Hardware, Software, Apps, Devices, AI : HealthTech driving Behaviour Change to improve Patient Outcomes

Nelson Advisors
Sep 27
15 min read

Executive Summary and Strategic Imperatives

Defining the Digital Therapeutic Paradigm: From Connectivity to Continuous Care

The digital transformation of healthcare represents a profound industry evolution, shifting the focus from episodic treatment to continuous, proactive intervention. This new paradigm, known broadly as Digital Health, utilises technology to facilitate active therapeutic engagement and deliver care outside traditional clinical settings, effectively realising the concept of "virtual hospital wards" and enabling remote diagnostics. This movement is supported by a massive and rapidly expanding ecosystem, currently comprising over 337,000 mobile health applications globally. Crucially, a growing segment of these applications focuses intensely on disease-specific management and chronic conditions, moving beyond general wellness tracking to deliver tangible clinical utility.

The market has identified key areas ripe for disruption through digital integration. Data-driven companies are increasingly leading innovation in historically underserved sectors, such as women's health, offering integrated, end-to-end care pathways that personalise treatment based on continuous data streams. Similarly, the increasing professional acknowledgement of mental health’s importance has driven the emergence of more specialised mental health and well-being solutions utilising digital tools for improved accessibility and efficacy.

Strategic Imperatives for MedTech Transformation

The successful deployment and maintenance of sophisticated, connected medical devices necessitate a comprehensive transformation for medtech companies. The traditional model, centred on one-time device sales, is becoming obsolete in a connected world where ongoing service delivery is not merely expected but is required for sustained clinical function and regulatory adherence.

This transformation requires a fundamental shift toward service-oriented business models that prioritise long-term value creation over immediate returns from device sales. Companies must invest in the foundational capabilities and infrastructure required to deliver continuous services throughout a device's operational life, including device maintenance, security monitoring, data management, and the crucial ability to provide over-the-air (OTA) software and firmware updates.

Payer recognition and the acceleration of approval and reimbursement for digital tools are directly predicated on the demonstration of clinical utility and verifiable cost savings. Because Software as a Medical Device (SaMD) and Digital Therapeutics (DTx) rely on continuous software refinement, for security patches, bug fixes,and algorithmic improvements, the traditional device-only model is incompatible with the criteria required for continuous market access and reimbursement justification.

Therefore, the transition to a service-oriented model is not an elective business change, but a strategic and commercial prerequisite to ensure continuous operational compliance and the necessary sustained behavioural intervention efficacy.

Technological Pillars of Behavioural Intervention

Connected Devices and Wearable Architecture: The Foundation of Continuous Data Capture (Hardware)

The physical layer of digital health is built upon Sensor-based Digital Health Technology (sDHT). These devices, ranging from advanced wearables to smart inhalers and sophisticated glucose monitors, form the foundation for behaviour change interventions by providing a constant stream of real-time feedback and actionable data. This continuous monitoring empowers patients to effectively manage complex conditions such as asthma and diabetes. The US. Food and Drug Administration (FDA) maintains a registry of authorised sDHT medical devices, providing crucial insights into the current regulatory landscape and setting clear safety and effectiveness expectations for innovators.

Remote Patient Monitoring (RPM) tools are a core component of this infrastructure. These systems utilise sensors to track critical physiologic and behavioural data, augmented by the collection of electronic Patient Reported Outcomes (ePROs) delivered via mobile applications. This data flow is essential not only for chronic condition management but also for aiding in population risk management and enabling highly personalised care delivery through continuous feedback loops. For example, Biofourmis provides FDA-approved digital therapeutics solutions for chronic heart conditions, leveraging wearable devices and its sophisticated, personalised AI platform, Biovitals.

The fundamental value proposition of the hardware transitions from simple monitoring to providing the essential feedstock for algorithmic intervention. While RPM traditionally focuses on clinical metrics (eg. blood pressure, blood glucose), the key mechanism for behaviour change intervention (BCI) relies heavily on the sensor-based tracking of behavioural data, such as steps, sleep quality, and active minutes. This rich, continuous stream of data, often collected implicitly, becomes the necessary input for the personalisation engines (AI/ML) that generate targeted nudges and feedback.

Software as a Medical Device (SaMD) and Digital Therapeutics (DTx)

The software layer defines the intervention itself. AI/ML-based software intended to treat, diagnose, cure, mitigate, or prevent disease is classified under the FD&C Act as Software as a Medical Device (SaMD). This classification subjects the software to stringent regulatory oversight consistent with traditional medical devices.

The market for Digital Therapeutics (DTx), defined by these regulated software solutions, is demonstrating robust financial growth. The DTx market was valued at approximately $4.68 Billion in 2024 and is projected to continue growing at a Compound Annual Growth Rate (CAGR) of 16.61%. This expansion reflects successful clinical adoption; out of over 360 commercially available software-based digital therapies, 140 prescription DTx are approved for patient use at home.

Furthermore, the technological pillars are advancing rapidly from monitoring capabilities to diagnostic capabilities. Over 103 digital diagnostics, frequently enabled by Artificial Intelligence (AI) and Machine Learning (ML), are now commercially available. These tools go beyond tracking the response to treatment; they are used for risk assessment, accelerating diagnosis, and providing prognostic value, thereby influencing patient behaviour earlier in the care pathway through tools like risk screening and diagnosis support.

The Role of the Application Layer: Designing User Interfaces as Behaviour Change Conduits

The mobile application serves as the primary conduit for the patient experience, delivering personalised care instructions and Behaviour Change Techniques (BCTs). The design of this layer is critical, as connected medical devices must operate seamlessly in real-world environments and serve a diverse user set, many of whom are not healthcare professionals.

Effective application design must embed core principles of usability and user-centricity. The application must simplify complex interactions and present continuous data in an accessible manner. The quality of the user interface directly impacts the ability of the SaMD to facilitate behavior change, as a poorly designed app can introduce friction that undermines motivation and adherence.

Applied Behavioral Science and Engagement Design

Foundational Models for Behaviour Change Interventions (BCIs)

The efficacy of digital health solutions hinges on the meticulous application of established psychological and behavioral science frameworks. Building digital tools around these models maximises patient engagement and the probability of sustained behavioural conversion.

The Fogg Behavior Model (FBM) provides a practical explanation of how behaviors occur in the context of digital health solutions. It posits that a behaviour (B) happens only when an individual possesses sufficient Motivation (M), high Ability (A), and a timely Trigger (T). Digital health applications leverage the FBM by simplifying tasks and interfaces (increasing Ability) and delivering personalised, context-aware prompts (Triggers) at optimal times to encourage action.

The Transtheoretical Model (TTM) offers a framework for staging interventions based on a user's readiness for change. TTM recognises distinct stages: Pre contemplation (no intention of change), Contemplation (recognising the need for change),Preparation (planning action), and Action (new behaviour initiated).Designing interventions informed by TTM ensures that personalised content and challenges align with the user’s current stage, thereby maximising the relevance of the nudge and increasing the probability of conversion to sustained action.

Practical Behaviour Change Techniques (BCTs)

Systematic reviews of Digital Behaviour Change Interventions (DBCIs) for physical activity confirm that successful techniques revolve around three core methods: self-monitoring of behaviour, goal setting and the provision of prompts and cues. Habit formation within these systems is systematically encouraged through positive reinforcement and cues, often using automatic monitoring, descriptive feedback, self-set goals and virtual rewards.

Mechanisms of Sustained Engagement: Gamification and Adherence

Gamification is a powerful tool for transforming necessary, but often monotonous, health routines into intrinsically motivating experiences. By integrating game mechanics such as rewards, level progression, and challenges, patients are encouraged to consistently complete tasks and monitor their progress toward health goals.

The clinical impact of this approach is measurable. Research indicates that gamified health applications improved medication adherence by 30% when compared to non-gamified counterparts, leading patients to be more consistent with medication schedules and developing a deeper understanding of their condition.

Behaviour change interventions for chronic disease management frequently face challenges related to delayed feedback; for instance, it takes months to observe a change in HbA1c or long-term blood pressure reduction. Gamification provides an immediate emotional reward loop through virtual rewards and points systems, serving as a critical 'fast feedback' mechanism that maintains motivation and adherence in the absence of rapid clinical feedback. This mechanism accelerates habit formation, reinforcing desired actions in the short term to sustain engagement toward long-term clinical goals.

Despite these advances, a recognised gap exists in design strategy: the lack of research and focus on implicit interaction. While many existing studies rely on explicit user effort (eg., manually logging data), high adherence is only sustainable when the intervention requires minimal conscious effort (maximising Fogg Ability). Overcoming this scalability bottleneck requires leveraging advanced AI and ubiquitous sensors to infer user context and deliver behavioural interventions ambiently, thereby solving the challenge of long-standing adherence noted in early clinical evaluations.

Operationalising Behaviour Change Models in Digital Health Design

Behavioural Model	Core Principle	Mechanism/Technique Used in Apps	Design Goal
Fogg Behaviour Model (FBM)	Behaviour = Motivation Ability Trigger	Contextual, timely prompts/cues (Triggers); simplified task flows (Ability)	Maximising action completion by reducing friction and providing instantaneous cues.
Transtheoretical Model (TTM)	Stages of Readiness (Pre contemplation to Action)	Personalised content delivery and staggered challenges based on current readiness stage	Increasing the probability of conversion to the sustained action stage.
Operant Conditioning / Gamification	Positive Reinforcement, Rewards	Virtual rewards, points systems, level progression, badges	Improving adherence rates and creating short-term motivation loops to sustain long-term habit formation.
Social Learning Theory (MINDSPACE)	Social Influence and Modelling	Community forums, peer comparisons, shared progress (Implicitly suggested by the data)	Building social motivation and reinforcing positive behavioural norms.

Advanced AI and Algorithmic Nudging Systems

Personalised Interventions at Scale: The Necessity of AI

Scaling behaviour change interventions to large populations requires sophistication far beyond static application logic. AI and Machine Learning provide the necessary complexity to move from broad intervention strategies to hyper-personalised, context-aware digital nudging. AI not only optimises behavioural triggers but also enables complex diagnostic functions within the SaMD framework, including risk screening, diagnosis support, and prognostic analysis.

Deep Dive: Knowledge Graph Neural Networks (KGNN) in Health Nudging

The application of advanced AI architecture is exemplified by production level systems like NudgeRank. This algorithmic nudging system utilises a novel combination of Graph Neural Networks (GNN) augmented with an extensible Knowledge Graph (KG) to deliver personalised, context-aware nudges to over 1.1 Million care recipients daily. This represents one of the largest enterprise deployments of AI dedicated to health behaviour change.

The core of the architecture lies in the NudgeRank Knowledge Graph Constructor, which dynamically forms a heterogeneous graph where users and nudges are represented as nodes. Directed edges capture the historical history of user-item interactions (eg. ratings, opens). User knowledge is captured through over

130 binary markers detailing attributes and behaviours (eg., "age: 30s," "steps: 2.5k"). Nudge knowledge is also encoded, detailing the specific target segments and desired behavioural outcomes. This structure enables personalisation by mitigating the cold-start problem, allowing the system to recommend new nudges to relevant target segments based on the graph connections and shared behavioural data with similar users.

A critical technical feature of the KGNN is the Diversity Mechanism. Controlled by a configurable parameter pdiversity, this mechanism manages the proportion of nudges that are randomly sampled from the user's candidate list. This forced algorithmic exploration strategy is essential for maximising long-term learning utility by continually gathering new user-nudge interaction data, broadening user preference knowledge and preventing the model from becoming stagnant or clinically rigid over time. The system's objective is to manage the tension between maximising immediate prediction accuracy (exploitation) and gathering new data to handle changing contexts (exploration).

To protect user experience and maintain long-term adherence, the Constraints Filter enforces critical business rules. This includes the Negative Rating Filter, which removes all nudges a user has actively disliked for a specified period and the Nudge Budget Filter, which limits the number of daily nudges to a configurable parameter kdaily to actively mitigate "nudge fatigue". By optimising the trigger timing based on real-time data and KG connections, the AI ensures the intervention occurs at the precise moment of maximum receptivity, transforming a generic manual cue into an effective, context-aware trigger.

NudgeRank™ AI Architecture: Key Components and Behavioural Function

Component	Technical Description	Behavioral Function/Value
Knowledge Graph Neural Network (KGNN)	Heterogeneous graph with User/Nudge nodes and interaction edges.	Enables hyper-personalisation by connecting users to relevant nudges based on similar users’ behaviour; optimises the timing and context of the Trigger.
Diversity Mechanism (pdiversity)	Configurable parameter for forced random sampling of candidate nudges.	Ensures algorithmic exploration, broadens learning of user preferences, and prevents stagnation in recommendation patterns, crucial for long-term adaptation.
Constraints Filter	Enforces Negative Rating Filter and Nudge Budget Filter (kdaily).	Mitigates "nudge fatigue" and respects user feedback, safeguarding the user experience and long-term acceptance.
Distributed Infrastructure	Runs on Kubernetes with parallel batch scoring pipelines.	Ensures performance and resilience, allowing the system to scale personalisation to over 1.1 million users without contextual latency.

Quantifiable Impact of Personalised AI

Rigorous evaluation of the NudgeRank system, deployed in collaboration with the Health Promotion Board of Singapore, demonstrated statistically significant efficacy at improving physical activity across the population. Key behavioural improvements included a 6.17% increase in daily steps and a 7.61% increase in exercise minutes. Furthermore, the system achieved a substantial increase in user engagement, evidenced by a 13.1% open rate, compared to a baseline system's 4%. This increase in open rate signals the clinical advantage of context-aware, hyper-personalised delivery.

Clinical Validation and Outcome Evidence

Chronic Disease Management Case Studies

Empirical evidence confirms the transformative potential of integrated digital solutions across major chronic disease categories, demonstrating a clear causal pathway from technology-driven behaviour change to improved patient outcomes.

For patients with Diabetes Mellitus, the use of real-time Continuous Glucose Monitoring (CGM) has been associated with dramatic clinical improvements. One study revealed that the proportion of patients with intensively managed diabetes who achieved an HbA1c level of <7% nearly doubled, rising from 24.6% in pre study samples to 50.8% after at least 12 weeks of CGM usage. Participants with higher baseline HbA1c levels experienced even larger absolute reductions, exceeding

1%.

In Hypertension management, digital health interventions incorporating remote monitoring have achieved impressive results, doubling the success rates for blood pressure control. This improvement was particularly significant among historically underserved patient populations, demonstrating the capacity of digital delivery to overcome traditional structural barriers to care and promote health equity.

Digital solutions are also effective in Weight Management. Mobile applications providing evidence-based weight loss programs have been shown to be a cost-effective and accessible alternative to intensive in-person programs. Pooled results from meta-analyses show that app usage led to a mean weight reduction of 0.84 kg, with significant decreases in BMI (median 1.8 kg/m) and waist circumference (median −3.8cm).

In one study, groups receiving the intervention lost 2.6% of their body weight within the first three months.

Expanding Clinical Utility: Mental Health and Integrated Pathways

Beyond traditional chronic physical diseases, specialised mental health and well-being solutions continue to gain prominence. Companies in this space, such as Lyra Health and Headspace, leverage digital tools to provide scalable access to care. Furthermore, data driven approaches are increasingly used to develop integrated, end-to-end care pathways, particularly in complex fields like women’s health.

Challenges in Clinical Evaluation and Frameworks

Despite demonstrable positive clinical outcomes, rigorous clinical review identifies key deficiencies in evaluation methodology. Few studies adequately address the usability of these technological interventions, and the explicit reason for not utilising or citing specific behaviour theories often remains unclear. This presents a structural risk: while clinical outcomes (HbA1c, BP) are improving, the process validation for sustained engagement is weak.

A critical need exists for a common assessment framework that employs a broad range of measurements.This framework should focus not just on immediate clinical metrics, but on process measurements related to motivation for health behaviour change, long-standing adherence, expenditure and patient satisfaction. The focus on immediate efficacy rather than sustained maintenance leaves a high-risk gap in understanding long-term patient retention and future development reproducibility.

Clinical Efficacy of Digital Interventions: Summary of Key Outcomes

Condition	Technology Intervention	Key Outcome Measure	Observed Improvement / Change
Type 1/2 Diabetes	Real-Time Continuous Glucose Monitoring (CGM)	HbA1c Levels	50.8% of patients achieved HbA1c <7%; >1% absolute reduction for many.
Hypertension	Remote Patient Monitoring (RPM) + Intervention	Blood Pressure Control Success	Doubled success rates in underserved patients via remote monitoring.
Physical Activity	AI Nudging System (KGNN-RecSys)	Daily Steps & Exercise Minutes (MVPA)	6.17% increase in daily steps; 7.61%increase in exercise minutes.
Overweight/Obesity	Mobile Apps (Evidence-Based)	Weight/BMI Reduction	Pooled results showed 0.84 kg reduction; 2.6% body weight reduction in 3 months.

Regulatory, Reimbursement and Ethical Frameworks

Navigating the Regulatory Landscape: FDA Frameworks

The oversight of digital health solutions, particularly those employing AI/ML, must contend with the rapid iteration cycles of software development. AI/ML-based software intended to treat or diagnose is classified as SaMD and is subject to the FDA’s risk categorization framework.

Acknowledging that software can rapidly respond to glitches, adverse events, and safety concerns, the FDA has proposed adaptive regulatory processes, such as the Pre-Cert program. These proposals recognise that the speed of software iteration has surpassed the capacity of traditional, static regulatory pathways. This gap in oversight creates a high commercial risk for developers of continuously learning SaMD, necessitating strategic investment in regulatory compliance focused on adaptive pathways to ensure safety and effectiveness standards are maintained.

Market Access and Reimbursement

Market access is accelerating as payers increasingly recognise the clinical utility and documented cost savings provided by DTx solutions. This acceleration is inextricably linked to the commercial shift away from device-only sales toward the service-oriented model. Securing consistent, ongoing reimbursement streams necessitates that developers satisfy payer requirements for long-term maintenance, cybersecurity, and, most critically provide proven, durable patient outcomes over the full lifecycle of the intervention.

Ethical Governance of AI-Driven Behaviour Change

The deployment of sophisticated AI in health promotion introduces critical ethical considerations spanning issues of fairness, bias, and patient autonomy.

AI tools carry the inherent risk of causing unintended harm through biased algorithms. If training data is unrepresentative, or if algorithms prioritise certain outcomes, this can promote discrimination or inaccurate decision-making, potentially leading to systematic errors that exacerbate existing health inequality.

Algorithmic fairness is therefore not merely an ethical requirement, but a technical one for achieving efficacy at population scale. When an AI system fails for a demographic due to bias, resulting in ineffective nudges, it fundamentally fails its clinical mandate for that group.

A robust debate exists regarding the ethical implications of using personalised nudges, derived from complex AI models, particularly concerning manipulation and infringement upon patient autonomy. However, an alternative viewpoint suggests that targeted health promotion, even if perceived as manipulative, can be justified as counter-manipulation. In this framework, the use of AI to encourage healthy choices acts against the widespread, manipulative commercial "illness promotion" inherent in the marketing of unhealthy products. The argument suggests that by neutralising pre-existing negative influences, manipulative health promotion may paradoxically enhance or restore the target’s autonomy.

Strategic Recommendations and Future Outlook

Strategic Recommendations for Development and Deployment

Mandate Behavioral Science Integration: Future development efforts must formally begin and continuously cite established behavioural models (FBM, TTM) throughout the design and clinical validation process. This integration is the only way to systematically address the pervasive, long-standing challenge of adherence and ensures the methodological reproducibility of clinical success.

Invest in Implicit Sensing and Interaction: Resources must be strategically channeled toward advanced sensor integration and AI inference engines that drastically minimise the need for explicit user interaction. By reducing the cognitive load required (increasing Fogg Ability), reliance on implicit data collection and ambient delivery of interventions will significantly reduce friction, which is the key determinant of long-term patient retention.

Adopt "Ethics as Utility": Developers of AI-driven behavior change systems must recognize that algorithmic fairness is integral to clinical utility. Implementing rigorous bias auditing tools and promoting transparency in how the AI utilises demographic and behavioural data (eg, in the Knowledge Graph structure) is essential. This ensures that the system works reliably across all patient groups, protecting against systematic errors and maintaining clinical integrity.

Future Trends: Digital Biomarkers and Integrated Care Pathways

The digital health market is poised for continued robust expansion driven by the maturation and adoption of digital therapeutics.

The next generation of innovation will center on the development and validation of sensor-based digital biomarkers. These markers track nuanced aspects of patient health and behaviour in both clinical care and research settings, ultimately leading to the approval of novel digital endpoints by regulatory bodies in the US and Europe. These biomarkers will provide objective, high-resolution data that traditional clinical measures cannot capture.

The industry is converging toward the provision of comprehensive, integrated, end-to-end care pathways.Platforms utilising sophisticated AI, such as Biofourmis’ Biovitals, will increasingly leverage predictive analytics to deliver highly personalised interventions across complex acute and chronic conditions, establishing a new standard for precision medicine and population health management. This future state envisions technology that not only monitors disease but proactively predicts and intervenes against clinical exacerbations, transforming the ability of physicians to meet patients’ needs precisely where they occur.

Nelson Advisors > MedTech and HealthTech M&A

Nelson Advisors specialise in mergers, acquisitions and partnerships for Digital Health, HealthTech, Health IT, Consumer HealthTech, Healthcare Cybersecurity, Healthcare AI companies based in the UK, Europe and North America. www.nelsonadvisors.co.uk

Nelson Advisors regularly publish Healthcare Technology thought leadership articles covering market insights, trends, analysis & predictions @ https://www.healthcare.digital

We share our views on the latest Healthcare Technology mergers, acquisitions and partnerships with insights, analysis and predictions in our LinkedIn Newsletter every week, subscribe today! https://lnkd.in/e5hTp_xb

Founders for Founders > We pride ourselves on our DNA as ‘HealthTech entrepreneurs advising HealthTech entrepreneurs.’ Nelson Advisors partner with entrepreneurs, boards and investors to maximise shareholder value and investment returns. www.nelsonadvisors.co.uk

#NelsonAdvisors #HealthTech #DigitalHealth #HealthIT #Cybersecurity #HealthcareAI #ConsumerHealthTech #Mergers #Acquisitions #Partnerships #Growth #Strategy #NHS #UK #Europe #USA #VentureCapital #PrivateEquity #Founders #BuySide #SellSide #Divestitures #Corporate #Portfolio #Optimisation #SeriesA #SeriesB #Founders #SellSide #TechAssets #Fundraising #BuildBuyPartner #GoToMarket #PharmaTech #BioTech #Genomics #MedTech