Smaller Paddles – Let’s think big about AI in Healthcare

There’s an episode of That 70’s Show in which the characters Kelso and Red bond over the game Pong. Eventually, they get bored and want to up the ante. They both agree there’s only one solution: smaller paddles. We, in the future, laugh at how simplistic these two were thinking. Video games went on to become complex narratives and billion-dollar franchises. How could they be so naïve?

AI (or ML or LLM – whichever you want to use) has come along and promised to revolutionize our lives. Actually, nothing has been promised – the human populace just all got so collectively excited and/or scared that it feels inevitable (see also: flying cars in the 1960s and hoverboards in the 1980s). We don’t know what AI will ultimately do for us, but in the Healthcare sector companies are trying out the tools in a few predictable ways.

• Prep a response to a patient’s message in the EHR. Generative AI can predict how a provider might respond to a patient’s inquiry and whip up a quick, thorough response. A review by the human provider, some slight tweaks, and off the message goes. One study even found that AI responses outperform providers when it comes to empathy – but that’s more a condemnation of the operating environment and societal expectations of providers than praise for AI.
• Write notes for providers. By listening to the conversation between a provider and a patient generative AI can throw that information on top of the patient’s chart and come up with a decent note. A review by the human provider, some minimal changes, and off the note goes.
• Enhance enterprise searches. Finding information in the maze of Sharepoint pages, shared drives, web pages, and medical databases within a healthcare organization is difficult. AI is being piloted to help pull relevant information from these disparate sources quickly for providers.

You’ll notice two themes throughout these initial use cases: they are aimed at providers, and they are all about increasing productivity. From today’s point of view that makes total sense. Providers are reporting record burnout, and profit margins for healthcare systems are just now breaking even. These cases will impact what’s already in front of us.

We collectively looked at this exciting new tool and came up with the healthcare equivalent of smaller paddles. Is that good enough?

Making providers’ lives easier is all well and good, but providers and administrators have two different opinions on how that time saved will be used. Providers’ first thought is “Now I can spend more time with my patients.” Administrators’ first thought is “More productivity means more patients seen means positive profit margins.” Ultimately those two desires will be at odds and no amount of AI will resolve that dispute. So how should we approach AI in healthcare?

Let’s start with how we frame the discussion. Forget the provider for a moment – how might we use AI to move to true patient-centric care? From the patient’s point of view, symptom checkers are about all the AI we offer at this point. They help, but patients could also use help searching vast amounts of health information even within a single healthcare system. Have you tried finding an attachment you were issued 6 months ago in the sea of past encounters in your portal? Patients must write the messages providers are responding to – how might we help them generate more concise requests? A tool to help patients put their requests into medically relevant terms might help them feel empowered and more engaged with their health.

Even those examples may be small in the long run. We know AI tools work best off robust data sets, but our health and care data remains fractured. AI, like EHRs and other tools before it shines a light on inefficient processes we humans have cobbled together over time. We didn’t take the opportunity to update how we operate when EHRs took off. Let’s take this opportunity to think bigger. Let’s create the systems needed for tomorrow’s tools to truly transform healthcare.

Originally published on The Journal of mHealth.