Artificial-Intelligence informed exercise prescriptions in primary care: Perspectives from people with long-term conditions, their carers, and healthcare professionals

Artificial-Intelligence informed exercise prescriptions in primary care: Perspectives from people with long-term conditions, their carers, and healthcare professionals | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Artificial-Intelligence informed exercise prescriptions in primary care: Perspectives from people with long-term conditions, their carers, and healthcare professionals Jacob Keast, Lucy Smith, Suzan Ghannam, Hajira Dambha-Miller This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7500570/v1 This work is licensed under a CC BY 4.0 License Status: Under Revision Version 1 posted 10 You are reading this latest preprint version Abstract Background Exercise is an important component of managing Long-Term Conditions (LTCs) and can improve health outcomes. Artificial intelligence (AI)-based exercise prescription software has the potential to support exercise amongst people with LTC. However, its adoption in primary care requires consideration of the perspectives of patients, carers, and healthcare professionals, particularly in relation to benefits, concerns, and practical issues for implementation. Methods A qualitative study was conducted with four online focus groups. This included two groups of people living with LTCs and/or their informal carers (n = 7), and two groups with healthcare professionals (n = 6). Semi-structured topic guides were used, and discussions were transcribed verbatim, and analysed thematically following Braun and Clarke’s six-phase framework. Results Six themes were identified: (1) understanding and trusting AI in physical activity tools; (2) positioning AI tools within everyday primary care; (3) reaching and engaging those at risk of being overlooked; (4) making AI a practical partner in self-management; (5) the role of AI in health and physical fitness; and (6) personalisation and safety in AI. Participants recognised benefits including reduced administrative burden, integration with existing systems, and personalised motivational feedback. Concerns were raised about safety, governance, and data transparency. Adoption was seen as reliant on clinician and patient buy-in, minimal workflow disruption, clear referral pathways, and hybrid human–digital support. Inclusive, user-friendly design—such as multilingual options, large fonts, and simplified interfaces—was considered essential to address barriers such as language needs, digital literacy, and socioeconomic constraints. Prior experiences with apps and wearables influenced expectations, with fragmented systems driving calls for a centralised hub. Conclusions AI-based exercise prescription software could enhance exercise support for people with LTCs if designed with transparency, safety, and personalisation. Effective integration into primary care will need to streamline workflows, provide hybrid human–digital support, and use inclusive design to engage diverse users. Artificial Intelligence primary care general practice exercise prescription digital health chronic disease management qualitative study Background Exercise has long been recognised as a key determinant of health, with evidence of benefit in many Long-Term Conditions (LTCs) such as cardiovascular and respiratory diseases, metabolic syndrome, dementia, amongst others [ 1 – 5 ]. For example, regular exercise has been shown to reduce systolic blood pressure by up to 11mmHg, and have a similar or greater efficacy as psychotherapy in the management of depression or anxiety [ 6 , 7 ], in addition to improving function, quality of life, and clinical outcomes [ 8 ]. Exercise prescriptions are an additional strategy in primary care and rehabilitation to promote more structured, evidence-based exercise engagement and activity. However, planning exercise prescriptions to support specific, sometimes multiple LTC whilst accounting for general health status, age, lifestyle, and activity history is complex and challenging [ 9 , 10 ]. Further, use of exercise prescriptions in primary care is variable despite its known benefits, often due to constraints on consultation time, competing priorities, and inconsistent access to exercise specialists [ 11 ]. Clinicians also report variable confidence and training in exercise prescriptions, contributing to fragmented and inconsistent provision [ 12 ]. The use of artificial intelligence (AI) in exercise prescriptions is an emerging innovation, that combines patient-specific data with algorithmic processing to deliver more streamlined and personalised exercise recommendations [ 13 ]. AI-assisted tools are increasingly considered as a means of supporting people with LTC with a specialist standard of support, by providing exercise recommendations that are responsive to an individual’s current health status, functional ability, and personal health or fitness goals. By adjusting plans dynamically, AI tools are capable of delivering more precise and sustainable interventions than static prescriptions. AI-based platforms have been proposed as a way to streamline exercise assessment, automate elements of an exercise prescription, provide real-time feedback, and integrate with existing clinical systems [ 14 , 15 ]. Early evaluations indicate technical feasibility and positive user perceptions, particularly when behaviour change techniques such as goal setting, reminders, and progress feedback are embedded [ 16 ]. These features align with current NHS priorities for digital innovation, prevention, and personalised care [ 14 ]. Despite growing interest, limited qualitative research has explored how AI-driven exercise prescription tools are perceived by those living with LTCs, their informal carers, and also the healthcare professionals who will be using or recommending them. Earlier studies in related fields highlight uncertainties around integration into current practice, the preservation of patient–practitioner relationships, and accountability for AI-generated recommendations [ 17 ]. Digital exercise tools more broadly, have shown mixed results for adherence and engagement, with evidence emphasising the importance of co-design to promote sustained use and meaningful outcomes [ 18 , 19 ]. In this study, we address this gap by exploring the perspectives of people with LTCs, their carers, and primary care professionals on the potential use of AI-based exercise prescription software. We focus on perceived benefits, concerns, and practical considerations for implementation and adoption within primary care. Methods Study design A qualitative study with online focus groups. Recruitment and Participants Participants included individuals with LTCs, informal carers, and professionals in health, social care, and exercise prescription delivery. Recruitment involved direct invitations via email or telephone to people who had previously expressed interest in research participation in earlier MLTC studies, snowball sampling through participant networks, and targeted outreach to relevant professionals via professional networks supported by an electronic poster that was circulated. Interested individuals completed a Qualtrics survey capturing demographic details, professional background, LTC status (if applicable), and—for informal carers—confirmation of their caregiving role. Eligible participants were aged ≥ 18 years and either diagnosed with an LTC, employed in health or social care, or providing informal care for someone with an LTC, and able to give informed written consent. We used purposive sampling to capture diversity in experience, background, and roles. Participants received an electronic consent form to complete in advance, including agreement to audio-recording. Ethics approval and consent to participate Ethical approval was obtained through the University of Southampton Faculty of Medicine ethics committee ERGO reference: 105919. All participants provided written consent to participate. Informed consent was obtained from all participants, and we adhered to the declaration of Helsinki . Data collection and analysis Focus groups were conducted online via Microsoft Teams (moderator: LS), recorded, and transcribed verbatim using the platform’s transcription function. Transcripts were anonymised and imported into Nvivo version 15.2.1 for data management and analysis. Semi-structured topic guides were tailored for either patient/carer or professional groups, with language adapted to ensure clarity and accessibility for all participants. Discussion topics included: Perceived benefits and limitations of AI-based exercise software Potential of digital exercise prescription platforms using an illustrative example (VITOVA) [20] Integration with clinical systems and workflows Usability, implementation and digital access Challenges and concerns in AI adoption Barriers affecting specific populations Thematic analysis followed Braun and Clarke’s six-phase framework: familiarisation, coding, theme development, theme review, theme definition, and reporting [21]. Coding was inductive and grounded in participants’ accounts, with LS and SG undertaking primary coding, and HDM providing independent review. The analysis was conducted reflexively, acknowledging the interpretive role of the researchers in shaping themes rather than treating them as passively “emerging” from the data. This approach, consistent with Braun and Clarke’s guidance, treats thematic analysis as a theoretically flexible yet rigorous method for identifying patterns of shared meaning across the dataset, linked by a central organising concept. Thematic analysis was chosen over other approaches because the aim was not to generate new theory or to apply a predetermined coding structure, but to provide a rich, detailed, and flexible account of stakeholder perspectives. Credibility and trustworthiness were strengthened through iterative discussions of coding frameworks, peer debriefing within the research team, and the maintenance of an audit trail documenting analytic decisions. Member checking was undertaken to enable participants to confirm, challenge, and refine the researchers’ interpretations of the data. We used the SRQR reporting checklist when editing to offer clarity on review of the manuscript, included in the supplementary materials [22,23]. The datasets generated and/or analysed during the current study are not publicly available due consent not being explicitly gained from each participant, but are available from the corresponding author on reasonable request subject to ethics approval and data sharing agreements. Results Participant characteristics A total of four focus groups were conducted: two groups with patients and informal carers (n = 7) and two groups with healthcare professionals (n = 6). Participant characteristics are described in Table 1 and the LTC reported amongst those with LTCs included: Hypertension, Hypercholesterolaemia, Peripheral Neuropathy, Fibromyalgia, Chronic Fatigue Syndrome, Bechet’s Disease, Chronic Migraines, Functional Neurological Disorder, Depression, Anxiety, and Osteoarthritis. Table 1: Summary of participant characteristics Professionals (n=6) Patient/Carer (n=7) Total (n=13) Age (SD) 32 (10) 56 (14) 44 (16) Gender (%) M F 3 (50%) 3 (50%) 3 (43%) 4 (57%) 6 (46%) 7 (54%) Level of Education (%) Secondary Tertiary Unknown 0 (0%) 5 (83%) 1 (17%) 1 (14%) 3 (43%) 3 (43%) 1 (8%) 8 (61%) 4 (31%) Ethnicity (%) White Asian 5 (83%) 1 (17%) 6 (86%) 1 (14%) 11 (85%) 2 (15%) Thematic analysis generated six interconnected themes capturing perspectives on the role of AI-enabled exercise prescriptions in the context of LTCs. These encompassed understandings of, and trust in, AI technologies; integration into routine primary care; strategies to reach and engage diverse groups; features to support self-management; the influence of previous experiences with digital fitness tools; and the importance of personalisation and safety. Theme 1: Understanding and trusting AI in physical activity Participants described a spectrum of interpretations of AI. For some, it conjured futuristic images of robots or autonomous systems; for others, it was already present in everyday devices such as smartphones, voice assistants, and wearables. Some questioned whether the illustrative case study represented “real AI” or was simply a rules-based matching system: “I think, well, do we really need AI or not? … it’s not really AI because all you’re doing is matching things to certain things and coming up with something else.” (FG3, P8, patient/carer) Others identified clear value, particularly if AI could integrate multiple data sources, automate routine information retrieval, and enhance communication: “…you can maybe say to AI, ‘What’s my referral status?’ and it could pull up that information from different sources once you’ve given it permission.” (FG3, P10, patient/carer) Across all groups, trust hinged on transparency over data access, storage, and sharing, and on alignment with national guidance for physical activity and LTCs. Healthcare professionals highlighted governance and accountability concerns, particularly the potential for “hallucinations” to generate unsafe advice: “If there’s harm due to hallucination, who would be accountable? … in healthcare, there’s a healthcare professional who is accountable.” (FG2, P5, professional) Fears about misuse of personal activity data, for example by insurers, were also raised: “If I don’t control it, I don’t trust it.” (FG4, patient/carer) Theme 2: Positioning AI prescriptions within everyday primary care Embedding AI- based exercise prescriptions within existing care pathways was seen as essential for uptake and sustained use. Clinicians emphasised the need for minimal disruption to workflows, clear referral pathways, and short, focused training to demonstrate functionality and benefits: “…the conversation [about] exercise is very limited… the app needs to be very simple summary statistics.” (FG2, P5, professional) Both groups highlighted the importance of human support alongside AI. The inclusion of monthly contact with a health coach was well received: “…it will be somebody who would support you and get in touch with you… something that can motivate people to continue.” (FG1, P3, patient/carer) Professionals suggested linking AI proposed exercise prescription outputs to existing patient records for oversight, while patients stressed the need for reassurance that recommendations were reviewed by a trusted clinician. Referral routes were debated—self-referral was viewed as empowering, but clinician referral was favoured for people with complex LTCs to ensure safe onboarding: “…should this be downloadable before they see a clinician or… only ever come afterwards as a recommendation?” (FG2, P4, professional) Carer involvement was suggested as a means of supporting adherence, though some participants voiced concerns about potential impacts on independence and privacy. Theme 3: Reaching and engaging those at risk of being overlooked Participants identified multiple barriers to engagement with AI based exercise prescriptions, including language needs, low digital literacy, cognitive impairments, LTC-related fatigue, rural location, limited or unstable internet access, low income, limited social support, and cultural scepticism toward technology: “…communication needs… making it as accessible as possible because you don’t want to be excluding people.” (FG1, P3, patient/carer) “People… have a preconceived idea of tech… that will be a barrier.” (FG3, patient/carer) To address these, participants recommended designing accessibility in from the outset. Suggestions included hybrid delivery (digital and print), large fonts, high-contrast displays, voice recognition, offline functionality, multilingual content, and simplified interfaces: “…printing out also works really well… and… having different languages… because English is not the first language in my practice.” (FG4, P13, professional) Visual guidance such as animations or diagrams was preferred by some over text-heavy instructions, and culturally relevant examples were seen as important for engagement. Healthcare professionals proposed “digital champions” within primary care teams to help patients navigate the new technology. Theme 4: Making AI-based exercise prescriptions a practical partner in self-management People living with LTCs frequently described feeling overwhelmed by health-related tasks. AI tools were seen as valuable when they reduced administrative burden, integrated with existing systems, and avoided duplication: “…better if it can actually be combined into one… because if not, you have hundreds of apps… it’s not a chore, it’s not a pressure.” (FG3, P7, patient/carer) Clinicians noted that integration with electronic health records could automate data entry and reduce manual tracking. Motivation and self-efficacy were regarded as critical to sustained use, with personalised progress updates and clinician feedback loops viewed as helpful for maintaining engagement: “…great to have oversight with multiple healthcare professionals… just as you would with the patient record.” (FG4, professional) Safeguards were seen as essential to prevent harm from overexertion: “…there would have to be some safeguards in place to stop people overdoing it… physically or mentally.” (FG2, P4, professional) Theme 5: Role of AI in health and physical fitness Expectations for AI were shaped by prior experiences with digital health and fitness tools, including MyFitnessPal, NHS online services, wearable trackers, military fitness apps, and physiotherapy platforms. These tools were valued for structure, accountability, progress tracking, and convenience, but some participants preferred in-person classes for the social benefits: “…you get more from the gym because you’re with people, not just in your room.” (FG2, patient/carer) Limitations cited included drop-off in use over time, subscription costs, poor integration between systems, and the isolating nature of home-based exercise. Wearables and physiotherapist-linked video plans were seen as effective by some, but juggling multiple platforms led to “app fatigue.” Participants called for a single, centralised hub combining exercise plans, wearable data, dietary tracking, medication reminders, and health records. Theme 6: Personalisation and safety in AI-based prescriptions Personalisation was repeatedly emphasised as a strength of AI, provided it was implemented effectively. Participants expected AI to adapt to daily fluctuations in health, particularly for conditions such as ME/CFS, diabetes, and arthritis: “Some days I can do more, some days nothing — it needs to ask, ‘How are you today?’ first.” (FG3, patient/carer) Safe prescription required integration of comorbidity profiles, medication schedules, blood sugar data, and mobility limitations. AI that could “know your body” like a trusted physiotherapist and tailor recommendations accordingly was viewed as ideal. Inclusive design principles—clear, uncluttered interfaces; supportive, non-judgemental language; multiple language options; and offline access—were regarded as critical for long-term engagement: “A daily, ‘you’re doing really well… 10% ahead this week,’ not, ‘you’re fat.’” (FG1, patient/carer) Integration with trusted healthcare platforms such as the NHS App or existing physiotherapy systems was seen as increasing credibility and uptake: “If it’s in the NHS App it might be easier to trust and use.” (FG2, professional). Discussion In this study we aimed to explore the perspectives of people with LTCs, their informal carers, and healthcare professionals on the use of AI-based exercise prescription software, focusing on perceived benefits, concerns, and practical considerations for implementation and wide-scale adoption. Six themes emerged: understanding and trusting AI in physical activity tools; positioning AI tools within everyday primary care; reaching and engaging those at risk of being overlooked; making AI a practical partner in self-management; the role of AI-based prescription in health and physical fitness; and personalisation and safety in AI. Participants expressed varied understandings of what qualifies as AI in the context of exercise prescription. Some viewed the case study example as little more than a rules-based system, while others recognised the novelty of AI tools capable of integrating multiple health and lifestyle data streams to tailor prescriptions dynamically. This uncertainty reflects a broader lack of clarity in healthcare about what constitutes “AI” and may undermine trust when patients and professionals are unclear about how personalised exercise recommendations are generated. Concerns about safety—particularly the possibility of AI “hallucinations” producing inaccurate or unsafe prescriptions—were directly linked to calls for clear governance, alignment with NICE recommendations for physical activity and LTCs, and explicit clinical accountability. These findings are consistent with literature emphasising that transparent processes, robust oversight, and professional responsibility are essential for safe use of AI in healthcare [24]. Positioning AI-based exercise prescriptions within primary care was seen as contingent on clinician and patient buy-in, seamless integration into existing workflows, and minimal disruption during time-limited consultations. Participants emphasised the importance of concise, actionable summaries that could support exercise conversations, targeted training for clinicians to understand AI-generated outputs, and continued human oversight to validate and adapt prescriptions where necessary. The value placed on blended models—AI-generated prescriptions supported by periodic human contact—reinforces evidence that fully automated solutions are less acceptable in primary care. Reaching and engaging those at risk of being overlooked was seen as a critical challenge for equitable adoption of AI-driven exercise prescriptions. Barriers such as language needs, limited digital literacy, socioeconomic disadvantage, unstable internet, and cultural scepticism were viewed as likely to exclude many with LTCs who might benefit most. These reflect the well-documented dimensions of the digital divide and the risk of deepening health inequalities if left unaddressed [25,26]. Participants advocated for inclusive design strategies such as multilingual interfaces, printable prescription summaries, voice navigation, simplified versions for low digital literacy, and co-design with seldom-heard groups. Such approaches are consistent with inclusive digital health design principles promoted by NHS Digital and the World Health Organisation [14,27]. Making AI-based prescriptions a practical partner in self-management was tied to reducing, not adding to, the workload of managing LTCs. Participants highlighted automation of monitoring tasks, interoperability with existing health records, and personalised exercise plans that adapted to progress and capacity as essential to avoid duplication and “app fatigue.” These findings resonate with treatment burden theory, which emphasises balancing workload with capacity to sustain engagement [28]. Motivation and self-efficacy were seen as crucial, with requests for embedded behaviour change techniques, progress feedback, and clinician monitoring. Such strategies align with existing behavioural science frameworks. Prior experiences with apps and wearables also shaped expectations of AI-enabled exercise prescriptions. While such tools were appreciated for providing structure and accountability, frustrations over fragmented systems and subscription costs led to calls for a centralised hub that could integrate exercise, health, and lifestyle data. This aligns with wider digital health research emphasising the need for interoperable, user-centred platforms to support long-term engagement. Finally, personalisation and safety were regarded as non-negotiable features of AI-based exercise prescriptions. Participants expected tools to adapt to daily health fluctuations and to account for comorbidities, medications, and mobility limitations, echoing principles of precision health [29]. They stressed the importance of inclusive, user-friendly design and integration with trusted platforms such as the NHS App to enhance credibility, uptake, and sustained use. Comparison to existing literature The six themes highlighted in this study align with, and extend, the growing literature on AI-assisted digital tools in exercise prescription for LTCs. As previously emphasised, AI-assisted digital tools hold strong potential for personalised exercise prescription, overcoming many challenges faced by generic and non-specialist support. But these tools must overcome recognised barriers, akin to those expressed in previous studies where professional and patient trust, AI transparency, and clinical safety are important concerns in relation to “hallucinations” and accountability [30]. Similar to Shawli et al. (2024), healthcare professionals in this study expressed scepticism toward AI’s role in rehabilitation, highlighting the need for governance and human oversight, alongside the transparent development of AI-assisted tools [16]. Participants’ call for embedding AI within primary care and linking outputs to existing records resonates with NHS priorities for digital integration and prevention outlined in the Long Term Plan [14]. The previously identified barriers of rurality, deprivation, and digital literacy, reinforce the importance of co-design and reaching those at risk of being overlooked, developed here with the suggestion of multilingual tools and hybrid delivery of exercise support. Complementing these findings, recent systematic reviews provide additional context demonstrating that digital health interventions can reduce sedentary behaviour among people with LTCs, though outcomes varied across populations [31], while others highlight the effectiveness of digital tools in maintaining physical activity but noted limitations in evidence quality and long-term follow-up [32]. Wilson et al. (2024) further emphasised that advancing digital health equity requires tackling structural and contextual barriers [33]; our participants’ recommendations for multilingual interfaces and hybrid delivery offer practical strategies to operationalise such equity within exercise support. Together, these insights position AI-assisted tools as promising but conditional on being transparent, clinically safe, and equitably accessible. Strengths and limitations A key strength of this study is the inclusion of multiple stakeholder perspectives, incorporating both lived experience of LTCs and professional viewpoints from providers. This enabled triangulation of insights, adding depth and enhancing credibility. The purposive sampling approach ensured diversity in condition types, professional roles, and care contexts, allowing exploration of a broad range of experiences. Conducting focus groups online increased accessibility for participants with mobility challenges, caring responsibilities, or geographical constraints, potentially reducing barriers to participation. The use of thematic analysis with independent coding review, combined with member checking, strengthened the rigour and trustworthiness of the findings. However, several limitations should be acknowledged. The sample size was modest (n=13) and drawn from a limited geographical area, which may constrain transferability to other settings. However, as is typical in qualitative research, the focus was on achieving depth and richness of insight rather than numerical representativeness, allowing detailed exploration of participants’ experiences and perspectives. Participants were self-selecting and may have had a pre-existing interest in digital health, potentially leading to over-representation of more digitally literate or positively inclined individuals. Those most at risk of digital exclusion such as those with limited internet access, low literacy, or high social vulnerability, may have been under-represented, limiting the study’s ability to fully capture the breadth of barriers faced by these groups. The online format, while accessible for some, may have excluded individuals without adequate digital skills or resources. Conclusion and implications for policy makers Our findings suggest that AI-based exercise prescription tools could enhance physical activity support for people with LTCs, provided they are developed and implemented with transparency, alignment to evidence-based guidelines, clinical oversight, and inclusive design. Successful adoption in primary care will require minimal disruption to existing workflows, clearly defined referral pathways, and hybrid delivery models that accommodate differing levels of digital literacy and access. Future work should prioritise co-production with larger, more diverse user groups, including those most at risk of digital exclusion, and rigorously evaluate the safety, acceptability, and effectiveness of such tools in real-world healthcare settings. Declarations Funding HDM receives funding from the National Institute for Health and Care Research (NIHR) Multiple Long-Term Conditions (MLTC) Cross NIHR Collaboration (CNC) (NIHR207000) and the NIHR Artificial Intelligence for Multiple Long-Term Conditions (AIM) programme (NIHR202637). The views expressed in this publication are those of the author(s) and not necessarily those of the NHS, the NIHR or the Department of Health and Social Care. Competing Interests None declared. Acknowledgements We would like to thank all the patients and members of the public who contributed to the focus group. Consent for publication Not applicable References Kraus WE, Powell KE, Haskell WL, Janz KF, Campbell WW, Jakicic JM, et al. Physical Activity, All-Cause and Cardiovascular Mortality, and Cardiovascular Disease. Med Sci Sports Exerc. 2019; 51(6):1270–81. doi: 10.1249/MSS.0000000000001939. Rejbi IB, Trabelsi Y, Chouchene A, Ben Turkia W, Ben Saad H, Zbidi A, et al. Changes in six-minute walking distance during pulmonary rehabilitation in patients with COPD and in healthy subjects. Int J Chron Obstruct Pulmon Dis. 2010; 5:209–15. Epub 20100809. doi: 10.2147/copd.s7955. Activity against health risks associated with type 1 diabetes: “Health benefits outweigh the risks”. World J Diabetes. 2022; 13(3):161–84. doi: 10.4239/wjd.v13.i3.161. Pearce M, Garcia L, Abbas A, Strain T, Schuch FB, Golubic R, et al. Association Between Physical Activity and Risk of Depression: A Systematic Review and Meta-analysis. JAMA Psychiatry. 2022; 79(6):550–9. doi: 10.1001/jamapsychiatry.2022.0609. Jia RX, Liang JH, Xu Y, Wang YQ. Effects of physical activity and exercise on the cognitive function of patients with Alzheimer disease: a meta-analysis. BMC Geriatr. 2019; 19(1):181. Epub 20190702. doi: 10.1186/s12877-019-1175-2. Singh B, Olds T, Curtis R, et alEffectiveness of physical activity interventions for improving depression, anxiety and distress: an overview of systematic reviews. British Journal of Sports Medicine. 2023;57:1203-1209. doi: 10.1136/bjsports-2022-106195. Börjesson M, Onerup A, Lundqvist S, Dahlöf B. Physical activity and exercise lower blood pressure in individuals with hypertension: narrative review of 27 RCTs. Br J Sports Med. 2016 Mar;50(6):356-61. doi: 10.1136/bjsports-2015-095786. Dergaa I, Ben Saad H. Using artificial intelligence for exercise prescription in personalised health promotion: A critical evaluation of OpenAI's GPT-4 model. J Med Internet Res. 2024;26:e51308. doi:10.2196/51308. Rooney D, Gilmartin E, Heron N. Prescribing exercise and physical activity to treat and manage health conditions. Ulster Med J. 2023; 92(1):9–15. Festa RR, Jofré-Saldía E, Candia AA, et al. Next steps to advance general physical activity recommendations towards physical exercise prescription: a narrative review: BMJ Open Sport & Exercise Medicine 2023;9:e001749. Pedersen BK, Saltin B. Exercise as medicine – evidence for prescribing exercise as therapy in 26 different chronic diseases. Scand J Med Sci Sports. 2015;25 Suppl 3:1–72. doi:10.1111/sms.12581. Moore GF, Audrey S, Barker M, et al. Process evaluation of complex interventions: Medical Research Council guidance. BMJ. 2015;350:h1258. doi:10.1136/bmj.h1258 Tanaka H, Miyachi M, Nakamura Y. Artificial intelligence in exercise prescription: current applications and future directions. Sports Med. 2021;51(4):545–557. doi:10.1007/s40279-020-01382-9. NHS England. The NHS Long Term Plan. London: NHS England; 2019. Available from: https://www.longtermplan.nhs.uk/. Strain T, Milton K, Dall PM, et al. The impact of rurality and deprivation on physical activity and sedentary behaviour in the UK: a cross-sectional analysis from the Health Survey for England. BMJ Open. 2021;11(8):e048270. doi:10.1136/bmjopen-2020-048270. Shawli L, Alsobhi M, Chevidikunnan MF et al. Physical therapists’ perceptions and attitudes towards artificial intelligence in healthcare and rehabilitation: A qualitative study. Musculoskeletal Science and Practice. 2024;73:103152. Doi.org/10.1016/j.msksp.2024.103152. Argent R, Daly A, O’Leary E. Patient Involvement With Home-Based Exercise Programs: Can Connected Health Interventions Influence Adherence? JMIR mhealth uhealth. 2018;6(3):e47. doi: 10.2196/mhealth.8518. Teo JL, Zheng Z, Bird SR. Identifying the factors affecting ‘patient engagement’ in exercise rehabilitation. BMC Sports Science, Medicine, Rehabilitation. 2022; 14:18. doi.org/10.1186/s13102-022-00407-3. Vassilakopoulou, P., Hustad, E. Bridging Digital Divides: a Literature Review and Research Agenda for Information Systems Research. Inf Syst Front 2023; 25, 955–969. https://doi.org/10.1007/s10796-020-10096-3. VITOVA. (n.d.). Retrieved August 16, 2025. Available from: https://vitovaltd.com/ Braun, V., & Clarke, V. Reflecting on reflexive thematic analysis. Qualitative Research in Sport, Exercise and Health, 2019;11(4), 589–597. https://doi.org/10.1080/2159676X.2019.1628806. O’Brien BC, Harris IB, Beckman TJ, etal. Standards for reporting qualitative research: A synthesis of recommendations. Academic Medicine [Internet]. 2014;89(9):1245–51. Available from: https://journals.lww.com/academicmedicine/fulltext/2014/09000/Standards_for_Reporting_Qualitative_Research__A.21.aspx O’Brien BC, Harris IB, Beckman TJ, et al. The SRQR reporting checklist. The EQUATOR network reporting guideline platform [Internet]. The UK EQUATOR Centre; 2025. Available from: https:/resources.equator-network.org/reporting-guidelines/srqr/srqr-checklist.docx Saenz AD, Mass General Brigham AI Governance Committee, Centi A. et al. Establishing responsible use of AI guidelines: a comprehensive case study for healthcare institutions. npj Digit. Med. 2024;7, 348. https://doi.org/10.1038/s41746-024-01300-8 Gunkel, DJ. Second thoughts: toward a critique of the digital divide. New Media & Society. 2003; 5(4), 499–522. Choudrie J, Pheeraphuttranghkoon, S., & Davari, S. The digital divide and older adult population adoption, use and diffusion of mobile phones: a quantitative study. Information Systems Frontiers. 2018; 22, 673–695. https://doi.org/10.1007/s10796-018-9875-2. Murphy K, Di Ruggiero E, Upshur R, Willison DJ, Malhotra N, Cai JC, Malhotra N, Lui V, Gibson J. Artificial intelligence for good health: a scoping review of the ethics literature. BMC Med Ethics. 2021;22(1):1–7. May C, Eton DT, Boehmer K, et al. Rethinking the patient: Using Burden of Treatment Theory to understand the changing dynamics of illness. BMC Health Services Research. 2014;14, 281. https://doi.org/10.1186/1472-6963-14-281 Denny JC, Collins FS, Mehta JP. Precision health: The next generation of personalized medicine. The Lancet. 2021;398(10318), 1417–1426. https://doi.org/10.1016/S0140-6736(21)01808-0. Ahmed MI, Spooner B, Isherwood J, et al. A Systematic Review of the Barriers to the Implementation of Artificial Intelligence in Healthcare. Cureus. 2023; 4:15(10):e46454. doi: 10.7759/cureus.46454. Zhang Y, Ngai F, Yang Q, Xie Y Effectiveness of Digital Health Interventions on Sedentary Behavior Among Patients With Chronic Diseases: Systematic Review and Meta-Analysis JMIR Mhealth Uhealth 2025;13:e59943 DOI: 10.2196/59943 Howes S, Stephenson A, Grimmett C, Argent R, Clarkson P, Khan A, Lait E, McDonough LR, Tanner G, McDonough SM. The effectiveness of digital tools to maintain physical activity among people with a long-term condition(s): A systematic review and meta-analysis. Digit Health. 2024 Dec 20;10:20552076241299864. doi: 10.1177/20552076241299864. Wilson, S., Tolley, C., Mc Ardle, R. et al. Recommendations to advance digital health equity: a systematic review of qualitative studies. npj Digit. Med. 7, 173 (2024). https://doi.org/10.1038/s41746-024-01177-7 Additional Declarations No competing interests reported. Cite Share Download PDF Status: Under Revision Version 1 posted Editorial decision: Revision requested 12 Nov, 2025 Reviews received at journal 03 Nov, 2025 Reviewers agreed at journal 25 Oct, 2025 Reviews received at journal 02 Oct, 2025 Reviewers agreed at journal 11 Sep, 2025 Reviewers agreed at journal 11 Sep, 2025 Reviewers invited by journal 11 Sep, 2025 Editor assigned by journal 10 Sep, 2025 Submission checks completed at journal 10 Sep, 2025 First submitted to journal 10 Sep, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7500570","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":513813598,"identity":"9af2f0c0-a2db-41c5-83cd-c22af3755e59","order_by":0,"name":"Jacob Keast","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA1UlEQVRIiWNgGAWjYBACxh72Awwf/9ggCfEQ1MKTwDizIY2BgY1YLUAFBsy8DYdJ0MLccyBNgnfHeXt++eZjDxhq7BgMzhwg4LDexmMSkmduJ85sY0s3YDiWzGBwtoGAln6GNAkDttsJBsd4zCQY2A4wGJwn4DCgFjOJBLZz9hAt/4jR0ttgJnGw7QDjBpAWxrYDRDis50yyZcOZZKBf0tIkEvuSeSQJed+wJ/3g7T8Vdvb8zIePSXz4ZifHdyaBgBYUVyQQEZEM8gRVjIJRMApGwSgAAOM0QKTSnjOdAAAAAElFTkSuQmCC","orcid":"","institution":"University of Southampton","correspondingAuthor":true,"prefix":"","firstName":"Jacob","middleName":"","lastName":"Keast","suffix":""},{"id":513813604,"identity":"5e87ada8-4035-4402-8547-a4b000a5cfe6","order_by":1,"name":"Lucy Smith","email":"","orcid":"","institution":"University of Southampton","correspondingAuthor":false,"prefix":"","firstName":"Lucy","middleName":"","lastName":"Smith","suffix":""},{"id":513813609,"identity":"dc42221a-c5fd-4f1d-ba32-b9249de9e5d2","order_by":2,"name":"Suzan Ghannam","email":"","orcid":"","institution":"University of Southampton","correspondingAuthor":false,"prefix":"","firstName":"Suzan","middleName":"","lastName":"Ghannam","suffix":""},{"id":513813613,"identity":"6e3ac045-61b4-4bef-b332-309b2aa18145","order_by":3,"name":"Hajira Dambha-Miller","email":"","orcid":"","institution":"University of Southampton","correspondingAuthor":false,"prefix":"","firstName":"Hajira","middleName":"","lastName":"Dambha-Miller","suffix":""}],"badges":[],"createdAt":"2025-08-31 12:38:11","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7500570/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7500570/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":91641864,"identity":"3815e4c7-aeef-480c-91dc-a9fe53cbaf6e","added_by":"auto","created_at":"2025-09-18 15:04:00","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":719817,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7500570/v1/13aeee41-c40d-4e59-a908-e4e238d580f8.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Artificial-Intelligence informed exercise prescriptions in primary care: Perspectives from people with long-term conditions, their carers, and healthcare professionals","fulltext":[{"header":"Background","content":"\u003cp\u003eExercise has long been recognised as a key determinant of health, with evidence of benefit in many Long-Term Conditions (LTCs) such as cardiovascular and respiratory diseases, metabolic syndrome, dementia, amongst others [\u003cspan additionalcitationids=\"CR2 CR3 CR4\" citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]. For example, regular exercise has been shown to reduce systolic blood pressure by up to 11mmHg, and have a similar or greater efficacy as psychotherapy in the management of depression or anxiety [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e], in addition to improving function, quality of life, and clinical outcomes [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]. Exercise prescriptions are an additional strategy in primary care and rehabilitation to promote more structured, evidence-based exercise engagement and activity. However, planning exercise prescriptions to support specific, sometimes multiple LTC whilst accounting for general health status, age, lifestyle, and activity history is complex and challenging [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e, \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]. Further, use of exercise prescriptions in primary care is variable despite its known benefits, often due to constraints on consultation time, competing priorities, and inconsistent access to exercise specialists [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. Clinicians also report variable confidence and training in exercise prescriptions, contributing to fragmented and inconsistent provision [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eThe use of artificial intelligence (AI) in exercise prescriptions is an emerging innovation, that combines patient-specific data with algorithmic processing to deliver more streamlined and personalised exercise recommendations [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. AI-assisted tools are increasingly considered as a means of supporting people with LTC with a specialist standard of support, by providing exercise recommendations that are responsive to an individual\u0026rsquo;s current health status, functional ability, and personal health or fitness goals. By adjusting plans dynamically, AI tools are capable of delivering more precise and sustainable interventions than static prescriptions. AI-based platforms have been proposed as a way to streamline exercise assessment, automate elements of an exercise prescription, provide real-time feedback, and integrate with existing clinical systems [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]. Early evaluations indicate technical feasibility and positive user perceptions, particularly when behaviour change techniques such as goal setting, reminders, and progress feedback are embedded [\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]. These features align with current NHS priorities for digital innovation, prevention, and personalised care [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]. Despite growing interest, limited qualitative research has explored how AI-driven exercise prescription tools are perceived by those living with LTCs, their informal carers, and also the healthcare professionals who will be using or recommending them. Earlier studies in related fields highlight uncertainties around integration into current practice, the preservation of patient\u0026ndash;practitioner relationships, and accountability for AI-generated recommendations [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]. Digital exercise tools more broadly, have shown mixed results for adherence and engagement, with evidence emphasising the importance of co-design to promote sustained use and meaningful outcomes [\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e, \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]. In this study, we address this gap by exploring the perspectives of people with LTCs, their carers, and primary care professionals on the potential use of AI-based exercise prescription software. We focus on perceived benefits, concerns, and practical considerations for implementation and adoption within primary care.\u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003eStudy design\u003c/h2\u003e\u003cp\u003eA qualitative study with online focus groups.\u003c/p\u003e\u003c/div\u003e\n\u003ch3\u003eRecruitment and Participants\u003c/h3\u003e\n\u003cp\u003eParticipants included individuals with LTCs, informal carers, and professionals in health, social care, and exercise prescription delivery. Recruitment involved direct invitations via email or telephone to people who had previously expressed interest in research participation in earlier MLTC studies, snowball sampling through participant networks, and targeted outreach to relevant professionals via professional networks supported by an electronic poster that was circulated. Interested individuals completed a Qualtrics survey capturing demographic details, professional background, LTC status (if applicable), and\u0026mdash;for informal carers\u0026mdash;confirmation of their caregiving role. Eligible participants were aged\u0026thinsp;\u0026ge;\u0026thinsp;18 years and either diagnosed with an LTC, employed in health or social care, or providing informal care for someone with an LTC, and able to give informed written consent. We used purposive sampling to capture diversity in experience, background, and roles. Participants received an electronic consent form to complete in advance, including agreement to audio-recording.\u003c/p\u003e\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate\u003c/strong\u003e\u003cbr\u003e\u0026nbsp;Ethical approval was obtained through the University of Southampton Faculty of Medicine ethics committee ERGO reference: 105919. All participants provided written consent to participate. Informed consent was obtained from all participants, and we adhered to the declaration of Helsinki .\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData collection and analysis\u003c/strong\u003e\u003cbr\u003e\u0026nbsp;Focus groups were conducted online via Microsoft Teams (moderator: LS), recorded, and transcribed verbatim using the platform\u0026rsquo;s transcription function. Transcripts were anonymised and imported into Nvivo version 15.2.1 for data management and analysis. Semi-structured topic guides were tailored for either patient/carer or professional groups, with language adapted to ensure clarity and accessibility for all participants. Discussion topics included:\u003c/p\u003e\n\u003cul\u003e\n \u003cli\u003ePerceived benefits and limitations of AI-based exercise software\u003c/li\u003e\n \u003cli\u003ePotential of digital exercise prescription platforms using an illustrative example (VITOVA) [20]\u003c/li\u003e\n \u003cli\u003eIntegration with clinical systems and workflows\u003c/li\u003e\n \u003cli\u003eUsability, implementation and digital access\u003c/li\u003e\n \u003cli\u003eChallenges and concerns in AI adoption\u003c/li\u003e\n \u003cli\u003eBarriers affecting specific populations\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eThematic analysis followed Braun and Clarke\u0026rsquo;s six-phase framework: familiarisation, coding, theme development, theme review, theme definition, and reporting [21]. Coding was inductive and grounded in participants\u0026rsquo; accounts, with LS and SG undertaking primary coding, and HDM providing independent review. The analysis was conducted reflexively, acknowledging the interpretive role of the researchers in shaping themes rather than treating them as passively \u0026ldquo;emerging\u0026rdquo; from the data. This approach, consistent with Braun and Clarke\u0026rsquo;s guidance, treats thematic analysis as a theoretically flexible yet rigorous method for identifying patterns of shared meaning across the dataset, linked by a central organising concept. Thematic analysis was chosen over other approaches because the aim was not to generate new theory or to apply a predetermined coding structure, but to provide a rich, detailed, and flexible account of stakeholder perspectives. Credibility and trustworthiness were strengthened through iterative discussions of coding frameworks, peer debriefing within the research team, and the maintenance of an audit trail documenting analytic decisions. Member checking was undertaken to enable participants to confirm, challenge, and refine the researchers\u0026rsquo; interpretations of the data. We used the SRQR reporting checklist when editing to offer clarity on review of the manuscript, included in the supplementary materials [22,23]. The datasets generated and/or analysed during the current study are not publicly available due consent not being explicitly gained from each participant, but are available from the corresponding author on reasonable request subject to ethics approval and data sharing agreements.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003e\u003cstrong\u003eParticipant characteristics\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eA total of four focus groups were conducted: two groups with patients and informal carers (n = 7) and two groups with healthcare professionals (n = 6). \u003cstrong\u003eParticipant characteristics are\u0026nbsp;\u003c/strong\u003edescribed in Table 1 and the LTC reported amongst those with LTCs \u0026nbsp; included: Hypertension, Hypercholesterolaemia, Peripheral Neuropathy, Fibromyalgia, Chronic Fatigue Syndrome, Bechet\u0026rsquo;s Disease, Chronic Migraines, Functional Neurological Disorder, Depression, Anxiety, and Osteoarthritis.\u003c/p\u003e\n\u003cp\u003e\u003cu\u003eTable 1: Summary of participant characteristics\u003c/u\u003e\u003c/p\u003e\n\u003cdiv align=\"center\"\u003e\n \u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 20%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eProfessionals\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e(n=6)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20%;\"\u003e\n \u003cp\u003e\u003cstrong\u003ePatient/Carer\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e(n=7)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTotal\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e(n=13)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 20%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eAge (SD)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20%;\"\u003e\n \u003cp\u003e32 (10)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20%;\"\u003e\n \u003cp\u003e56 (14)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20%;\"\u003e\n \u003cp\u003e44 (16)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 20%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eGender (%)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eM\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eF\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20%;\"\u003e\n \u003cp\u003e3 (50%)\u003c/p\u003e\n \u003cp\u003e3 (50%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20%;\"\u003e\n \u003cp\u003e3 (43%)\u003c/p\u003e\n \u003cp\u003e4 (57%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20%;\"\u003e\n \u003cp\u003e6 (46%)\u003c/p\u003e\n \u003cp\u003e7 (54%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 20%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eLevel of Education (%)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eSecondary\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eTertiary\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eUnknown\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20%;\"\u003e\n \u003cp\u003e0 (0%)\u003c/p\u003e\n \u003cp\u003e5 (83%)\u003c/p\u003e\n \u003cp\u003e1 (17%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20%;\"\u003e\n \u003cp\u003e1 (14%)\u003c/p\u003e\n \u003cp\u003e3 (43%)\u003c/p\u003e\n \u003cp\u003e3 (43%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20%;\"\u003e\n \u003cp\u003e1 (8%)\u003c/p\u003e\n \u003cp\u003e8 (61%)\u003c/p\u003e\n \u003cp\u003e4 (31%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 20%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eEthnicity (%)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eWhite\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eAsian\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20%;\"\u003e\n \u003cp\u003e5 (83%)\u003c/p\u003e\n \u003cp\u003e1 (17%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20%;\"\u003e\n \u003cp\u003e6 (86%)\u003c/p\u003e\n \u003cp\u003e1 (14%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20%;\"\u003e\n \u003cp\u003e11 (85%)\u003c/p\u003e\n \u003cp\u003e2 (15%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n\u003c/div\u003e\n\u003cp\u003eThematic analysis generated six interconnected themes capturing perspectives on the role of AI-enabled exercise prescriptions in the context of LTCs. These encompassed understandings of, and trust in, AI technologies; integration into routine primary care; strategies to reach and engage diverse groups; features to support self-management; the influence of previous experiences with digital fitness tools; and the importance of personalisation and safety.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTheme 1: Understanding and trusting AI in physical activity\u0026nbsp;\u003c/strong\u003e\u003cbr\u003e\u0026nbsp;Participants described a spectrum of interpretations of AI. For some, it conjured futuristic images of robots or autonomous systems; for others, it was already present in everyday devices such as smartphones, voice assistants, and wearables. Some questioned whether the illustrative case study represented \u0026ldquo;real AI\u0026rdquo; or was simply a rules-based matching system: \u0026ldquo;I think, well, do we really need AI or not? \u0026hellip; it\u0026rsquo;s not really AI because all you\u0026rsquo;re doing is matching things to certain things and coming up with something else.\u0026rdquo; (FG3, P8, patient/carer) Others identified clear value, particularly if AI could integrate multiple data sources, automate routine information retrieval, and enhance communication: \u0026ldquo;\u0026hellip;you can maybe say to AI, \u0026lsquo;What\u0026rsquo;s my referral status?\u0026rsquo; and it could pull up that information from different sources once you\u0026rsquo;ve given it permission.\u0026rdquo; (FG3, P10, patient/carer) Across all groups, trust hinged on transparency over data access, storage, and sharing, and on alignment with national guidance for physical activity and LTCs. Healthcare professionals highlighted governance and accountability concerns, particularly the potential for \u0026ldquo;hallucinations\u0026rdquo; to generate unsafe advice: \u0026ldquo;If there\u0026rsquo;s harm due to hallucination, who would be accountable? \u0026hellip; in healthcare, there\u0026rsquo;s a healthcare professional who is accountable.\u0026rdquo; (FG2, P5, professional) Fears about misuse of personal activity data, for example by insurers, were also raised: \u0026ldquo;If I don\u0026rsquo;t control it, I don\u0026rsquo;t trust it.\u0026rdquo; (FG4, patient/carer)\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTheme 2: Positioning AI prescriptions within everyday primary care\u003c/strong\u003e\u003cbr\u003e\u0026nbsp;Embedding AI- based exercise prescriptions within existing care pathways was seen as essential for uptake and sustained use. Clinicians emphasised the need for minimal disruption to workflows, clear referral pathways, and short, focused training to demonstrate functionality and benefits: \u0026ldquo;\u0026hellip;the conversation [about] exercise is very limited\u0026hellip; the app needs to be very simple summary statistics.\u0026rdquo; (FG2, P5, professional) Both groups highlighted the importance of human support alongside AI. The inclusion of monthly contact with a health coach was well received: \u0026ldquo;\u0026hellip;it will be somebody who would support you and get in touch with you\u0026hellip; something that can motivate people to continue.\u0026rdquo; (FG1, P3, patient/carer) Professionals suggested linking AI proposed exercise prescription outputs to existing patient records for oversight, while patients stressed the need for reassurance that recommendations were reviewed by a trusted clinician. Referral routes were debated\u0026mdash;self-referral was viewed as empowering, but clinician referral was favoured for people with complex LTCs to ensure safe onboarding: \u0026ldquo;\u0026hellip;should this be downloadable before they see a clinician or\u0026hellip; only ever come afterwards as a recommendation?\u0026rdquo; (FG2, P4, professional) Carer involvement was suggested as a means of supporting adherence, though some participants voiced concerns about potential impacts on independence and privacy.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTheme 3: Reaching and engaging those at risk of being overlooked\u003c/strong\u003e\u003cbr\u003e\u0026nbsp;Participants identified multiple barriers to engagement with AI based exercise prescriptions, including language needs, low digital literacy, cognitive impairments, LTC-related fatigue, rural location, limited or unstable internet access, low income, limited social support, and cultural scepticism toward technology: \u0026ldquo;\u0026hellip;communication needs\u0026hellip; making it as accessible as possible because you don\u0026rsquo;t want to be excluding people.\u0026rdquo; (FG1, P3, patient/carer) \u0026ldquo;People\u0026hellip; have a preconceived idea of tech\u0026hellip; that will be a barrier.\u0026rdquo; (FG3, patient/carer) To address these, participants recommended designing accessibility in from the outset. Suggestions included hybrid delivery (digital and print), large fonts, high-contrast displays, voice recognition, offline functionality, multilingual content, and simplified interfaces: \u0026ldquo;\u0026hellip;printing out also works really well\u0026hellip; and\u0026hellip; having different languages\u0026hellip; because English is not the first language in my practice.\u0026rdquo; (FG4, P13, professional) Visual guidance such as animations or diagrams was preferred by some over text-heavy instructions, and culturally relevant examples were seen as important for engagement. Healthcare professionals proposed \u0026ldquo;digital champions\u0026rdquo; within primary care teams to help patients navigate the new technology.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTheme 4: Making AI-based exercise prescriptions a practical partner in self-management\u003c/strong\u003e\u003cbr\u003e\u0026nbsp;People living with LTCs frequently described feeling overwhelmed by health-related tasks. AI tools were seen as valuable when they reduced administrative burden, integrated with existing systems, and avoided duplication: \u0026ldquo;\u0026hellip;better if it can actually be combined into one\u0026hellip; because if not, you have hundreds of apps\u0026hellip; it\u0026rsquo;s not a chore, it\u0026rsquo;s not a pressure.\u0026rdquo; (FG3, P7, patient/carer) Clinicians noted that integration with electronic health records could automate data entry and reduce manual tracking. Motivation and self-efficacy were regarded as critical to sustained use, with personalised progress updates and clinician feedback loops viewed as helpful for maintaining engagement: \u0026ldquo;\u0026hellip;great to have oversight with multiple healthcare professionals\u0026hellip; just as you would with the patient record.\u0026rdquo; (FG4, professional) Safeguards were seen as essential to prevent harm from overexertion: \u0026ldquo;\u0026hellip;there would have to be some safeguards in place to stop people overdoing it\u0026hellip; physically or mentally.\u0026rdquo; (FG2, P4, professional)\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTheme 5: Role of AI in health and physical fitness\u003c/strong\u003e\u003cbr\u003e\u0026nbsp;Expectations for AI were shaped by prior experiences with digital health and fitness tools, including MyFitnessPal, NHS online services, wearable trackers, military fitness apps, and physiotherapy platforms. These tools were valued for structure, accountability, progress tracking, and convenience, but some participants preferred in-person classes for the social benefits: \u0026ldquo;\u0026hellip;you get more from the gym because you\u0026rsquo;re with people, not just in your room.\u0026rdquo; (FG2, patient/carer) Limitations cited included drop-off in use over time, subscription costs, poor integration between systems, and the isolating nature of home-based exercise. Wearables and physiotherapist-linked video plans were seen as effective by some, but juggling multiple platforms led to \u0026ldquo;app fatigue.\u0026rdquo; Participants called for a single, centralised hub combining exercise plans, wearable data, dietary tracking, medication reminders, and health records.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTheme 6: Personalisation and safety in AI-based prescriptions\u003c/strong\u003e\u003cbr\u003ePersonalisation was repeatedly emphasised as a strength of AI, provided it was implemented effectively. Participants expected AI to adapt to daily fluctuations in health, particularly for conditions such as ME/CFS, diabetes, and arthritis: \u0026ldquo;Some days I can do more, some days nothing \u0026mdash; it needs to ask, \u0026lsquo;How are you today?\u0026rsquo; first.\u0026rdquo; (FG3, patient/carer) Safe prescription required integration of comorbidity profiles, medication schedules, blood sugar data, and mobility limitations. AI that could \u0026ldquo;know your body\u0026rdquo; like a trusted physiotherapist and tailor recommendations accordingly was viewed as ideal. Inclusive design principles\u0026mdash;clear, uncluttered interfaces; supportive, non-judgemental language; multiple language options; and offline access\u0026mdash;were regarded as critical for long-term engagement: \u0026ldquo;A daily, \u0026lsquo;you\u0026rsquo;re doing really well\u0026hellip; 10% ahead this week,\u0026rsquo; not, \u0026lsquo;you\u0026rsquo;re fat.\u0026rsquo;\u0026rdquo; (FG1, patient/carer) Integration with trusted healthcare platforms such as the NHS App or existing physiotherapy systems was seen as increasing credibility and uptake: \u0026ldquo;If it\u0026rsquo;s in the NHS App it might be easier to trust and use.\u0026rdquo; (FG2, professional).\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eIn this study we aimed to explore the perspectives of people with LTCs, their informal carers, and healthcare professionals on the use of AI-based exercise prescription software, focusing on perceived benefits, concerns, and practical considerations for implementation and wide-scale adoption. Six themes emerged: understanding and trusting AI in physical activity tools; positioning AI tools within everyday primary care; reaching and engaging those at risk of being overlooked; making AI a practical partner in self-management; the role of AI-based prescription in health and physical fitness; and personalisation and safety in AI.\u003c/p\u003e\n\u003cp\u003eParticipants expressed varied understandings of what qualifies as AI in the context of exercise prescription. Some viewed the case study example as little more than a rules-based system, while others recognised the novelty of AI tools capable of integrating multiple health and lifestyle data streams to tailor prescriptions dynamically. This uncertainty reflects a broader lack of clarity in healthcare about what constitutes \u0026ldquo;AI\u0026rdquo; and may undermine trust when patients and professionals are unclear about how personalised exercise recommendations are generated. Concerns about safety\u0026mdash;particularly the possibility of AI \u0026ldquo;hallucinations\u0026rdquo; producing inaccurate or unsafe prescriptions\u0026mdash;were directly linked to calls for clear governance, alignment with NICE recommendations for physical activity and LTCs, and explicit clinical accountability. These findings are consistent with literature emphasising that transparent processes, robust oversight, and professional responsibility are essential for safe use of AI in healthcare [24].\u003c/p\u003e\n\u003cp\u003ePositioning AI-based exercise prescriptions within primary care was seen as contingent on clinician and patient buy-in, seamless integration into existing workflows, and minimal disruption during time-limited consultations. Participants emphasised the importance of concise, actionable summaries that could support exercise conversations, targeted training for clinicians to understand AI-generated outputs, and continued human oversight to validate and adapt prescriptions where necessary. The value placed on blended models\u0026mdash;AI-generated prescriptions supported by periodic human contact\u0026mdash;reinforces evidence that fully automated solutions are less acceptable in primary care.\u003c/p\u003e\n\u003cp\u003eReaching and engaging those at risk of being overlooked was seen as a critical challenge for equitable adoption of AI-driven exercise prescriptions. Barriers such as language needs, limited digital literacy, socioeconomic disadvantage, unstable internet, and cultural scepticism were viewed as likely to exclude many with LTCs who might benefit most. These reflect the well-documented dimensions of the digital divide and the risk of deepening health inequalities if left unaddressed [25,26]. Participants advocated for inclusive design strategies such as multilingual interfaces, printable prescription summaries, voice navigation, simplified versions for low digital literacy, and co-design with seldom-heard groups. Such approaches are consistent with inclusive digital health design principles promoted by NHS Digital and the World Health Organisation [14,27].\u003c/p\u003e\n\u003cp\u003eMaking AI-based prescriptions a practical partner in self-management was tied to reducing, not adding to, the workload of managing LTCs. Participants highlighted automation of monitoring tasks, interoperability with existing health records, and personalised exercise plans that adapted to progress and capacity as essential to avoid duplication and \u0026ldquo;app fatigue.\u0026rdquo; These findings resonate with treatment burden theory, which emphasises balancing workload with capacity to sustain engagement [28]. Motivation and self-efficacy were seen as crucial, with requests for embedded behaviour change techniques, progress feedback, and clinician monitoring. Such strategies align with existing behavioural science frameworks. Prior experiences with apps and wearables also shaped expectations of AI-enabled exercise prescriptions. While such tools were appreciated for providing structure and accountability, frustrations over fragmented systems and subscription costs led to calls for a centralised hub that could integrate exercise, health, and lifestyle data. This aligns with wider digital health research emphasising the need for interoperable, user-centred platforms to support long-term engagement.\u003c/p\u003e\n\u003cp\u003eFinally, personalisation and safety were regarded as non-negotiable features of AI-based exercise prescriptions. Participants expected tools to adapt to daily health fluctuations and to account for comorbidities, medications, and mobility limitations, echoing principles of precision health [29]. They stressed the importance of inclusive, user-friendly design and integration with trusted platforms such as the NHS App to enhance credibility, uptake, and sustained use.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eComparison to existing literature\u003c/strong\u003e\u003cbr\u003e\u0026nbsp;The six themes highlighted in this study align with, and extend, the growing literature on AI-assisted digital tools in exercise prescription for LTCs. As previously emphasised, AI-assisted digital tools hold strong potential for personalised exercise prescription, overcoming many challenges faced by generic and non-specialist support. But these tools must overcome recognised barriers, akin to those expressed in previous studies where professional and patient trust, AI transparency, and clinical safety are important concerns in relation to \u0026ldquo;hallucinations\u0026rdquo; and accountability [30]. Similar to Shawli et al. (2024), healthcare professionals in this study expressed scepticism toward AI\u0026rsquo;s role in rehabilitation, highlighting the need for governance and human oversight, alongside the transparent development of AI-assisted tools [16]. Participants\u0026rsquo; call for embedding AI within primary care and linking outputs to existing records resonates with NHS priorities for digital integration and prevention outlined in the Long Term Plan [14]. The previously identified barriers of rurality, deprivation, and digital literacy, reinforce the importance of co-design and reaching those at risk of being overlooked, developed here with the suggestion of multilingual tools and hybrid delivery of exercise support.\u003c/p\u003e\n\u003cp\u003eComplementing these findings, recent systematic reviews provide additional context demonstrating that digital health interventions can reduce sedentary behaviour among people with LTCs, though outcomes varied across populations [31], while others highlight the effectiveness of digital tools in maintaining physical activity but noted limitations in evidence quality and long-term follow-up [32]. Wilson et al. (2024) further emphasised that advancing digital health equity requires tackling structural and contextual barriers [33]; our participants\u0026rsquo; recommendations for multilingual interfaces and hybrid delivery offer practical strategies to operationalise such equity within exercise support. Together, these insights position AI-assisted tools as promising but conditional on being transparent, clinically safe, and equitably accessible.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eStrengths and limitations\u003c/strong\u003e\u003cbr\u003e\u0026nbsp;A key strength of this study is the inclusion of multiple stakeholder perspectives, incorporating both lived experience of LTCs and professional viewpoints from providers. This enabled triangulation of insights, adding depth and enhancing credibility. The purposive sampling approach ensured diversity in condition types, professional roles, and care contexts, allowing exploration of a broad range of experiences. Conducting focus groups online increased accessibility for participants with mobility challenges, caring responsibilities, or geographical constraints, potentially reducing barriers to participation. The use of thematic analysis with independent coding review, combined with member checking, strengthened the rigour and trustworthiness of the findings.\u003c/p\u003e\n\u003cp\u003eHowever, several limitations should be acknowledged. The sample size was modest (n=13) and drawn from a limited geographical area, which may constrain transferability to other settings. However, as is typical in qualitative research, the focus was on achieving depth and richness of insight rather than numerical representativeness, allowing detailed exploration of participants\u0026rsquo; experiences and perspectives. Participants were self-selecting and may have had a pre-existing interest in digital health, potentially leading to over-representation of more digitally literate or positively inclined individuals. Those most at risk of digital exclusion such as those with limited internet access, low literacy, or high social vulnerability, may have been under-represented, limiting the study\u0026rsquo;s ability to fully capture the breadth of barriers faced by these groups. The online format, while accessible for some, may have excluded individuals without adequate digital skills or resources.\u003c/p\u003e"},{"header":"Conclusion and implications for policy makers","content":"\u003cp\u003eOur findings suggest that AI-based exercise prescription tools could enhance physical activity support for people with LTCs, provided they are developed and implemented with transparency, alignment to evidence-based guidelines, clinical oversight, and inclusive design. Successful adoption in primary care will require minimal disruption to existing workflows, clearly defined referral pathways, and hybrid delivery models that accommodate differing levels of digital literacy and access. Future work should prioritise co-production with larger, more diverse user groups, including those most at risk of digital exclusion, and rigorously evaluate the safety, acceptability, and effectiveness of such tools in real-world healthcare settings.\u003c/p\u003e\n"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eHDM receives funding from the National Institute for Health and Care Research (NIHR) Multiple Long-Term Conditions (MLTC) Cross NIHR Collaboration (CNC) (NIHR207000) and\u0026nbsp;the NIHR Artificial Intelligence for Multiple Long-Term Conditions (AIM) programme (NIHR202637). The views expressed in this publication are those of the author(s) and not necessarily those of the NHS, the NIHR or the Department of Health and Social Care.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting Interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNone declared.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe would like to thank all the patients and members of the public who contributed to the focus group.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eKraus WE, Powell KE, Haskell WL, Janz KF, Campbell WW, Jakicic JM, et al. Physical Activity, All-Cause and Cardiovascular Mortality, and Cardiovascular Disease. Med Sci Sports Exerc. 2019; 51(6):1270\u0026ndash;81. doi: 10.1249/MSS.0000000000001939.\u003c/li\u003e\n\u003cli\u003eRejbi IB, Trabelsi Y, Chouchene A, Ben Turkia W, Ben Saad H, Zbidi A, et al. Changes in six-minute walking distance during pulmonary rehabilitation in patients with COPD and in healthy subjects. Int J Chron Obstruct Pulmon Dis. 2010; 5:209\u0026ndash;15. Epub 20100809. doi: 10.2147/copd.s7955.\u003c/li\u003e\n\u003cli\u003eActivity against health risks associated with type 1 diabetes: \u0026ldquo;Health benefits outweigh the risks\u0026rdquo;. World J Diabetes. 2022; 13(3):161\u0026ndash;84. doi: 10.4239/wjd.v13.i3.161.\u003c/li\u003e\n\u003cli\u003ePearce M, Garcia L, Abbas A, Strain T, Schuch FB, Golubic R, et al. Association Between Physical Activity and Risk of Depression: A Systematic Review and Meta-analysis. JAMA Psychiatry. 2022; 79(6):550\u0026ndash;9. doi: 10.1001/jamapsychiatry.2022.0609.\u003c/li\u003e\n\u003cli\u003eJia RX, Liang JH, Xu Y, Wang YQ. Effects of physical activity and exercise on the cognitive function of patients with Alzheimer disease: a meta-analysis. BMC Geriatr. 2019; 19(1):181. Epub 20190702. doi: 10.1186/s12877-019-1175-2.\u003c/li\u003e\n\u003cli\u003eSingh B, Olds T, Curtis R, et alEffectiveness of physical activity interventions for improving depression, anxiety and distress: an overview of systematic reviews. British Journal of Sports Medicine. 2023;57:1203-1209. doi: 10.1136/bjsports-2022-106195.\u003c/li\u003e\n\u003cli\u003eB\u0026ouml;rjesson M, Onerup A, Lundqvist S, Dahl\u0026ouml;f B. Physical activity and exercise lower blood pressure in individuals with hypertension: narrative review of 27 RCTs. Br J Sports Med. 2016 Mar;50(6):356-61. doi: 10.1136/bjsports-2015-095786.\u003c/li\u003e\n\u003cli\u003eDergaa I, Ben Saad H. Using artificial intelligence for exercise prescription in personalised health promotion: A critical evaluation of OpenAI\u0026apos;s GPT-4 model. J Med Internet Res. 2024;26:e51308. doi:10.2196/51308.\u003c/li\u003e\n\u003cli\u003eRooney D, Gilmartin E, Heron N. Prescribing exercise and physical activity to treat and manage health conditions. Ulster Med J. 2023; 92(1):9\u0026ndash;15.\u003c/li\u003e\n\u003cli\u003eFesta RR, Jofr\u0026eacute;-Sald\u0026iacute;a E, Candia AA, et al. Next steps to advance general physical activity recommendations towards physical exercise prescription: a narrative review: BMJ Open Sport \u0026amp; Exercise Medicine 2023;9:e001749.\u003c/li\u003e\n\u003cli\u003ePedersen BK, Saltin B. Exercise as medicine \u0026ndash; evidence for prescribing exercise as therapy in 26 different chronic diseases. Scand J Med Sci Sports. 2015;25 Suppl 3:1\u0026ndash;72. doi:10.1111/sms.12581.\u003c/li\u003e\n\u003cli\u003eMoore GF, Audrey S, Barker M, et al. Process evaluation of complex interventions: Medical Research Council guidance. BMJ. 2015;350:h1258. doi:10.1136/bmj.h1258\u003c/li\u003e\n\u003cli\u003eTanaka H, Miyachi M, Nakamura Y. Artificial intelligence in exercise prescription: current applications and future directions. Sports Med. 2021;51(4):545\u0026ndash;557. doi:10.1007/s40279-020-01382-9.\u003c/li\u003e\n\u003cli\u003eNHS England. The NHS Long Term Plan. London: NHS England; 2019. Available from: https://www.longtermplan.nhs.uk/.\u003c/li\u003e\n\u003cli\u003eStrain T, Milton K, Dall PM, et al. The impact of rurality and deprivation on physical activity and sedentary behaviour in the UK: a cross-sectional analysis from the Health Survey for England. BMJ Open. 2021;11(8):e048270. doi:10.1136/bmjopen-2020-048270.\u003c/li\u003e\n\u003cli\u003eShawli L, Alsobhi M, Chevidikunnan MF et al. Physical therapists\u0026rsquo; perceptions and attitudes towards artificial intelligence in healthcare and rehabilitation: A qualitative study. Musculoskeletal Science and Practice. 2024;73:103152. Doi.org/10.1016/j.msksp.2024.103152.\u003c/li\u003e\n\u003cli\u003eArgent R, Daly A, O\u0026rsquo;Leary E. Patient Involvement With Home-Based Exercise Programs: Can Connected Health Interventions Influence Adherence? JMIR mhealth uhealth. 2018;6(3):e47. doi: 10.2196/mhealth.8518.\u003c/li\u003e\n\u003cli\u003eTeo JL, Zheng Z, Bird SR. Identifying the factors affecting \u0026lsquo;patient engagement\u0026rsquo; in exercise rehabilitation. BMC Sports Science, Medicine, Rehabilitation. 2022; 14:18. doi.org/10.1186/s13102-022-00407-3.\u003c/li\u003e\n\u003cli\u003eVassilakopoulou, P., Hustad, E. Bridging Digital Divides: a Literature Review and Research Agenda for Information Systems Research. Inf Syst Front 2023; 25, 955\u0026ndash;969. https://doi.org/10.1007/s10796-020-10096-3.\u003c/li\u003e\n\u003cli\u003eVITOVA. (n.d.). Retrieved August 16, 2025. Available from: https://vitovaltd.com/\u003c/li\u003e\n\u003cli\u003eBraun, V., \u0026amp; Clarke, V. Reflecting on reflexive thematic analysis. Qualitative Research in Sport, Exercise and Health, 2019;11(4), 589\u0026ndash;597. https://doi.org/10.1080/2159676X.2019.1628806.\u003c/li\u003e\n\u003cli\u003eO\u0026rsquo;Brien BC, Harris IB, Beckman TJ, etal. Standards for reporting qualitative research: A synthesis of recommendations. Academic Medicine [Internet]. 2014;89(9):1245\u0026ndash;51. Available from: https://journals.lww.com/academicmedicine/fulltext/2014/09000/Standards_for_Reporting_Qualitative_Research__A.21.aspx\u003c/li\u003e\n\u003cli\u003eO\u0026rsquo;Brien BC, Harris IB, Beckman TJ, et al. The SRQR reporting checklist. The EQUATOR network reporting guideline platform [Internet]. The UK EQUATOR Centre; 2025. Available from: https:/resources.equator-network.org/reporting-guidelines/srqr/srqr-checklist.docx\u003c/li\u003e\n\u003cli\u003eSaenz AD, Mass General Brigham AI Governance Committee, Centi A. et al. Establishing responsible use of AI guidelines: a comprehensive case study for healthcare institutions. npj Digit. Med. 2024;7, 348. https://doi.org/10.1038/s41746-024-01300-8\u003c/li\u003e\n\u003cli\u003eGunkel, DJ. Second thoughts: toward a critique of the digital divide. New Media \u0026amp; Society. 2003; 5(4), 499\u0026ndash;522.\u003c/li\u003e\n\u003cli\u003eChoudrie J, Pheeraphuttranghkoon, S., \u0026amp; Davari, S. The digital divide and older adult population adoption, use and diffusion of mobile phones: a quantitative study. Information Systems Frontiers. 2018; 22, 673\u0026ndash;695. https://doi.org/10.1007/s10796-018-9875-2.\u003c/li\u003e\n\u003cli\u003eMurphy K, Di Ruggiero E, Upshur R, Willison DJ, Malhotra N, Cai JC, Malhotra N, Lui V, Gibson J. Artificial intelligence for good health: a scoping review of the ethics literature. BMC Med Ethics. 2021;22(1):1\u0026ndash;7.\u003c/li\u003e\n\u003cli\u003eMay C, Eton DT, Boehmer K, et al. Rethinking the patient: Using Burden of Treatment Theory to understand the changing dynamics of illness. BMC Health Services Research. 2014;14, 281. https://doi.org/10.1186/1472-6963-14-281\u003c/li\u003e\n\u003cli\u003eDenny JC, Collins FS, Mehta JP. Precision health: The next generation of personalized medicine. The Lancet. 2021;398(10318), 1417\u0026ndash;1426. https://doi.org/10.1016/S0140-6736(21)01808-0.\u003c/li\u003e\n\u003cli\u003eAhmed MI, Spooner B, Isherwood J, et al. A Systematic Review of the Barriers to the Implementation of Artificial Intelligence in Healthcare. Cureus. 2023; 4:15(10):e46454. doi: 10.7759/cureus.46454.\u003c/li\u003e\n\u003cli\u003eZhang Y, Ngai F, Yang Q, Xie Y Effectiveness of Digital Health Interventions on Sedentary Behavior Among Patients With Chronic Diseases: Systematic Review and Meta-Analysis JMIR Mhealth Uhealth 2025;13:e59943 DOI: 10.2196/59943\u003c/li\u003e\n\u003cli\u003eHowes S, Stephenson A, Grimmett C, Argent R, Clarkson P, Khan A, Lait E, McDonough LR, Tanner G, McDonough SM. The effectiveness of digital tools to maintain physical activity among people with a long-term condition(s): A systematic review and meta-analysis. Digit Health. 2024 Dec 20;10:20552076241299864. doi: 10.1177/20552076241299864.\u003c/li\u003e\n\u003cli\u003eWilson, S., Tolley, C., Mc Ardle, R. et al. Recommendations to advance digital health equity: a systematic review of qualitative studies. npj Digit. Med. 7, 173 (2024). https://doi.org/10.1038/s41746-024-01177-7\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"bmc-primary-care","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"famp","sideBox":"Learn more about [BMC Primary Care](https://bmcprimcare.biomedcentral.com/)","snPcode":"","submissionUrl":"https://author-welcome.nature.com/12875","title":"BMC Primary Care","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Artificial Intelligence, primary care, general practice, exercise prescription, digital health, chronic disease management, qualitative study","lastPublishedDoi":"10.21203/rs.3.rs-7500570/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7500570/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cb\u003eBackground\u003c/b\u003e\u003c/p\u003e\u003cp\u003eExercise is an important component of managing Long-Term Conditions (LTCs) and can improve health outcomes. Artificial intelligence (AI)-based exercise prescription software has the potential to support exercise amongst people with LTC. However, its adoption in primary care requires consideration of the perspectives of patients, carers, and healthcare professionals, particularly in relation to benefits, concerns, and practical issues for implementation.\u003c/p\u003e\u003cp\u003e\u003cb\u003eMethods\u003c/b\u003e\u003c/p\u003e\u003cp\u003eA qualitative study was conducted with four online focus groups. This included two groups of people living with LTCs and/or their informal carers (n\u0026thinsp;=\u0026thinsp;7), and two groups with healthcare professionals (n\u0026thinsp;=\u0026thinsp;6). Semi-structured topic guides were used, and discussions were transcribed verbatim, and analysed thematically following Braun and Clarke\u0026rsquo;s six-phase framework.\u003c/p\u003e\u003cp\u003e\u003cb\u003eResults\u003c/b\u003e\u003c/p\u003e\u003cp\u003eSix themes were identified: (1) understanding and trusting AI in physical activity tools; (2) positioning AI tools within everyday primary care; (3) reaching and engaging those at risk of being overlooked; (4) making AI a practical partner in self-management; (5) the role of AI in health and physical fitness; and (6) personalisation and safety in AI. Participants recognised benefits including reduced administrative burden, integration with existing systems, and personalised motivational feedback. Concerns were raised about safety, governance, and data transparency. Adoption was seen as reliant on clinician and patient buy-in, minimal workflow disruption, clear referral pathways, and hybrid human\u0026ndash;digital support. Inclusive, user-friendly design\u0026mdash;such as multilingual options, large fonts, and simplified interfaces\u0026mdash;was considered essential to address barriers such as language needs, digital literacy, and socioeconomic constraints. Prior experiences with apps and wearables influenced expectations, with fragmented systems driving calls for a centralised hub.\u003c/p\u003e\u003cp\u003e\u003cb\u003eConclusions\u003c/b\u003e\u003c/p\u003e\u003cp\u003eAI-based exercise prescription software could enhance exercise support for people with LTCs if designed with transparency, safety, and personalisation. Effective integration into primary care will need to streamline workflows, provide hybrid human\u0026ndash;digital support, and use inclusive design to engage diverse users.\u003c/p\u003e","manuscriptTitle":"Artificial-Intelligence informed exercise prescriptions in primary care: Perspectives from people with long-term conditions, their carers, and healthcare professionals","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-09-18 14:55:50","doi":"10.21203/rs.3.rs-7500570/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-11-12T10:59:12+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-11-03T19:56:55+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"59882702463184329111699565706266210927","date":"2025-10-25T14:52:10+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-10-02T15:33:39+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"293050711034380986942798222852337549809","date":"2025-09-11T18:03:17+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"60431036585558002064190713342273045392","date":"2025-09-11T15:24:50+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-09-11T15:15:36+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-09-10T12:05:41+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-09-10T11:51:15+00:00","index":"","fulltext":""},{"type":"submitted","content":"BMC Primary Care","date":"2025-09-10T11:47:50+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"bmc-primary-care","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"famp","sideBox":"Learn more about [BMC Primary Care](https://bmcprimcare.biomedcentral.com/)","snPcode":"","submissionUrl":"https://author-welcome.nature.com/12875","title":"BMC Primary Care","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"b029f22c-a12e-4194-8417-f61bb0869de2","owner":[],"postedDate":"September 18th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"in-revision","subjectAreas":[],"tags":[],"updatedAt":"2026-02-16T06:54:37+00:00","versionOfRecord":[],"versionCreatedAt":"2025-09-18 14:55:50","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7500570","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7500570","identity":"rs-7500570","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}