Pulmonology Coding Alert

Will the price point deter future adopters?

In the first part of this article, Pulmonology Coding Alert examined some key questions medical coders have about introducing artificial intelligence (AI) coding solutions into their healthcare organization. Here in part two, you’ll learn the importance of auditing the technology’s code choices, if the AI can scale, and how to vet potential vendors.

Before reading part two, check out part one, which appeared in Pulmonology Coding Alert, volume 25, number 10.

Will You Need to Audit the AI’s Code Choices?

One issue that arises with using AI to suggest medical codes is who do you contact if the code recommendations are incorrect?

“There isn’t really anybody to talk to, but there is somebody working the algorithm. There was somebody that [was] trained [on] that data set,” said Jessica Miller, MHA, CPC, CPC-I, subject matter expert at MediCodio, during her session “Successful AI Integrations: Case Study” at AAPC’s HEALTHCON 2024. She added, “We’re not without a resolution of finding out why the presented answer didn’t meet again that expectation.”

Auditing the AI’s code suggestions is key to ensuring the codes submitted are correct. Reviewing the technology’s recommendations will add quality assurance to the process. The human coder needs to make a note that they’ve reviewed the AI’s code recommendations. If the technology has gone through the process and nobody reviewed the codes, then your organization can find out when the system suggested the codes and who needs to review them.

How Will the AI Continue to Learn?

Similar to human coders, having the AI system review previous results, guideline changes, and annual coding updates will help it learn to be more accurate and efficient in its coding recommendations.

“What I want the system to do is analyze the report and give me all the potential solutions, and then over time, I get to train it and tell it, ‘this is the right answer,’” Miller said. “This is actually a great way to reduce those biases and also to transfer in that real-time processing of the information because it’s going to get your feedback,” she added.

The crux of AI is data-in and data-out because the data output is dependent on the quality of the data input. If you have a poor dataset to start with, you’ll receive incorrect answers.

By tracking and watching what the technology generates, you can fine-tune the accuracy of the system by providing better data to start with over time.

Can the Technology Be Scaled to Meet Your Needs?

Yes. Regardless of whether you have a small rural practice or operate a hospital system in a large metropolitan area, AI can be sized up or down to suit your processing needs without requiring additional overhead or other technology solutions.

“Scalability can make sure that your artificial intelligence solution accommodates those large volumes. And it has to be able to again handle those processing times,” Miller explained.

Will the Interfaces Be Easy to Use?

A user interface (UI) is one of the trickiest aspects of a technology to develop because whether it’s user friendly or not is subjective. Each user has a different level of comfort with technology, and what is easy to use for one person may be incredibly complicated or too simple for another. The UI’s design comes down to personal preferences.

By building an AI coding solution with a customizable interface, you’ll allow the human coders to change aspects, move buttons, or adjust windows to suit their comfort level and efficiency. When the technology vendor visits your organization to discuss the AI, they’ll work with you to understand your needs.

“We have a conversation and we know what it is you’re doing. We can probably figure out a way to use technology to do it faster, and that doesn’t mean taking the responsibility away from the coder,” Miller said.

Will the AI Technology Be Secure and Compliant?

This is healthcare, and there are rules surrounding every technology that enters the industry to ensure it maintains HIPAA compliance and meets strict security requirements to protect the patient’s data.

Beyond the industry standards, if your organization has additional compliance and security requirements, you’ll want to do your due diligence and ask the vendor to ensure the software solution meets your needs.

How do You Inspect Vendor Reputation and Support?

AI companies are popping up all over the map right now as the industry is growing quickly. This sharp growth can make it hard to know which vendors are reliable and trustworthy.

“Looking at vendors and how they’re going to be analyzed is still yet to be seen because we don’t have any real government oversight on how we’re writing these programs, other than what is standard guideline for us and how we have to address those biases, address that security, address the potential for these bots to go rogue,” Miller explained.

The best option currently is to review the vendors’ customer base. See who is reviewing and researching their technology, check how the public is receiving the product, how and where the AI is being used, and what people using it often are saying.

Is AI too Expensive for Your Organization?

When personal computers (PCs) first entered the market, they were very expensive. As time went on and the technology advanced, purchasing a PC seemed more feasible. We are at the start of the AI revolution, and the technology appears to be expensive or may seem cost prohibitive for your healthcare organization.

“Like all technology, as we move forward, we’re going to see that price point decrease over time,” Miller added.

If you’re hesitant about adding AI to your workflow due to cost right now, as the price point decreases, we’ll see improvements made to the technology in the same time.

Mike Shaughnessy, BA, CPC, Development Editor I, AAPC