If you set something up and it just goes into a black hole and no one ever sees it you have problems. Not only are clinicians not interested in it, but the data you get will be rubbish. You need that loop, you need that feedback, so that the people who are entering the data can see the value of it and they start acting on it, and they start making sure the data is of a high quality when it goes inFrom “Leaders’ perspectives on learning health systems” BMC Health Services Research 2020
Changing practice is resource intensive and error prone and not to be taken lightly, even when it has the potential to improve care. For more than half a century, AI systems have been developed to change clinical behaviour and failed to do so. Evidence is the established pathway for changing clinical practice. Evidence that can motivate and guide worthwhile change, must be reproducible and transparent.
Any rapidly learning health system undoubtedly uses multiple AI algorithms to measure, analyze and make recommendations for change. Indeed, AI are good at finding complex mathematical relationships among elements in a dataset. Such AI-based systems are designed by data scientist for data scientists, and so rarely require, or even have room for, clinical input. The result is that recommendations made by AI are not trustworthy. They are hard to interpret because of their complexity, their behaviour is unpredictable when applied to data other than the data they were trained on, and they are not explicit about the assumptions that went into their design. One might think that AI were made to replace clinicians, not support them.
Contrast that with systems that put the clinician at the centre of clinical care. They provide explicit, high quality evidence, in a language that doesn't require years of data science training. Clinicians can read such evidence because they can understand it. Understanding evidence is a prerequisite to changing clinical practice. Understanding is key to trust, and trust is key to behaviour change.
Evidentli’s analytics platform puts the clinician at the centre of the analysis: clinicians design analyses by choosing patient populations, outcomes and the clinical events to study; they may even choose to use AI as part of the analysis. The methodology is instantly available for peer review. The results are ready in minutes and can be interpreted, not just by data scientists, but also by other clinicians.
Leave a Reply
Comments, questions or queries? Let us know!"