Notes from the AI Medicine Trenches. Part 2 – data and apps

This post was written by Brendan DUNPHY (CEO of C-BIA Consulting) and first published on AIMed

The following are notes penned by an AIMed Europe conference attendee, who listened intently as Dr Anthony Chang and other experts in the field of medical artificial intelligence explained how the emerging paradigm of medical AI would be powered by structured (and unstructured) data combined with healthcare apps – these are his observations:

Does a lack of structured data and notes slow progress?

The biggest problem domain is that much of the data captured today is unstructured notes, probably as much as 80-90% of all data. Clinicians’ notes may be subjective and may not be complete or accurate. Medical notes perform several purposes and are not simply a factual record of the patient/clinician interaction, diagnosis or treatment.

Extracting meaningful data from patients itself can be hard, much may not be relevant akin to a stream of consciousness. Implicit assumptions and clinician bias hide between the lines so extracting meaning from these notes (which may also contain patient identifiers…) is challenging. Natural Language Processing (NLP) is developing and domain-specific medical dictionaries exist, but manual cleansing is still necessary to ensure notes are accuratly interpreted by machines.

“AI is an accelerator to give us escape velocity”

It is clear from AIMed Europe a few weeks ago that a growing number of clinicians are more than willing to integrate data-driven tools in to their decision-making workflow, whether AI or ML based. This could be to fill information gaps, better spot conditions, verify diagnosis, validate treatment plans or a myriad of other ‘data light’ areas under served by reliable and timely data at the point of need today.  Many are young and starting out on their medical careers, digital natives coming to terms with a largely analogue professional world slow to adapt to digital trends and expectations. The NHS is encouraging their interest through a Clinical Entrepreneur program, soon to be open to its 4th cohort which will for the first time include non-clinical entrepreneurs.

Healthcare: just another app?

Patient-pressure is on the rise as digital natives used to social media connectivity, YouTube content and Amazon service levels and responsiveness struggle to adapt their behaviours to the ‘one size fits all’ NHS. For them Healthcare should be just another app. This pressure is amplified by the trickle of data from digital wearables and lifestyle apps and the 24/24 real-time health monitoring these enable. This will soon become a torrent as prices fall, sensors become wireless and accuracy becomes medical grade. Go to a surgery full of sick people and wait indefinitely? Book an appointment by phone? Receive a letter cancelling an appointment? These are analogue processes ill-suited to a digital age and a major driver for a slew of apps designed to bring primary care onto the smartphone and into the 21C. In China the uberisation of primary care is already well developed with smartphone access to remote GP’s flipping the existing place-based system, creating transparency by encouraging patient feedback and ranking of doctors and making some stars, more akin to You Tube gaming stars than GP’s. Could it happen here? Should it happen here?

Whose data is it anyway?

‘Data natives’ go further, demanding access to ‘their’ personal medical data and freeing themselves from the perceived tyranny of the monolithic NHS, enabling choice and carrying their records with them wherever they go, presenting whenever they need. No more repeated tests or X-rays because the clinician doesn’t have access to test results conducted elsewhere, second opinions may soon become the norm.

This is a review of AIMed Europe 2018, learn more about the leading global event for clinicians in AI in healthcare here.

Leave a Reply