OR WAIT 15 SECS
A review of important factors to consider before implementing the solution that will transform EDC.
The use of Electronic Data Capture (EDC) systems for the capture of clinical trial data is increasing.1,2 The estimates of the extent of adoption are varied and confounded by the definition of what constitutes EDC (e.g., electronic diary data is included in some of the estimates). The true percentage is somewhat irrelevant, as EDC is clearly replacing paper for the majority of trials.
The speculated benefits of EDC are many, including improved quality, faster access to data, reduced effort, and faster database lock. So if this is the case, why has it taken more than 20 years for Remote Data Entry (RDE)/EDC to gain acceptance? Whatever the reasons, be they technical, process, and/or cost related, the uptake has been slow and there are still drawbacks—the primary one being re-entry of data by the site staff. Others include delays in data accessibility, albeit shorter delays, and the continued need for source data verification, which is on the critical path to database lock with EDC.
The global emergence of the availability of patient care data electronically, albeit a slow emergence in the United States, presents the opportunity to address all of these issues by transferring data electronically from the electronic source.
In a nonscientific survey conducted on the ACT Web site, 71% of 407 respondents felt that electronic health records (EHR) and EDC would merge within the next five years—a very short time frame when one considers the comparatively slow adoption of EDC. Of those that disagreed, it is reasonable to speculate that the primary reason was not a belief that it will not happen but rather that it will not happen within five years.
There have been successful eSource pilots (e.g., CDISC's Starbrite study at Duke and Siemens' implementation at the Technical University of Munich). CDISC is coordinating a demonstration of EHR/EDC interoperability at the HIMSS show this month. The eClinical Forum and PhRMA have joined forces to produce a white paper on "The Future Vision of EHRs as eSource for Clinical Research."3 ExL Pharma has organized two EHR/EDC meetings. So the momentum is there.
The purpose of this article is not to describe the perfect solution nor to lay out a road map for EHR/EDC adoption. Its intent is to set out a list of "points to consider" before embarking on the path to data capture nirvana. The scope is multicenter pharmaceutical-sponsored global trials.
Before discussing the use of EHR data for clinical trials, it is important to define what is meant by EHR in this context. There are a number of terms used interchangeably to describe systems that computerize patient health care data. The terms Electronic Health Record (EHR), Electronic Medical Record (EMR), Electronic Patient Record (EPR), Clinical Patient Record (CPR), and Lifetime Clinical Record (LCR) are all terms that are used by various individuals and organizations at times to mean the same thing and at other times to mean different things.
The most widely used term today in the context of the use of data for clinical research is EHR, but the same term is often used to describe a series of interlocking solutions that are tied to a series of care delivery activities (including nonclinical ones such as billing) and that track a patient across multiple institutions such as hospitals and physician practices and even across different health systems.4 It is not necessary to wait for such systems to be widespread before electronic data can be automatically transferred to clinical research systems used by the biopharmaceutical industry.
Also, these "electronic" systems do not necessarily contain data in a format that can be used for electronic transfer of data. An EMR can be comprised of scanned documents with nothing more than a retrieval capability based on patient name or some other identifier. This may still qualify as an EHR, but would not contain source data and would not enable the automatic transfer of data for clinical trials.
In the remainder of this article, the term EHR will be used to mean source data captured in a format that enables it to be electronically transferred in a structured way to clinical research systems, with the qualification that some manipulation and mapping may be required during the transfer. The automated transfer of data from EHRs to an EDC system will be referred to as EHR/EDC.
The care of patients is the number one priority for health care providers, be they hospitals or physician practices. Therefore, research requirements are going to be secondary. The implications of this simple truth are manifold.
When choosing new information systems, the decision makers are going to be focused on solutions that optimize the management of patient care. Features to facilitate clinical research are going to rank low on the selection criteria, if they appear at all. The corollary of this is that Healthcare Information Technology (HIT) vendors will therefore focus on functionality to support health care provision and not clinical research—so features to support clinical trials such as the ability to collect and store trial specific data not required for routine care will have low priority on their development plans. A particular implication is that HIT vendors will be reluctant to invest in the effort to evaluate 21 CFR 11 compliance and rectify any shortcomings. This does not mean that vendors will not develop functionality to enable clinical trials, but any scalable implementation of EHR/EDC will have to be able to cope with systems that are not optimized for clinical trials.
Health care providers will not replace their existing health care information systems purely to facilitate research. Therefore, any EHR/EDC solution must work with a large variety of systems, with varying degrees of sophistication.
Because of the focus on patient care, clinical trial processes and procedures must cause minimal disruption to the existing processes and procedures in place for standard clinical practice. One implication of this: Any EHR/EDC solution that requires health care data for patients enrolled in trials to be captured and managed differently than data for patients not enrolled in trials is unlikely to be successful. Staff do not have the time or inclination to collect the same data for two patients differently, based on trial participation. Thus, it is most likely that solutions that automate the capture of trial data will receive data directly from the EHR rather than from an intermediate source that populates both the EHR and EDC system.
The adoption of new data standards to make data collection and transfer in clinical trials more efficient is unlikely to be acceptable to health care providers. Historically, there has been little reason to share data outside of an institution, so the data structures have been developed to optimize the internal operations of each institution. The collaboration between HL7 and CDISC and the work of Duke Clinical Research Institute to standardize the collection of cardiovascular data across both health care and research are two examples of positive steps forward. However, there has to be a realization that data will have to be mapped in some way for each institution—although this mapping can be used across multiple trials. An example of how this is achieved is shown in Figure 1.
Data Mapping Makes EHR/EDC Possible
Data recorded for individual patient care does not have to be highly structured nor coded for subsequent integration for summarization purposes. There are exceptions (e.g., disease coding for billing purposes), but for some data types the data is not collected in ways that directly enable transfer of trial data. Today's EHR systems do offer the ability to collect data in a structured and coded way, but often the data comes from dictated notes and is stored in an unstructured way that is not amenable to automated extraction. Therefore, to increase the percentage of readily usable data, it will be necessary for the users of the system to record the data in ways that, although possible, are not currently utilized routinely.
An example of this is medical history data. If recorded by a physician, it will most likely be in a dictated, unstructured record. Related to this is the requirement of regularly scheduled visits for clinical trials—a concept that does not exist, at least not in the same disciplined way, within EHRs. Therefore, methods have to be developed that capture only the data that is required for particular visits.
IT staff within Health Care Providers (HCP) are focused on systems that support the provision of health care and the associated administration. With EDC studies, they are minimally involved, assisting with any initial evaluation of the suitability and requirements of the site. EHR/EDC is perceived as being disruptive, taking them away from their primary tasks. Therefore, solutions must minimize the site's IT department's implementation and maintenance effort, and it is essential to involve the appropriate IT staff in any discussions, as a motivated investigator will not be able to participate without the buy-in of the IT staff.
All of the above relate to the relative priority of health care and research. The final points to consider are security and data privacy. With EDC, no data is transferred electronically from HCP systems and there is no connection between such systems and those used by the pharmaceutical company. However with EHR/EDC there is a concern about the security risk, as systems have to be opened up to allow for the transfer of data. This concern should be addressable through appropriate security measures, and the automated transfer of data outside of the HCP's firewall is done today for applications such as point of admission insurance coverage verification.
A related concern is data privacy and the need to ensure that only data that is required for the clinical trial is transferred to the pharmaceutical company. This implies that data manipulation and mapping needs to be done within the firewall or at a trusted third party. The privacy issue is further complicated by regional, national, and state differences.
Overall, the implementation of an EHR/EDC solution at a healthcare provider is technically feasible today but needs to be carefully considered and requires in-depth discussions with multiple stakeholders at the HCP, including investigators, the IT department, data privacy officers, and, most importantly, the administration. Given the novelty of automated transfer of data, the administration will be concerned on many different fronts, and any potential implementation needs to be discussed in detail with them long before any anticipated trial start date.
Although the majority of issues in implementing EHR/EDC will focus on the site, there are considerations that focus on the use of such an approach by the pharmaceutical industry.
As described earlier, implementation at a particular site is going to be time consuming, and therefore it is very likely that multicenter studies in which such an approach is used will also involve sites that use "traditional" EDC. Therefore, to minimize disruption, the data should flow into the company's existing EDC system so that existing edit checks and discrepancy management can be used.
The alternative is for the data to be checked and discrepancies managed in some other system—a major disruption. One implication of this is that the industry is unlikely to be interested in receiving data from EDC systems used at HCPs unless it flows directly into the sponsor's EDC system. Also, this approach is very likely to lose the advantage to be had from eSource, as the data in the HCP's EDC system will most likely have been transferred from the EHR, either automatically or, more likely, manually.
It is very likely that any prospective clinical trial will require data to be collected that is not part of the data routinely collected as part of patient care, and this data will most likely not be entered into the EHR. Therefore, in order for the "hybrid" approach described earlier to be viable, it is important that trial data that is not available in the EHR can be merged with data automatically transferred within the electronic case report form (eCRF). Otherwise, a different database would have to be developed for sites using EHR/EDC.
In order to maximize the benefits of automated transfer of source data, careful consideration should be given to changes to the data once it has been transferred. Changes made at the source should be propagated to the sponsor's system, ideally automatically, along with the necessary audit trail information. Also, site staff should not be able to make changes to eCRF data if the data has been transferred from an EHR. This capability may not be possible with all implementations, so processes need to be put in place—similar to those used with central lab data—to ensure that changes are only made at the source.
The transfer of data automatically to an EDC system is going to require investigation with the vendor and may require additional configuration and changes to the system. If this is the case, it may be possible to do proof-of-concept studies with workarounds to ensure that the investment in changes is justified. Also, it is likely that as the automated transfer of EHR data becomes adopted, EDC vendors will need to adapt their systems to optimize such transfers.
Another consideration is regulatory acceptance. CDISC's eSDI group, which was commissioned by the FDA and included FDA representation, has issued a document that describes the requirements for using eSource in regulated clinical trials. It also describes five scenarios that, if implemented appropriately, could meet those requirements.5 As the direct use of EHR data is still very much in its infancy, it is advisable to have specific discussions with the FDA and other regulatory bodies before using the approach in a clinical trial that might be used to support a submission.
Implementing the infrastructure to enable the automated transfer of EHR data is not a small undertaking and may require investment in sites at which a sponsor will run multiple trials. It will certainly require time-consuming discussions with multiple stakeholders at the site.
The electronic transfer of EHR data for clinical trials is expected to be the future of data capture at some point in time and offers significant advantages over today's EDC systems, where manual data entry is the primary input method. There are multiple challenges, but these challenges are well characterized and all surmountable. Before beginning implementation there are many points to consider, but it is indeed possible today.
1. D. Borfitz, "Conspiring Forces Behind EDC Adoption," CenterWatch, 10 (2) (2003).
2. J. Paul, R. Seib, T. Prescott, "The Internet and Clinical Trials: Background, Online Resources, Examples and Issues," Journal of Medical Internet Research, 7 (1) p. e5 (2005).
3. "The Future Vision of EHRs as eSource for Clinical Research" (Release 1, 14 September 2006). Available at: http://www.eclinicalforum.com/frameknowledge.htm.
4. T. Handler, R. Holtmeier, J. Metzger, M. Overhage, S. Taylor, C. Underwood, "HIMSS Electronic Health Record Definitional Model Version 1.0," 6 November 2003. Available at: http://www.himss.org/content/files/EHRattributes.pdf.
5. Electronic Source Data Interchange (eSDI) Group, "Leveraging the CDISC Standards to Facilitate the use of Electronic Source Data within Clinical Trials," 20 November 2006. Available at: http://www.cdisc.org/eSDI/eSDI.pdf.
Hugh Donovan is general manager, clinical trials business, with Siemens Medical Systems, 51 Valley Stream Parkway, MS E56, Malvern, PA 19355, (610) 219-5140, email: firstname.lastname@example.org.