Natural Language Processing Model for Accurate Medical Billing Coding System

We created a real-time natural language processing (NLP) algorithm to analyze and accurately determine medical codes in order to make our client’s medical billing process more time and cost efficient.

Medical billing is an important component of practices and is key to ensuring the smooth operation of the revenue cycle for our client. Proper medical billing codes, along with supporting information, are necessary to accurately charge health insurance providers and ensure accurate reimbursements. For example, each billing must include the code and reasons for suggesting the particular procedure with diagnostic evidence in order for the insurance provider to make a maximum claim payment.

As one of the leading radiology groups in the market, our client needed to set in place a more error-free and timely computing method for processing medical bills to relevant insurance companies. In order to optimize claims, the existent process required each medical code to be validated and supported by information provided by the physicians about each case. With over 300,000 dictations to be processed each month, this was time-consuming and was prone to manual errors of omissions or misinterpretations.

Problems That Needed To Be Addressed

Data submitted by physicians were often insufficient to narrow down the exact medical code. Hence, there was a lot of time spent on submitting Requests for Information (RFI) and validation. Improper assignment of codes also led to reduced claims, leading to an increase in the overall cost for the company.

The Solution We Provided

We built an NLP algorithm that identified key information required by medical coders relevant to the procedures described in the reports. Automated prompts were set up to retrieve missing information from the concerned physician.

The Impact and Long-term Results

Our billing solution was able to quickly learn the various parameters that defined the complete billing information required and could assess and categorize over 30,000 dictations as “complete” or “incomplete with physician’s input needed” in fewer than 10 minutes.

Proper medical billing codes, along with supporting information, are necessary to accurately charge health insurance providers and ensure accurate reimbursements.

So, Here’s How We Did It

NLP Algorithm: While ensuring the protection of PHI and HIPAA compliance, our Director of Data Science lead the team that built a real-time NLP algorithm to ingest and analyze dictated reports. This analysis identified key information required by medical coders relevant to the procedures described. If key details were omitted, based on for instance the exam modality (i.e., X-ray, MRI), the physician was prompted to add specific details before sending off the report.