Observations: Webinar on Artificial Intelligence (AI) for Health Care Delivery, March 24th

29 March, 2022

AI and Machine Learning for Health Care Delivery live web event on Thursday, March 24th.

https://events.catalyst.nejm.org/events/ai-and-machine-learning-for-heal...

I really enjoyed this webinar and can highly recommend it to HIFA members. I have summarised the presentations as well as possible though I have not been able to link each fact to each presentation as I do not have shorthand and the webinar was very full of talkers and presentations.

Here are the comments that I summarised from the webinar

"Paper systems of medical records were not fit for purpose for pulling public health data together for covid management whereas digital systems of records were. AI was not all important. Involving reluctant stakeholders and other groups was. AI did not make clinical decisions - clinicians did often utilising the analysed data created by the AI. There was no direct transfer from algorithm suggestions to action decisions.

"Successful AI implementations were co-designed with health providers and consumers sharing information and data with users all of the way along. Government level and University level decisions aided the management of privacy and sharing data. In some ways AI separating data from identity increases security.

"Scientists and data analysts spent time considering how to balance simplicity against maximum complexity (not least to obtain valid informed agreement from data providers. Explaining using simple models makes it easier to explain and fully inform and to obtain and maintain trust in the data analysis). Clinicians wanted to know how significant the AI model was and if the clinicians understood the AI achi8eved greater buy in. Data gathering continues after the original data collection period to allow remodelling and evaluation.

"An analysis of radiology reports and follow ups. 5% of all reports required a follow up. 33% were not followed up. This study was on 570,000 reports, was 77% sensitive and 99% specific. AI prevented missed follow ups.

"Some myths about AI:

“We don’t have the data for AI to work on.” Wrong all clinical systems do, now.

“Its really hard to do.” No, it is not. There are plenty of AI programs which can be designed to fit over an HER.

“You ned a fancy computer”. Yes, but you can hire fancy computers for $5.15 an hour.

"AI is not the solution. Start with “I have this process to improve or this problem to solve.How could machine learning enable a solution?”

“AI empowers human systems, it does not replace them.”

[...]

AI in Healthcare: https://catalyst.nejm.org/topic/topics/ai-in-healthcare

HIFA profile: Richard Fitton is a retired family doctor - GP. Professional interests: Health literacy, patient partnership of trust and implementation of healthcare with professionals, family and public involvement in the prevention of modern lifestyle diseases, patients using access to professional records to overcome confidentiality barriers to care, patients as part of the policing of the use of their patient data

Email address: richardpeterfitton7 AT gmail.com