Medical coding continues to be one of the most challenging parts of the revenue cycle management (RCM). It is now at the cusp of a revolutionary change—one so profound that it can transform the way RCM works. This change is brought about by AI-enabled auto coding.
Computer-assisted coding (CAC)—which generates codes directly from clinical documents—has come a long way in the past few years. It helps coders reduce errors and improve accuracy. Auto-coding encompasses a variety of computer-based approaches and goes beyond merely assigning codes.
Accuracy in coding has a direct impact on reimbursements for the providers. AI-enabled auto coding solutions reduce improper payments and enable healthcare organizations to become more compliant, while also improving the bottom line.
According to the 2021 report by U.S. Department of Health & Human Services, CMS improper payment rate was 6.26% to the tune of USD 25.03 billion. Out of this, around 24.2% were due to medical necessity and incorrect coding and 64.1% were due to missing documentation. The findings over the years are quite startling. To measure the improper payments in the Medicare fee-for-service (FFS) program, CMS implemented the comprehensive error rate testing (CERT) program, which is designed to comply with the payment integrity information act of 2019 (PIIA).
Reporting year |
Total expenditures (in billions) |
Improper payment rate |
95% Improper payment rate confidence interval |
Improper payment amount (in billions) |
2017 |
USD 380.76 |
9.51% |
8.9% - 10.1% |
USD 36.21 |
2018 |
USD 389.30 |
8.12% |
7.6% - 8.6% |
USD 31.62 |
2019 |
USD 398.62 |
7.25% |
6.9% - 7.6% |
USD 28.91 |
2020 |
USD 410.81 |
6.27% |
5.8% - 6.7% |
USD 25.74 |
2021 |
USD 399.78 |
6.26% |
5.8% - 6.7% |
USD 25.03 |
Table 1: National CERT improper payment rate by reporting year
AI-powered auto-coding solutions will improve accuracy and consistency.
Strategic applications of AI can significantly reduce human error and improper payments and improve the physician query system. The benefits of AI-powered medical coding are -
Medical coding depends on the coder's proficiency. Not all coders have the same skill level, hence the probability of error is high either due to knowledge gap or analytical gap—the difference where a coder is unable to properly analyze the medical records to interpret the documentation and apply coding guidelines accordingly to derive accurate codes.
As medical coding guidelines get updated, coders must remember the latest guidelines and the coding clinics. The proposed auto coding solution will consider the latest coding clinics and guidelines while assigning codes, thereby improving adherence to compliance requirements.
AI-enabled auto coding will considerably reduce the time it takes for coders to go through each medical record and extract the diagnosis and procedures performed and deliver accurate results.
To make this happen, the AI-enabled solution will have to overcome a few key challenges. The final code selection is based on the coders’ knowledge of the guidelines, clinical concepts, and compliance regulations. These guidelines provide clarity but are also open to interpretation. The solution needs to learn to identify the importance of specific documentation and its relation to ICD-10-CM and PCS codes. As more data is processed by this auto coding solution, the accuracy will improve. Also, these solutions need to be flexible to seamlessly integrate with the different electronic medical record (EMR) systems used by the providers.
The solution can also incorporate the patient’s history and current medical issues. It can include all the medical services provided to the patient, bringing in value-based patient-centric medical care. With the capability of the auto coding solutions and potential future application, it is certain that AI-enabled auto coding has arrived, and its impact cannot be overlooked.
