ISO/IEC AI foundational standards

Hi Peter,

Your comment "vendors can already easily use old-fashioned data queries to determine the most likely needed codes from the appointment type and patient history" is where part of the problem is. While most EMR applications can do queries the problem is the way that the data was entered in the first place. Appointment 'types' are not going to provide much clinical insight. "Patient history" covers a lot of concepts. Reviewing physician entered visit notes is pretty sketchy. A visit reason if coded with ICPC has been reported to be useful. The visit diagnosis may or may not be helpful because of the use of free text or the use of a small selection of ICD-9 codes. This is where a well designed user interface and using a good SNOMED subset would be extremely useful. The Problem List is a crucial place where good data would be very useful. Physicians find it a real chore to keep the problem list up to date and this is one area where things could be improved by the EMR vendors but in my experience vendors have not been interested in improving on what the problem list looks like or how it functions.

Octo Barnett paraphrased "There is no free lunch." by saying "There is no free text".

Ray

Do you know if Videa collects a consent from dental patients to include their x-rays in the training database?

Thanks for the thoughtful reply!

The domain of AI and ML into health services is of considerable interest to me for my teaching and research.

One example that I have learnt a bit from is www.videa.ai/ .. In my teaching I use an interesting real life case study on how Videa has evolved its business model and the challenges it (still perhaps) faces in actual adoption. There are certainly specific other stories around -- and admittedly many of them will be more tech-influenced that others.

My two cents is that it is a "process" of innovation and adaptation -- and some leaders will emerge in this space eventually. And one absolutely essential part of the process is perhaps a collaboration-focused Executive Education program centred on senior medical professionals who have more insights on how "actual intelligence" can be (possibly, if at all) combined with artificial one . The Videa example may provide some initial answers...

Despite being in the tech-business domain I am also often surprised at the human entrepreneurial spirit. Try the driver assist mode on your car on long road trips, for example. It's a relief on winding hilly terrain or freeway jams !!

It will be interesting to see what machine learning is incorporated into EMRs when vendors can already easily use old-fashioned data queries to determine the most likely needed codes from the appointment type and patient history.

I have a dim view of putting artificial intelligence before actual intelligence.

As a physician I really hate working as a data entry clerk. My experience is that EMR vendors by an large have very little idea of how to make things easier and quicker to accomplish good data entry. I have seen incredible resistance to incorporating features that would dramatically speed up accurate data entry. There are things that an entry level computer programmer could do. Unfortunately, if these features are not built in when an EMR is initially developed I have seen incredible resistance by vendors to change things and they say that 'no one is asking for these changes'. My experience is that physicians are not aware of some simple things that could be done to facilitate good data entry and they are not asking for these changes. Each time I have described these things or demonstrated these things, physicians would say 'what a great idea' but after that they don't bother to ask their EMR vendor for things to be changed. There have been a few times when these physicians talked to their vendor the typical reply would be ' no one else has asked for this' or ' why would you want to do that'.

There seems to be no effective venue where issues like this can be presented to EMR / EHR vendors. There seems to be no effective way that users can get significant changes incorporated into the applications that we use for many hours every day.

Totally agree, Steven. We want our clinicians to deliver patient care and not be stuck looking at computer screens while having patient interactions.

As you noted, automated data capture will help improve data quality. Others include natural language processing, consistent and widespread use of coding and other standards, etc.