Artificial Intelligence can fail without the proper health context
We are at an inflection point in healthcare technology. Vast data resources have been unlocked through the transition to electronic medical records, value-based care is requiring sophisticated analysis of patient outcomes and machine learning, and artificial intelligence (AI) has evolved rapidly. All the tools are ready - and the difficult part is what comes next.
Just a pause before we continue, did you know that the Ninsaúde Apolo clinic and office management system is perfect for new clinics? Click here and discover exclusive advantages.
In my career in health and oncology, I am a chemist, pharmacologist, entrepreneur, analyst, academic, researcher, venture capitalist, and technologist. I saw this "difficult part" from many angles. Changes don't happen easily in healthcare. The decision-making systems in our industry today are inefficient, full of flaws, and human prejudices. This is evident in oncology, my company's focus area, and also in my brilliant wife's specialty.
I realized that the problem is not technology, but context. We need to share a vision. We need to translate information back and forth between doctors, researchers, patients, and computers to ask better questions and find better answers. The way to bring health leaders together around AI is to invite them into the right contextual setting for the findings.
How do we contextualize this? Earlier this year, I joined the new National Academy of Medicine Healthcare Intelligence/Machine Learning working group. Together with 35 other health leaders, we are outlining the promise, development, deployment, and use of AI for policymakers, suppliers, payers, pharmacists, technology companies, and patients. Each part of our health system needs a better translation:
- Doctors: Doctors are communicators, putting their knowledge into decisions about patient care and expectations. AI needs to understand and evolve within that framework, using a physician's experience to ask informed questions from vast data sets. It shouldn't stop there: AI must present physicians with complex statistical recommendations in an easy-to-use format and close the feedback loop with an analysis of what worked. As health care evolves, the value of medical translation expands. Doctors will translate increasingly complex concepts for patients, as well as translate how medical expertise is applied to machine learning and how medical practice is transformed based on real-world data.
- Patients: Dr. Google is almost always a second opinion in the exam room. With the advent of machine learning, patient data literacy should also be a focus. Patients must and can be involved in care decisions: weighing risks, cost, and discomfort based on real-world data on what works in their exact situation. It is essential that patients can contextualize with their providers what matters to them.
- Payers: has the patient improved? Was the treatment we approved the most economical? Where can we reduce risk while still innovating? Health plans know that the rising costs of current health care are not sustainable. Adding the payer's context around the reimbursement goals in science can help reduce costs and improve results. Payers could certainly do a better job of translating why they make decisions about denials and costs, than the cryptic explanation of previous benefit letters or refusals of authorization sent today.
- Researchers: how to improve our drug discovery and clinical trial process is the subject of a book, not a point, but it is fair to say that there are vast opportunities for better translation in the pharmaceutical, device, and biotechnology sectors. I am very excited to improve how researchers intelligently select safe patient groups for clinical trials that reduce adverse events and use real-world data to show that a small control experience translates to broad and diverse patient segments.
- Politicians: regulators play a crucial role in the responsible adaptation of AI to the health ecosystem. They translate when these models are medical devices and when they should be reimbursable. They will monitor issues related to safety, liability, and even unintended inclination if the models are built based on biased data. Like doctors, politicians are crucial interpreters and need to translate needs across all parts of the health care system effectively.
- Technology: Technology leaders share the burden of translation here as well. Healthcare technologies are often developed in a vacuum - far from hearing what doctors, patients, payers, and pharmacists need. The first comments about AI promised miraculous cures and produced few results. To be effective, AI needs to be carefully trained to contextualize results in more practical “so-what” actions. It is a conversation and it is not easy.
Until we can respond with clarity and consistency, "Should this patient take this medication, will it work better than another, and is it worth the price?", There is more context needed to do so. Artificial intelligence is just a tool to help us get there. It will only work well when it is very well trained and understood.
Source: Brigham Hyde para Forbes