Opportunities for advancing clinical outcomes research are on the rise. This is largely thanks to electronic health records (EHRs)—a patient care tool that is already capturing meaningful clinical information on an unprecedented scale.
But the sheer existence of EHR databases does not automatically lead to significant, generalizable, and actionable research findings. This largely results from the patient-centric (versus research-centric) nature of EHRs, and the varying degrees to which they are used. Specifically, research productivity is threatened by the increased data management and transformation burden EHRs place on clinical outcomes researchers. This burden involves:
- Quality concerns: Recorded by practitioners, EHR data face classic human error issues.
- Collection inconsistency: EHRs are designed to record relevant patient information. However, what is relevant in an ICU differs from what is relevant in a family doctor’s office. Furthermore, the type, amount, and precision of information recorded in EHRs vary across individual practitioners and whole care networks.
- Generalizability: Younger practitioners and wealthier healthcare centers rely on EMR software more consistently. Are the conclusions derived using EHRs from such sources really generalizable? Or are the findings confounded by geographic, socioeconomic, and other factors?
- Construct validity: Researchers must consider what metrics actually represent their phenomena of interest. In the absence of pain reports on well-validated and quantified scales, for example, can a prescription dosage or diagnostic code be proxy variables for pain levels?
- Information accessibility: Up to 70% of useful information is captured in EHRs’ unstructured text fields (i.e., practitioner notes). Determining what text is meaningful and having staff review all text is labor intensive and unscalable. Algorithmic approaches to automate this are increasingly available (e.g., natural language processing), but are computationally demanding.
With these factors in mind, the question becomes how to minimize the extra data management and transformation burden EHR data places on researchers.
One major opportunity involves merging datasets. For example, connecting EHR and prescription claims data helps validate a treatment strategy noted in EHRs. Similarly, matching EHRs with genetic or family medical histories stored in a medical registry database, for example, can offer a fuller picture of patients of interest.
Relatedly, centralizing the EHR data aggregation and curation task can help ease the burden on individual investigators. Notably, an Architecture for Research Computing in Health (ARCH) strategy centralizes EHR, biobank, claims, electronic data capture, and other available data from diverse sources at an institutional level. Managing and transforming data at this level pools data, resources, and expertise. This ultimately improves research efficiency overall, which is a major goal of the clinical research field generally.
Intuitively, actually pursuing these strategies requires a well-rounded informatics infrastructure. Specifically, such a database platform must be flexible (conducive to both structured and unstructured data), scalable, able to curate data from diverse sources (e.g., both medical registry and EHR data), and able to manage complex data quality issues.
RexDB® offers such a platform. Not only is RexDB a proven pipeline for clinical research database merging, curation, and centralization, but the Prometheus Research team behind RexDB are informatics experts with specific experience developing database architecture, customizable reporting options, querying tools, and data transformation methods for clinical data. Altogether, Prometheus Research can help researchers truly leverage the power of EHR data through software toolkits, clinical informatics expertise, and proven experience transforming raw data into research-grade information.
Do you know of any other strategies that effectively promote researcher productivity in the clinical environment? Learn more here.