Over the last decade, sponsors, sites and vendors have made significant progress automating the transfer of data from the site EHR into the sponsor EDC.
But the initiative only partially addresses the industry’s goal of seamless eSource-to-EDC data transfer. Specifically, we might think of two dimensions that determine to what degree an eSource-to-EDC initiative succeeds:
- How much of the EDC data does the EHR map to? We might call this the mappability dimension.
- How many sites utilize an EHR system in the first place? We might call this the prevalence dimension.
The two dimensions can be visualized as follows:
First dimension: Challenges with EHR Mappabilty
EHR data does not always map directly to the EDC. First, many study procedures are protocol-specific and not standard-of-care; therefore, they do not have analogous templates in the EHR. A good example is Adverse Events. While some EHRs incorporate the Adverse Event concept, many sites may not utilize an EHR with this feature, or may not be trained to utilize it. As a result, many Adverse Events relate to observations captured in unstructured notes, such as Urgent Care or Emergency Room encounters.
Second, many structured data sets in the EDC are not always accurately and completely covered in the most analogous EHR data domain. A good example is Medical History. An oncology protocol, for instance, may require the site document an extensive list of a subject’s diagnoses, but the formal “Diagnoses” section of the EHR might contain only a fraction of the data required – other medical history listings may be reflected in unstructured text, or in structured data fields such as Labs, Problem Lists, Vitals, etc.
For examples of how the Medical History and Concomitant Medications eCRF forms often map poorly to their analogous data domains in the EHR, see this study by CRIO, which showed that 98% of Medical History and Concomitant Medication records in the EDC required some sort of reconciliation from the EHR.
Estimates of mappability vary from 15% to 70%; the wide range reflects not only the emerging maturity of the software, but also the variety of eCRF requirements encountered, the therapeutic area (e.g., higher concordance between EHR and EDC on oncology studies where there are more standard-of-care procedures) and EHR documentation practices by sites.
Second dimension: EHR prevalence is limited
Of course, if a site is not utilizing an EHR, then an EHR-to-EDC integration is of no use. For many studies, especially those outside oncology, research is conducted at community-based sites, such as physician practices, professional research sites or, increasingly, mobile providers.
As CRIO’s research shows, 75% of non-oncology research sites do not utilize EHR for source data collection. They may utilize the EHR for feasibility, eligibility determination, or scheduling, but they often use either paper worksheets or industry-specific eSource tools for source data collection.
How CRIO’s Central eSource fills the gap
For studies where community-based sites are used, CRIO’s Central eSource can scale in both dimensions.
First, because CRIO’s eSource template is centrally designed, CRIO’s study design team can ensure that the eSource template maps to the eCRF from the outset. Because the template is study specific, the mappability between the templates is, by definition, very high, with a few exceptions:
- CRIO does not have dynamic logic that opens and closes certain forms and visits; because it’s an eSource system, CRIO does not have the prompt question of whether a blank form is blank because it didn’t occur or because the site hasn’t completed data entry.
- Some formats do not directly translate. For instance, CRIO does not support a time-only field; instead, CRIO’s time stamps are always captured as date-time fields.
Beyond that, CRIO’s eSource is designed for contemporaneous and granular data collection, whereas eCRF forms are often designed for secondary and summary-level data collection. CRIO optimizes for site usage, and sometimes that results in question structures that differ from their analogous eCRF counterpart.1
When a conscious effort is made to harmonize study designs, CRIO’s central eSource can achieve 90%+ data mappability.
Second, CRIO’s centrally designed eSource can be delivered to as many of the sites on the study who elect to participate. Today, CRIO clients represent about 25% of all U.S. sites on studies involving chronic conditions, and non-CRIO sites can be set up with their own re-usable account.
With CRIO’s support, sponsors that clearly communicate the benefits of CRIO’s central eSource program can achieve 70% or higher site participation.
Summary: Central eSource completes and extends the EHR-to-EDC initiative
The following table summarizes the benefits and limitations of EHR vs. Central eSource in integrating with the EDC:
| EHR | Central eSource | |
| Mappability of system | Limited to certain data domains; AI required for additional mapping | Limited mainly by the ability to harmonize designs during the study setup process |
| Prevalence of system | Most AMCs, but few community-based sites; sites not using an EHR cannot “opt in” to use one for the study | 25% of US-based community sites are CRIO clients, and any non-CRIO site may opt in for the study |
| Mappability (in practice) | 15-70% | 90%+ |
| Prevalence on study (in practice) | To date, limited to a small number of AMCs that have implemented EHR-to-EDC software systems | 70% attainable on a study; prior implementation of eSource not required |
Clearly, any eSource-to-EDC initiative needs to consider CRIO Central eSource as a serious option to achieve meaningful scale across the sponsor’s full study portfolio.
1A good example might be the physical examination. The source might list every body system required to be examined and ask for the PI’s assessment of each, whereas the eCRF might simply ask for a free-text entry listing of any clinically significant abnormal observations.