As Medicare Advantage Organizations (MAOs) adopt and integrate clinically trained AI technology into their medical coding processes, it’s vital to discern its capabilities but also be realistic about its limitations.
Within the last few years, artificial intelligence (AI) has become a big buzzword in the health care industry (for good reason). Within risk adjustment and medical coding, AI technology has been utilized in an attempt to increase the accuracy and efficiency of coding projects, thereby improving population health management, ensuring regulatory compliance and accurate reimbursement.
Unfortunately, natural language processing (NLP), the subset of AI that’s been utilized in medical coding for years, has repetitively come up short in reaching its potential, resulting in a general negative reaction when medical coders hear the words, “NLP” or “AI.”
Enter: clinically trained AI technology. To effectively support medical coders, AI technology must be customized, have domain-specific training, undergo ongoing refinement, and integrate expertise from subject matter experts to ensure accurate and compliant coding practices.
As Medicare Advantage Organizations (MAOs) adopt and integrate clinically trained AI technology into their medical coding processes, it’s vital to discern its capabilities but also be realistic about its limitations.
What clinically trained AI can do for your MAO’s coding practices
Increase coding production: Clinically trained AI technology can organize a chart and automate the coding processes by extracting and pre-populating relevant information including data such as, the suggested risk adjusted ICD-10-CM code, encounter date, provider name, signature, note type, section, provider type, and the corresponding clinical support. By supplying this information automatically, it reduces the keystrokes and mouse clicks needed to review a chart. Additionally, this information can be presented to the coder only when pertinent, allowing the coder to focus on the diagnosis being reviewed thereby speeding up the coding process and reducing manual efforts.
Increase coding accuracy: Clinically trained AI models use information from standard content (e.g. ICD-10-CM) for any given year, therefore it doesn't allow for typos which happen during manual coding processes. Clinically trained AI aligns with coding guidelines to consistently ensure compliant coding practices and identify all risk adjusted codes in structured and unstructured text enabling complete code capture. Additionally, it can identify and link relevant clinical supporting documentation such as symptoms, procedures, medications, and laboratory information
Increase visibility: Clinically trained AI can help increase visibility into coding projects in three ways:
- Produce detailed reporting on coding results that can aid in identifying trends for coder and physician education initiatives.
- Assist in coding project management to include detailed production and accuracy metrics, allowing managers to have real time insights into coding activities.
- Assign confidence scoring to each diagnosis prior to a human coder reviewing the chart which will assist in chart prioritization, saving valuable time when determining which charts need to be reviewed.
What clinically trained AI cannot do for your MAO’s coding practices
Replace coders: Medical records are complex and no two medical records are alike. Each EMR template displays information differently, each physician has a unique documentation style, and each medical record will contain a diverse range of information. Clinically trained AI technology has limitations when it comes to understanding a medical record, particularly in tasks requiring common sense reasoning because variability in documentation is where a human coder is absolutely required. Once a risk adjusted diagnosis has been identified, the coders must apply the Official Guidelines for Coding and Reporting in the ICD-10-CM book to ensure compliance.
Let’s face it, with the recent regulatory changes concerning RADV repayment methodologies, there is simply too much on the line financially to not get it right. Clinically trained AI is intended to complement and empower the coder to do their important work better, faster, and more accurately.
Be perfect: We use AI in our daily lives, and we know AI doesn’t always get it right. How often do the ads or suggestions on Instagram or Amazon get it right? How often does it get it wrong? The answer probably isn’t 100 percent right or 100 percent wrong but somewhere in between. This applies to clinically trained AI as well. But by working collaboratively with AI technology, coders can leverage both their own expertise with the technology intelligence to achieve optimal outcomes in coding processes, in the most efficient way possible.
While clinically trained AI is still an emerging and evolving technology, it holds immense promise for transforming risk adjustment programs by enhancing productivity, accuracy, and visibility. However, it is essential to acknowledge its limitations and the ongoing need for human expertise and iterative refinement. By harnessing the collaborative power of clinically trained AI and human coders, MAO’s can achieve optimal outcomes in their risk adjustment programs.
We would love to connect with you to share more about the Wolters Kluwer Risk Adjustment solutions, integrated with our differentiated clinically trained AI technology. Reach out to us today to learn more about the Health Language Coder Workbench and recently released Regulatory Audit Module.
Make sure to visit our risk adjustment team at the upcoming RISE Risk Adjustment Forum, May 20-22 in New Orleans, and attend our Track A: Tools & Tech Spotlight presentation on Tuesday, May 21st at 12:20 p.m.-12:35 p.m.
About the author
Melissa James, CPC, CPMA, CRC, risk adjustment SME, senior consultant, Wolters Kluwer, Health Language, supports the company's Health Language solutions with content maintenance.
She has more than 20 years of health care experience in coding, risk adjustment, billing, physician and coder education, accounts receivable management, regulatory and compliance, and consulting. She received her associate degree from Pueblo Community College.