Patient notes contain substantial information but are difficult for computers to analyse due to their unstructured format. Large-language models (LLMs), such as Generative Pre-trained Transformer 4 (GPT-4), have changed our ability to process text, but we do not know how effectively they handle medical notes. We aimed to assess the ability of GPT-4 to answer predefined questions after reading medical notes in three different languages.

Longitudinal digital health studies combine passively collected information from digital devices, such as commercial wearable devices, and actively contributed data, such as surveys, from participants. Although the use of smartphones and access to the internet supports the development of these studies, challenges exist in collecting representative data due to low adherence and retention. We aimed to identify key factors related to adherence and retention in digital health studies and develop a methodology to identify factors that are associated with and might affect study participant engagement.

Perimembranous ventricular septal defect (PMVSD) is a prevalent congenital heart disease, presenting challenges in predicting spontaneous closure, which is crucial for therapeutic decisions. Existing models mainly rely on structured echocardiographic parameters or restricted data. This study introduces an artificial intelligence (AI)-based model, which uses natural language processing (NLP) and machine learning with the aim of improving spontaneous closure predictability in PMVSD.

Palliative spine radiation therapy is prone to treatment at the wrong anatomic level. We developed a fully automated deep learning-based spine-targeting quality assurance system (DL-SpiQA) for detecting treatment at the wrong anatomic level. DL-SpiQA was evaluated based on retrospective testing of spine radiation therapy treatments and prospective clinical deployment.

Chatbots are artificial intelligence (AI) programs designed to simulate conversations with humans that present opportunities and challenges in scientific research. Despite growing clarity from publishing organisations on the use of AI chatbots, researchers' perceptions remain less understood. In this international cross-sectional survey, we aimed to assess researchers' attitudes, familiarity, perceived benefits, and limitations related to AI chatbots. Our online survey was open from July 9 to Aug 11, 2023, with 61 560 corresponding authors identified from 122 323 articles indexed in PubMed. 2452 (4·0%) provided responses and 2165 (94·5%) of 2292 who met eligibility criteria completed the survey. 1161 (54·0%) of 2149 respondents were male and 959 (44·6%) were female. 1294 (60·5%) of 2138 respondents were familiar with AI chatbots, and 945 (44·5%) of 2125 had previously used AI chatbots in research. Only 244 (11·4%) of 2137 reported institutional training on AI tools, and 211 (9·9%) of 2131 noted institutional policies on AI chatbot use. Despite mixed opinions on the benefits, 1428 (69·7%) of 2048 expressed interest in further training. Although many valued AI chatbots for reducing administrative workload (1299 [66·9%] of 1941), there was insufficient understanding of the decision making process (1484 [77·2%] of 1923). Overall, this study highlights substantial interest in AI chatbots among researchers, but also points to the need for more formal training and clarity on their use.

Neuroprosthetics research has entered a stage in which animal models and proof-of-concept studies are translated into clinical applications, often combining implants with artificial intelligence techniques. This new phase raises the question of how clinical trials should be designed to scientifically and ethically address the unique features of neural prostheses. Neural prostheses are complex cyberbiological devices able to acquire and process data; hence, their assessment is not reducible to only third-party safety and efficacy evaluations as in pharmacological research. In addition, assessment of neural prostheses requires a causal understanding of their mechanisms, and scrutiny of their information security and legal liability standards. Some neural prostheses affect not only human behaviour, but also psychological faculties such as consciousness, cognition, and affective states. In this Viewpoint, we argue that the technological novelty of neural prostheses could generate challenges for technology assessment, clinical validation, and research ethics oversight. To this end, we identify a set of methodological and research ethics challenges specific to this medical technology innovation. We provide insights into relevant ethical guidelines and assess whether oversight mechanisms are well equipped to ensure adequate clinical and ethical use. Finally, we outline patient-centred research ethics requirements for clinical trials involving implantable neural prostheses.

There is a growing global burden of liver disease with the current management for complications of liver cirrhosis being reactive as opposed to proactive, affecting outcomes. Management can often be suboptimal in overburdened health-care systems with considerable socioeconomic and geographical disparity existing, which was exacerbated by the COVID-19 pandemic, highlighting the need for sustainable care pathways to be delivered remotely. To this end, digital health care could be the key and, in this Review, we highlight the principal studies that have explored the use of digital technology in the management of cirrhosis complications. While digital health care is a somewhat new field, considerable advances have been made in various domains, particularly in the development of remote monitoring and risk modelling. We aim to provide a balanced perspective of the opportunities for and barriers to the integration of digital technology into established liver-care pathways. Lastly, we reflect on the current acceptability of digital health care and the required future directions to ensure the digital transformation of hepatology is a success.

Without careful dissection of the ways in which biases can be encoded into artificial intelligence (AI) health technologies, there is a risk of perpetuating existing health inequalities at scale. One major source of bias is the data that underpins such technologies. The STANDING Together recommendations aim to encourage transparency regarding limitations of health datasets and proactive evaluation of their effect across population groups. Draft recommendation items were informed by a systematic review and stakeholder survey. The recommendations were developed using a Delphi approach, supplemented by a public consultation and international interview study. Overall, more than 350 representatives from 58 countries provided input into this initiative. 194 Delphi participants from 25 countries voted and provided comments on 32 candidate items across three electronic survey rounds and one in-person consensus meeting. The 29 STANDING Together consensus recommendations are presented here in two parts. Recommendations for Documentation of Health Datasets provide guidance for dataset curators to enable transparency around data composition and limitations. Recommendations for Use of Health Datasets aim to enable identification and mitigation of algorithmic biases that might exacerbate health inequalities. These recommendations are intended to prompt proactive inquiry rather than acting as a checklist. We hope to raise awareness that no dataset is free of limitations, so transparent communication of data limitations should be perceived as valuable, and absence of this information as a limitation. We hope that adoption of the STANDING Together recommendations by stakeholders across the AI health technology lifecycle will enable everyone in society to benefit from technologies which are safe and effective.

Large language models (LLMs) offer promising applications in mental health care to address gaps in treatment and research. By leveraging clinical notes and transcripts as data, LLMs could improve diagnostics, monitoring, prevention, and treatment of mental health conditions. However, several challenges persist, including technical costs, literacy gaps, risk of biases, and inequalities in data representation. In this Viewpoint, we propose a sociocultural-technical approach to address these challenges. We highlight five key areas for development: (1) building a global clinical repository to support LLMs training and testing, (2) designing ethical usage settings, (3) refining diagnostic categories, (4) integrating cultural considerations during development and deployment, and (5) promoting digital inclusivity to ensure equitable access. We emphasise the need for developing representative datasets, interpretable clinical decision support systems, and new roles such as digital navigators. Only through collaborative efforts across all stakeholders, unified by a sociocultural-technical framework, can we clinically deploy LLMs while ensuring equitable access and mitigating risks.

Since the COVID-19 pandemic, considerable advances have been made to improve epidemic preparedness by accelerating diagnostics, therapeutics, and vaccine development. However, we argue that it is crucial to make equivalent efforts in the field of outbreak analytics to help ensure reliable, evidence-based decision making. To explore the challenges and key priorities in the field of outbreak analytics, the Epiverse-TRACE initiative brought together a multidisciplinary group of experts, including field epidemiologists, data scientists, academics, and software engineers from public health institutions across multiple countries. During a 3-day workshop, 40 participants discussed what the first 100 lines of code written during an outbreak should look like. The main findings from this workshop are summarised in this Viewpoint. We provide an overview of the current outbreak analytic landscape by highlighting current key challenges that should be addressed to improve the response to future public health crises. Furthermore, we propose actionable solutions to these challenges that are achievable in the short term, and longer-term strategic recommendations. This Viewpoint constitutes a call to action for experts involved in epidemic response to develop modern and robust data analytic approaches at the heart of epidemic preparedness and response.

Longitudinal patient registries generate important evidence for advancing clinical care and the regulatory evaluation of health-care products. Most national registries rely on data collected as part of routine clinical encounters, an approach that does not capture real-world, patient-centred outcomes, such as physical activity, fatigue, ability to do daily tasks, and other indicators of quality of life. Digital health technologies that obtain such real-world data could greatly enhance patient registries but unresolved challenges have so far prevented their broad adoption. Based on our experience implementing digital health technologies in registries and observational studies, we propose potential solutions to three practical challenges we have repeatedly encountered: determining what to measure digitally, selecting the appropriate device, and ensuring representativeness and engagement over time. We describe the example of a hypothetical patient registry for valvular heart disease, a condition for which there is substantial variation in treatment selection and postintervention outcomes, and for which patient-centred outcome data are urgently needed to inform clinical care guidelines and health-service commissioning.

Microbiology reference laboratories perform a crucial role within public health systems. This role was especially evident during the COVID-19 pandemic. In this Viewpoint, we emphasise the importance of microbiology reference laboratories and highlight the types of digital data and expertise they provide, which benefit national and international public health. We also highlight the value of surveillance initiatives among collaborative international partners, who work together to share, analyse, and interpret data, and then disseminate their findings in a timely manner. Microbiology reference laboratories have substantial impact at regional, national, and international levels, and sustained support for these laboratories is essential for public health in both pandemic and non-pandemic times.

Synthetic data, generated through artificial intelligence technologies such as generative adversarial networks and latent diffusion models, maintain aggregate patterns and relationships present in the real data the technologies were trained on without exposing individual identities, thereby mitigating re-identification risks. This approach has been gaining traction in biomedical research because of its ability to preserve privacy and enable dataset sharing between organisations. Although the use of synthetic data has become widespread in other domains, such as finance and high-energy physics, use in medical research raises novel issues. The use of synthetic data as a method of preserving the privacy of data used to train models requires that the data are high fidelity with the original data to preserve utility, but must be sufficiently different as to protect against adversarial or accidental re-identification. There is a need for the development of standards for synthetic data generation and consensus standards for its evaluation. As synthetic data applications expand, ongoing legal and ethical evaluations are crucial to ensure that they remain a secure and effective tool for advancing medical research without compromising individual privacy.