Three ways artificial machine learning will impact medical coding

Computational resources are catching up with all of the healthcare data that’s been accumulating.

Over the past few years, there has been a lot of talk about artificial intelligence and machine learning in healthcare, much of it geared toward drawing attention and eliciting a reaction, like how technology could replace physicians. But, if 2020 and a global pandemic have taught us anything, it’s that we have a lot of gaps in our healthcare system, much of which begin with how we perform foundational administrative tasks like documentation and coding.

No one pretends this part of the job is their favorite. In fact, these administrative activities are a leading cause for physician burnout, which was at 42% before the pandemic

Despite those troubling numbers, we are in a really exciting time within the industry, where computational resources are catching up with all of the healthcare data that’s been accumulating. That means we will be seeing more and more technologies that will reduce the administrative burden and enable providers and clinical staff to get back to the reason they went into medicine in the first place: treating patients.

Machine learning and medical coding

This is especially true in the world of medical coding, where the number of annual errors hovers at about 30%, with billing errors reaching as high as 80%. While that number is alarming in a good year, billing errors can make or break a medical business in 2020, one that is reeling from the significant financial impacts of COVID-19.

With 2020 in the rearview, the industry still faces a tough road ahead, but there are many reasons to be optimistic about the future—starting with how machine learning will impact medical coding, allowing providers to spend more time with patients than on their computers.

Augmentation

The ability, and technology, to augment the coding process is available – and starting to be implemented. These tools give clinical and billing teams “super powers” by providing code suggestions based on notes they have entered into the EHR, improving the accuracy of the process at the frontlines.

Autonomous coding

Humans are a critical part of the coding process, especially as we build AI systems. Autonomous coding without having learned from human providers and coders during the augmentation phase will lack the efficacy we need from the technology. Over the next few years, as we learn from those closest to these systems, we will begin to see tools that codify charts without any user intervention or user input.

Audit

While this is a need for organizations, it simply isn’t an option due to resource constraints (e.g. staff and money). As we improve existing tools, we will begin to see the ROI and cost/benefit tip toward widespread accessibility across both provider and payor organizations.

Today, we have such a complex coding system that adds to providers’ administrative burden—an average of 16 minutes per patient—which is becoming unsustainable for many in medicine. But, we can see the bend in the road up ahead, which leads to a much more sustainable and “super-powered” future, where the machines do the work necessary to enable providers to do what we need them—and they love—to do.