The EHR Data Mine

Recently, EHRIntelligence explored how data from electronic health records can be used to support clinical informatics. Informatics involves using data to advance the delivery of patient care. While this used to be an arduous task of collecting de-identified information and manual data entry, the widespread adoption of EHRs in healthcare has led to an explosion of immediately actionable data. The Office of Health and Human Services reports that 96% of hospitals are using EHRs (an adoption rate consistent for nearly a decade now) along with 78% of office-based physicians.

For internally generated data, EHRs have been a vehicle for tracking and trending data, improving decision support for patients. External data is a helpful supplement but can require investment in digital interfaces or coordination with health information exchanges (HIEs) to receive useful data.

THE EHR’S DATA IMPACT

Population health can be a difficult area of study in healthcare because of how manual and time-consuming previous methods like surveys can be. So long as a hospital’s patient population can produce a representative sample, data found in the EHR allows for real-time analysis, an exciting change that can be applied to public health threats like pandemics ore more regional phenomena like how geographies or demographics affect health outcomes.

An example of how this is being used is in the Department of Veterans Affairs with a cohort that exposed how, across a variety of different community types, a patient’s neighborhood food environment increases their risk of type 2 diabetes. The VA has the benefit of being a nationwide system and its nationwide, up-to-date data set is a great fit for studies like this.

Another area that EHR data can have a direct impact on is clinical trials. Clinical trials have been a challenge for years, both in meeting recruitment goals and in creating properly diverse cohorts for the studies. It’s commonplace in any recent article published about the impact of AI on healthcare to see that clinical trials are an area in which researchers expect advanced computing models to make a large difference. The data needed to make these AI programs effective is found in the EHR. Clinicians can search their EHR for patients that match a trial’s profile and even notify patients of the opportunity through their chart as well. By examining the larger population of matching candidates rather than using traditional recruitment methods, researchers can actively seek out underrepresented groups needed to generalize results more accurately.

What about local, community, or rural hospitals? These institutions can overcome their naturally smaller datasets if they’re using a major EHR like Epic or Cerner’s offering. These large technology vendors’ software is collecting data spanning thousands of locations and millions of patients, giving them the opportunity to offer massive de-identified datasets for use in surveillance, reporting, and clinical research applications.

?

WHAT’S MISSING?

There’s no doubt that the EHR is a valuable tool for improving patient care and in finding insights from data that’s already being gathered. And even though it’s miles ahead of attempting the same processes without an EHR, there are still challenges to be overcome. For one thing, despite many hospitals using the same EHR software, a researcher can’t simply query all hospitals using Meditech to find a desired population. Hospital data is proprietary, confidential, and uses unique nomenclature, so big data access usually comes at the expense of identified patients and fewer data fields.

But before standards like LOINC and SNOMED or exchange formats like FHIR can be used to normalize data and create agreed upon interoperability principles, healthcare institutions need to be sure that their own EHR data is complete and accurate. Controlling processes within a hospital’s walls are one thing, but healthcare offices receive a lot of files from the outside.

How can hospitals ensure they have the best data possible in the EHR?

• Interfaces: It doesn’t get better than having an interface with an outside provider that can send you every piece of discrete data you need, properly matched, instantly to your EHR. Integrations are more valuable the more concentrated your outside providers are.The problem with interfaces is that they are costly and time-consuming to implement while only linking a single outside entity. Organizations must balance current and expected volumes from their partners, avoiding wasteful integrations that might only service a handful of documents while still recognizing the value for more prolific document senders.
• Staffing: It’s getting more difficult to manage both staffing and burnout in both the clinical and administrative sides of healthcare. Regardless, it’s an area in which hospital leadership has direct access and control to improve outcomes. EHRs don’t cause burnout in and of themselves and the concerns about them doing so are overcome with training, customization, and the right people for the job, all areas a hospital can impact.
• Vendors: When integrations are impractical and documentation is becoming backlogged, departments may resort to indexing documents and posting an image rather than keying discrete data; they may also seek out third-party services. Vendors in this space may act as an outsourced labor force for simple tasks like indexing or can be used to replace a staff entirely with premium services like data abstraction. While this requires a certain level of trust, giving clinicians discrete data versus image files to scour saves time and produces better outcomes.Some organizations opt for automation software which allows them to control costs at the same time as they ingest more data more quickly and accurately. Software like Extract’s classifies, indexes, matches, identifies, and delivers data and documents to an EHR or other systems like an LIS or DMS. Your own staff who are familiar with your processes verify that we got things correct, then move on to the next document. Automation software turns the unstructured, non-interfaced, pain in the butt documents into the fastest and easiest you’ll ingest this side of an interface.

If you’d like to see how you can implement a solution to get you caught up, manage fluctuating volume, or deliver your clinicians more complete data, please reach out and we’d be happy to show you how it all works.