Generative artificial intelligence and ethical considerations in health care: a scoping review and ethics checklist.

被引量：- 发表：1970

Effect of wearable activity trackers on physical activity in children and adolescents: a systematic review and meta-analysis.

Physical inactivity in children and adolescents has become a pressing public health concern. Wearable activity trackers can allow self-monitoring of physical activity behaviour and promote autonomous motivation for exercise. However, the effects of wearable trackers on physical activity in young populations remain uncertain. In this systematic review and meta-analysis, we searched PubMed, Embase, SPORTDiscus, and Web of Science for publications from database inception up to Aug 30, 2023, without restrictions on language. Studies were eligible if they were randomised controlled trials or clustered randomised controlled trials that examined the use of wearable activity trackers to promote physical activity, reduce sedentary behaviours, or promote overall health in participants with a mean age of 19 years or younger, with no restrictions on health condition or study settings. Studies were excluded if children or adolescents were not the primary intervention cohort, or wearable activity trackers were not worn on users' bodies to objectively track users' physical activity levels. Two independent reviewers (WWA and FR) assessed eligibility of studies and contacted authors of studies if more information was needed to assess eligibility. We also searched reference lists from relevant systematic reviews and meta-analyses. Systematic review software Covidence was used for study screening and data extraction. Study characteristics including study setting, participant characteristics, intervention characteristics, comparator, and outcome measurements were extracted from eligible studies. The two primary outcomes were objectively measured daily steps and moderate-to-vigorous physical activity. We used a random-effects model with Hartung-Knapp adjustments to calculate standardised mean differences. Between-study heterogeneity was examined using Higgins I2 and Cochran Q statistic. Publication bias was assessed using Egger's regression test. This systematic review was registered with PROSPERO, CRD42023397248. We identified 9619 studies from our database research and 174 studies from searching relevant systematic reviews and meta-analyses, of which 105 were subjected to full text screening. We included 21 eligible studies, involving 3676 children and adolescents (1618 [44%] were female and 2058 [56%] were male, mean age was 13·7 years [SD 2·7]) in our systematic review and meta-analysis. Ten studies were included in the estimation of the effect of wearable activity trackers on objectively measured daily steps and 11 were included for objectively measured moderate-to-vigorous physical activity. Compared with controls, we found a significant increase in objectively measured daily steps (standardised mean difference 0·37 [95% CI 0·09 to 0·65; p=0·013]; Q 47·60 [p<0·0001]; I2 72·7% [95% CI 53·4 to 84·0]), but not for moderate-to-vigorous physical activity (-0·08 [-0·18 to 0·02; p=0·11]; Q 10·26 [p=0·74]; I2 0·0% [0·0 to 53·6]). Wearable activity trackers might increase daily steps in young cohorts of various health statuses, but not moderate-to-vigorous physical activity, highlighting the potential of wearable trackers for motivating physical activity in children and adolescents. More rigorously designed trials that minimise missing data are warranted to validate our positive findings on steps and to explore possible long-term effects. The Hong Kong University Grants Committee and Seed Fund for Basic Research of the University of Hong Kong.

被引量：- 发表：1970

Medical artificial intelligence for clinicians: the lost cognitive perspective.

The development and commercialisation of medical decision systems based on artificial intelligence (AI) far outpaces our understanding of their value for clinicians. Although applicable across many forms of medicine, we focus on characterising the diagnostic decisions of radiologists through the concept of ecologically bounded reasoning, review the differences between clinician decision making and medical AI model decision making, and reveal how these differences pose fundamental challenges for integrating AI into radiology. We argue that clinicians are contextually motivated, mentally resourceful decision makers, whereas AI models are contextually stripped, correlational decision makers, and discuss misconceptions about clinician-AI interaction stemming from this misalignment of capabilities. We outline how future research on clinician-AI interaction could better address the cognitive considerations of decision making and be used to enhance the safety and usability of AI models in high-risk medical decision-making contexts.

被引量：- 发表：2024

Human mobility patterns in Brazil to inform sampling sites for early pathogen detection and routes of spread: a network modelling and validation study.

Detecting and foreseeing pathogen dispersion is crucial in preventing widespread disease transmission. Human mobility is a fundamental issue in human transmission of infectious agents. Through a mobility data-driven approach, we aimed to identify municipalities in Brazil that could comprise an advanced sentinel network, allowing for early detection of circulating pathogens and their associated transmission routes. In this modelling and validation study, we compiled a comprehensive dataset on intercity mobility spanning air, road, and waterway transport from the Brazilian Institute of Geography and Statistics (2016 data), National Transport Confederation (2022), and National Civil Aviation Agency (2017-23). We constructed a graph-based representation of Brazil's mobility network. The Ford-Fulkerson algorithm was used to rank the 5570 Brazilian cities according to their suitability as sentinel locations, allowing us to predict the most suitable locations for early detection and to track the most likely trajectory of a newly emerged pathogen. We also obtained SARS-CoV-2 genetic data from Brazilian municipalities during the early stage (Feb 25-April 30, 2020) of the virus's introduction and the gamma (P.1) variant emergence in Manaus (Jan 6-March 1, 2021), for the purposes of model validation. We found that flights alone transported 79·9 million (95% CI 58·3-101·4 million) passengers annually within Brazil during 2017-22, with seasonal peaks occurring in late spring and summer, and road and river networks had a maximum capacity of 78·3 million passengers weekly in 2016. By analysing the 7 746 479 most probable paths originating from source nodes, we found that 3857 cities fully cover the mobility pattern of all 5570 cities in Brazil, 557 (10·0%) of which cover 6 313 380 (81·5%) of the mobility patterns in our study. By strategically incorporating mobility patterns into Brazil's existing influenza-like illness surveillance network (ie, by switching the location of 111 of 199 sentinel sites to different municipalities), our model predicted that mobility coverage would have a 33·6% improvement from 4 059 155 (52·4%) mobility patterns to 5 422 535 (70·0%) without expanding the number of sentinel sites. Our findings are validated with genomic data collected during the SARS-CoV-2 pandemic period. Our model accurately mapped 22 (51%) of 43 clade 1-affected cities and 28 (60%) of 47 clade 2-affected cities spread from São Paulo city, and 20 (49%) of 41 clade 1-affected cities and 28 (58%) of 48 clade 2-affected cities spread from Rio de Janeiro city, Feb 25-April 30, 2020. Additionally, 224 (73%) of the 307 suggested early-detection locations for pathogens emerging in Manaus corresponded with the first cities affected by the transmission of the gamma variant, Jan 6-16, 2021. By providing essential clues for effective pathogen surveillance, our results have the potential to inform public health policy and improve future pandemic response efforts. Our results unlock the potential of designing country-wide clinical sample collection networks with mobility data-informed approaches, an innovative practice that can improve current surveillance systems. Rockefeller Foundation.

被引量：- 发表：2024

The diagnostic and triage accuracy of the GPT-3 artificial intelligence model: an observational study.

Artificial intelligence (AI) applications in health care have been effective in many areas of medicine, but they are often trained for a single task using labelled data, making deployment and generalisability challenging. How well a general-purpose AI language model performs diagnosis and triage relative to physicians and laypeople is not well understood. We compared the predictive accuracy of Generative Pre-trained Transformer 3 (GPT-3)'s diagnostic and triage ability for 48 validated synthetic case vignettes (<50 words; sixth-grade reading level or below) of both common (eg, viral illness) and severe (eg, heart attack) conditions to a nationally representative sample of 5000 lay people from the USA who could use the internet to find the correct options and 21 practising physicians at Harvard Medical School. There were 12 vignettes for each of four triage categories: emergent, within one day, within 1 week, and self-care. The correct diagnosis and triage category (ie, ground truth) for each vignette was determined by two general internists at Harvard Medical School. For each vignette, human respondents and GPT-3 were prompted to list diagnoses in order of likelihood, and the vignette was marked as correct if the ground-truth diagnosis was in the top three of the listed diagnoses. For triage accuracy, we examined whether the human respondents' and GPT-3's selected triage was exactly correct according to the four triage categories, or matched a dichotomised triage variable (emergent or within 1 day vs within 1 week or self-care). We estimated GPT-3's diagnostic and triage confidence on a given vignette using a modified bootstrap resampling procedure, and examined how well calibrated GPT-3's confidence was by computing calibration curves and Brier scores. We also performed subgroup analysis by case acuity, and an error analysis for triage advice to characterise how its advice might affect patients using this tool to decide if they should seek medical care immediately. Among all cases, GPT-3 replied with the correct diagnosis in its top three for 88% (42/48, 95% CI 75-94) of cases, compared with 54% (2700/5000, 53-55) for lay individuals (p<0.0001) and 96% (637/666, 94-97) for physicians (p=0·012). GPT-3 triaged 70% correct (34/48, 57-82) versus 74% (3706/5000, 73-75; p=0.60) for lay individuals and 91% (608/666, 89-93%; p<0.0001) for physicians. As measured by the Brier score, GPT-3 confidence in its top prediction was reasonably well calibrated for diagnosis (Brier score=0·18) and triage (Brier score=0·22). We observed an inverse relationship between case acuity and GPT-3 accuracy (p<0·0001) with a fitted trend line of -8·33% decrease in accuracy for every level of increase in case acuity. For triage error analysis, GPT-3 deprioritised truly emergent cases in seven instances. A general-purpose AI language model without any content-specific training could perform diagnosis at levels close to, but below, physicians and better than lay individuals. We found that GPT-3's performance was inferior to physicians for triage, sometimes by a large margin, and its performance was closer to that of lay individuals. Although the diagnostic performance of GPT-3 was comparable to physicians, it was significantly better than a typical person using a search engine. The National Heart, Lung, and Blood Institute.

被引量：21 发表：2024