The challenge is solving the good data problem. I don't think we want our highly trained and burnt out clinicians to be downgraded to data entry clerks.
Nice areas of incremental improvement with automated data capture during workflows is happening but not going to be on the order of magnitude to enable wide spread data quality improvement.
You're absolutely right, Ray. According to estimate, 80% of effort to get insights from data is spent on data cleansing (that's assuming there is data in the first place). Let's see what comes out of the pan-Canadian Health Data Strategy in moving this forward!
Finnie,
Thanks for the links. ML is the type of functionality that will enable the long awaited paradigm shift in the administration of healthcare. The sad bit is that there is little understanding that ML absolutely depends on good data. There has been little progress to remediate the way that organizations and physicians record data. EMR applications continue to ignore good practices with regard to the user interface and things to facilitate good data entry. I recently had a demo of a new EMR and it was obvious that they had fallen into the same traps that earlier EMRs had over the past 20 years. It would be nice to get organizations to understand that there is a problem that remains to be solved.
The joint technical committee set up by the IEC and ISO has published two new foundational standards for artificial intelligence. They represent important milestones in the standardization of AI. See link for further details.
Improving the quality of patient care through the effective sharing of clinical information among health care organizations, clinicians and their patients.
Print
Email
ISO/IEC AI foundational standards
Twitter