The health care market is one of the fastest growing markets in the digital universe. But this growth presents the problem of how to effectively and efficiently analyze and understand this data.

The health care market is one of the fastest growing markets in the digital universe. From 153 billion gigabytes of data collected in 2013, the health care market is on track to collect close to 2.3 trillion gigabytes of data in 2020 with an annual growth rate of 48 percent1.  

But this growth presents the problem of how to effectively and efficiently analyze and understand this data. By some estimates, more than 80 percent of the electronic medical records (EMR) data is unstructured. Given the resource and time constraints, it is physically impossible for a human to read through that amount of data and draw insights in a timely manner. That is why the health care industry is rapidly turning to artificial intelligence (AI) to help understand this data and drive better health outcomes for patients. The AI in the health care market is slated to expand from $2.1 billion in 2018 to $36.1 billion in 2025–a growth rate of 50.2 percent2. With all this money being invested, how will the use of AI help improve the health of members?

Today’s Medicare Advantage, Medicare Part D, and ACA Exchange plans operate in a challenging world.  Accurate risk adjustment data on these members submitted to the Centers for Medicare & Medicaid Services (CMS) is vital to health plan success. Most Medicare enrollees are over age 65 and typically have more chronic conditions than younger people who may participate in the Affordable Care Act (ACA) exchanges. This does not reduce the need for accurate and timely risk adjustment data on the ACA exchange members because many may not have been previously insured due to pre-existing conditions and/or high plan premiums and their conditions may not have been diagnosed or treated.

Revenue earned by health plans is risk adjusted using their membership health status levels and other factors. On the Medicare side, health plans are paid per member per month in proportion to the health status level of each member. On the commercial exchange side, premium revenue is redistributed in the market pool within a state and weighted by the overall health status levels of each plan’s membership. In both cases, health status levels (or risk scores) are determined by the number of distinct medical conditions the members have which are identified by diagnosis codes submitted on claims by the providers who take care of them.

Given the many constraints faced by providers, the diagnosis codes often do not give the complete picture of a patient’s health. As a result, plans do not receive documentation of all the diagnoses tied to their members and thus are unable to share that information with CMS and the states. The lack of information leads to inaccurate calculation of the members’ risk scores and the potential for lower reimbursements for an individual’s care. That translates into less funding to care for the member’s health.

Natural language processing and machine learning

This whole cycle can be transformed if we mine the hidden information in patient medical records and use it to create a complete picture of each member’s health and what diagnoses he or she have. Natural Language Processing (NLP) and Machine Learning (ML), two main building blocks of AI, are already helping us do that and enabling plans to get a more accurate picture of members’ health conditions and better coordination of their care while generating more accurate risk scores and more accurate payments.

NLP analyzes human language. It provides the key capability to extract insights from unstructured text. ML can “learn” from data to find patterns and make decisions with minimal human intervention. NLP algorithms can process a vast number of medical records by combing through unstructured texts, identifying ICD-10 diagnosis codes and tagging sections of interest in EMRs. This process provides documented proof of new diagnoses or removes incorrect diagnoses from a member’s record. The algorithms help validate the medical records so that they can go through the regulatory scrutiny applied by CMS in Risk Adjustment Data Validation (RADV) audits.

A medical record can be reviewed within a few minutes since the sections that a coder needs to look at are tagged and presented seamlessly. The coder either approves the outcome suggested by an NLP algorithm or rejects it. These decisions help in training ML algorithms to improve over time to accurately prioritize workload by presenting medical records with a higher probability of the need to update the member’s diagnosis profile and allowing the health plan to take care of the member in a holistic manner while receiving accurate payments.

Utilizing these AI technologies, plans can dramatically improve their coder’s efficiency and the risk adjustment department’s effectiveness in retrospective chart reviews as well as in auditing charts to find unknown conditions and improve prospective diagnosis code capture.

Diana Benli, vice president of government solutions product management at Cognizant®, explains, “The adoption of artificial intelligence (AI) and analytics to optimize the business of health care is here today.  Cognizant has made focused and considerable investments in AI to improve health care outcomes, increase intelligent automation and identify key patterns to optimize effectiveness in program areas such as risk adjustment and quality.”

Cognizant’s TriZetto® Risk Adjustment Manager (RAM) solution is tailored to the needs of each health plan and their population. RAM runs a series of algorithms to identify possible additional diagnoses that are appropriate for members based on their claims’ history. With that data, health plans can act by understanding whether diagnoses code additions or deletions are needed. Armed with that information, health plans can submit data to CMS to accurately reflect their members’ risk scores.

Click here to learn more about Cognizant’s solutions for risk adjustment and other government solutions.

  1. Source EMC Digital Universe.

About the authors

Amol Joshi has over 15 years of experience in the health care and technology fields. He is an expert in Medicare and Part D with a significant focus on risk adjustment. He manages the Risk Adjustment Manager, Risk Score Manager, Prescription Drug Event Manager and Rx Reporting Manager solutions within the TriZetto Elements product portfolio at Cognizant.

Stacy Duhon has over 20 years of experience in health plan operations and risk adjustment with more than 27 years in healthcare management. He works with Cognizant clients to assess and improve their Government Programs business processes.

© 2019 Cognizant.  All rights reserved.