For a long time treatment approaches in primary care were quite simple with a range of medications and therapies. The medications targeting a specific set of conditions or symptoms, while therapies treat some conditions such as certain mental health conditions and MSK. In reality medications cause side effects which means a broader range of medicines are available complicating the simplicity but the British National Formulary helps to navigate the increasing variety of medications, medicinal products, and side effects. But with new approaches and technologies maturing things are about to get a lot more complex. New systems and approaches need to be implemented to navigate this complexity here artificial intelligence could demonstrate tremendous value and unlock new approaches.
Social prescribing, apps, genomics and...more?
Healthcare systems in the UK, Canada, America, Australia and Scandinavia are taking an increasing interest in the use of non-medical interventions to address peoples social needs - termed social prescribing. This brings a whole range of community activities into play that can be potential interventions to address social determinants of health. For example, a patient comes to the clinician with a mental health condition after several interactions we find that they are very anxious due to financial difficulties and so are struggling to support their family - rather than medicating to address the mental health symptoms the patient chooses art classes and community support services which helps them to manage their anxiety and finances improving their health and situation. Fortunately social prescribers are becoming established as a new role in health systems to help identify the right non-medical intervention. This means a clinician doesn’t need to know the specifics of what’s available but needs to be aware of the potential of social prescribing as an intervention.
Great - so there’s medications, therapies and social prescribing, but what about apps? The NHS apps library and others have now been created so those who might have been considered for therapy (e.g. CBT) might benefit from a mindfulness app. These libraries incorporate an assessment process to check apps are safe and secure. So maybe interventions through apps need to be considered in consultations.
We also shouldn’t forget the genomic revolution, after all the cost of genome testing is so cheap it’s becoming a popular birthday present, which has it’s own ethical issues! A persons DNA can affect how they metabolise medications. In the near future we will be able to match a persons genomic markers to medications to improve efficacy. For example it's estimated 5-10% of people with Caucasian descent poorly metabolise codeine into morphine so it acts as a placebo; while at the other extreme a small subset of people rapidly metabolise codine leading to quick pain relief shortly followed by no relief. To be effective medication dosage and frequency needs varying to match a persons genetic makeup. 3D printing of medications could better personalise medications informed by a persons DNA provide optimal dosage and frequency.
Some of the above emerging technologies and approaches are already impacting health care systems but all are expected to be in reach within 5-10 years. A complex picture emerges and we’ve not even considered the variety of medical devices, point of care systems and testing which are similarly becoming more complex and capable. Where previously there was a relatively simple treatment selection we now have a whole gamut of options. This is fantastic to enable better personalisation and choice to support meeting health goals. But how should these options be configured and arranged to best suit the needs and preferences of the person alongside the advise and medical expertise of the clinician?
Truly Personalised
Often AI articles, discussions and utilisation focuses on AI as a convergence tool - by this we mean software honed to give specific decisions. This might be a clinical decision on a diagnosis, triaging or treatment e.g. does this person have a tumour, does this person need to see a GP, where should the person be referred. Of course this is natural considering the pressures on health systems globally - increasing demand, complex conditions, increasing costs and decreasing numbers of clinicians. However there is little consideration given to the potential for divergent AI - by this we mean software that works across many information sources to give the range of potential scenarios many of which may not naturally have been considered.
The concept of divergent AI means a tool that creates a list of all potential combinations of interventions from medication, apps and social prescribing including genetic information to further personalise. This is AI that augments clinicians by providing options of interventions - options that consist of combinations of technology and approaches to create truly personalised pathways. Clinician expertise and knowledge combined with the patients knowledge and preferences makes it possible to come to an agreed approach that works for them. The AI member of the clinical team does the admin task of searching the information to augment clinicians and enabling patient choice.
AI is only as good as the data that informs it, but we are already in a strong position in this case. The BNF is already available digitally so all the medication information is digitalised and validated. Apps track key parameters (adherence, outcomes etc) and have a verification process. Genomics information infrastructure is being developed but it is all digital, information is gathered with high precision instrumentation providing a level of quality assurance. Social prescribing activities are potentially the lowest quality data but we are already seeing localities developing databases of community activities. 3D printing of medications is still in its infancy and there is still significant progress required for medication metabolisation rates to be fully mapped to the genome, but some mapping has started. The data for each of these potential treatment methodologies is standardised or easily standardised but in silos. This information can potentially be of sufficient quality to power AI tools, but needs validation and examination for biases.
AI powered planning
To create a step change in the delivery of care new approaches need to be developed. Currently consideration of AI applications remain mostly focused on converging on an answer. However this limits us to thinking about creating software that replicates current processes carried out by individuals. The real change will be unlocked by the utilisation of AI to do things that are currently not possible. In the near future with such a dizzying array of potential interventions AI is an ideal tool to unlock truly personalised and optimised care - not by generating a single answer but a range of potential options upon which clinical expertise and patient preference can be applied.