How to extract data from EMRs: Automating structured and unstructured data retrieval

Electronic medical records (EMRs) provide a rich data source for clinical outcomes research, including vital signs, laboratory values, and other care protocols and outcomes. Much of this information, however, is frequently embedded in unstructured narrative text and, therefore, unusable or inaccessible.

Health record data collection and extraction – including information from radiology reports, discharge summaries and dosing, for example – is possible, but how does it benefit payer organizations?

Captured but trapped data, which often sits inside data repositories and silos like EMRs without assignment or value to health systems and payer organizations, can be used to measure care quality, care coordination, and even plan design.

How to extract EMR data

Natural language processing (NLP) is a solution that can extract and interpret data insights from EMRs, extrapolating numeric data from narrative text to identify medical concepts. Payer organizations that customize NLP tools to perform numerical data extraction gain a significant advantage toward better plan design and reducing costs to members and the organization.

EMR data extraction can be a complex task. This is partly because each EMR provides a wide range of data collection points, data repositories, and rules for how information is gathered throughout the point of care. This simply means there are options and strategies for how the data entering these systems comes in. EMRs contains a mixture of structured and unstructured data in nearly every instance. The interpretation phase follows data collection and impacts the path forward.

So, while most of the data that gets entered about a patient are structured initially, as data enters from multiple sources – intake, physicians and other caregivers, revenue teams, and payer-generation information ̶ there’s still copious amounts of unstructured information that begins to compile for each patient. Yet, as cumbersome as unstructured data may be, it has a significant value to health systems and their payer partners.

NLP is a highly effective means to extract data from an EMR. So, while structured data is well worth its weight, unstructured data can be where the treasure lies. There’s a plethora of information in unstructured data that can be interpreted through algorithms and other technology-enabled solutions, specifically when open text and free text data present an issue in an unstructured form.

Free-text, free-form fill notes often contain essential data for payers, including utilization information that can be used for plan design. For example, the free-fill form usually includes clinicians’ information to determine, predict, and treat diseases or conditions and the recommended action. Other ways NLP is beneficial to healthcare data extraction is finding relevant or exact keywords or phrases in a text and other documents.

Once the data is extracted from the EMR, it can easily be placed into a single structured record, and patterns identified.

APIs for EMR data extraction

Perhaps most appropriate for the work of EMR data extraction is an application programming interface (API). APIs are mechanisms that enable two software components to communicate with each other using a set of definitions and protocols. Therefore, payers can utilize API technology to simplify EMR data extraction and collection.

With APIs, providers, health systems, and payer organizations can extract data from any EMR and move it wherever best suits their data needs. Likewise, payers wanting or needing to report specific data to patients can do so from across multiple locations and facilities into a single platform, document, or report.

Enabling efficient EMR data extraction with the help of technology interfaces like APIs still requires automated process and a standardized manner of accessing the health record attainable. An established format where all available patient information is synoptically summarized in a standard manner is also required.

Types of data extraction available to payers

Payer organizations seeking a technology solution to extract data from EMRs must consider the types of data they want to collect, how the data will be used, how it will be consumed, and by whom once structured.

For example, some vendors excel at collection and extraction, while others are better at working only in the practice space, with health systems, or with just payers. Identifying a vendor equally adept at serving each sector is a rarity. Payers, of course, encounter and require their own level of complexity, which requires a partner that understands these specific challenges.

Perhaps the most significant challenge in EMR data extraction is not the technology or processes required to access this data but the cleanliness in which the data is delivered to the subscribing organization. Ultimately, the value of EMR data extraction is not mining the data but mining the gems in the data and delivering them as finely cut, clean nuggets of information to inform decisions throughout the payer’s ecosystem. Thus, a data extraction tool that makes medical data easier to derive value from and ensures accuracy is paramount.

How Inovalon helps

No matter the APIs used or the NPL programs employed, EMR data extraction must replace manual record retrieval so that payer organizations can access relevant medical records digitally, cleanly, and actionably.

The Inovalon ONE® Platform provides automated clinical data extraction to more than 375,000 providers connected who can access relevant digital patient medical records 10 times faster than manual retrieval. In addition, the solution automatically captures comprehensive, rich data — for clinical quality and risk score accuracy improvement, clinical trial support, and pharmacy-related prior authorization data aggregation.

To date, Inovalon has delivered more than $35.5 million in risk adjustment value to its customers by leveraging Electronic Record On Demand machine learning technology.

To learn more about Inovalon’s EMR data extraction solutions, discover the Inovalon ONE® Platform.

Inovalon and design^® and Inovalon^® are trademarks of Inovalon, Inc.