Kepro improves healthcare outcomes with fast and accurate insights from Text Analytics for health

Supporting priority populations

Ensuring the quality and integrity of government healthcare programs has always been at the center of Kepro’s mission. With more than 35 years of experience as a health services company offering a diverse portfolio of care management, quality oversight, and clinical assessment services for Medicare and state Medicaid agencies, Kepro is a trusted partner and advocate.

Kepro provides technology-enabled services for priority populations to help healthcare recipients access care in the home or community of their choice. However, making sure that beneficiaries get the right care can take significant time and effort from Kepro’s team of highly trained nurses and physicians. Kepro conducts quality and medical necessity reviews for government healthcare programs, and its ability to deliver value depends upon Kepro’s clinical teams getting accurate, contextualized medical information quickly.

Nurses and physicians focused on clinical reviews need timely access to beneficiaries’ medical information so they can verify that a specific healthcare procedure is warranted and aligned to nationally recognized criteria. After that, procedures can be approved, care delivered, and payers billed appropriately. But in many cases, typically for complex medical conditions, clinical staff must sift through reams of medical documents to find specific pieces of information to confirm that a procedure is medically necessary and will likely achieve the best healthcare outcome.

Slow, manual reviews

High-quality patient care requires and produces a staggering array of clinical documentation. It includes medical records, clinical summaries, clinical progress and discharge notes, medication lists, and other medical documents, all from multiple sources in a wide range of structured and unstructured formats, such as PDFs, Excel worksheets, Word documents, faxed forms, and even handwritten notes.

To variability and complexity, add sheer volume. Kepro review teams often analyze hundreds of pages of medical records daily. The volume of unstructured text that needs to be carefully evaluated, organized, and summarized continues to rapidly grow. This information is rarely standardized, and its accurate processing adds significant time for Kepro’s highly trained experts, as they conduct their critical clinical reviews.

AI-powered automation

Kepro wanted to give reviewers an easier way to identify and extract vital information from large, complex, and mostly unstandardized datasets. With it, they could make clinically important healthcare decisions—such as discharge readiness or eligibility for specific types of care—faster and more accurately. The company needed a solution that would help review teams unify data across technology platforms, automate information search processes, and understand and contrast case details against national healthcare criteria to promote stronger, more robust healthcare reviews. Kepro used Text Analytics for health, a healthcare-specific AI offering that’s part of Microsoft Azure Cognitive Services, to automate the review of clinical documentation at scale.

With Text Analytics for health, the company takes advantage of natural language processing to parse unstructured clinical documents quickly and accurately and then identify, categorize, and link entities to a dictionary of more than 100 medical headings that align to standard payer definitions. It built neural networks to analyze medical documentation for multiple specialties and automatically populate forms for Medicare and Medicaid reviewers across twelve different document templates.

Early on in its automation journey, Kepro understood the significant amount of effort, data, and time that it would take to train machine learning models to recognize and deliver accurate results aligned to a broad and nuanced range of clinical entities. Because Text Analytics for health comes pre-trained on a wide array of medical and clinical data, documentation, and formats, Kepro spent far less effort, time, and cost developing a solution that can use AI to recognize and highlight crucial relationships in clinical documentation.

“Building and training all the models on our own would have been a big mountain to climb,” observes Joshua Dominick, Kepro’s Senior Manager of Business Intelligence & Outcomes. “The sheer volume of clinical entities offered in Text Analytics for health far surpassed anything comparable on the market.”

As Kepro got deeper into its exploration and use of Text Analytics for health, it discovered the additional value and power of the entity relation extraction capabilities within the text-mining AI service. These quickly expose key relationships between entities—relationships that can be hard to find manually in 1,000-page documents.

The right information faster

Kepro developers used Azure App Service to quickly build an intuitive UI where reviewers upload and choose pertinent files. Reviewers can select and deselect entities, based on review requirements or chart assessments, and instantly identify the entities and the relationships between them.

For example, when reviewers search for references to a specific medication, instead of combing through hundreds of pages of unstructured information, they use Text Analytics for health to automatically scan the documentation and identify and highlight every reference to that medication, including data about dosage and when and how it has been administered.

Drawing on the flexibility and versatility of Text Analytics for health, Kepro’s UI organizes and presents all the entities and relationships identified by the scan in an easily consumable manner—a table of contents, scannable data tables, highlighted entities, and links to their specific instances in the document—so that clinicians can review and understand a specific entity’s existence and context in a beneficiary’s medical history.

“With Text Analytics for health, we can find key indicators in clinical documents instantly and then abstract and display all pertinent information to improve the speed and quality of our reviews,” says Andrea Browman, Vice President of Utilization Management at Kepro.

Higher efficiency, better healthcare

Kepro is developing collaborative intelligence solutions that combine the clinical skills of our staff with the speed and automation capabilities of AI.  By incorporating Text Analytics for health in the clinical review workflow, Kepro staff conduct better, more efficient, and more thorough quality and medical necessity reviews; and they do it much faster, which helps patients get the right healthcare when they need it. Healthcare providers get their authorizations sooner, so they can focus less on administration and more on providing needed care. It all adds up to positively impacting patient outcomes.

“We see a direct link between the quality of clinical information providers can access and the quality of care they’re able to deliver,” says Sean Harrison, Senior Vice President of Product Development and Health Intelligence at Kepro. “While extracting key clinical information from very large documents can seem like a proverbial ‘needle-in-a-haystack’ exercise, with the right tools, it’s actually quite doable. Privacy, security, and safety are also critical to Kepro AI initiatives.  Microsoft’s commitment to responsible AI principles is another reason we developed our solution around Text Analytics for health.“

“Kepro is one of the most innovative healthcare companies using machine learning and natural language processing to improve clinical reviews and quality assessments” says Parashar Shah, AI Platform Product Manager at Microsoft. “They have been an early adopter of Microsoft Cognitive Services, applying Text Analytics for health across a wide range of clinical use cases that have a tangible impact on improving health outcomes.”

By using Text Analytics for health, Kepro will more efficiently meet its contracted service-level agreements (SLAs), sustain its status as a trusted partner and create future operational benefits. Offloading laborious and inefficient manual document searches from review teams helps prevent bottlenecks and backlogs, minimizes overtime, and reduces the need to onboard additional staff to meet demand. And because reviewers now spend less time hunting for the information that they need, they have more time available for analysis, which yields faster turnaround, cost savings, and, most importantly, higher quality healthcare.

Find out more about Kepro on Twitter and LinkedIn.