Autonomous Coding for Radiology 101: Boost accuracy with NLP

February 1, 2022 |
February 1, 2022

Close doesn’t count in medical coding. ICD-10-CM and CPT codes must always be correct to ensure proper reimbursement, compliance, and accurate data analytics. But achieving high levels of accuracy—that’s the hard part.

For the past few years, healthcare organizations heard the promise of computer-assisted coding (CAC) as a way to improve accuracy. Yet today, nearly 4 in every 10 hospitals has yet to implement CAC. And while CAC offers some successes, including a reported 30% boost in coder productivity (specifically for outpatient services), even those organizations’ most advanced in CAC still struggle with fulfilled productivity improvement promises, accuracy, denials, and coder satisfaction.

Quite simply, CAC just doesn’t cut it. That’s why savvy healthcare organizations are now embracing autonomous coding. It uses innovative technology—Artificial Intelligence (AI), Clinical Language Understanding (CLU), and Natural Language Processing (NLP) and rules engine—to machine code simple visits, like radiology.

In this Autonomous Coding for Radiology 101 blog post, we’ll take a closer look at what autonomous coding is, how it can help your health information management (HIM) and revenue cycle team improve coding accuracy, and how you can get started.

What is autonomous coding?

Let’s start by clearing up this myth: CAC is not autonomous coding. With CAC, a computerized learning engine uses a rules engine to identify suggested medical codes from clinical documentation. However, it’s not fully automated. In CAC, coders must review every suggested code and choose whether to accept it, modify it, or deny it based on the context of the documentation provided. In addition, CAC can only work with structured data, and analytics show that most EHR data (80%) is unstructured.

Autonomous coding is much more sophisticated than CAC, applying computer algorithms to do most of the work behind the scenes. Using AI, CLU, NLP, and a rules engine, autonomous coding converts both structured and unstructured data to the correct codes. For those accounts with a high confidence level (typically over 70%), the codes are transmitted directly to billing. The accounts with a low confidence level (typically under 30%) are routed to a human coder. When autonomous coding is applied to simple, niche service types, human coders can be redirected to the more complex service types.

For a deeper dive into the difference between autonomous coding and CAC, read this white paper.

Why focus on autonomous coding for radiology?

When it comes to autonomous coding, most health systems don’t know where to start. Radiology makes an ideal launching-off point because the automation confidence rate is high (achieving an average of 95% coding success through the machine) AND it is a very high-volume service area.

How does autonomous coding for radiology work?

Let’s say a radiology report comes in for a chest X-ray. Autonomous coding for radiology uses AI, CLU, NLP, and rules to first establish a confidence level for whether the machine can assign the codes with an accuracy of 95% or better. For the accounts with a high confidence level, the correct diagnosis codes, procedure codes, E/M levels, and modifiers are assigned and the account routes directly to billing. If the machine cannot code with high confidence or high accuracy, the software routes the account to a human for coding.

How accurate can I get with autonomous coding for radiology?

The industry standard for autonomous coding is achieving 95% accuracy, but the solution has proven to be as high as 98% accurate.

How will autonomous coding for radiology impact me and my staff?

Coding is complicated, and managing staff is a challenge. In addition, medical coders want to perform higher complexity of coding. According to the 2021 Libman Education HIM Professional Census, 87% of 1500+ HIM professionals surveyed want to increase their coding expertise, with nearly half (49%) saying they want to do so to advance their careers.

Yet with the coder shortage and high volume of patient visits, their ability to train on the more complicated service areas can be impeded. And COVID-19 hasn’t helped. Roughly 1 in 4 coders surveyed (27%) say the pandemic has increased their workload. When coding volumes increase, HIM leaders are forced to add coders to their team and often must hire contract coders. But that expense can run HIM leaders between $55 – $65 an hour or even more.

Autonomous coding for radiology helps right-size staffing budgets by eliminating the need for pricy contract coders. It also gives your in-house medical coders the time they need to sharpen their skills with complicated outpatient and inpatient encounters to advance their careers.

 What other challenges does autonomous coding for radiology solve?

Autonomous coding for radiology can supercharge the revenue cycle. When fully optimized, autonomous coding can complete radiology coding workflows within seconds, making cases available for billing an average of three days sooner. As a result, autonomous coding for radiology helps to improve cash flow, reduce DNFB and AR, and create more consistent and timely billing.

 Which EMRs can use autonomous coding for radiology?

DeliverHealth’s autonomous coding software is EHR-agnostic. We have worked with hundreds of EHR systems, including Epic, Allscripts, AthenaHealth, eClinicalWorks, Meditech, NextGen and Cerner.

 How can I get started with autonomous coding for radiology?

If you choose DeliverHealth for autonomous coding, we’ll start by getting to know your organization’s key stakeholders. We’ll ask the right questions, then use the answers to develop a customized implementation plan that will establish your ROI and defines the next steps for establishing workflow to enable the autonomous coding engine and direct billing.

 What’s my next best step with autonomous coding for radiology?

If you want to learn more about autonomous coding for radiology with DeliverHealth, read this sheet. Or, better yet, reach out to us today.