A.I.: Digital Demon or Darling wrap up

Those who attended our evening event in March on Artificial Intelligence were subjected to a deeply enlightening and educational discussion from Neville Board, Chief Digital Health Officer, State of Victoria; and Tara Donnelly, Chief Digital Officer, NHS England.

Thank you to our fantastic host Jason Micallef and sponsor The University of Melbourne and our moderator John Stanway, CEO, The Royal Children’s Hospital.

Here are the top three takeaways from the evening:

  • Artificial Intelligence doesn’t have to work at the bedside or operating table – think of the example of prioritising ambulances through traffic lights; 
  • Define the underlying problem you’re trying to solve and then focus on the solution.   Take a step back and understand / picture the whole model of care that’s at play; and
  • Don’t build it, create the market for it – others will come, partner and help build the solution! 

Here’s the wrap up:

  1. What is Artificial Intelligence?

It means many things to many people.  At Van Wynn’s base definition, it’s computers doing the tasks that humans do. Quite often, we think of it as highly complicated business rules with very large data sets.  However, the actual application is much broader than that. One example is the City of Liverpool and its use of A.I. to change traffic lights to allow prioritising ambulances or Andiamo in designing custom-made splints for children in days for immediate wearing (as opposed to taking months which, by that time, the child had already outgrown them).  Sometimes, it involves simple algorithms that reduce or prevent harm.  It’s in areas of diagnoses and identifying patterns where fatigue can play a part in diagnostic errors.  It’s proven helpful in allowing remote health experts to “plug in” to give opinions.  We can also use it to analyse reactions to certain medicines for people with multiple morbidities.  A.I. is often touted for its cost savings which is not very engaging for clinicians. Should we change the core narrative to focus on safety where the patient benefit is so clear?  There was also the example of a children’s hospital in Pittsburgh, USA which uses A.I. to flag “red, amber, green’ issues for full visibility for everyone to see:  clinicians, nurses and parents.

  • Privacy on the Use of Data and the Issue of Cyber Security 

The consensus was we haven’t done a great job on this in getting people to agree to use their data and addressing their trust and privacy concerns upfront.  We need to be better at gaining the “social licence”, acting responsibly and winning the public’s trust.  We need to improve our messaging and ask upfront so people offer their consent to participate.  How do we navigate this?  NHS UK uses their “app” as the main way to authenticate yourself and then you gain access and can connect to larger groups.  It seems the public is on board and their appetite for adoption is way ahead of the government:  they are not waiting.  In many cases, parents are already granting clinics temporary access to their child’s data.   So, is it about better engagement or is it simply because the government is lagging? The discussion then moved into the issues around the electronic medical record and the specific example of My Health Record in Australia and how messaging created a lack of assurance for the public.  

  • Electronic Medical Record (EMR)

Digital technology can play a key part in connecting and helping patients and doctors but there can be unintended consequences:  how much time are we spending on the computer vs. actual patient time?  We need more eye contact!  Quite often, the issue is a lack of workflow consideration.  Eye-tracking analysis on key tasks has shown the discharge department is often more concerned about the time it takes to discharge a patient rather than on the patient being discharged.  We need to get to the point where technology is demonstratively helping doctors but it’s sometimes hard to show.  “Voice applications” should improve things.  The consensus was high volume tasks should be the focus of A.I. and using digital technology to connect.  We need to shift “high cost, low access” areas to “low cost, high access”. EMR is not a panacea but it’s one step in the right direction.  

  • Today’s Pragmatic Assessment

It was noted long-term conditions consume 70% of our hospital resources.  Most of the time, we’re looking at the ex-post analysis with the focus on hospital-acquired complications (HAC’s) and their consequences.  Perhaps, we should be focusing on actual prevention.  NHS UK has flagged an interesting statistic on “digital therapeutics” based on over 350,000 applications:  there’s an inverse relationship between good practice and real beneficial impact vs. well-liked and high usage apps.  We should be focusing on the mental health space as this would be a game changer in today’s hectic world.  NHS provides a digital front-end for everyone which acts as a “front-door”, allowing people to book appointments and delivering substantive factual health information (as opposed to simply well-liked apps).  NHS was now looking into digitising other market segments like maternity, child services and the fringe gap (i.e.  homeless) to ensure nobody was left behind.   The next step is moving into long-term health issues.

  • The Market, its Forces and Digital Adoption

Great innovations often start with and are driven by the market. Sometimes it’s better to get excited about the technology and focus there whilst letting the implementation side figure things out.  However, we need to ensure the supporting care system is in place.  One example given was that it’s great to have people remotely monitoring their own blood pressure but consider the support when there’s an actual event and possible stroke.  Can we get better at spending investment dollars?  We should conduct assessments whether we could have done things better, ex-post.  In some areas, we should change the settings and accessibility for frequent users. However, in other areas, we need to be mindful whether the end segment can use / access the technology.  There was a feeling we should be focusing on digital screening tools and practical applications (and keeping things simple – the “KISS” acronym) but with an understanding and appreciation trials were always easier than full market implementation as we need to improve testing at scale.  We should examine other impediments to digital adoption including medical paternalism. The reality is digital convergence is happening all around us but the health sector seems to be the last bastion. Next segments on which to focus include mental health, frail /elderly and complex patients.  In the end, it’s still all about the “care equation”.  

  • Funding Model

Within some hospitals, the top 300 complex patients consume 15% of all resources.  One key issue is the current funding model doesn’t support the care model needed: we need a different form of funding. One possibility is to consider 3 or 4 different models of care.  The challenge though would be the associated workforce required to implement and the amount of work to have the appropriate resources to support it.  Another idea is to provide these complex patients with a care co-ordinator and try to reduce their admissions.  Departments of Health are also evaluating different funding models including providing financial incentives.  In Germany, the government is considering providing a share of the savings for improving health and reducing costs.  Other ideas being mooted included models that provide a payment for keeping patients well.  Some discussions include offering financial incentives for accelerating digital adoption with even possible disincentives for slow adoption.  The bottom line: it’s a very complex equation and we should try to focus on solving one problem at a time.  

  • The Role of Government and what NHS England is Accomplishing Digitally

The government’s role should be to set out and establish the general Code of Conduct with commensurate rules.  From a practical standpoint, we need to show the clear evidence for effectiveness and articulate the associated benefits.  Things don’t always require a randomised control trial (“RCT”). However, we need to understand and balance any potential harm vs. unlimited parameters.  One example is the “brush DJ” app where it has proven 88% of kids will brush their teeth longer when listening to a song they like.  Does the government really need to conduct, and spend all that money on, an RCT for this?    

NHS England has defined its role and delineated 10 aspects as it rolls out its new digital platform and patient interface to the entire country – its “digital front door”:

  1. NHS website provides trusted quality information and comprehensive content / app library to allow people to search medical conditions and perform symptom checks;
  2. Used to validate the individual before granting access to other groups and information;
  3. “111 online” allows people to get urgent healthcare online through their laptop, smartphone ort other device;
  4. You can book doctor appointments and easily change / reschedule them;
  5. Access your summary care record;
  6. “Digital Maternity” delivers a digital pathway of care from conception to post-natal; 
  7. (“Digital Child Health”, with a child’s key health information, will be added to allow appropriate sharing with those involved in the care;) 
  8. Delineate organ donation preferences;
  9. Refill / repeat prescriptions;
  10. Seek to partner with quality solution providers rather doing it themselves in-house; and 
  11. Ensure a wide and universal digital participation by helping those who will benefit the most by accessing it – for example, the homeless (giving them actual mobile telephones) and those with dementia.

Please take the time to review Tara’s full NHS presentation on “Digital Innovation” on our Events insights page.  It’s in three parts:

            Digital Innovation and How to Make It Work For You At Scale: part 1

            Digital Innovation and How to Make It Work For You At Scale: part 2

Digital Innovation and How to Make It Work For You At Scale: part 3