4 Physician Insights about Evaluating AI Healthcare Solutions

Posted in Clinical AI, Machine Learning, NLP, Innovation, Pieces Predict, Healthcare By admin Posted on

How can Clinical Information Officers be confident about which AI technology will truly enable clinicians to focus more on patients and less on administrative tasks? AI is ready and able to make a deeper impact on the day-to-day work of care teams, however many healthcare organizations aren't sure how to make AI work for their teams, and don't feel fully equipped to make these decisions - and they are not alone. According to Jim Ritchie, MD, Consultant Renal Physician and CCIO at The Northern Care Alliance NHS Foundation Trust, the discussion around AI can feel confusing and overwhelming. Yet, that doesn’t mean healthcare organizations should not proceed.

August Socials (1)-1

Here are 4 insights from his experience:

  1. Understanding the goals of an AI solution and the outcome is critical: “So much of the discussion about AI has been focused on predictive modeling or risk scores but AI must also be about helping me to get work done rather than giving me another number to think about,” said Dr. Ritchie.
  2. Considering strictly clinical AI as a separate animal is a must-do: “Helping clinicians to get work done, like streamlining the clinical handover process, improves operational efficiency and much more,” he said. “It’s an easier fix than the strictly clinical use cases and more appreciated in many ways.”
  3. Lacking a process and standards for evaluating AI is challenging: “Most healthcare organizations are not adept at evaluating AI technology, ”said Dr. Ritchie. ”They often lack a structured approach to evaluating features and usability outside of foundational technical requirements like cybersecurity essentials. But standards still cannot inform whether an AI technology will work in your environment, if it is the right tool for your technology infrastructure, or whether the training group is representative of the population you serve."
  4. Playing in the sandbox can help: For Dr. Ritchie, this problem requires action. He says, “That’s why we’re exploring building a sandbox environment for AI vendor testing. Even though that is costly and time-consuming, it is important.”

Dr. Ritchie is right. It can be hard to evaluate new technology like AI and natural language generation but there are tools to provide guidance and make it easier. That’s why we developed our Essential AI Technology Checklist. It outlines considerations for  how to pick the right technology to solve the right problems – for physicians, nurses, patients, and the bottom line.