By Arianna Poli (Linköping University)
Introduction: why social and behavioural data matter
In the evolving landscape of healthcare, the move toward personalised medicine is driven by advanced technologies. But its true impact lies in how it helps us understand and care for people as individuals, taking into account their unique social and behavioural contexts. While clinical data such as biomarkers, imaging, and diagnostics have traditionally dominated predictive models, there is growing recognition that health outcomes are deeply influenced by social and behavioural factors. These include education, income, housing, lifestyle, stress levels, and access to care. Together, they are known as the social determinants of health.
In rehabilitation, where long-term recovery and quality of life are central, these factors are especially critical. The PREPARE project embraces this complexity by integrating socio-behavioural and sociodemographic data into AI-driven patient stratification models, aiming to deliver truly personalised and holistic rehabilitation strategies.
The role of socio-behavioural data in predictive healthcare
Health and rehabilitation outcomes are shaped by a complex interplay of clinical, behavioural, and social factors. While medical data provides essential insights into a patient’s condition, it often tells only part of the story. Social and behavioural dimensions, such as daily routines, social contexts and living conditions, can significantly influence how individuals respond to rehabilitation and how well they recover.
These factors are central to understanding patient needs and tailoring interventions accordingly. By incorporating socio-demographic and behavioural data into predictive models, healthcare systems can move beyond generic treatment plans and toward more responsive, person-centred rehabilitation strategies.
Enhancing risk stratification with AI
Artificial intelligence is increasingly being used to support decision-making in healthcare, including rehabilitation. One of its most promising applications is in identifying patients who may benefit from additional support or tailored interventions. By analyzing patterns across diverse data sources, AI can help clinicians anticipate which individuals are at higher risk of complications, delayed recovery, or readmission.
Importantly, models that include socio-demographic and behavioural information, such as living conditions, employment status, or access to care, can offer a more complete picture of a patient’s situation. This allows healthcare providers to move beyond purely clinical indicators and consider the broader context that may influence recovery.
For example, a patient with a stable medical condition but limited social support may face challenges that are not immediately visible in their health records. AI tools that account for these factors can flag such cases early, enabling timely interventions like home-based care or community support services.
This kind of risk stratification helps ensure that resources are directed where they are most needed, improving outcomes and making rehabilitation more efficient and equitable.
Challenges in integrating socio-behavioural data
Despite its importance, integrating socio-behavioural data into healthcare models presents several challenges:
- Data collection and standardisation. Unlike clinical data, socio-behavioural information is often unstructured, self-reported, or context-dependent. In addition, this type of data is not always prioritised in daily clinical practice, which means it may be inconsistently collected or missing for some patients. Collecting reliable and comprehensive socio-behavioural data across diverse populations requires robust methodologies, standardised workflows, and strong ethical safeguards.
- Privacy and ethics. Social individual data can be sensitive. Ensuring patient confidentiality while enabling meaningful analysis demands advanced privacy-preserving techniques, such as federated learning and differential privacy.
- Contextual relevance and cultural sensitivity. Socio-behavioural factors are deeply embedded in cultural and local contexts. Variables such as family structure, community support, and health beliefs may differ significantly across regions and populations. Ensuring that models account for these differences without oversimplifying or misrepresenting them is a key challenge in developing equitable and generalisable predictive tools.
Real-world applications and impact
Across healthcare systems, the integration of socio-behavioural data is beginning to reshape how care is delivered and how patients are supported throughout their recovery journeys.
In many hospitals, predictive models are now being used to identify patients at risk of readmission. These models are more accurate when they include variables like housing stability, employment status, and mental health indicators. For example, a patient recovering from a stroke who lives alone and has limited mobility may benefit from additional home-based services, which can be flagged by a model that understands their social context.
Public health initiatives are also leveraging behavioural data to design community-based rehabilitation programs. In areas with high rates of chronic conditions, local governments and health providers are using data on lifestyle patterns, such as physical activity levels and dietary habits, to create targeted interventions that support long-term recovery and prevent relapse.
These examples show that when healthcare systems embrace the full complexity of human lives, they can deliver care that is not only more effective but also more compassionate. The integration of socio-behavioural data helps move rehabilitation from a one-size-fits-all approach to one that truly meets people where they are.
How PREPARE is addressing the challenge
The PREPARE project is at the forefront of this transformation. By integrating socio-behavioural and sociodemographic data into its AI models, PREPARE ensures that rehabilitation strategies are not only clinically sound but also socially and behaviourally informed.
Key innovations include:
- Federated learning frameworks. These allow data from multiple sources to be used without compromising patient privacy, enabling robust model training across diverse populations.
- Multi-domain data integration. PREPARE combines clinical, social and behavioural data to create comprehensive patient profiles.
- Pilot studies across nine diseases. These pilots test the effectiveness of socio-behavioural data integration in real-world rehabilitation scenarios, helping refine models and validate outcomes.
Looking ahead: a more human-centric future
As healthcare continues to embrace AI and data-driven decision-making, the inclusion of socio-behavioural data will be essential for achieving equity, effectiveness, and empathy in care. PREPARE’s approach exemplifies how technology can be harnessed to understand the full spectrum of human health, beyond biology, into behaviour and society.
By embedding these insights into predictive models, we move closer to a future where rehabilitation is not just personalised, but truly person-centred.
.
Posted by: Nayara Güércio (Trilateral Research)