Originally posted on The Horizons Tracker.
One of the prime use cases of early forays of AI into healthcare was in helping clinicians keep on top of the huge volume of medical knowledge being pumped out each day. Vendors such as IBM promised to scour this torrent and make it accessible to doctors.
Whilst it could be argued that this promise has still not really translated into reality, it is still very much a need of the industry. A recent paper1 published in the Journal of General Internal Medicine highlights both the challenge and a potential solution that will help to accelerate the transition of new evidence into standard practice.
Transforming medical knowledge
Whilst many industries face profound challenges in terms of knowledge creation, the medical sector faces more than most. It takes many years to train a doctor so they are ready to safely practice, but the rate of change in best practice makes it hard to keep up with the latest evidence. The authors cite the 20 year delay in incorporating the latest clot-busting treatments for heart attacks as an example of the difficulties in incorporating the latest thinking.
“There are lots of reasons why new knowledge isn’t being rapidly incorporated into practice,” the authors say. “If you have to read it in a journal, understand it, figure out what to do based on it, and fit that process into your busy day and complicated work flow, for a lot of practitioners, there’s just not enough room for this.”
At the moment, systematic reviews are conducted by the likes of the Cochrane Collaboration, but despite their best efforts the movement of knowledge into medical practice is very slow. It’s a challenge that technology needs to solve. The paper argues that the first step is to ensure that all human readable knowledge is transformed into computable forms.
“A lot of scientific studies result in some kind of model: an equation, a guideline, a statistical relationship, or an algorithm. All of these kinds of models can be expressed as computer code that can automatically generate advice about a specific patient,” they argue.
There are those who are doing better than others in transforming biomedical knowledge into more computable forms that are not only accessible but open to all. The paper cites the University of Michigan Medical School’s Department of Learning Health Sciences as a notable example, with their Knowledge Grid platform storing computable knowledge in digital libraries that is then used to generate patient-specific advice.
“The value of Big Data is to generate Big Knowledge,” the authors say. “The power of Big Data is to provide better models. If all those models do is sit in journal articles, no one’s going to be any healthier.”
You sense that it’s a journey we are only really at the start of, but it’s a journey that the industry very much needs to make, especially with the amount of data available to support clinical decision making also increasing en masse.
Article source: Putting Clinical Evidence Into Practice.
- Guise, J. M., Savitz, L. A., & Friedman, C. P. (2018). Mind the Gap: Putting Evidence into Practice in the Era of Learning Health Systems. Journal of general internal medicine, 1-3. ↩