Participant/EHR Characteristics

There were 64 responses in the survey representing over 98% of the CTSA Consortium; 122 individuals contributed to the responses. The largest group of respondents (48%) self-identified as informatics specialists while 16% did so as recruitment specialists, 11% as regulatory specialists, and 25% as other.

When asked what EHR program their institution/hospital uses, 42 respondents said Epic, 13 said Cerner, and 3 said they use a homegrown system. Fourteen indicated “other” (including most commonly Allscripts and Centricity), with 1 of these reporting they were in the planning stages of identifying a system. Five CTSAs reported using 2 or more EHR systems.

Methods of EHR-Based Recruitment

Although the use of patient portals (e.g., Epic MyChart) to notify patients about research opportunities was reported as initially implemented or fully operational at only 20% of responding institutions; however, 70% were either exploring or planning the use of such tools (Table 1). Demand for these tools was reported as frequent or very frequent by 25% of respondents, and another 30% described it as “occasional” (see Fig. 1).

Fig. 1 Summary of responses to questions regarding methods of electronic health records (EHR)-based cohort identification and recruitment. (a) Current implementation. (b) Perceived demand. Brokered data warehouse (DW), access to data warehouse by staff members; Self-serve Qry, use of self-service tools to run de-identified queries; EHR registry, use of EHRs to build patient lists to aid in recruitment; Patient portal, use of EHR patient portals to notify patients of research opportunities; Provider alerts, use of electronic alerts to care providers of patients in clinic meeting eligibility requirements; Research associate (RA) alert, use of electronic alerts to the research team if patients in clinic meet eligibility requirements; Business intelligence (BI) tools access, research given direct query access to data warehouse through business intelligence tools.

Table 1 Responses to questions regarding current implementation of methods of electronic health records (EHR)-based cohort identification and recruitment

Patient portal, use of EHR patient portals to notify patients of research opportunities; Provider alerts, use of electronic alerts to care providers of patients in clinic meeting eligibility requirements; Research team alerts, use of electronic alerts to the research team if patients in clinic meet eligibility requirements; Brokered data warehouse, access to data warehouse by staff members; Self-Service query, use of self-service tools to run de-identified queries; Business intelligence tools, research given direct query access to data warehouse through business intelligence tools; EHR registry, use of EHRs to build patient lists to aid in recruitment.

EHR alerts to care providers about patients in clinic meeting eligibility criteria were reported as in exploratory or planning stages by 44% and initially or fully implemented by 45% of respondents. Demand for these tools was reported as frequent or very frequent by 22% and “occasional” by nearly half (47%) of the respondents. Similarly, EHR alerts addressed to research teams, when an eligible patient is scheduled or attends a clinic visit, were described as exploratory or planning by 39% and some form of implementation at 48%. However, reported demand for this method was notably higher, with 39% describing it as frequent/very frequent.

With regard to access to EHR data via a data warehouse, most institutions (94%) said they were in initial implementation of, or had fully operational processes for providing access via a staff member or analyst (Table 1). This high level of implementation was consistent with perceived demand for such services; 88% of respondents described demand as frequent/very frequent. Similarly, most institutions (92%) said they had initially or fully implemented self-service tools (e.g., i2b2 [Reference Murphy8]), allowing researchers to run queries on de-identified aggregate data. Demand was also high, with two-thirds describing it as frequent/very frequent. Use of off-the-shelf business intelligence tools that allow researchers to run more complex queries on EHR data was much less prevalent with 45% reporting initial or full implementation. Perceived demand for these tools was mixed, with 30% reporting it as frequent or very frequent.

Finally, nearly two-thirds of institutions (64%) had initially or fully implemented the use of EHR data to build registries to aid in recruitment (Table 1). Demand for this approach was strong, with 30% describing it as occasional and 42% as frequent/very frequent.

When asked whether there were other EHR-related approaches they had considered, piloted, or implemented to facilitate research recruitment, 50 respondents (78%) said yes. Among those who provided additional details, common elements included:

• ∙ National networks of data sharing, either industry based (such as TriNetX^®, or Cerner PowerTrials) or funded programs such as the PCORnet and SHRINE-based networks.

• ∙ Vendor-based or homegrown tools to assist in matching patients with clinical trials eligibility criteria.

• ∙ Other outreach, for example, asking patients directly for permission to be contacted for recruitment purposes (including both opt-in and opt-out approaches); the use of patient portals and Web sites to allow patients to indicate preferred method of contact; the use of direct mail, electronic communication (email, text, apps), and phone; point of patient registration or clinic visit procedures; and community-based efforts.

Nearly three-fourths (72%) of respondents said they gather metrics or other evidence of the impact of EHR-based recruitment. Among those who provided additional details, descriptions fell into 3 broad categories: (1) system utilization measures, such as numbers of hits, requests, or logons; (2) recruitment measures, such as numbers of patients who were identified as potentially eligible, responded after being contacted, screened and/or enrolled; and (3) other user-related measures, such as the numbers of grants, publications, or the results of surveys and gathered anecdotes.

Workflow and Regulatory Process for Cohort Recruitment

When asked, “Does your institution have 1 or more established workflow processes for cohort recruitment into clinical research which leverages the power of EHR to identify large numbers?”, nearly two-thirds (64%) said yes, and another 22% said they were piloting such programs. Given the option to provide additional details, respondents described a variety of workflow elements, such as regulatory and related review/approval processes, including the use of committees that review requests; informatics tools and processes (e.g., i2b2/SHRINE, ACT, tools that link EHRs to CTMS or REDCap); data sources (e.g., databases, registries, data warehouses); tools and processes to assess eligibility; and tools and processes associated with initiating patient contact, including use of MyChart portal for recruitment purposes. They also described general processes and approaches, including the use of dedicated/specialized teams and services, including honest brokers, data analysts, and recruitment specialists; self-services approaches; and standard operating procedures on issues like direct mail recruitment, guidebooks, and consultations to improve recruitment letters. Examples of specific processes and tools respondents mentioned included:

• ∙ A listing of patients who can be contacted after committee approval (as opposed to having to go through their direct provider).

• ∙ An automated interface that streamlines the load of EHR data into REDCap; template implementation for data extraction to make the process easier and more efficient for analysts.

• ∙ Using patient portals and tablets to engage volunteers to opt-in for research creating a flag in the EHR to allow subsequent identification.

Nearly one-third (30%) of respondents reported substantive restrictions on workflow processes, beyond regulatory reviews and approvals. These often included logistical constraints, such as limited resources (e.g., time, staff) to carry out recruitment processes, and the need for more training and awareness regarding the tools and systems available. In addition, respondents often described limitations related to collaborations and data sharing, such as intrainstitutional challenges (e.g., sharing data between the university and hospital); the need to identify a collaborator at the institution; and, with regard to multisite studies, additional steps such as contracting, data harmonization, network review panels and governance, and IRB approval and local Principal Investigator at each site. Somewhat less commonly, respondents noted:

• ∙ Challenges associated with the nature of the study and/or data, such as studies that require chart review, study topics that are seen to be sensitive, and data availability in complex cases.

• ∙ Additional regulatory requirements, such as compliance review of data security plans, and the need for data use agreements.

• ∙ Challenges in the process of contacting patients, such as required provider involvement, creating burden for the provider as well as delays for the researcher, and the restrictions on the number of times a patient can be contacted for recruitment purposes.

A similar number of respondents (30%) reported that alternative approaches of regulatory and workflow processes had been implemented or piloted at their institution. Among those who offered addition details, descriptions commonly included workflow elements, such as:

• ∙ Regulatory and related review/approval processes, for example enabling the informatics program to sign off on compliance/security issues for studies that meet criteria for routine data uses, and the use of a centralized or delegated IRB.

• ∙ Informatics tools and processes, such as building a library of queries and computational phenotypes to streamline queries and enhance consistency across defined diseases.

• ∙ Data sources, such as use of an external registry linked to EHRs of enrolled subjects.

• ∙ Tools and processes to assess eligibility, for instance the informatics and participant core working together using honest brokers to develop lists of patients meeting inclusion/exclusion criteria.

• ∙ Tools and processes associated with initiating patient contact, such as the expanded use of EHRs to engage patients in recruitment through MyChart.

Many also described general processes and approaches, such as the use of dedicated teams and services (e.g., a recruitment team that contacts providers on behalf of the study to obtain approval to contact the patient, and creation of a Participant Recruitment Center); the use of self-services tools that have built-in honest broker capability; and workflows tailored to the type of project, data needs and sources.

Finally, over one-third (36%) of respondents said they had additional insights to offer on using EHRs to facilitate research recruitment that might be useful. Workflow elements were commonly mentioned, including:

• ∙ Regulatory and related review/approval processes, such as ensuring that HIPAA waivers for permission to contact for research recruitment become a standard part of the patient intake workflow; minimizing the time for regulatory review by the use of standard data use agreements developed in collaboration with the IRB and legal office; use of data review committees to relieve the burden of regulatory compliance assessment on the honest broker.

• ∙ Informatics tools and processes, such as the importance of evaluating data quality and provenance for each query; strict segregation of identified Versus de-identified data in data warehouse; use of open standards, non-proprietary software, and data extraction tools designed by programmers who are trained in statistics, and closer collaboration with academic informaticians and software engineers.

• ∙ Data sources, such as addressing the physical, technological, and cultural separation between the operational and research sides of the institution; integrating EHRs from other clinical partners (with different EHR platforms) into data warehouse; tools and processes to assess eligibility, including the necessity of allowing flexibility by working in “what if” mode, and dialog with investigators to help them clarify their ideas.

• ∙ Tools and processes for initiating patient contact, including: obtaining consent from patients for direct researcher contact in the future (including mention of a regional registry built by having admission clerks ask about willingness to be contacted); an automated algorithm to identify prospective participants de-identified to researchers, allowing a message to go directly to patients without releasing PHI to the investigator; the use of new technologies such as ResearchKit to aid in recruitment; development of templates and guidance for how to initiate contact; collection of data on satisfaction among patients who receive research invitations.

  Respondents also offered insights on general processes and approaches, such as:

• ∙ The use of dedicated teams and services (e.g., offering a suite of recruitment services with a mix of technology and human-supported options), and the importance of having “people available who have a good understanding of the tools, clinical contexts, etc. to help users develop reproducible and trustworthy best practices for obtaining and using data for research”.

• ∙ The importance of creating a business model that justifies institutional support.

• ∙ Other ideas, such as leveraging PCORnet experiences and recognition that communication, trust, effective process, and data sharing agreements are the key to working with community partners in collaborative clinical and translational research.

Recruitment Practices

Once a researcher receives a list of potential research participants, respondents reported a wide range of recruitment practices allowed at their institution (Table 2). Practices involving an intermediary were allowed at over half of institutions, including approaching potential participants in clinic who were previously identified (55%) and contact only after introduction by the care provider or clinic (53%). Direct approaches were less commonly allowed; for example, a letter sent from the researcher explaining how s/he got the potential participant’s name (34%).

Table 2 Distribution of responses to the question “Once the researcher receives the list of participants, what recruitment practices are allowed, with IRB approval?”Footnote *

EHR, electronic health records; IRB, Institutional Review Board; PCP, primary care provider.

* Respondents were instructed to identify all the practices that were allowed at their institution.

Twenty-three percent of respondents indicated “other” recruitment practices, many of whom offered details on provider involvement, such as making initial contact to introduce the study or obtain consent for researcher contact, providing a signed letter for researchers to use to make initial contact, or providing approval to make contact. Many also noted that their recruitment practices were study specific, for example, depending on whether the study topic is considered sensitive or specific IRB approvals. A few described recruitment processes carried out by others, such as recruitment specialists.

Variations on the Hypothetical Scenario

When asked whether workflow processes would substantially differ if the research involved a rare disease, cancer, or pediatrics, only a minority of participants said yes. One respondent indicated that, for a rare disease, they would extend the search to include patients from other medical centers within their larger academic system. For a cancer study, a few (8%) respondents identified additional review and approval processes (e.g., cancer center specific processes) and the use of cancer registries for recruitment. About one-fifth (19%) of respondents noted substantial differences for pediatric studies, based on separation between health care entities focused on children Versus adults, initial contact with surrogates (e.g., parents, guardians), and separate IRBs.