I appreciate the perspective of Christopher Longhurst in his new NEJM AI article. Two key words: local and outcomes. 1. Local: If we only validate models generically in a laboratory setting, we will miss out on the true promise and potential of these new tools. 2. Outcomes: Testing an algorithm is important, but what is most meaningful is whether it can improve clinical outcomes. Remember to measure what matters. Regardless the type of AI you are using, how is it helping providers or patients? Is it improving healthcare? As payers, government agencies, and technology companies continue to focus on AI, they need to commit to improving patient outcomes by supporting, funding, and investing in AI models where outcomes are locally assessed. Please take a few minutes to read Chris’ article and share it with the AI stakeholders and decision makers at your organizations, the future health of our patients and communities will be improved as a result.
💯% agree with Chris. He spoke about those two things at the recent AMDIS conference too!
Going to include this in the next newsletter! Good stuff!