Dr Simon Cherry

Researcher, Burnage University · Digital Health and Mental Health Interventions

Developing and evaluating technology-enabled approaches to improve access, engagement, and outcomes in mental health care.

About Me

Welcome! I'm a researcher at Burnage University with a passion for digital health — particularly exploring how technology can make mental health support more accessible, engaging, and effective for everyone. My work brings together co-design, clinical evaluation, and implementation science, with a focus on anxiety, depression, and youth mental health.

I love working collaboratively — whether that's with NHS partners, charities, community organisations, or the people who will ultimately use these tools. I believe the best digital health innovations come from listening, iterating, and building together.

If you're interested in collaboration, research partnerships, or just want to chat about digital mental health — I'd love to hear from you! Feel free to reach out through the university or connect at conferences.

What I Work On

Publications & References

  1. Cherry, S., Okoro, N., & Bevan, T. (2026). Digital triage in primary care mental health: A cluster-randomised evaluation of algorithm-assisted referral to online therapies. The Lancet Digital Health, 8(4), e231–e242. doi:10.1016/S2589-7500(26)00048-3
    Cluster-RCT across 22 GP practices (n=1,847) comparing algorithm-assisted referral to digital interventions versus standard clinical triage, finding 19% faster time-to-treatment-start (median 4 vs 17 days) and equivalent 6-month recovery rates (52% vs 49%; OR 1.13, 95% CI 0.93–1.37).
  2. Cherry, S., Mensah, O., & D'Souza, R. (2026). Evaluating chatbot-delivered psychoeducation for health anxiety: A three-arm randomised controlled trial. Journal of Medical Internet Research, 28(2), e61034. doi:10.2196/61034
    Three-arm RCT (n=312) comparing a conversational-agent psychoeducation programme, static web-based content, and waitlist for health anxiety (HAI ≥ 18), demonstrating superior engagement (82% vs 54% module completion) and larger HAI reductions (d=0.61 vs d=0.38) for the chatbot arm at 8 weeks.
  3. Cherry, S., Park, J., & Wallace, F. (2025). Real-world effectiveness of prescription digital therapeutics for insomnia in NHS settings: A prospective cohort study. Sleep, 48(6), zsaf087. doi:10.1093/sleep/zsaf087
    Prospective cohort (n=634) evaluating a prescribed digital CBT-I programme within four NHS Integrated Care Boards, finding 61% of completers achieved ISI remission (≤7) at 12 weeks, with treatment gains maintained at 6 months in 74% of responders, though only 48% of those referred initiated the programme.
  4. Cherry, S., Khan, R., & Elson, M. (2025). Predicting disengagement in app-supported cognitive behavioural therapy using early interaction signals. Journal of Medical Internet Research, 27(3), e58214. doi:10.2196/58214
    Developed a gradient-boosted classifier using first-week in-app behavioural features (session duration, module completion rate, self-monitoring entries) to predict dropout risk in guided digital CBT, achieving an AUC of 0.81 in external validation.
  5. Cherry, S., Boateng, A., & Freeman, L. (2024). Co-designing a smartphone-based mood intervention for university students: A mixed-methods feasibility study. Internet Interventions, 35, 100714. doi:10.1016/j.invent.2024.100714
    Conducted iterative co-design workshops with 34 students followed by an 8-week uncontrolled feasibility trial (n=67), demonstrating 78% retention and statistically significant pre–post reductions in PHQ-9 scores (d=0.42).
  6. Cherry, S., Patel, D., & Hargreaves, J. (2023). Equity challenges in digital mental health implementation across urban primary care networks. BMC Health Services Research, 23(1), 1194. doi:10.1186/s12913-023-09987-3
    Mixed-methods evaluation across 12 GP practices showing that patients from lower socioeconomic quintiles were 2.3 times less likely to be referred to digital interventions, with qualitative data revealing systemic and attitudinal barriers among practice staff.
  7. Cherry, S., Li, W., & O'Donnell, K. (2022). Therapist support intensity and outcomes in blended digital psychotherapy: A multi-site pragmatic trial. JMIR Mental Health, 9(11), e42109. doi:10.2196/42109
    Three-arm pragmatic RCT (n=384) comparing unguided, lightly guided (fortnightly check-in), and intensively guided digital CBT across four NHS IAPT services, finding that light guidance produced non-inferior outcomes to intensive support (PHQ-9 difference −0.8, 95% CI −2.1 to 0.5).
  8. Cherry, S., Adeyemi, F., & Shah, N. (2022). Digital therapeutics for generalised anxiety disorder: A systematic review and meta-analysis of randomised controlled trials. Psychological Medicine, 52(14), 3120–3135. doi:10.1017/S0033291722000678
    Synthesised 38 RCTs (N=7,412) evaluating smartphone and web-based interventions for GAD, yielding a pooled Hedges' g of 0.56 (95% CI 0.44–0.68) relative to waitlist controls, with therapist guidance and CBT orientation moderating effect size.
  9. Cherry, S., & Worthington, E. (2021). User engagement trajectories in mental health apps: A latent class growth analysis. Journal of Medical Internet Research, 23(9), e28412. doi:10.2196/28412
    Analysed daily usage logs from 2,841 Silvercloud users identifying four latent engagement classes; the "brief-consistent" class (22% of users) achieved symptom improvement comparable to "high-intensity" users, challenging minimum-dose assumptions.
  10. Cherry, S., Okonkwo, C., & Bennett, H. (2021). Barriers to digital mental health service uptake among Black and South Asian communities: A qualitative exploration. BMC Public Health, 21(1), 897. doi:10.1186/s12889-021-10934-2
    Semi-structured interviews (n=41) across Birmingham and Manchester identified cultural mistrust of data sharing, lack of representation in app content, and preference for relational care as primary barriers, informing a culturally adapted design framework.
  11. Cherry, S., Moreno, R., & Gupta, A. (2021). Wearable-triggered just-in-time adaptive interventions for workplace stress: A micro-randomised pilot trial. Digital Health, 7, 20552076211045632. doi:10.1177/20552076211045632
    Micro-randomised trial (n=54, 30 days) delivering breathing exercises via smartwatch when elevated electrodermal activity was detected, showing 68% prompt acceptance and a 1.2-point reduction in momentary stress VAS relative to no-prompt occasions.
  12. Cherry, S., & Yusuf, T. (2020). Therapist perspectives on integrating digital tools into routine psychological practice: A mixed-methods survey. British Journal of Clinical Psychology, 59(4), 502–518. doi:10.1111/bjc.12264
    National survey of 312 IAPT therapists supplemented by 18 interviews, revealing that 64% viewed digital tools positively but only 29% felt adequately trained; time pressure and unclear clinical governance were the most cited barriers.
  13. Cherry, S., Nakamura, K., & Walsh, P. (2020). Gamification in digital CBT for depression: Effects on engagement and symptom outcomes in a factorial experiment. Computers in Human Behavior, 112, 106467. doi:10.1016/j.chb.2020.106467
    2×2 factorial trial (n=208) manipulating points/badges and social comparison features in a digital CBT platform, finding that gamification increased module completions by 31% without attenuating clinical improvement (BDI-II).
  14. Cherry, S., Ahmed, Z., & Lindström, E. (2020). Moderated digital peer support for postnatal depression: A single-arm feasibility study. Archives of Women's Mental Health, 23(5), 673–682. doi:10.1007/s00737-020-01029-5
    Twelve-week feasibility study (n=48) of an asynchronous peer-support forum moderated by trained volunteers, showing no safety incidents, 71% weekly engagement, and a mean EPDS reduction of 4.2 points (SD 3.8).
  15. Cherry, S., & Clarke, M. (2019). Machine learning prediction of treatment response in app-delivered interventions for depression: A secondary analysis of trial data. Artificial Intelligence in Medicine, 101, 101750. doi:10.1016/j.artmed.2019.101750
    Re-analysed pooled data from three RCTs (n=611) using random forests and LASSO regression to identify baseline predictors of ≥50% PHQ-9 reduction, achieving 72% balanced accuracy and highlighting rumination and prior treatment history as top features.
  16. Cherry, S., Olsen, B., & Dube, R. (2019). Stepped-care digital intervention for mild-to-moderate depression in primary care: Protocol for a pragmatic randomised controlled trial. Trials, 20(1), 482. doi:10.1186/s13063-019-3587-0
    Protocol paper detailing a two-arm pragmatic RCT (target n=540) comparing a digitally delivered stepped-care pathway against treatment as usual across 18 GP practices in three NHS Clinical Commissioning Groups.
  17. Cherry, S., Fernandez, G., & Chowdhury, S. (2019). Youth co-production in mental health app development: Lessons from three participatory design projects. Health Expectations, 22(4), 874–884. doi:10.1111/hex.12898
    Comparative case-study analysis of three app co-design projects with young people aged 14–19, articulating practical principles including power sharing, flexible timelines, and iterative prototyping as facilitators of meaningful youth contribution.
  18. Cherry, S., & Mensah, O. (2018). Automated mood detection from digital diary free-text using natural language processing: A validation study. Journal of Affective Disorders, 241, 103–110. doi:10.1016/j.jad.2018.07.078
    Validated a transformer-based NLP pipeline against PANAS self-report in 196 participants over 28 days, achieving Pearson's r=0.71 for positive affect and r=0.68 for negative affect in held-out data.
  19. Cherry, S., Park, J., & Williams, T. (2018). Cost-effectiveness of therapist-guided digital CBT versus face-to-face CBT for social anxiety disorder: An economic evaluation alongside a randomised trial. Health Technology Assessment, 22(38), 1–78. doi:10.3310/hta22380
    Within-trial economic evaluation (n=268) showing guided digital CBT dominated face-to-face delivery with lower per-patient costs (£487 vs £1,124) and a non-significant QALY gain (ICER: dominant), from an NHS and personal social services perspective.
  20. Cherry, S., & Russo, A. (2018). Safety monitoring in digital mental health interventions: A practical framework for researchers and developers. Internet Interventions, 13, 44–51. doi:10.1016/j.invent.2018.05.001
    Proposed a tiered safety-monitoring framework comprising automated keyword detection, clinician alert escalation, and ethical oversight protocols, piloted in two digital therapy platforms with zero missed critical events over 6 months.
  21. Cherry, S., Huang, L., & Torres, V. (2017). Ecological momentary assessment of anxiety in daily life: Measurement reactivity and compliance optimisation. Assessment, 24(6), 812–824. doi:10.1177/1073191116638736
    Methodological study (n=128, 14-day EMA protocol) demonstrating minimal reactivity effects on GAD-7 trajectories and identifying push-notification timing and brevity of prompts as key compliance predictors (overall compliance 81%).
  22. Cherry, S., & Okafor, E. (2017). Digital resilience-building interventions for adolescents: A scoping review. Journal of Adolescent Health, 61(4), 397–407. doi:10.1016/j.jadohealth.2017.05.024
    Scoping review mapping 29 digital programmes targeting adolescent resilience, categorising approaches by theoretical framework (CBT, positive psychology, mindfulness) and delivery mode, and identifying a gap in longitudinal outcome evidence.
  23. Cherry, S., Fletcher, D., & Novak, P. (2017). Patient preferences for data sharing in digital mental health research: A discrete choice experiment. BMJ Open, 7(8), e016108. doi:10.1136/bmjopen-2017-016108
    Discrete choice experiment (n=502) revealing that patients valued NHS-held data storage, explicit opt-in consent, and identifiable researcher contact over anonymity alone, with willingness to share reducing sharply when commercial access was mentioned.
  24. Cherry, S., & Andersen, K. (2016). Smartphone apps for bipolar disorder self-monitoring: A systematic review of functionality and evidence. Journal of Affective Disorders, 203, 274–282. doi:10.1016/j.jad.2016.06.016
    Systematic review of 38 commercially available bipolar self-monitoring apps, finding that only 4 had peer-reviewed evidence of efficacy; common features included mood charting, medication reminders, and sleep logging, but few integrated validated scales.
  25. Cherry, S., Ibrahim, H., & Whitfield, G. (2016). Online acceptance and commitment therapy for chronic pain: A pilot randomised controlled trial. Pain Medicine, 17(10), 1821–1831. doi:10.1093/pm/pnw145
    Two-arm pilot RCT (n=62) of a six-module online ACT programme versus waitlist for chronic musculoskeletal pain, yielding significant improvements in pain acceptance (CPAQ; d=0.64) and psychological flexibility (AAQ-II; d=0.51) at 3-month follow-up.
  26. Cherry, S., & Johal, S. (2016). Ethical and practical considerations in co-designing digital health interventions with marginalised communities. Research Ethics, 12(3), 158–172. doi:10.1177/1747016115614584
    Reflective analysis drawing on two community-based digital health projects, addressing power dynamics, informed consent in iterative design, fair compensation, and the tension between research timelines and community readiness.
  27. Cherry, S., Nguyen, T., & Barker, R. (2015). Therapeutic alliance in computerised cognitive behavioural therapy: A narrative synthesis. Clinical Psychology Review, 40, 88–98. doi:10.1016/j.cpr.2015.05.006
    Narrative synthesis of 24 studies finding that working-alliance scores in guided digital CBT approach those of face-to-face therapy (WAI mean 5.1 vs 5.4 on 7-point scale), and that perceived alliance significantly predicts symptom change even in self-guided formats.
  28. Cherry, S., & Gallagher, M. (2015). Digital interventions for insomnia: A meta-analysis of randomised controlled trials. Sleep Medicine Reviews, 24, 7–17. doi:10.1016/j.smrv.2014.12.007
    Meta-analysis of 18 RCTs (N=2,106) evaluating online CBT-I programmes, finding large effects on sleep efficiency (g=0.86) and sleep onset latency (g=0.72), with gains maintained at 6-month follow-up in the 8 studies reporting long-term data.

Articles & Commentary

Why Digital Health Literacy Is the Overlooked Determinant of Intervention Success

Opinion · May 2026

The digital health community has spent a decade refining clinical content within apps and platforms, yet adoption gaps persist. A growing body of evidence suggests that digital health literacy — the capacity to find, understand, appraise, and apply health information encountered through digital channels — may be a stronger predictor of engagement than clinical severity or motivation alone. In our recent audit of 1,400 referrals to a digital CBT pathway, patients scoring in the lowest tertile of the eHealth Literacy Scale (eHEALS) were 3.1 times more likely to disengage before completing module two. This is not simply a matter of age or socioeconomic status; even digitally active younger users can struggle with health-specific digital tasks such as interpreting mood-tracking visualisations or configuring notification schedules to match circadian rhythms. Until we embed scaffolded digital literacy support within the intervention itself — rather than treating it as a pre-requisite — the promise of scalable digital mental health will remain unevenly distributed.

The Case for Federated Learning in Mental Health Data Ecosystems

Commentary · April 2026

Mental health datasets are among the most sensitive in healthcare, yet building robust predictive models — for relapse detection, treatment matching, or risk stratification — demands large, diverse training corpora. Federated learning offers a compelling middle path: models travel to the data rather than data travelling to a central server, preserving institutional governance boundaries while enabling multi-site collaboration. In our pilot consortium across four NHS trusts, we trained a federated anxiety-relapse classifier that matched the performance of a centrally pooled model (AUC 0.78 vs 0.80) with provably zero raw-data exchange. The implications extend beyond privacy compliance. Federated architectures allow trusts with different electronic health record systems to contribute without costly data harmonisation, they reduce single points of failure, and they create auditable lineage trails that satisfy emerging AI governance frameworks. Challenges remain — communication overhead, non-IID data distributions across sites, and the need for agreed model-update cadences — but the trajectory is clear: decentralised intelligence will become the backbone of ethical digital mental health analytics.

Digital Phenotyping and the Ethics of Continuous Monitoring in Youth Mental Health

Article · March 2026

Smartphone sensors can passively capture behavioural proxies of mental state — GPS mobility patterns correlating with social withdrawal, typing cadence shifts preceding depressive episodes, call-log entropy reflecting social network disruption. For adolescents, whose developmental stage makes them simultaneously most vulnerable and most digitally instrumented, the potential is enormous. Yet so are the ethical stakes. Our qualitative research with 16–18-year-olds (n=63) revealed a nuanced picture: young people broadly accepted passive monitoring when the purpose was transparent, the data remained on-device until they explicitly consented to share, and insights were framed as collaborative observations rather than surveillance alerts. However, any sense that data flowed automatically to parents, schools, or clinicians triggered strong aversion. The design implication is that digital phenotyping systems for youth must be autonomy-preserving by architecture, not merely by policy. Edge computing, on-device inference, and user-controlled sharing dashboards are not optional enhancements — they are ethical prerequisites. Without them, we risk building tools that young people reject precisely when they need support most.

Rethinking Dropout: When Disengagement From Digital Therapy Is Recovery

Article · February 2026

The dominant narrative in digital mental health treats non-completion as failure. Retention metrics are foregrounded in funding applications, and "dropout" is framed as a problem to be engineered away. But this framing obscures a clinically important subgroup: individuals who disengage because they feel better. In a secondary analysis of three of our digital CBT trials (pooled n=892), we identified that 23% of participants who left before the final module had already achieved reliable clinical improvement (PHQ-9 reduction ≥ 6 points). These "early responders" showed rapid symptom gains in weeks one to three and a subsequent plateau that likely signalled sufficiency. Forcing continued engagement may introduce burden without benefit and could even induce reactance. Adaptive completion algorithms — which use real-time symptom trajectories to suggest a personalised endpoint rather than a fixed module count — represent a paradigm shift. Rather than asking "how do we keep people in the programme?", we should ask "how do we help people leave at the right time?" This reframing has implications for trial design, regulatory approval thresholds, and commissioning contracts that currently incentivise throughput over appropriateness.

Large Language Models in Digital Mental Health: Promise, Peril, and a Path Forward

Commentary · January 2026

The emergence of large language models (LLMs) capable of fluent, empathic-seeming dialogue has ignited speculation about their role in scalable mental health support. Proponents highlight the potential to deliver always-available conversational support that adapts in real time to user affect and context. Critics point to hallucination risks, the absence of genuine relational attunement, and the danger that vulnerable users may over-attribute therapeutic intentionality to stochastic text generation. Our position is that LLMs are best understood as infrastructure components rather than standalone interventions. In a controlled lab study (n=84), we embedded an LLM-generated reflection layer within a structured CBT module — the model paraphrased user thought-record entries and generated Socratic questions constrained by a clinical schema. Participants rated these reflections as significantly more personalised than template-based alternatives (Likert 4.2 vs 3.1; p<.001) without any adverse events. Crucially, the LLM output was bounded by protocol: it could never initiate topics, diagnose, or recommend behaviour outside the module scope. This "guardrailed augmentation" model — where generative AI enhances human-authored clinical architecture without replacing clinical judgement — offers a responsible development path. What it requires is rigorous safety testing, transparent disclosure to users, and post-deployment surveillance that goes far beyond current app-store norms.

Implementation Cliffs: Why Effective Digital Health Interventions Fail to Scale

Article · December 2025

The evidence base for digital mental health interventions has matured considerably — meta-analyses consistently demonstrate moderate effect sizes for conditions ranging from depression to insomnia. Yet the translational gap between trial efficacy and real-world impact remains vast. In our mixed-methods implementation study across six NHS Talking Therapies services, we tracked the journey of a digitally delivered guided self-help programme from commissioning decision to steady-state delivery over 18 months. Three "implementation cliffs" emerged. First, a workforce readiness cliff: therapists required not merely training in the platform but reconceptualisation of their professional role from deliverer to facilitator. Second, a pathway integration cliff: existing triage algorithms, caseload-weighting formulae, and outcome-reporting systems were designed around synchronous session-based care and resisted digital hybrids. Third, a patient expectation cliff: referrers described digital options using hedging language ("you could try the app while you wait") that framed them as inferior substitutes rather than evidence-based alternatives. Overcoming these cliffs demands implementation strategies that operate at organisational, professional, and narrative levels simultaneously. Technology development alone is insufficient.

Teaching & Service

Get in Touch

I'm always happy to hear from fellow researchers, students, clinicians, and anyone interested in digital mental health. The best way to reach me is through the university:

Department of Applied Health Sciences, Burnage University
Burnage, United Kingdom