AI-based clustering offers hope for overburdened NHS primary care

The study explores the concept of using artificial intelligence to develop “clusters” of patients with similar health and social care profiles. These clusters, built from large-scale data sets such as the Clinical Practice Research Datalink (CPRD), aim to prioritize individuals most at risk of worsening SCNs and to tailor support more efficiently.

Rising levels of chronic illness in the UK are placing unprecedented strain on primary care. Patients living with multiple long-term conditions (MLTCs) often face a complex mix of health and social challenges that existing healthcare systems are not equipped to manage effectively.

A new study published in the Journal of Multimorbidity and Comorbidity titled "Value of using artificial intelligence derived clusters by health and social care need in primary care: A qualitative interview study with patients living with multiple long-term conditions, carers and health care professionals," investigates how AI can be leveraged to identify and support patients with MLTCs who often experience unmet social care needs (SCNs).

The research presents qualitative findings from 44 interviews involving patients, carers, and health and social care professionals (HSCPs). It addresses pressing concerns about care inefficiencies in primary healthcare and offers recommendations on how AI-derived clustering can enhance holistic, patient-centered interventions.

Is primary care equipped to address social care needs for people with MLTCs?

Primary care was widely recognized by participants as the logical first point of contact for discussing social care needs. Individuals living with MLTCs, as well as carers and health professionals, described the complexity of navigating fragmented services and determining where to seek support. Many participants expressed that while general practitioners (GPs) often serve as the entry point for discussions, the growing scarcity of time and resources has left both patients and providers struggling to address SCNs effectively.

The study highlighted structural barriers, including short appointment durations and difficulty accessing GP services. These factors often forced patients to manage their needs independently. Respondents suggested that expanding roles in primary care, such as employing social prescribers or dedicated care navigators, could help alleviate this burden and create more opportunities for meaningful conversations about SCNs. Such roles were seen as essential, especially when GPs lack the capacity to engage deeply with patients’ non-medical needs.

Moreover, carers and professionals acknowledged that many individuals perceive social care as applicable only in extreme cases, leading to reluctance in seeking support. This stigmatization, combined with logistical hurdles, exacerbates unmet needs in an already vulnerable population.

Can AI-derived clusters offer a realistic solution for holistic care?

The study explores the concept of using artificial intelligence to develop “clusters” of patients with similar health and social care profiles. These clusters, built from large-scale data sets such as the Clinical Practice Research Datalink (CPRD), aim to prioritize individuals most at risk of worsening SCNs and to tailor support more efficiently.

Participants responded positively to this approach in principle. Many believed that clustering could streamline processes, enable early interventions, personalize care pathways, and improve the allocation of resources. For overstretched professionals, AI was seen as a welcome assistive tool that could complement existing methods and increase operational efficiency.

However, several concerns emerged around the reliability and completeness of the data feeding these AI models. Participants feared that not all relevant SCN information is consistently recorded, especially given that such needs often go undocumented in clinical consultations. Additionally, patients and carers expressed apprehension about the possibility of being miscategorized or underserved due to data inaccuracies or systemic biases.

There was also unease about data security and the potential misuse of sensitive health records, particularly by private entities like insurance companies. All groups emphasized the importance of safeguarding patient information and ensuring transparency in AI-driven decision-making.

Despite these concerns, the overwhelming view was that AI could serve as a valuable support system, so long as it does not replace the human elements of care.

How should AI tools be implemented to preserve person-centered care?

Participants emphasized that AI tools must never substitute the nuanced, empathetic dialogue between a patient and their clinician. There was strong consensus that AI-derived clusters should function as aids, not authorities, in care planning. Human judgment, contextual awareness, and emotional intelligence were deemed irreplaceable in providing meaningful care.

Interviewees stressed that care should be tailored to individual circumstances and preferences. Rigid adherence to algorithmic outputs without dialogue risks dehumanizing patients and undermining trust. Moreover, when discussing future risks of SCNs, participants urged a communication approach that is positive, non-alarming, and responsive to each patient's communication style.

The study calls for balanced integration: AI tools should be brief, adaptable for use by multiple care roles, contain up-to-date information, and facilitate, rather than hinder, effective, person-centered interactions. Rather than automating decision-making, these tools should empower clinicians with better information and support patients in making informed, shared decisions about their care.

The researchers offer several key recommendations for implementing AI clustering tools in primary care:

• Ensure interventions are concise and user-friendly to avoid adding to clinician workloads.
• Make tools adaptable across multidisciplinary teams.
• Keep signposting information current and accessible.
• Use AI clusters as a risk-identification aid, not as a final arbiter.
• Train professionals in having supportive conversations that respect patient autonomy.
• Address potential reluctance on both sides, clinicians and patients, in discussing SCNs.