Insights from explainable AI in oesophageal cancer team decisions.

BACKGROUND: Clinician-led quality control into oncological decision-making is crucial for optimising patient care. Explainable artificial intelligence (XAI) techniques provide data-driven approaches to unravel how clinical variables influence this decision-making. We applied global XAI techniques to examine the impact of key clinical decision-drivers when mapped by a machine learning (ML) model, on the likelihood of receiving different oesophageal cancer (OC) treatment modalities by the multidisciplinary team (MDT). METHODS: Retrospective analysis of 893 OC patients managed between 2010 and 2022 at our tertiary unit, used a random forests (RF) classifier to predict four possible treatment pathways as determined by the MDT: neoadjuvant chemotherapy followed by surgery (NACT + S), neoadjuvant chemoradiotherapy followed by surgery (NACRT + S), surgery-alone, and palliative management. Variable importance and partial dependence (PD) analyses then examined the influence of targeted high-ranking clinical variables within the ML model on treatment decisions as a surrogate model of the MDT decision-making dynamic. RESULTS: Amongst guideline-variables known to determine treatments, such as Tumour-Node-Metastasis (TNM) staging, age also proved highly important to the RF model (16.1 % of total importance) on variable importance analysis. PD subsequently revealed that predicted probabilities for all treatment modalities change significantly after 75 years (p < 0.001). Likelihood of surgery-alone and palliative therapies increased for patients aged 75-85yrs but lowered for NACT/NACRT. Performance status divided patients into two clusters which influenced all predicted outcomes in conjunction with age. CONCLUSION: XAI techniques delineate the relationship between clinical factors and OC treatment decisions. These techniques identify advanced age as heavily influencing decisions based on our model with a greater role in patients with specific tumour characteristics. This study methodology provides the means for exploring conscious/subconscious bias and interrogating inconsistencies in team-based decision-making within the era of AI-driven decision support.

Bridging the gap from medical to psychological safety assessment: consensus study in a digital mental health context.

BACKGROUND: Digital Mental Health Interventions (DMHIs) that meet the definition of a medical device are regulated by the Medicines and Healthcare products Regulatory Agency (MHRA) in the UK. The MHRA uses procedures that were originally developed for pharmaceuticals to assess the safety of DMHIs. There is recognition that this may not be ideal, as is evident by an ongoing consultation for reform led by the MHRA and the National Institute for Health and Care Excellence. AIMS: The aim of this study was to generate an experts' consensus on how the medical regulatory method used for assessing safety could best be adapted for DMHIs. METHOD: An online Delphi study containing three rounds was conducted with an international panel of 20 experts with experience/knowledge in the field of UK digital mental health. RESULTS: Sixty-four items were generated, of which 41 achieved consensus (64%). Consensus emerged around ten recommendations, falling into five main themes: Enhancing the quality of adverse events data in DMHIs; Re-defining serious adverse events for DMHIs; Reassessing short-term symptom deterioration in psychological interventions as a therapeutic risk; Maximising the benefit of the Yellow Card Scheme; and Developing a harmonised approach for assessing the safety of psychological interventions in general. CONCLUSION: The implementation of the recommendations provided by this consensus could improve the assessment of safety of DMHIs, making them more effective in detecting and mitigating risk.

The #longcovid revolution: A reflexive thematic analysis.

Research has identified long COVID as the first virtual patient-made condition (Callard and Perego, 2021). It originated from Twitter users sharing their experiences using the hashtag #longcovid. Over the first two years of the pandemic, long COVID affected as many as 17 million people in Europe (WHO, 2023). This study focuses on the initial #longcovid tweets in 2020 (as previous studies have focused on 2021-2022), from the first tweet in May to August 2020, when the World Health Organization recognised the condition. We collected over 31,000 tweets containing #longcovid from Twitter. Using Braun and Clarke's reflexive thematic analysis (2020), informed by the first author's experience of long COVID and drawing on Ian Hacking's perspective on social constructionism (1999), we identified different grades of social constructionism in the tweets. The themes we generated reflected that long COVID was a multi-system, cyclical condition initially stigmatised and misunderstood. These findings align with existing literature (Ladds et al., 2020; Rushforth et al., 2021). We add to the existing literature by suggesting that Twitter users raised awareness of long COVID by providing social consensus on their long COVID symptoms. Despite the challenge for traditional evidence-based medicine to capture the varied and intermittent symptoms, the social consensus highlighted that these variations were a consistent and collective experience. This social consensus fostered a collective social movement, overcoming stigma through supportive tweets and highlighting their healthcare needs using #researchrehabrecognition. The #longcovid movement's work was revolutionary, as it showed a revolutionary grade of social constructionism, because it brought about real-world change for long COVID sufferers in terms of recognition and the potential for healthcare provisions. Twitter users' accounts expose the limitations of traditional evidence-based medicine in identifying new conditions. Future research on novel conditions should consider various research paradigms, such as Evidence-Based Medicine Plus (Greenhalgh et al., 2022).

Preventive digital mental health interventions for children and young people: a review of the design and reporting of research.

Digital health interventions (DHIs) have frequently been highlighted as one way to respond to increasing levels of mental health problems in children and young people. Whilst many are developed to address existing mental health problems, there is also potential for DHIs to address prevention and early intervention. However, there are currently limitations in the design and reporting of the development, evaluation and implementation of preventive DHIs that can limit their adoption into real-world practice. This scoping review aimed to examine existing evidence-based DHI interventions and review how well the research literature described factors that researchers need to include in their study designs and reports to support real-world implementation. A search was conducted for relevant publications published from 2013 onwards. Twenty-one different interventions were identified from 30 publications, which took a universal (n = 12), selective (n = 3) and indicative (n = 15) approach to preventing poor mental health. Most interventions targeted adolescents, with only two studies including children aged ≤10 years. There was limited reporting of user co-design involvement in intervention development. Barriers and facilitators to implementation varied across the delivery settings, and only a minority reported financial costs involved in delivering the intervention. This review found that while there are continued attempts to design and evaluate DHIs for children and young people, there are several points of concern. More research is needed with younger children and those from poorer and underserved backgrounds. Co-design processes with children and young people should be recognised and reported as a necessary component within DHI research as they are an important factor in the design and development of interventions, and underpin successful adoption and implementation. Reporting the type and level of human support provided as part of the intervention is also important in enabling the sustained use and implementation of DHIs.