Can AI Help Solve America’s Medical Records Problem?
by John Werner · ForbesIf you don’t work in a medical office, you might be asking yourself, “what medical records problem?”
However, if you’ve ever asked a hospital to send medical records to some other practice or healthcare facility, you might have an inkling of what the country’s medical professionals are dealing with.
The Old Machine: Faxes in Medical Offices
Essentially, medical offices are unable to send each other information in a direct and straightforward way. There’s a physical reminder of this in the back of many offices – making noise, spitting out paper documents.
That’s right — it’s a fax machine.
To many first-time visitors, it might seem strange that 21st-century doctors’ offices are relying on a technology that’s been replaced in most other industries.
But there’s another problem that makes this a unique challenge for the medical world — the stringent privacy standards of HIPAA mean that systems like Docusign or various types of PDFs may not be secure enough to send through email. And if there’s not a safe digital conduit, then it’s back to the fax.
MORE FOR YOU
Trump Vs. Harris 2024 Polls: Harris Leads In 4 Latest Surveys
NSA Tells iPhone And Android Users: Reboot Your Device Now
Sean ‘Diddy’ Combs’ Kids Defend Him In Joint Statement: Here Are All The Major Allegations And Latest News
This Vox article goes into detail about the reasons for the fax machine’s ubiquity in American medicine (you can also read this article at SentinelOne to understand the ramifications of HIPAA).
Those who were paying attention remember how the HHS HITECH act involved “meaningful use” standards set by the Obama administration, incentivizing the adoption of medical health records across the board.
The federal office was adept at getting offices to adopt EHRs and EMRs. However, the government failed to plan for the problem of interoperability, and many medical software systems don’t talk to each other, so humans have to take information from one system and funnel it into another.
The Potential of Artificial Intelligence Models
Let’s turn to a new way to potentially solve this problem.
Theoretically speaking, if AI systems can take data and make it change from any format into any other format, they should be able to help bridge the gap between different medical software systems and thereby automate nearly all of the labor-intensive processes are now done by clerks.
An article at Andreessen Horowitz talks about solving the “messy inbox problem” by synchronizing different types of unstructured or poorly structured data into a lean, mean identification machine, so to speak.
A few diagrams provide suggestions of how to build this kind of architecture, leading to a “native-AI system of record.”
“By being able to replace human labor with LLMs, you’re effectively stepping in front of any existing downstream software systems,” writes David Haber. “The ability to synthesize unstructured voice or text data and systematically extract relevant information earns you the right to initiate (and own) other downstream workflows.”
That idea of downstream systems is relevant. Getting things right up top leads to a tighter funnel, a better way of distributing information.
At the end of the day, using the power of LLMs, we may be able to finally get past some of the data formatting that’s been bedeviling medical records offices for decades, despite the digital changeover years ago.
But that’s by no means the only use case for AI in this way. From finance to law, and beyond, business leaders are revolutionizing business processes by “cleaning up” data systems in these ways. Let’s keep looking at what companies are doing to reinvent their office practices. Soon, the ancient fax may be a thing of the past.