An Interview with Paul McGinness, Lenus Health

Ok – let’s assume we are now proactively diagnosing and treating chronic conditions in the community…. We can take a wider lens on the global healthcare challenges faced today to explore prevention. One way of doing this is using AI to help tackle health inequalities – the avoidable, unfair and systematic differences in health between different groups of people. Risk models built with fairness and explainability can and should be used to help surface inequalities and understand better how they can be more effectively addressed.

 

Most predictive algorithms in healthcare are used purely academically, but we spent the last 4 years developing the data pipelines and technical infrastructure required to bring AI to direct care of patients. This led to a ground-breaking deployment in March of this year for the direct care of patients with COPD at NHS Greater Glasgow and Clyde as part of an MHRA-approved clinical investigation. The potential for explainable risk stratification in clinical workflows to reorientate chronic condition care from reactive to proactive is huge. COPD care alone costs the NHS more than £2 billion each year and is growing – that’s unsustainable.

 

Healthcare, for many reasons – many of them valid, is hard to change. But for it to become more sustainable, we must shift the balance of resources from one that is aimed at episodic care to one that matches the needs of the population for proactive, integrated, and ultimately preventive care pathways. AI can support this, but governments (and providers more directly) must invest in technology that transforms rather than digitises the existing. Policy, guidance, and funding must strategically align, or these existing problems will only slow - never reverse.

 

I think everyone reading this believes AI should increase meaningful human contact. Instead of patiently pausing during an appointment while notes or various physiological metrics are taken, that time can be used for a continuous and more informed clinician-patient dialogue. Our experience has been that putting AI-supported clinical workflows in place empowers patients – their care teams can focus the conversation around the key drivers putting that particular patient at risk to optimise guideline-directed care.

 

The workforce will undergo a dramatic transformation in the next decades, affecting both low and high-skilled jobs. Automation will take over many manual and some knowledge-based tasks, but that is not the only change. For example, care teams will have access to more digital biomarkers, and many aspects of clinical information gathering will also be automated. If AI is to become a daily part of healthcare provision, then the ability to evaluate things like model thresholds and being savvy to potential data biases will be important skills for clinicians to have, in a similar way to peer review.

 

There are a growing number of people doing remarkable work in this field. Personally, I have the pleasure of working daily with clinical and technical leaders spearheading digital transformation in their organisations. These are the projects that put Lenus on our path to operationalise predictive AI in clinical workflows. In the case of COPD, I will single out the work of Professors Chris Carlin and David Lowe. Early on, they saw the potential of machine-learning risk stratification, and together we’ve done ground-breaking work on model fairness and explainability that has led to a world-first deployment by one of the largest health systems in the UK, NHS Greater Glasgow and Clyde.