Since the American Recovery and Reinvestment Act of 2009 incentivized the adoption and use of electronic health records (EHRs), EHRs have become ubiquitous in the health care industry. Recent federal reports show about 84% adoption in hospitals and about 86% adoption in office-based practices.
Patient information that was once captured on paper is now being regularly recorded and stored in EHRs, creating new opportunities for analyzing and drawing insights from these increasingly rich data sets. While EHRs can support easier access to and sharing of information by individual providers and patients, larger efforts (e.g. population health interventions or the use of EHR data for predictive analytics) struggle to make sense of the newly digital information due to limited technologies available to process complex EHR data.
Why is EHR data complex?
A recent article in Medical Care by Dan Zeltzer and colleagues focused on the use of predictive analytics, and reviewed some of the challenges researchers and other users encounter when processing large amounts of EHR data compared to administrative claims data. First, health data is sensitive and the sharing of EHR data presents security concerns due to issues around privacy and confidentiality. Second, EHR data varies between facilities due to the multitude of EHR vendors and/or health care providers using different EHR features. For instance, identical lab tests might not display in a similar manner across two hospitals with different EHR vendors.
The authors of the Medical Care article further discuss how these variations in data result in predictive models that may not be generalizable because the models were “trained” with EHR data from a single site or single health system. To expand the functionality and applicability of these tailored predictive models would be costly and time consuming. In comparison to EHR data, claims data is less messy and easier to process due to the use of uniform codes for billing purposes. Claims data usually provides fewer patient details, but those details are more structured and standardized across various health care systems.
Why should these challenges be addressed?
EHR data presents numerous challenges and it’s imperative these are addressed now. In early 2019, the Office of the National Coordinator and the Centers for Medicare & Medicaid Services released proposed rulemaking calling for the expansion of EHR interoperability. Expanding interoperability would exponentially increase the amount of accessible EHR-based data throughout the health care industry. As opposed to functioning in a silo, interoperability allows researchers, providers, and patients access to much larger amounts of data. While interoperability can improve care continuity, there needs to be a way to make sense of the newly increased amounts of accessible EHR data.
What’s being done now?
To help combat the daunting challenges that processing large amounts of EHR data present to researchers, providers, and other analysts, a recent surge of innovation in artificial intelligence (AI) has started to emerge. Success stories are beginning to showcase the benefits and possibilities of this new technology. In 2018, the University of Iowa Hospitals and Clinics combined machine learning technology (a form of AI) with EHR data to prevent surgical site infections by providing clinical decision support to nurses and providers at the point of care. To date, this work has led to a 74% reduction in surgical site infections across a three-year period, which is a cost savings of about $1.2 million.
Another example is from the University of Maryland Medical System, where researchers have developed a machine learning model that can review health data variables in real-time, which can then be used to predict the likelihood of a patient’s readmission. Upon evaluation, the researchers discovered their model could better predict readmission than frequently used readmission scores. Considering that readmissions occur for almost 20% of US patients, this could be a valuable tool in identifying and preventing costly hospital readmissions.
Next steps
There are many insights EHR data can provide, and coupled with the capabilities of AI these insights will only continue to grow. The individual success stories are numerous but, as mentioned above, there is a need for additional work on how to expand these success stories to make them usable across multiple institutions. Further, with any new advancement in health care technology, more research is needed not only to investigate its potential advantages, but also to investigate its potential drawbacks. Concerns about whether technology will benefit populations equitably and whether these advances improve patient care and outcomes (and not just organizational bottom lines) are important additional considerations.
To read more about how EHR data compares to claims data when used for predictive modeling, check out this earlier Medical Care blog post by Lisa Lines.
Alexa Ortiz