How to Improve Risk Adjustment with NLP

The following is a guest article by Carey Ketelsen and Andy Kumar from Ciox.

NLP, as you probably already know, stands for natural language processing. The technology has expanded its use for the processing of health data over the last four or five years, and companies around the health technology and health plan provider ecosystems are using it in a multitude of ways to not only mine relevant clinical information but also to improve and streamline operations.

At its essence, NLP is a form of Artificial Intelligence that we can implement to extrapolate. From a risk adjustment standpoint, NLP can read physician documentation, identify the vital clinical facts, and map those facts back to ICD-10 codes without human intervention. Human coders validate the resulting codes and return them to health plans tackling risk adjustment, which are further evaluated and submitted directly to CMS.

Paired with AI and ML

NLP has been critical in streamlining operations by improving human coders’ productivity at a rate of 4-8x. It is beneficial as well in improving coder accuracy. But where NLP shines brightest is when combined with machine learning technology as it enables the NLP engine to learn based on coding patterns and behaviors automatically.

By pairing your NLP technology with AI and machine learning, it is possible to take risk adjustment and quality programs into a new phase of technological capability. NLP enhances your ability to detect conditions based on coding patterns or leading health indicators while providing additional visibility into the patient populations based on clinical documentation. You can spot trends and find critical clinical information in what would otherwise be a stack of contemporaneous notes or an aggregated data set far too large for one person’s consumption.

Another use case for NLP is second-level review (or 2LR)

There are two crucial reasons health plans are using NLP to conduct a second-level review. 2LR gives organizations a complete understanding of your member’s risk profiles, without additional provider contact or abrasion. It is also a complement to existing coding efforts and brings confidence to the integrity of a risk adjustment coding operation.

No first-pass coding effort is perfect, and the industry-standard is generally acceptable at 95 percent accuracy. Despite high-quality first-pass coding, organizations still accept that approximately 5 percent of potential conditions remain uncoded.

NLP-enabled second-level review helps to identify new conditions, miscodes, and documentation gaps in the member data for existing conditions. So, coders can process higher chart volumes with increased accuracy and efficiency.

The checklist for success in risk adjustment today might look something like this: First, perform a first pass review with human coders, then audit with a combination of coders and NLP, followed by an NLP-enabled second-level review, with a 2LR scope that includes both adds and deletes. This approach delivers the current market’s most efficient use case to achieve a balance of the cost of operation, productivity, accuracy, and completeness.

NLP in Risk Adjustment

As alluded to above, NLP is playing a significant role in the future of risk adjustments efforts. From chase list targeting, to risk condition presence and grouping, chart value and segmentation, coding accuracy review and risk factor accuracy review, the tool is involved in nearly every step of the efficient risk adjustment operation of tomorrow.

NLP provides great insight around chase list targeting as it enables identification of charts with potential before they are pulled. You reduce the total number of charts pulled because you are targeting the chart with the highest potential. In the absence of claim data, NLP can look out in the clinical and medical record and identify members that you may not have suspected without claims information. Thus, the methodology can produce anywhere from 15-20 percent increase in the average revenue per chart.

Another area where NLP can provide significant value is around risk condition presence and grouping. NLP can quickly identify charts with no codes or charts with many codes. When you are running a coding operation, you have certain skill sets of coders. With this functionality, you can tie a higher skillset coder to a more complex code set, while routing low complexity code reviews to lower scale coders.

NLP, as outlined earlier, provides enormous value in auditing and coding accuracy review. It can also help ensure that the work done is the only work that needs doing. With NLP, instead of recoding the entire chart, you can focus on the net new codes for adds and deletes to make sure that your coding team is not spending time redoing entire charts.

The last area where NLP is reshaping the risk adjustment operation is risk factor accuracy review. NLP can be a great tool not just at the chart level, but at the member level as well. It supports your ability to understand the risk score associated with that member. A member could have information coming from multiple different sources, various charts and several different flavors of clinical records. With NLP, you can quickly process these multiple sources of data and gain insight into the potential risk score for a member with far greater analytical, far more informed data. As a result, health plans can match the risk they bear against financial reimbursement and see where the gap is.

In the age of all digital everything, our health data matters. By pairing tools like NLP, AI and Machine Learning, we’re arming our life sciences organizations, our health insurers, and one day soon, even our healthcare providers with real-world data, analyzed and processed for clues, trends and patterns. The power of NLP today is fast and cost-effective processing. The power tomorrow may be endless.

About Ciox
Ciox Health, a leading health technology company, simply and securely connects healthcare decision makers with the data and hidden insights in patient medical records. Combined with the industry’s most extensive network access to healthcare data, Ciox Health’s expertise, relationships, technology and scale make a difference for healthcare stakeholders and empower greater health for patients. Through its technology platform, which includes solutions for data acquisition, release of information, clinical coding, data abstraction, and analytics, Ciox helps clients securely and consistently solve the last mile challenges in clinical interoperability. Learn more about Ciox technology and solutions by visiting or Twitter and LinkedIn.