As more healthcare organizations begin to explore autonomous coding, it will become increasingly important to know how to accurately assess the technology and evaluate different vendors. With this in mind, we’ve compiled a list of the top 10 questions that we believe healthcare organizations should be asking autonomous coding vendors, covering everything from electronic health record (EHR) integration to metrics reporting.
By getting answers to these questions and understanding why they are important to ask, healthcare organizations can develop a better understanding of autonomous coding and be better positioned to select the vendor that is the right fit for their coding automation needs.
There are many different types of medical coding automation solutions available today. It is important that healthcare organizations distinguish between these solutions, especially in terms of the level of human intervention required, to better understand the potential impact on coding workflows and operations.
At Nym, we define autonomous coding as any solution that truly requires zero human intervention, from the moment the chart is received from the EHR to the chart being sent to billing. This is different from solutions like computer-assisted coding (CAC), which require medical coders to validate any CAC-suggested codes before the chart is sent to billing.
Autonomous coding is also distinct from AI-assisted coding, which is similar to CAC but leverages more advanced technology and AI to read structured/unstructured clinical data and suggest the appropriate medical codes. While AI-assisted coding solutions can improve the accuracy of assigned codes over time, they still require humans to validate all code suggestions before a chart is sent to billing.
If an autonomous coding solution cannot be fully integrated with your EHR system, then workflows can become disjointed and data can get siloed in different systems. This leads to coding-related challenges such as medical coders doing double the work because they have to re-enter data from the nonintegrated system, or claims getting lost in the process (and therefore revenue loss down the line).
One key innovation in healthcare interoperability has been the adoption of Fast Healthcare Interoperability Resources (FHIR) by both health systems and healthcare software vendors. At Nym, we leverage FHIR to retrieve the latest, most up-to-date clinical data from our customer’s systems in real-time. This ensures our engine’s coding is accurate and complete. FHIR also makes it much easier for customers to implement Nym’s coding engine, as IT teams do not have to build a custom report or interface specifically for Nym.
With most healthcare IT departments strained for resources, organizations should be aware of how much historical data is required by the autonomous coding solution to better assess the internal IT lift. Additionally, healthcare organizations should find out whether a standard data-pull method or a different method will be used, as this will impact the timeline and resources required.
Understanding how the historical claims data will be used is important for many reasons. First, healthcare organizations may need to allocate additional resources for tasks like mapping historical claims data to charts, or other similar exercises. Second, it’s important to understand if historical data will be used to build large data models for the vendor, which may be an area of concern.
Center for Medicare and Medicaid Services (CMS), payer, Current Procedural Terminology (CPT), and International Classification of Diseases (ICD) guidelines are constantly being updated and/or changed. To ensure compliance with effective dates and avoid denials/revenue loss, it’s important that autonomous coding solutions can be quickly updated to accurately reflect all guideline changes and in a timely manner. Additionally, as medical interventions, treatments, and services evolve, autonomous coding solutions need to be capable of coding encounters that include data related to new medical advancements in a timeline manner without requiring significant volumes of coding data to train the solution.
Accuracy is the top priority in medical coding as it directly reflects patient care, impacts reporting, drives revenue, and much more. Healthcare organizations should understand what processes the autonomous coding vendor uses to ensure that coding accuracy is maintained, as well as how the vendor audits any updates/enhancements made to the solution. Proper testing should always take place before production to assess the positive impact of new updates/enhancements and to avoid unintended consequences post-production.
One of the biggest fears around AI in healthcare is that some solutions are built with a “black-box” approach, meaning there is little to no visibility into how the solution generates its output. Healthcare organizations need to be able to show the rationale behind code selection when claims are denied or audits occur. Therefore, it’s important for healthcare organizations to partner with an autonomous coding solution that produces some sort of traceable documentation to explain how and why the technology assigned specific codes to each encounter. If a claim is denied or an audit occurs, having this documentation can be instrumental in refuting the denial or supporting the audit. Additionally, having this transparency may also help healthcare organizations identify opportunities for improved documentation, driving better revenue capture down the line.
Coding departments are used to tracking their medical coder's productivity metrics, such as charts coded per day, coding accuracy, discarded charts, not final coded (DNFC) charts, coding denials rate, and days in accounts receivable (A/R). When making the switch to autonomous medical coding, it is important to ensure that the vendor has a system in place for tracking and reporting all key metrics.
Ideally, vendors should also have a system in place to explain why charts couldn’t be coded by the autonomous coding solution (AKA “dropped” charts). This will help healthcare organizations identify areas where they can make adjustments, such as tweaking documentation practices, to reduce “dropped” charts. Additionally, understanding why charts couldn’t be coded by an autonomous coding solution will help vendors improve and tailor the autonomous coding solutions to provider needs, allowing for better coverage.
Different healthcare settings, from emergency departments to outpatient surgery clinics, follow different coding guidelines and regulations and can have drastically different coding workflows as well. Vendors with experience in the specific setting the healthcare organization is looking for are generally better equipped to both implement the autonomous coding solution and deliver the most value.
Experience with many different health systems, specifically, is also critical. Each health system has its own unique set of coding and documentation guidelines (distinct from those issued by CMS and the American Medical Association), and will often have different accuracy requirements, coding workflows, EHR vendors, and documentation practices (to name a few of the many differences). Vendors with a wider range of health system customers should be able to configure their solution to a new provider’s unique needs more successfully than a vendor with fewer health system customers.
Answering this question can help the provider and vendor set realistic expectations around key results and metrics (e.g. accuracy, turnaround time, etc.). It can also help to ask the vendor what types of steps they are taking to improve in areas where their solution struggles to assess whether it will be a short or long-term challenge.
Additionally, knowing where a vendor’s solution struggles may help healthcare organizations assess and prioritize the specialty, or specific coding type, such as facility or professional coding, that they would like to start with when they implement autonomous coding.
Autonomous medical coding delivers value by increasing coding speed, improving quality, reducing costs, and much more. However, it can also be a sizeable up-front investment in terms of time, resources, and costs associated with implementation. What’s more, the up-front investments can vary significantly from vendor to vendor.
Because of this, it is important that healthcare organizations evaluate the long-term return on investment (ROI) delivered by different autonomous coding vendors and decide whether each will offer the type of impact expected. For example, one organization may be prioritizing cost reduction while another is prioritizing accuracy improvements and lower denial rates. Knowing the long-term ROI of each vendor’s autonomous coding solutions can help healthcare organizations choose the vendor that fits their specific needs.