Automation and artificial intelligence are proving to be key to the future of risk adjustment and value-based care. By combining the expertise of human coders, with the speed and efficiency of machine learning we will be able to maximize the outcomes from risk adjustment programs.

The shift to value-based care and risk bearing contracts has necessitated the need to find ways to improve efficiency and financial performance of risk adjustment programs. Accurate representation of a member’s clinical risk is critical to a health plan’s performance. This has heightened the necessity for a plan to have greater understanding and visibility into their members’ medical conditions, medications, etc.

Patient data being collected today is exponentially growing, and resources are limited to review all that structured and unstructured data. It is estimated that over 80 percent of this resides in silos and often in unstructured formats. Given the sheer volume of data being generated, it is virtually impossible for humans to analyze all this information and derive timely insights to help improve patient care. In the world of risk adjustment, this leads to an overreliance on large coding teams (in-house or outsourced) and data analytics teams trying to make use of data residing in silos. In addition, factors like data lags and high seasonal-volume variability makes it difficult to handle large volumes in order to meet Centers for Medicare & Medicaid Services’ (CMS) prescribed timelines. 

This is where automation and artificial intelligence (AI) are proving to be key to the future of risk adjustment and value-based care. By combining the expertise of human coders, with the speed and efficiency of machine learning (ML) we will be able to maximize the outcomes from risk adjustment programs.

AI-driven risk adjustment solutions leverages natural language processing (NLP) and deep learning (DL) to maximize code extraction and help risk-bearing entities improve care quality and financial performance. Key aspects of such a solution include:

  • Digitization of various clinical documents that includes but not limited to progress notes, hospital admissions, imaging, CT scans, EKG to maximize information extraction.
  • High-precision NLP algorithms to comb through vast amounts of unstructured text to identify relevant ICD-10 diagnosis codes and other relevant indicators to validate Dx identification for risk adjustment, etc.
  • A consolidated view of all records available for a patient, so that the coders can easily validate the recommendations from the ML/NLP output.
  • An ability to seamlessly integrate with the client’s existing application and technologies.

Advanced analytics are then used to drive continuous improvement, including:

  • NLP-enabled member suspecting for better gap closure.
  • Bi-directional risk adjustment that helps minimize the risk of an audit with robust coding accuracy.
  • Derived metrics to profile coding behaviors at a provider/provider group level.
  • Leveraged insights to customize provider education and outreach.
  • Transitioning from retrospective to prospective coding.
  • Enhance care delivery.

The following are some ways to leverage AI within the framework of a payer’s operational team to derive more value for retrospective or prospective risk recapture programs:

Prioritize medical records: Using actionable information from an AI engine, the operations teams can quickly identify medical records that need most attention based on the types of diagnosis suspected.

Deep dives for members with no chronic conditions: Leveraging ML/NLP on charts where human coders and/or claims do not reveal any HCCs may increase the probability of unearthing suspect conditions that may have been missed.

Improve coding accuracy and efficiencies: In non AI-backed coding review programs, human coders perform first and second pass coding exercises. Increased quality assurance may improve risk recapture but it also introduces a second opportunity for human error and leaves you with the problem of what to do with the unmatched codes from the various passes. In an AI-driven coding review, the technology serves as the first pass coding on the medical records before a human coder reviews the output and validates it. This helps by a) reducing the manual coding effort, making it more efficient and b) providing human coders with indexed and highlighted chart. The review of NLP codes is also much faster than a traditional review.

There are numerous benefits of AI-driven solutions:

  • Machine learning in health care has unlocked the information in clinical documentation, allowing physicians to focus more on care delivery, improve patient care and reduce health care costs.
  • The ability to verify and identify medical codes at a large scale with high precision.
  • Improve the quality of services offered, due to the rising needs of the value-based care.
  • As opposed to manual coding, NLP-based coding can lower coding costs per chart and accelerate the time to value.
  • NLP can help assign patients a risk factor and use their score to predict the costs of health care.
  • Reduce the workload of medical practitioners.

To understand and analyze contextual unstructured information with high precision is still an evolving area and that’s a growing trend in the market.