September 23, 2021

Applying AI to Leverage the Value of Unstructured Claims Data

This article was originally published by WorkCompWire, featuring Healthe’s Director of Enterprise Analytics, Mike Theis, in their Leaders Speak forum.

It’s estimated that up to 80 percent of the information contained in an electronic health record is unstructured and not readily accessible. Whether in clinician notes, physical therapist notes, or other text-based records, much of the information needed to facilitate return to work and to health is there — if we have the means to tap into it.

Every interaction between a care provider and an injured worker yields information that could be a significant piece of the puzzle – whether it’s information about comorbidities, social determinants of health, psychosocial factors, or medication- or treatment-specific risks. So when we’re looking at an injury, how do we understand those other factors to bring context to the complexity of the claim or the risk that an individual patient may be facing?

Enhancing Human Capabilities

Natural language processing (NLP) leverages artificial intelligence (AI) to help analytics systems understand and organize unstructured data. But I like to say that it’s more about augmented intelligence, rather than artificial intelligence. In other words, how do we build the right tools to provide stakeholders with the right information, at the right time, to make the best decision?

We’ve seen the most success by marrying the AI aspect with human expertise to produce the best outcome. For example, a Healthesystems clinical pharmacist may read through a thousand pages in a medical record during an independent pharmacotherapy evaluation. This effort is valuable in that it has often uncovered opportunities to help steer the trajectory of a claim toward a more positive and cost-effective outcome. But it is also time consuming. Enter NLP, the application of which can drive efficiencies in that process and extract key data elements for the clinician so that we can reduce the time it takes to review those medical records and intervene. Or, taking it a step further, by quickly identifying claims that would derive the greatest benefit from a clinician’s review, helping to prioritize when there is limited time.

Informing Population Management

Not only are AI and NLP technologies effective in extracting key information to move the needle on individual complex and high-cost claims, but also in feeding that data back into the analytics system to inform models looking at population health, helping claims organizations and their partners to stratify and segment the population more effectively.

As an example, we know that social determinants of health can have a significant impact on health outcomes, and some may be especially impactful to injured workers and their recovery. Patient populations with otherwise straightforward musculoskeletal injuries may have significantly different outcomes when a history of substance abuse intersects with the prescribing of opioids. Populations without a social support system or access to transportation may have significantly higher costs due to transportation needs or language barriers. The key is being able to connect the dots before the claim becomes complex, and an augmented approach to analytics using AI technologies can help inform future decisions related to patient populations impacted by these factors – whether it’s timely referrals for urine drug screening, or the foresight to anticipate higher-than-average transportation services on the claim.

Enhancing Program Management

AI-extracted data can also help measure and inform stakeholder behaviors and other aspects of program performance. How is each role performing within the system? Are the vendors scheduling services on time – and if not, where are the breakdown points? Are the providers or claims examiners requesting or approving services that conflict with your clinical strategies? How are the patients reporting outcomes and service results? That data is valuable, not just at a micro-level, but especially in analyzing and developing program management strategies to improve performance for all stakeholders.

The opportunities for NLP and other AI-based technologies are ripe for augmenting and accelerating some processes that are already successful in workers’ comp claims management – as well as creating brand new opportunities. I’m excited to see the impacts of this technology as it becomes more prevalent within our own programs at Healthe as well as the industry at large.

View the original article at WorkCompWire.

lockenvelopephone-handsetmenucross-circle