How AI is Resolving Common Coding Mistakes
Roni Berlin, BSHCM, CPC, CPB, who serves as associate vice president at ExdionRCM Solutions, opens an article at Medical Economics with the harsh reality that, if a medical student’s classes reflected the reality of practicing medicine today, "I would venture that more than half of the curricula would be focused on performing administrative work, including the very complex task of coding."
Accurate coding is one of the most intricate and often frustrating tasks that doctors and staff must do. Countless studies and surveys are in consensus that administrative burden is the most cited contributing factor for physician burnout, an escalating problem that existed long before the COVID pandemic. Erroneous coding can bring significant financial duress to a medical practice, either in the form of decreased revenue, audits, or even clawbacks from private insurers and revenue audit contractors (RACs). Lower reimbursements, creating pressure for doctors to work longer hours and bill more, as well as compliance requirements, are also leading contributors to physician burnout.
Common Missteps in Medical Coding
She goes on to list the top five specific and common coding mistakes -- or, in other words, where you are losing revenue:
Coding for evaluation/management services is often either too aggressive or too passive, and these coding errors are largely attributed to misinterpretation of E/M coding guidelines and the frantic pace of the clinical environment. Aggressive coding occurs when there isn’t proper documentation to prove out what was done. On the other hand, passive coding doesn’t take the entirety of the work performed into account.
Oftentimes, this is the result of incomplete charting, typically due to provider distraction. Charts without follow up typically result in the claim being sent late or unbilled.
The confusion between whether a patient is new or established, which should usually be established at the front desk, can lead to lower payments if the patient’s status is not properly captured.
Providers often miss administrative procedure codes for minor treatments, which can be a significant amount of lost revenue. This includes codes for injections, immunizations, immobilization, etc. Administering injections is among the most routine services provided in a primary care or urgent care practice, but one inoculation includes two codes: a CPT code for the injection, and a separate code for the medication or vaccination provided. Modifier 25 may also be applied if other care is being given. Another example of oversight can occur when placing a splint on a limb. There are two codes to enter: one for the application and another code for the supply item, such as a splint or a cast.
The largest errors are the improper use of modifiers 25 and 59 to expand treatment, which can lead to audits and clawbacks. Modifier 25 should be appended to an E/M code to report a significant but separately identifiable additional service rendered during the encounter, such as an injection. Modifier 59 is used to identify procedures/services other than E/M services that are not normally reported together but were appropriate to render under the circumstances.
Artificial Intelligence to the Rescue -- Medical Coding and RCM
All is not lost, however, as Ms. Berlin explains:
Revenue Cycle Management (RCM) solutions, powered by artificial intelligence (AI) and machine learning (ML), are augmenting – not replacing – humans, helping them think, work smarter, and minimize costly mistakes. At the root of AI-powered RCM is automating coding, which allows practices to optimize their coding, helping eliminate errors and missed tasks where most revenue is lost. These solutions interface with a medical practice’s EMR and practice management systems either through a process automatic route using bots, or through application programming interfaces (APIs). RCM solutions can increase practice revenue by as much as 25 percent in some instances.
This issue is not unlike the one experienced by providers utilizing BPOs to process their paper-based remits and EOBs/EOPs. Both processes require a human to manually review information, interpret the information, and enter the information into the system. However, these process are less than efficient -- not only eating away at precious time, but also prone to errors in each of the three phases mentioned above.
That's why the healthcare industry is turning to artificial intelligence and machine learning technologies to remove the inefficiencies and errors that are made by humans. The AI models are specifically trained with millions of documents (whether for medical coding or EOBs/EOPs), and the technology is able to read the documents, extract and interpret information (for EOBs/EOPs, accurately interpret the information and match to the standard ANSI code), and produce quality output that is ingestible into their systems.
The technology is ready for providers, RCM companies, clearinghouses, and billers to deploy now, and those who wait only further exacerbate current problematic issues.
