a final voting conference to refine the data set. RESULTS: Two authors determined that variation in published data elements was reached after abstraction of 20 economic evaluations. Expert consultation refined the data set to 83 unique data elements, and a multidisciplinary sample of 46 experts participated in the modified-Delphi process. A total of 68 elements (81%) were selected as required, spanning demographics and clinical characteristics, genomic data, cancer treatment, health and quality of life outcomes, and resource use. CONCLUSIONS: Cost-effectiveness analyses will fail to reflect the real-world impacts of precision oncology without data to accurately characterize patient care trajectories and outcomes. Data collection in accordance with the proposed core data set will promote standardization and enable the generation of decision-grade evidence to inform reimbursement.

Assessing women's preferences towards tests that may reveal uncertain results from prenatal genomic testing: Development of attributes for a discrete choice experiment, using a mixed-methods design.

Prenatal DNA tests, such as chromosomal microarray analysis or exome sequencing, increase the likelihood of receiving a diagnosis when fetal structural anomalies are identified. However, some parents will receive uncertain results such as variants of uncertain significance and secondary findings. We aimed to develop a set of attributes and associated levels for a discrete-choice experiment (DCE) that will examine parents' preferences for tests that may reveal uncertain test results. A two phase mixed-methods approach was used to develop attributes for the DCE. In Phase 1, a "long list" of candidate attributes were identified via two approaches: 1) a systematic review of the literature around parental experiences of uncertainty following prenatal testing; 2) 16 semi-structured interviews with parents who had experienced uncertainty during pregnancy and 25 health professionals who return uncertain prenatal results. In Phase 2, a quantitative scoring exercise with parents prioritised the candidate attributes. Clinically appropriate levels for each attribute were then developed. A final set of five attributes and levels were identified: likelihood of getting a result, reporting of variants of uncertain significance, reporting of secondary findings, time taken to receive results, and who tells you about your result. These attributes will be used in an international DCE study to investigate preferences and differences across countries. This research will inform best practice for professionals supporting parents to manage uncertainty in the prenatal setting.

100,000 Genomes Pilot on Rare-Disease Diagnosis in Health Care - Preliminary Report.

BACKGROUND: The U.K. 100,000 Genomes Project is in the process of investigating the role of genome sequencing in patients with undiagnosed rare diseases after usual care and the alignment of this research with health care implementation in the U.K. National Health Service. Other parts of this project focus on patients with cancer and infection. METHODS: We conducted a pilot study involving 4660 participants from 2183 families, among whom 161 disorders covering a broad spectrum of rare diseases were present. We collected data on clinical features with the use of Human Phenotype Ontology terms, undertook genome sequencing, applied automated variant prioritization on the basis of applied virtual gene panels and phenotypes, and identified novel pathogenic variants through research analysis. RESULTS: Diagnostic yields varied among family structures and were highest in family trios (both parents and a proband) and families with larger pedigrees. Diagnostic yields were much higher for disorders likely to have a monogenic cause (35%) than for disorders likely to have a complex cause (11%). Diagnostic yields for intellectual disability, hearing disorders, and vision disorders ranged from 40 to 55%. We made genetic diagnoses in 25% of the probands. A total of 14% of the diagnoses were made by means of the combination of research and automated approaches, which was critical for cases in which we found etiologic noncoding, structural, and mitochondrial genome variants and coding variants poorly covered by exome sequencing. Cohortwide burden testing across 57,000 genomes enabled the discovery of three new disease genes and 19 new associations. Of the genetic diagnoses that we made, 25% had immediate ramifications for clinical decision making for the patients or their relatives. CONCLUSIONS: Our pilot study of genome sequencing in a national health care system showed an increase in diagnostic yield across a range of rare diseases. (Funded by the National Institute for Health Research and others.).

Cost-Effectiveness of Amphotericin B Deoxycholate Versus Itraconazole for Induction Therapy of Talaromycosis in Human Immunodeficiency Virus-Infected Adults in Vietnam.

BACKGROUND: Talaromycosis (penicilliosis) is an invasive fungal infection and a major cause of human immunodeficiency virus (HIV)-related deaths in Southeast Asia. Guidelines recommend induction therapy with amphotericin B deoxycholate; however, treatment with itraconazole has fewer toxic effects, is easier to administer, and is less expensive. Our recent randomized controlled trial in Vietnam found that amphotericin B was superior to itraconazole with respect to 6-month mortality. We undertook an economic evaluation alongside this trial to determine whether the more effective treatment is cost-effective. METHODS: Resource use, direct and indirect costs, and health and quality-of-life outcomes (measured using quality-adjusted life-years [QALYs]) were evaluated for 405 trial participants from 2012 to 2016. Both a Vietnamese health service and a broader societal costing perspective were considered. Mean costs and QALYs were combined to calculate the within-trial cost-effectiveness of amphotericin vs itraconazole from both perspectives. RESULTS: From a Vietnamese health service perspective, amphotericin increases costs but improves health outcomes compared to itraconazole, at a cost of $3013/QALY gained. The probability that amphotericin is cost-effective at a conventional (World Health Organization CHOICE) threshold of value for money is 46%. From a societal perspective, amphotericin is cost-reducing and improves outcomes compared to itraconazole, and is likely to be a cost-effective strategy at any value for money threshold greater than $0. CONCLUSIONS: Our analysis indicates that induction therapy with amphotericin is a cost-effective treatment strategy for HIV-infected adults diagnosed with talaromycosis in Vietnam. These results provide the evidence base for health care providers and policy makers to improve access to and use of amphotericin.

What Aspects of Illness Influence Public Preferences for Healthcare Priority Setting? A Discrete Choice Experiment in the UK.

BACKGROUND: Decisions on funding new healthcare technologies assume that all health improvements are valued equally. However, public reaction to health technology assessment (HTA) decisions suggests there are health attributes that matter deeply to them but are not currently accounted for in the assessment process. We aimed to determine the relative importance of attributes of illness that influence the value placed on alleviating that illness. METHOD: We conducted a discrete choice experiment survey that presented general public respondents with 15 funding decisions between hypothetical health conditions. The conditions were defined by five attributes that characterise serious illnesses, plus the health gain from treatment. Respondent preferences were modelled using conditional logistic regression and latent class analysis. RESULTS: 905 members of the UK public completed the survey in November 2017. Respondents generally preferred to provide treatments for conditions with 'better' characteristics. The exception was treatment availability, where respondents preferred to provide treatments for conditions where there is no current treatment, and were prepared to accept lower overall health gain to do so. A subgroup of respondents preferred to prioritise 'worse' health states. CONCLUSION: This study suggests a preference among the UK public for treating an unmet need; however, it does not suggest a preference for prioritising other distressing aspects of health conditions, such as limited life expectancy, or where patients are reliant on care. Our results are not consistent with the features currently prioritised in UK HTA processes, and the preference heterogeneity we identify presents a major challenge for developing broadly acceptable policy.

Public preferences for delayed or immediate antibiotic prescriptions in UK primary care: A choice experiment.

BACKGROUND: Delayed (or "backup") antibiotic prescription, where the patient is given a prescription but advised to delay initiating antibiotics, has been shown to be effective in reducing antibiotic use in primary care. However, this strategy is not widely used in the United Kingdom. This study aimed to identify factors influencing preferences among the UK public for delayed prescription, and understand their relative importance, to help increase appropriate use of this prescribing option. METHODS AND FINDINGS: We conducted an online choice experiment in 2 UK general population samples: adults and parents of children under 18 years. Respondents were presented with 12 scenarios in which they, or their child, might need antibiotics for a respiratory tract infection (RTI) and asked to choose either an immediate or a delayed prescription. Scenarios were described by 7 attributes. Data were collected between November 2018 and February 2019. Respondent preferences were modelled using mixed-effects logistic regression. The survey was completed by 802 adults and 801 parents (75% of those who opened the survey). The samples reflected the UK population in age, sex, ethnicity, and country of residence. The most important determinant of respondent choice was symptom severity, especially for cough-related symptoms. In the adult sample, the probability of choosing delayed prescription was 0.53 (95% confidence interval (CI) 0.50 to 0.56, p < 0.001) for a chesty cough and runny nose compared to 0.30 (0.28 to 0.33, p < 0.001) for a chesty cough with fever, 0.47 (0.44 to 0.50, p < 0.001) for sore throat with swollen glands, and 0.37 (0.34 to 0.39, p < 0.001) for sore throat, swollen glands, and fever. Respondents were less likely to choose delayed prescription with increasing duration of illness (odds ratio (OR) 0.94 (0.92 to 0.96, p < 0.001)). Probabilities of choosing delayed prescription were similar for parents considering treatment for a child (44% of choices versus 42% for adults, p = 0.04). However, parents differed from the adult sample in showing a more marked reduction in choice of the delayed prescription with increasing duration of illness (OR 0.83 (0.80 to 0.87) versus 0.94 (0.92 to 0.96) for adults, p for heterogeneity p < 0.001) and a smaller effect of disruption of usual activities (OR 0.96 (0.95 to 0.97) versus 0.93 (0.92 to 0.94) for adults, p for heterogeneity p < 0.001). Females were more likely to choose a delayed prescription than males for minor symptoms, particularly minor cough (probability 0.62 (0.58 to 0.66, p < 0.001) for females and 0.45 (0.41 to 0.48, p < 0.001) for males). Older people, those with a good understanding of antibiotics, and those who had not used antibiotics recently showed similar patterns of preferences. Study limitations include its hypothetical nature, which may not reflect real-life behaviour; the absence of a "no prescription" option; and the possibility that study respondents may not represent the views of population groups who are typically underrepresented in online surveys. CONCLUSIONS: This study found that delayed prescription appears to be an acceptable approach to reducing antibiotic consumption. Certain groups appear to be more amenable to delayed prescription, suggesting particular opportunities for increased use of this strategy. Prescribing choices for sore throat may need additional explanation to ensure patient acceptance, and parents in particular may benefit from reassurance about the usual duration of these illnesses.

A New Paradigm for Safety Data Signal Detection and Evaluation Using Open-Source Software Created by an Interdisciplinary Working Group.

Techniques to evaluate large amounts of safety data continue to evolve based on a greater understanding of how the brain processes visual information and the advancement of programing tools. The Interactive Safety Graphics Task Force of the American Statistical Association Biopharmaceutical Safety Working Group has assembled a multidisciplinary team of experts in a variety of domains to develop the next generation of open-source visual analytical tools for safety data based on these advances. The multidisciplinary approach resulted in the rapid development of the first tool, a novel interactive version of the familiar Evaluation of Drug-Induced Serious Hepatotoxicity (eDISH) graphic along with a unique clinical workflow to guide the reviewer through the data analysis. This now serves as the model for the team to expand the open-source platform into a suite of other interactive safety analysis tools.