Contributor: How Payers Apply AI to Solve Incomplete Data: 3 Use Cases

In this column, Calum Yacoubian, MD, discusses how artificial intelligence, specifically natural language processing, can help fill gaps in patient data.

In a business environment where regulations are constantly changing and costs continue to rise, maintaining member health is an ongoing challenge for payers. The need to manage messy and incomplete patient data adds burdens to payers.

For payers, complete and accurate member data is essential to assess patient health status, predict future risks, and identify gaps in care that need to be addressed. However, patient data that fits these criteria is scarce, and part of the challenge of leveraging healthcare data is its sheer volume: the typical hospital generates 50 petabytes of data per year, which is equivalent to to around 11,000 4K movies.

While most healthcare players are aware of the enormous amount of data created by the industry, few understand how unstructured or semi-structured this data is – up to 80%, according to a report published in Health informatics research. This means that a significant portion of industry data is effectively locked away in the notes sections of electronic health record (EHR) systems and not readily accessible to guide clinical decision-making. In many cases, semi-structured and unstructured data includes a range of important information, such as patient symptoms, disease progression, lifestyle factors, and laboratory tests.

Recently, payers have come under regulatory pressure to provide patients with seamless access to their own data, following the approval of the Interoperability and Patient Access Final Rule. Payers naturally focused initial compliance on access to structured data, but now they should include unstructured data to improve the usefulness of data for patients.

NLP automation, no manual chart reviews

Rather than devoting resources to tedious manual chart reviews, a growing number of payers are turning to artificial intelligence (AI)-powered tools such as natural language processing (NLP) to overcome search restrictions in mountains of data.

NLP augments the examination of graphics and the extraction of information by helping machines “read” text, simulating the human ability to understand natural language, enabling the analysis of unlimited amounts of textual data without fatigue, consistent and impartial manner. Essentially, NLP allows computers to understand the nuanced meaning of clinical language in a given body of text, such as identifying the difference between a patient who smokes, a patient who says he quit smoking 5 years ago, and a patient whose file indicates that he is trying to resign.

Payers have leveraged this previously hidden information to uncover hard evidence that powers predictive models that ultimately improve patient outcomes, reduce costs, and improve risk adjustment. Here are 3 use cases describing how.

Improved Medicare Advantage risk adjustment

Member health information that is used to ensure that Medicare Advantage plans receive the appropriate funds to care for their patients is captured in hierarchical condition categories or HCC codes. Inaccurate HCC coding can result in significant costs to health plans, prompting many payers to employ large groups of chart reviewers to manually sift through charts for new information about that data.

Faced with this challenge, Independence Blue Cross deployed NLP to augment its chart review program. The health plan has created 2 clear goals for its NLP initiative: first, to speed up the review process so that clinicians can review more documents per hour, and second, to capture diagnoses associated with HCC codes that may have been missed by case review teams.

An initial pilot of the project identified HCC code characteristics with over 90% accuracy, processing documents of 45 to 100 pages per patient. NLP helps Independence Blue Cross process hundreds of thousands of complex medical records, speeding up case reviews and enabling reviewers to increase their efficiency and productivity.

Predicting the risk of diabetic foot ulcers in patients

The financial challenges created by the COVID-19 pandemic have convinced the healthcare industry of the importance of predictive modeling. For example, a late 2020 PwC survey of healthcare executives found that nearly 75% of respondents said their organizations would invest more in predictive modeling in 2021. As one PwC executive noted: “The pandemic has amplified the presence of unprocessed data and the lack of effort to do enough.

A health plan used NLP to extract unstructured data to feed a model that predicted a patient’s risk of developing diabetic foot ulcers, a costly condition that can lead to amputation if left untreated . The payer’s data science team extracted patient EHR scores for cues of imminent risk, such as body mass index data, lifestyle factors, comments about medications and documented foot diseases. So far, the model has identified 155 at-risk patients who could be proactively managed, translating into potential annual savings of $1.5 million and $3.5 million from amputations averted, according to the payer’s internal data.

Addressing Social Determinants of Health (SDoH) Challenges

The social determinants of health, such as access to housing, food, transportation and employment, play a critical role in the overall health of patients. However, much of the data related to these topics is locked away in unstructured sources such as admissions, discharges, and progress notes.

One organization used NLP to search the unstructured notes of prostate cancer patient charts to identify those at risk of social isolation, with 90% accuracy. By adopting NLP to screen for this vital social characteristic, payers can establish awareness campaigns aimed at connecting with patients deemed to be at risk of missing appointments and suffering from uncontrolled disease progression.

Despite laudable industry-wide efforts to achieve greater interoperability of healthcare information systems, the industry is unlikely to escape the problem of fragmented and disparate data anytime soon. To take full advantage of the data at their fingertips, improve patient outcomes and reduce costs, payers will increasingly turn to AI-enabled technologies like NLP.

About John Tuttle

Check Also

Non-pharmacological treatments outperform opioid treatment for patients with chronic non-surgical pain

Updated prescribing guidelines published in 2014 and 2016 may explain a lower opioid use of …