.png)
Key Takeaway: Over the past few decades, new technologies such as computer-assisted coding (CAC) and autonomous coding have emerged to solve the challenges associated with manual coding. While CAC improves coder productivity, it falls short on coding accuracy and is highly dependent on human coders. Autonomous coding, on the other hand, ensures coding accuracy and drives greater efficiency, saving time that can be reallocated to patient care and critical operational activities.
Medical coding is an essential step of revenue cycle management that directly impacts the financial health of hospitals, health systems, and physician offices. Despite its importance, the coding process is outdated and has changed little since modern-day practices were established nearly 50 years ago. Current coding processes continue to rely largely on manual labor which is time-consuming, expensive, and prone to error.
These challenges have not gone unnoticed by healthcare executives. In a study conducted by BESLER and HIMSS Media, which surveyed over 100 leaders in finance, revenue cycle, reimbursement, and health information management-focused roles, eighty-four percent of respondents agreed that clinical documentation and coding posed medium to high risk to their revenue cycles. When coding operations are not running smoothly, healthcare organizations risk significant backlogs and delayed reimbursement, both of which negatively impact the health of the revenue cycle.
There is a clear need for innovation within the medical coding industry, and different technologies have emerged over the past few decades aimed at solving manual coding challenges. One of the first was encoders, a software that contains codebooks and enables coders to quickly search for key terms and phrases relevant to the medical chart. Following encoders was computer-assisted coding, which gained widespread adoption in 2015 with the transition from ICD-9 to ICD10.
Computer-assisted coding (CAC) was the first AI-based coding solution to hit the healthcare industry, and it continues to be used by healthcare organizations today. Powered by natural language processing (NLP), CAC software scans medical record documentation, identifies key terms, and suggests codes that support the treatment or service provided during a specific patient visit. By automating the analysis of physician notes for code assignment (the most time-consuming phase of the coding process), CAC software enables coders to process a larger volume of charts at a faster rate.
While CAC is an upgrade from manual coding, users continue to experience coding challenges that put them at risk of delayed reimbursement.
The technology that powers CAC can identify key terms and phrases in physician notes. However, it can run into challenges with subjectivity, abbreviations, and ambiguities that are characteristic of clinical language. This can result in suggested codes that don’t accurately reflect the patient encounter described in the medical chart.
In The Coder’s Guide to Physician Queries, Adrienne Commeree, CPMA, CCS, CEMC, CPIC, noted: “During AHIMA’s (the American Health Information Management Association’s) pilot testing of CAC software, the organization weighed in on some of the potential issues with using CAC software alone (with no human intervention). AHIMA noted that within specific areas of the pilot CAC testing in ICD-10, the coders did not accept 75 percent of the diagnosis codes presented, and they did not accept 90 percent of the procedure codes presented within the code sets.”
Further supporting these findings was a study conducted by AHIMA and the Cleveland Clinic which concluded that CAC alone- without the intervention of a credentialed coder- had a lower precision rate.
It is clear that human coders must validate CAC-suggested codes to ensure coding accuracy. Because this validation takes time, it limits the productivity increase promised by CAC vendors.
There are also challenges associated with CAC that arise before the software is even up and running. During implementation, staff must learn an entirely new coding workflow based on the specific CAC software they’re using. Modern CAC software is complex, including numerous workflow modules such as chart triage and different coding queues that make training especially challenging. Additionally, it can be difficult for IT staff to integrate CAC software into their organization’s existing IT stack. Together, these challenges create a CAC implementation process that has been described as time-consuming and difficult by healthcare providers and their staff.
As a result of these shortcomings, more and more healthcare organizations are turning to autonomous coding.
Autonomous coding leverages multiple subfields of artificial intelligence (machine learning, NLP, deep learning) to automate code assignment. The goal is to meet or exceed the accuracy thresholds of human coders so that autonomously coded charts can be sent directly to bill. Autonomous coding engines are also capable of quickly processing a large volume of charts, which can significantly accelerate a healthcare organization’s cash flow.
Nym has developed its own autonomous coding solution powered by the Company’s innovative clinical language understanding (CLU) technology. CLU is an entirely new form of NLP that combines medical knowledge and computational linguistics and applies it to clinical language. Powered by CLU, Nym’s coding engine takes provider notes within patient charts and uses them to accurately assign medical codes, all with zero human intervention.
Where CAC fell short of expectations, autonomous coding is revolutionizing the medical coding industry and filling gaps associated with other methods. In Part 2 of this blog series, we address key differences between autonomous coding and CAC and how those differences impact return on investment for healthcare organizations.
