Background and Significance
Since the passage of the Health Information Technology for Economic and Clinical Health
(HITECH) Act in 2009, the increased availability of routinely collected electronic
health data created the opportunity to conduct research more efficiently across diverse
populations. For example, researchers have increasingly shifted toward the design
of pragmatic clinical trials (PCTs), which are comparative effectiveness trials conducted
in a real-world setting and embedded within routine clinical care.[1]
[2] PCTs are designed to improve the applicability and generalizability of trial findings
to clinical practice with broad selection criteria, utilization of routine clinical
settings and personnel, few follow-up visits, and simple design.[1]
[3] The use of “real-world evidence” or information obtained outside of traditional
clinical research settings is also encouraged in the field of drug development and
within the 21st Century Cures Act.[4]
[5] Real-world evidence includes data from electronic health records (EHRs), claims
and billing data, and other data obtained through sensors and health applications.
Similarly, in the field of observational epidemiology, The Precision Medicine Initiative
(PMI) All of Us Research Program plans to use EHR data as part of follow-up of its 1,000,000 person–cohort
study.[6]
A challenge to both traditional and pragmatic research studies is the recruitment
of underrepresented populations. Many observational cohort studies and clinical trials
have traditionally been conducted within academic medical centers, hospital systems,
or other institutions focused on clinical care because clinical institutions typically
offer increased access to patients, clinical data, and research teams. This process
ignores patients who fall outside of clinical care and may benefit from participation
in research. Minority patients are generally underrepresented in clinical research
because of inadequate outreach, lack of awareness, and/or mistrust of the system.[7]
[8]
[9] This is problematic not only for recruitment, but also because much of the success
of medical research is dependent on positive public perception of the value of research.[10] Active community engagement in clinical research may help to increase enrollment
of study participants, including those of underrepresented groups, reduce the time
of the research project itself, and improve dissemination and adoption of research
findings among communities.[10] Additionally, community engagement and recruitment in research allows community
members to understand their own health issues, while informing researchers and policy
makers of the community priorities and the need for cultural sensitivity in research.[11]
[12]
Methods
Conventional clinical research studies often occur within the confines of health care
settings, as outlined in [Fig. 1]. Recruitment occurs through the identification of patients at their clinical encounters
or prior to visits via EHR data. Following enrollment, data are collected from the
patient directly and/or through historical EHR data. In this conventional workflow,
patients are either surveyed directly for follow-up or queries are sent to the EHR
to gather longitudinal patient data, a protocol incorporated into pragmatic research
studies.[13]
[14]
Fig. 1 Conventional stages of clinical research studies versus the new model for community-based
recruitment.
Our new model moves outside of the clinical setting and into the community to identify
participants for recruitment. Preliminary data are collected from participants at
the time of outreach with full consent for their participation and access to clinical
records. Utilizing identifiers from the collected data, participants are linked to
a record at a clinical site for eligibility verification and continued data collection
through EHR queries. Follow-up may occur directly from the patients, but through the
linkage to the clinical records, researchers may also query EHR data for longitudinal
information. This model facilitates the identification and recruitment of typically
underrepresented participants.
Linkage
Linkage of the community participant back to a record at the clinical site is an integral
component of the new recruitment model. During recruitment, participants are asked
a set of screening questions to ensure eligibility for the clinical trial. The purpose
of the linkage is twofold: (1) to verify the eligibility based on clinical evidence;
and (2) to enable long-term, passive follow-up of the participant via EHR data. This
implementation utilizes a privacy-preserving EHR linkage tool with hashing and matching
components, as described by Kho et al.[15] The hashing application deidentifies the user information, which is shared across
sites and linked for verification.
Workflow
[Fig. 2] presents an overview of the stages in the community-based recruitment workflow.
Fig. 2 Overall community-based recruitment workflow and follow-up.
Potential study participants are identified by community organizations via community
events or other outreach methods. Participants are asked a series of initial eligibility
questions and the collected data are encrypted and deidentified using the hashing
algorithm. If patients do not meet the initial eligibility, recruitment activities
stop and the participant is thanked for his/her time. If initially eligible, this
output is sent to the data hub/honest broker to conduct the linkage to clinical data
(eligibility criteria/variables of interest are included in the data) using the deidentified
hashing output from the participating clinical sites. If a match exists, the participant's
eligibility for the study is verified and the data hub/honest broker alerts the community
organization and the organization informs the community participant about successful
enrollment. Based on the study design, participants are then surveyed for primary
data collection directly via telephone, e-mail, or through an online portal. The participant
may be contacted directly for subsequent follow-up, or queries may be sent to the
clinical site by the clinical study hub to pull participant health information from
the EHR.
Below, we discuss two case studies utilizing the community recruitment method in the
following sections.
Case Study 1
Background for Aspirin Dosing: A Patient-centric Trial Assessing Benefits and Long-Term
Effectiveness (ADAPTABLE) Case Study
A. PCORI/PCORnet
In 2010, the Patient-Centered Outcomes Research Institute (PCORI) was established
as a nonprofit organization to help patients, providers, payers, and policy makers
make informed health care decisions and improve the quality of care by producing evidence-based
research supported by all participants in the health care community.[16] With the goal of advancing use of electronic health data in comparative effectiveness
research (CER), PCORI developed PCORnet: The National Patient-Centered Network.[17]
[18] The central goal of PCORnet is to maintain a “network of networks” of various health
care institutions and stakeholder groups and build partnerships for the purpose of
collecting and using data for improved CER.[17] PCORnet is a distributed research network conducting research across Clinical Data
Research Networks (CDRNs) and People-Powered Research Networks (PPRNs).[17] Each site follows the PCORnet Common Data Model (CDM), a standardized data structure
and format utilized by all CDRNs and PPRNs.[19]
B. CAPriCORN
The Chicago Area Patient-Centered Outcomes Research Network (CAPriCORN) is one of
the 13 CDRNs within the national PCORnet infrastructure. CAPriCORN is a collaboration
among 11 Chicago health care institutions, including private, county, and state hospitals
and health systems, Federally Qualified Health Centers (FQHCs) and two Department
of Veterans Affairs Hospitals, managed by one neutral nonprofit administrative entity
with a subcontract to one nonprofit public health agency to serve as the central hub.[20]
CAPriCORN is a distributed data network; each participating institution maintains
a relational data warehouse, including administrative and clinical data across inpatient
and outpatient settings.[20] Each institution maintains their own PCORnet CDM and the CAPriCORN CDM. The CAPriCORN
CDM includes all components of the PCORnet CDM and other additional data instruments
and elements. An independent nonprofit, the Medical Research Analytics and Informatics
Alliance (MRAIA) serves as CAPriCORN's data hub and honest broker.[21]
C. ADAPTABLE Background
ADAPTABLE is a demonstration project conducted through PCORnet.[14] This PCT is designed to compare the effectiveness of two aspirin doses, low (81
mg/day) versus high (325 mg/day), in preventing myocardial infarction and stroke among
20,000 individuals with coronary heart disease.[14] The project not only seeks to answer this clinical question, but to test the capabilities
of and refine the methods for conducting PCTs through PCORnet.[14] ADAPTABLE presents a new model for clinical trials, aiming to minimize the burden
of research activities on patients, clinicians, and institutions, while incorporating
patient-reported outcomes (PROs).[22]
The trial assesses patient eligibility through a computable phenotype run against
the EHR data mapped to the PCORnet CDM. Computable phenotypes are representations
of clinical conditions developed by querying the EHR using a standard set of data
elements or expressions.[23] Each site has developed its own methods of patient recruitment, which include direct
mail, phone calls, electronic messaging, and in-clinic recruitment. If patients are
willing to participate, they are consented and randomized via an online patient portal
or via traditional methods for patients without access to the Internet. To access
the portal, patients are provided with an access code or “Golden Ticket.” Subsequent
follow-up occurs from the patient directly through an online portal or phone surveys,
through queries of the PCORnet CDM for the patient's health information, and through
claims data if available.[24]
D. The CAPriCORN Approach to Community-Based Recruitment
A major aim of PCORI and PCORnet is to involve community organizations and patients
in research. On the recommendations of CAPriCORN's Patient-Clinician Advisory Council,
we designed and developed a novel recruitment strategy aimed at identifying underserved
individuals in nonclinical settings who may be eligible to participate in ADAPTABLE.
The strategy involves two community institutions, the Sinai Urban Health Institute
(SUHI) and PASTORS4PCOR (P4P), in Chicago.[25]
[26]
To be eligible for the ADAPTABLE trial, participants must have received care at an
institution within PCORnet that is participating in the trial. Potential participants
identified by the community site must be linked to a participating CAPriCORN institution
for future follow-up queries by the ADAPTABLE coordinating center. The linkage not
only allows for assessment of the institution eligibility criteria, but also eligibility
validation for patient-reported answers to the Health Screening Questionnaire.
There are five CAPriCORN sites participating in ADAPTABLE: Northwestern University,
University of Chicago, Rush University Medical Center, NorthShore University HealthSystem,
and the Cook County Health and Hospitals System (CCHHS). Each site maintains its own
pool of eligible patients based off of the computable phenotype. At four sites, research
assistants are responsible for identifying and inviting eligible patients to participate
in the trial. CCHHS provided data for their pool of eligible patients for eligibility
verification and follow-up of patients recruited in the community; active recruitment
at CCHHS is planned pending finalization of data sharing agreements. As with all participating
ADAPTABLE sites nationwide, individuals who are deemed eligible for the study will
receive a unique “Golden Ticket” number to access the secure Web-based patient portal
for consent and randomization, managed by the central ADAPTABLE Team. The tickets
are unique to the CDRN and clinical site within the CDRN where the patient was identified.
Example Workflow
[Fig. 3] presents the workflow for the implementation of the community-based recruitment
strategy within ADAPTABLE.
Fig. 3 ADAPTABLE community-based recruitment workflow.
Potential study participants are primarily identified at community events. Community
health workers from both SUHI and P4P interview potential participants by asking a
series of eligibility questions. The first question asked of the interested participant
is whether he/she receives clinical care at a participating CAPriCORN institution.
Recruitment ends if clinical care is received outside CAPriCORN, as this is a requirement
for eligibility. If care is received within CAPriCORN, the community health worker
proceeds with a verbal consent protocol. Once verbal consent is obtained, the potential
participant is asked a series of questions related to current aspirin dose, if any,
allergy to aspirin, age, cardiovascular health, current medications, and if female,
whether she is currently pregnant or nursing. The responses to the set of questions
in the initial screening questionnaire are tied to eligibility logic. If eligible,
the participant proceeds to the next questionnaire, which has two parts: (1) collection
of provider contact information and (2) personal information from the participant
including first and last name, date of birth, social security number, race/ethnicity,
gender, and contact information (address, phone number, and preferred method of contact).
After this questionnaire is administered, the eligibility interview with the participant
is complete.
Data collected during the interview are recorded in REDCap (Research Electronic Data
Capture), a secure data collection tool.[27] REDCap may be utilized either through a browser, if an Internet connection is available,
or via the REDCap mobile application on a tablet in offline mode with data synced
to the REDCap server once an Internet connection is available.
To link the potential participant with an existing record at a participating CAPriCORN
site, we used the EHR linkage tool described above.[15] The hashing software is conducted locally at the community institutions and requires
a one-time setup to configure the software. The software is installed on a local machine
with an environment accessible to the REDCap server. A script is executed directly
in the command line to abstract the data elements necessary from the REDCap database.
The five inputs needed to create the deidentified Hash ID and 18 hashes are collected
during the initial screening questionnaire.
The deidentified hashing output is sent via Secure File Transfer Protocol (SFTP) to
MRAIA, the CAPriCORN data hub and honest broker, for linkage to a participating CAPriCORN
site. Prior to implementation of the community-based recruitment strategy, the CAPriCORN
sites participating in ADAPTABLE generated their hashing output and sent it to MRAIA.
MRAIA matches the Hash IDs from the community institutions to the Hash IDs collected
from the participating CAPriCORN institutions.
If there is a match to the CAPriCORN Hash ID table, MRAIA assigns a “Golden Ticket”
to the participant record and shares the ticket number with the corresponding community
site. If the community participant matches multiple CAPriCORN institutions, MRAIA
has implemented informatics rules to assign ownership of the community participants,
which may vary by project. Examples of ownership rules may be:
-
– Based on patients' last primary care visit.
-
– Based on disease focus, last specialist visit.
-
– Based on equal distribution of credit; that is, if three health systems are matches
for one participant, and the first patient is assigned to System A, the next participant
will be assigned to System B, and so on.
For ADAPTABLE, MRAIA has opted to implement the third option based on equal distribution
of credit. Although community participants must be linked to a CAPriCORN institution
to participate, the community institution will receive the monetary credit per randomized
participant.
The community institution reidentifies the community participant following the alert
from MRAIA that a match was made. The community health workers proceed by informing
the participant of the successful match and eligibility verification and provide the
Golden Ticket information. After receipt of the Golden Ticket, the community participant
utilizes the ticket to login to the secure ADAPTABLE portal, managed by the Duke Clinical
Research Team and Mytrus, for consent and enrollment. The community institution is
required to follow-up with the community participant until the entire consent and
randomization procedure is completed in the portal. Following this step, all future
follow-up occurs by the ADAPTABLE Coordinating Center. Follow-up surveys are sent
to participants through the secure patient portal and queries are also sent to clinical
sites for longitudinal patient data.
At the time of writing the manuscript, ADAPTABLE work was still underway and final
outcomes were still being realized due to its novel design. We expect to have another
report/paper in the coming year about the implementation of the community recruitment
strategy within the ADAPTABLE case study.
Case Study 2
All of Us Background
The PMI was announced in early 2015 by President Obama and brings together the National
Institutes of Health (NIH), National Cancer Institute, and Food and Drug Administration
to bring precision medicine to health care.[28] A large focus of the PMI is to better utilize precision medicine to improve treatments,
prevention, and care. To expand our knowledge about precision medicine, the PMI developed
the All of Us Research Program, which will help facilitate the achievement of the PMI's long-term
objectives.[6] The goal of the All of Us program is to build a research cohort of one million or more Americans to better
understand the intersection of environmental, lifestyle, and biologic factors and
their impact on health.[6] The size of the All of Us cohort will allow researchers to detect associations between gene, environment, and
lifestyle factors, determine causes of differences in drug responses, discover biological
markers that lead to changes in risk for common diseases, and discover new classifications
and relationships between diseases.[6]
Example Workflow
The All of Us Research Program is another example of how researchers are leveraging a community-based
strategy to recruit typically underrepresented individuals, while also utilizing EHR
data for research.[29] The program utilizes two methods for participant recruitment, a community-based
direct volunteer (DV) and health care institution-based method.[29] In the DV method, any American can volunteer to participate in the cohort study
and consent for participation, reflecting the new community-based recruitment strategy
presented above. A major goal of the research team is to ensure that the All of Us cohort is representative of the U.S. population, including those from various age,
sex, race/ethnicity, socioeconomic, and geographic groups.[29] The institution-based method requires the collaboration of health care provider
organizations (HPOs) to recruit participants, similar to the conventional clinical
research workflow. The HPOs recruit, consent participants, and conduct the required
study activities.
Regardless of the method of recruitment, each participant of the All of Us cohort is required to provide a biologic specimen, behavioral data, and health data
for better understanding of diseases and their mechanism.[29] The eligibility criteria for potential participants are limited and center around
the core requirements above. Those recruited from the community provide health data
either from their EHR data, through direct transfer through the novel Sync for Science
tool, or by undergoing an initial exam from a provider. For those participants identified
at HPOs, it is the responsibility of the HPO to share the EHR data with the All of Us Program. At the time of consent, each participant will provide permission to be recontacted
for follow-up, through direct contact with the participant and/or longitudinal collection
of his/her EHR data.[29]
Data from various sources will be transmitted to the Data and Research Center (DRC)
with full consent from the participant. Data approaches considered involve consented
identified patient data and anonymized patient data from sources external to the EHR.
To aggregate records from different sources for an individual participant, the All of Us working group proposed the utilization of a unique identifier (the PMI ParticipantID)
and the use of record linkage strategies. The record linkage strategies may incorporate
privacy-preserving methods for sensitive information as needed, such as the utilization
of hashing algorithm for the patient identifiers prior to transmission to the DRC,[6] although with full patient consent, direct identity verification by participants
is anticipated to be the prime method for record linkage.
The project details and methods will be finalized during Phase I of the All of Us Research Program. The Research Program has already formed a national team including
Enrollment Sites, a DRC, a Biobank, and other partners to engage patients and communicate
findings.[30]
Discussion
We developed and piloted a new strategy to recruit participants for pragmatic clinical
studies in the community. Engaging patients directly from the community for research
is a step toward more efficient recruitment of underrepresented populations for clinical
research. This strategy demonstrates the potential for new, low-touch, pragmatic research
methods.
Prior studies have discussed the methodology and best practices for clinical research
recruitment directly from the community, but do not incorporate methods for eligibility
verification and automated follow-up with EHR data.[7]
[31]
[32] Pragmatic research studies simplify eligibility criteria and screening to improve
study generalizability in comparison to traditional clinical studies. Our strategy
utilizes direct recruitment of participants outside of the clinical setting and further
addresses the need to simplify screening and follow-up through the linkage of study
participants back to records at clinical sites.
In pragmatic clinical research, the intervention of interest often occurs in a real-world
setting and study visits are incorporated into routine clinical practice. Without
regular interaction through clinical care, follow-up for community-based participants
may be difficult. Automated follow-up via EHR data or direct patient contact is recommended
for good clinical practice.[33] ADAPTABLE and All of Us incorporate follow-up via existing EHR data and direct patient contact (phone or
e-mail). While other research studies have utilized automated follow-up via the EHR,[13] this approach uniquely gathers community participants' clinical information.
The community-based recruitment and data collection protocols could be made more efficient
by combining information collected directly from patients in their community with
clinical data to provide the complete clinical and social picture of a study participant
to researchers. The incorporation of PROs in PCTs is necessary to further increase
applicability and generalizability of research evidence to clinical practice. The
U.S. Food and Drug Administration defines PROs as “any report of the status of a patient's
health condition that comes directly from the patient, without interpretation of the
patient's response by a clinician or anyone else.”[34] The utilization of PROs as outcome measures in comparative effectiveness trials
has been limited thus far. PROs, combined with clinical data, can provide clinicians,
researchers, and policy makers with the complete picture when evaluating interventions
in research studies.
The community-partner strategy is not without limitations. As seen within ADAPTABLE,
in order for the linkage to occur, the honest broker must have all hashes of eligible
patients from the participating clinical sites. Without the hashes, the honest broker
will be unable to match the potential community participant data to an existing record
at the clinical site. Another limitation is that patients have to be seen at one of
the participating institutions or have EHR data there to participate in the study.
Study personnel may have to reach out to many individuals before identifying someone
with EHR data at a participating institution. Leveraging EHR data for multiple patients
and from multiple sites is also complex. Privacy and data ownership issues may arise
and need to be addressed. ADAPTABLE was implemented within an existing research network
with a common data model, making it easier to aggregate the EHR data. For All of Us, this infrastructure does not exist. Participants from the community must provide
their own electronic copies of EHR data. This requires their home health care institutions
to have technologies available to patients like “Blue Button” or Sync for Science
to view, download, and transmit health care data to the Coordinating Center. This
technology may not be available at all sites and may also come from multiple sites
if one participant receives regular care at multiple institutions. Once received at
the Coordinating Center, these data will be in multiple formats, making aggregation
difficult before data curation.
Another limitation is the potential for duplication of the patients approached by
both the clinical site and community site. To minimize this possibility, patients
are asked during recruitment if they have ever been approached about the ADAPTABLE
study by phone, letter, or during a visit with a health care provider. Additionally,
for the ADAPTABLE project, it is difficult for community partners to identify eligible
patients due to the specific eligibility criteria required. This community-based strategy
may be better suited for PCTs or cohort studies, such as All of Us, seeking a very diverse sample of participants with more general eligibility criteria.
