Skip to main content

Risk Adjustment

Many payer and provider organizations are challenged with comprehensively and accurately identifying and documenting risk-adjusted conditions. Accurate risk adjustment leads to more appropriate preventative care initiated to ensure the most cost-efficient care can be delivered.

  1. The Growing Medicare Challenge
  2. The Core of Risk Adjustment
  3. Clinical Documentation and MEAT
  4. Healthcare NLP for Risk Adjustment
  5. Improving Coding Accuracy
  6. Risk Adjustment Case Study
  7. Human in the loop (HITL) AI for Risk Adjustment
  8. Deploying Risk Adjustment Workflows
  9. Beyond Medicare Risk Adjustment

The Growing Medicare Challenge

With around 10,000 people enrolling in Medicare every day, accurate hierarchical condition categories (HCC) and risk adjustment are essential to predict future healthcare costs and the needs of individuals (based on diagnosis and demographics). Inaccurate or inconsistent documentation can leave healthcare organizations exposed to much higher levels of financial risk.

The Core of Risk Adjustment

HCC codes are at the heart of the risk adjustment process:

  • ICD-10 (diagnostic) codes are mapped to HCC codes to assign each patient's risk adjustment factor score;
  • A risk adjustment factor score (RAF score) refers to a medical risk adjustment model employed by the Centers for Medicare & Medicaid Services (CMS) to represent the status of a patient's health. The RAF score allows payers and providers to identify potential patient complications and provides them with a more complete picture of each patient;
  • With a better understanding of patients and their members, healthcare organizations can anticipate future financial requirements and predict the appropriate reimbursement.

Clinical Documentation and MEAT

To support an HCC, clinical documentation is required to substantiate the presence of a disease or condition (appearance on a problem list in not sufficient). This is in addition to including the clinical provider’s assessment and management plan for the disease or condition.

Most healthcare organizations use the Monitoring, Evaluation, Assessment, Treatment (MEAT) criteria as their documentation standard, together with the ICD-10-CM diagnosis coding and HCC assignments.

Healthcare NLP for Risk Adjustment

Linguamatics NLP enables payers and providers to connect technology with subject matter expertise in an easy-to-use workflow. This more accessible approach increases ensures a higher level of confidence in an organization's risk adjustment submissions. 

Our NLP platform also enables organizations to transform how they identify risk adjustable comorbid diagnoses by providing automated and semi-automated disease coding.

NLP also supports population health and risk stratification with publication grade accuracy. In both risk adjustment or precision medicine, our NLP platform has been proven to uncover diagnoses with > 90% precision and recall, and to significantly reduce medical chart review time.

Improving Coding Accuracy

Linguamatics NLP can help organizations improve their coding accuracy:

  • Differentiate between the MEAT and appearance of conditions on a problem list;
  • Distinguish between family and personal histories;
  • Distinguish the past history of a condition with its current presentation;
  • Distinguish between the first and subsequent Myocardial Infarctions (MI);
  • Conflicting documentation: for example, bilateral pedal pulses in a patient with a below-knee-amputation

Risk Adjustment Case Study

A large US healthcare payer serving ~200K Medicare Advantage members wanted to move to a more efficient automated and digital system. The diagram below illustrates how they used NLP to maintain risk scores for family members, and to submit reimbursement claims to CMS.

Using NLP they were able to improve medical chart processing by 25 – 50%, helping nurses identify conditions they might have previously missed. This payer is using NLP in other areas including analyzing transcribed call center notes to look for conditions such as colon and breast cancer, and in population health, assessing social determinants of health such as loneliness and lack of mobility. This company considers NLP a core capability with the payer community.

Risk adjustment case study - Existing Payer Customer

Increase Accuracy of Risk Adjustment

The Human Assisted Review Tool (HART) puts facts into the hands of experts.

Having combed through lengthy medical records (often hundreds of pages long), HART identifies conditions, codes, and risk categories and presents these to the user to accept or reject.

With the medical record presented in an easy to consume format (see fig. 1), coders can also highlight additional text and add new codes that may have been missed – or alter NLP presented codes to ones of their choice. This is where expert and technology combine in true augmented intelligence.

HART risk adjustment HITL

Figure 1: Each review consists of a left-hand panel of individual results to curate, alongside a large view of the complete document with color-coded highlights

With each code linked to associated text in the medical record, highlighted and presented to the user, an audit trail exists to back up each code submitted. If coders disagree with suggested codes, a note taking function lets them capture their reasoning should this be challenged in the future.

Deploying Risk Adjustment Workflows

Not every organization has the same requirements when it comes to deployment – therefore, our solution is available in a variety of deployment models. Using the NLP Data Factory, volume is not an issue, and our NLP risk adjustment workflows can be configured to our customer’s needs.

Because healthcare data is complex and often messy, there are a suite of pre-processors to deal with the different sources and formats in which payers receive data: Optical Character Recognition (OCR) correction for scanned PDFs, region detection for Consolidated Clinical Document Architecture (CCDAs) or Health Level Seven (HL7), we have the tools to ensure that the highest accuracy results are obtained.

Beyond Risk Adjustment

NLP can help healthcare organizations beyond risk adjustment:

  • extracting member level social determinants of health (SDoH) data
  • providing additional features for predictive models
  • sentiment analysis in call center transcripts
  • mining broader datasets such as FDA drug labels to gather information on reported adverse events.

Our NLP platform offers flexibility across multiple use cases, allowing organizations to adopt an enterprise strategy for unstructured data.

Ready to get started?

Request a Demo

Questions? Ask our experts