The Use of Electronic Health Records for Clinical Research: Recommendations to Advance the Field

Transcription

1 The Use of Electronic Health Records for Clinical Research: Recommendations to Advance the Field Information Technology Roundtable Clinical Research Forum 4/4/14 Draft

2 Contents Introduction... 1 High-Priority Tasks and Recommendations... 1 A. Identify Eligible Research Participants... 1 B. Enable Data Collection for Case Report Forms and Registries... 4 C. Engage Patients in Research... 5 D. Reuse of Clinical Data for Research... 6 E. Management of Data Access... 8 F. Integrate Data from Multiple Sources... 9 G. Create Governance Models H. Make the Case for the Value of EHRs to Support Clinical Research I. Facilitate Registration of Research Participant Consent and Preferences J. Create Metadata Surrounding Research Data Vendor Perspective AHC Perspective Next Steps ii

3 The Use of Electronic Health Records for Clinical Research: Recommendations to Advance the Field Introduction Electronic health records (EHRs) are a valuable tool for clinical research. EHR data can be used to identify eligible patients for clinical trials, monitor adverse drug reactions, and follow patient outcomes. 1 But because EHRs are designed to meet clinical practice needs, clinical investigators who want to use EHRs for their research must overcome numerous challenges. These challenges include the need to integrate information technology (IT) systems designed to support research with EHRs, the use of different terminologies in research and EHRs, and regulatory and institutional requirements that do not support EHR use for clinical research. 2 To explore these challenges and develop recommendations to address them, the IT Roundtable of the Clinical Research Forum sponsored a conference, EHRs in Support of Clinical Research, on October 14 16, 2013, in Washington, DC. The meeting brought together more than 90 participants with expertise in clinical research, health IT, and bioinformatics representing 38 academic health centers (AHCs). The conference also included 11 representatives of EHR and clinical trial management system (CTMS) vendors and 5 representatives of pharmaceutical companies. The meeting featured small- and large-group facilitated discussions of needs pertaining to the use of EHRs for clinical research. Based on the discussions at the IT Roundtable conference and further consideration of the issues by the IT Roundtable leaders, this document identifies the highest priority needs of clinical research, including multicenter clinical trials, with respect to EHRs and offers recommendations for meeting these needs. These recommendations are broad ranging and address needs for both innovative IT solutions and creative approaches to IT governance. High-Priority Tasks and Recommendations The tasks described in this section are high priority because they are important for advancing clinical research. The proposed solutions should be implementable in EHR systems within the near term (defined in this document as approximately 3 years). A. Identify Eligible Research Participants Task: Identify and match patients based on their characteristics to clinical research studies and efficiently assess their willingness to participate in research. 1 Yamamoto K, Sumi E, Yamazaki T, Asai K, Yamori M, Teramukai S, Bessho K, Yokode M, Fukushima M. A pragmatic method for electronic medical record-based observational studies: developing an electronic medical records retrieval system for clinical research. BMJ Open Oct 31;2(6). 2 Prokosch HU, Ganslandt T. Perspectives for medical informatics. Reusing the electronic medical record for clinical research. Methods Inf Med 2009;48:

4 Background EHRs provide an excellent tool for identifying and recruiting eligible patients to different types of clinical research studies. Institutional review boards (IRBs) have different policies about whether investigators may prescreen patients for their potential eligibility for a study. Some IRBs require patient consent for this purpose unless the study uses de-identified data only. Other IRBs have determined that patient agreement to receive care from the institution constitutes default consent to use their data in limited ways for research. When screening patients who might meet a trial s inclusion criteria, clinical researchers should be able to determine whether the patient has indicated that they do not want to be contacted about participating in any research studies. EHR systems need the capacity to indicate the different types of research activities to which patients give their consent. This information must be collected in a standardized way, so that patients are contacted in consistent ways and are not deluged with inquiries about their preferences regarding research. Existing EHRs lack some important basic information, such as patient ethnicity, that could make the matching process for patients to studies more accurate. New patients may be more likely to agree to joining research studies but for most EHRs they will not have any data to use for matching. Using some process for matching during a visit would be helpful. Some protocols are limited to individuals who are likely to adhere to protocols. Instead of just relying on providers perspective on adherence it is possible to utilize indices of adherence based on missed appointments and medication adherence. Instead of learning about potential research studies from health systems, some patients prefer to control their interactions with AHCs about research through an online portal. They could indicate their willingness to be contacted about studies and which types of studies they are willing to learn about. However, only a minority of patients is likely to use a self-service portal, and matching at the time of medical visits should be the foundation. Recommendations Recommendation A1: Develop a standardized process for requesting patients authorization to be contacted about participating in clinical research. Recommendation A2: Ensure that all patients understand that their entry into a given health care system constitutes consent to allow their de-identified data to be used for observational studies in a way that protects their confidentiality while advancing new medical discovery Recommendation A3: Make EHRs useful for screening patients for clinical studies by enabling them to integrate information from several resources, such as: Biorepositories Laboratory test results Commented [EP1]: Many studies will need to use identified data with HIPAA Waivers, these should not be excluded from this recommendation. 2

5 eprescribing Systems Clinical Notes from the EHR Clinical Trial Management Systems Patient Self-reported Outcomes Recommendation A4: Include the following minimal dataset in EHRs to help researchers identify and screen potential study participants: Basic patient characteristics, such as age, sex, and race Metadata (such as who determined a patient s ethnicity) to assess the data s validity Smoking status and diagnoses with date of onset are also valuable Patient authorizations to be contacted about participating in research and ability to track changes in this authorization As a standard, common data elements used frequently for different clinical research studies on patient characteristics becomes more widespread, study sponsors would be encouraged to match eligibility criteria to this common dataset Recommendation A5: Create systems to store computable consent forms that can be queried to determine whether patients have agreed to be contacted about clinical research participation, might be willing to participate in such research, and have agreed to allow their biospecimens and/or data to be used in research (research using PHI and supervised by IRB). Recommendation A6: Create systems and redesign workflow to efficiently manage communications with patients who have agreed to be contacted about participation in a study. Recommendation A7: Create efficient EHR alerts at time of visit reminding clinicians to ask patients meeting preliminary eligibility criteria about willingness to be contacted by research staff about participating in a study. Recommendation A8: Create a governance structure within the AHC that manages the use of EHR research reminders that is respectful of clinician s time and ability to attend to multiple messages, the patient s primary focus on addressing clinical issues, and the public needs to complete research studies and advance medical knowledge. Recommendation A9: Import CCD documents into CTMSs and Clinical trial data warehouses which per MU2 EHRs are required to be capable of producing this structured HL7 output Recommendation A10: Vendors and Research data warehouses should be capable of including in their design SNOMED CT, LOINC and RxNorm codes While the EHR has the promise to assist with recruitment of patients into studies, it will require collaboration among multiple stakeholders to be effective. In terms of policy, national and local leaders will need to determine what is possible in terms of use of data for prescreening and recruitment based on their interpretation of HIIPA regulations. Working with patients to determine acceptable ways to recruit is also necessary. Leaders of the EHRs will need to work with their IT experts to create algorithms that narrow the net of eligibility as small as possible to maintain the cooperation of clinicians who are receiving multiple alerts to improve quality and 3

6 process of care. False positives need to be minimized for the system to be successful. Researchers will need to work with clinicians so the clinicians efforts to recruit patients to research is appreciated and made as easy as possible. B. Enable Data Collection for Case Report Forms and Registries Task: Minimize duplication of effort in data collection; allow pre-population of case report forms (CRFs) and registries through the use of EHRs; and maximize the use of existing standards, including templates and ontologies relevant to the clinical research problem. Background Most research case report forms cannot be easily populated with data collected through the EHR. The data information for most fields that must be filled in for registries is in an incompatible format in EHRs. For example, EHRs do not have adverse event fields, and finding data on adverse events in EHRs requires looking at several fields. Research nurses usually review EHRs and manually key the appropriate information into the research database. A common minimal dataset of both primitives and derived data that are important for clinical research should be identified (e.g., based on the Continuity of Care Document standards for clinical care). Vendors could create ways to view these data and certify that their systems can collect these data. CRFs and research databases should be prepopulated with data already available in EHRs and other sources to prevent the need to enter the same information more than once and protect research data quality. For example, much of the information needed to complete adverse event reports is in EHRs, and these data should be able to be more easily incorporated into adverse event reporting systems. Research data, if collected to clinical standards, such as quality-of-life measures, might be useful for clinical care and should be entered into EHRs. Issues related to proprietary patient reported outcome measures would have to be addressed. Data collected in EHRs need to be useful and computable for clinical research. Data in text notes, for example, are difficult to use for research. In addition, some derived variables, such as body mass index and cancer stage, should be calculated automatically in EHRs. However, some of the research data require external data, are not appropriate for clinical use, or might not belong in an EHR system. Clinicians are being asked to complete many different forms, some of which have preloaded information. These forms are not well organized or thoughtfully inserted into the workflow. Recommendations Recommendation B1: To minimize duplication of effort, support pre-population of CRFs and clinical research registries with data from EHRs. Recommendation B2: Promote development and use of a set of common data elements (both primitives and derived values) in EHRs that are important for clinical research and can be transferred to CRFs in the following domains: 4

7 Demographics Vital signs Smoking status Medications Laboratory test results Adverse events Recommendation B3: Require vendors to incorporate the minimal dataset into their EHRs. Recommendation B4: Maximize the use of standard EHR templates, ontologies, and vocabularies, including standardized terms for transfer of derived variables from EHRs. Recommendation B5: Create a process for easily exporting data from EHRs to national and international research databases. Recommendation B6: Develop efficient systems to collect patient-reported outcomes data in EHRs that can be used both for research and clinical care. Summary The learning health system will not be achieved without better integration of EHR data for both clinical care and research. Patients and health care providers will be more likely to fully participate in meaningful research when barriers to research are minimized and the probability of the research changing clinical care is maximized. C. Engage Patients in Research Task: Utilize EHR when possible to engage research participants in the entire clinical research process, including identifying relevant research questions, designing studies that can recruit and retain participants successfully, and keeping participants informed of study results. Background The EHR provides new opportunities to engage patients as research participants in clinical research. Most patients prefer to participate in research with their clinical physicians to minimize burden and work with a health care team they trust. Patients should be encouraged to continue care with their usual team during a study along with the research team. Patients are study participants, stakeholders, and owners of their data, and researchers need to recognize patients rights. Researchers must share the results of their studies with participants and allow participants to make decisions about the use of their data. As members of the research team, participants should provide input into study designs, such as evaluating the feasibility of recruiting and maintaining participants. They should also be called on to identify research questions that are relevant to their concerns. 5

8 Advocacy groups, commercial enterprises (such as 23andMe and PatientsLikeMe), and individual patients have created websites with information on clinical research for the public. ResearchMatch is an example of a portal that provides information about studies that are enrolling patients. However most patients and providers are primarily interested in the local research environment and how research is conducted in the health system where they elect to receive care. We understand that communication with patients through electronic channels is most efficient. However alternate channels must be available for patients who do not have the means to communicate electronically. Commented [EP2]: Given the FDA action do you want to keep them on this list? Recommendations Recommendation C1: Enable patient research participants to directly enter their preferences to be contacted about research and use of their EHR data for research without filtering this information through nurses or physicians. Recommendation C2: Provide potential clinical research participants, through online portals or other mechanisms, with: Lists of current clinical studies and eligibility criteria Informed consent forms Access to patient advocacy websites Ability to enter some of their own data, such as patient-reported outcomes Ability to track their own data Advice on what to tell health care providers about their participation in a study Contact information for study personnel Access to research results Recommendation C3: Advocate to make clinicaltrials.gov more patient friendly so EHRs can link with the site and not have to duplicate descriptions of each protocol. Recommendation C4: Educate and support clinical nurses and physicians in how to best engage patients in research and identify studies for which their patients might be eligible. D. Reuse of Clinical Data for Research Task: Make it easy for clinicians to enter high-quality data into EHRs that can be reused in clinical research. Background Ensuring that providers enter the needed data will require incentives, such as physician report cards and ensuring that by entering data correctly, providers will be able to manage their patients, meet meaningful use and regulatory requirements, submit data to national registries, and support their own research. Data in proprietary EHRs are often not easily accessible to researchers because vendors consider their data models to be intellectual property. Security of the data is an important consideration for both the supplier and user of these data. Protected 6

9 health information must be managed carefully and should remain primarily under the control of a data warehouse team. Key Stakeholders In order for data entered into EHRs to be of high quality for research, there must be alignment between the needs of those entering data and those doing research on the data. Such alignment may occur naturally for physician-researchers, but it is common in the course of clinical activity that much data entry is done to meet clinical needs with little regard for later secondary use. Aligning incentives is largely the responsibility of the organization in which data collection occurs. Alignment might occur through incentives, such as providing reports on the data that those doing the data entry find valuable (e.g. PCMH or MU reporting). EHRs should be designed and configured so as to encourage the entry of data in ways that maximize secondary use (e.g. simple, well-designed, structured data entry). Validation methods should be used wherever possible to minimize erroneous entries. Free text should be minimized and critical common data elements important for clinical research should be identified and requested during data collection. Further, EHRs should be as open as possible, with tools to easily extract data from the live clinical systems for further processing for secondary use. Design and configuration of EHRs are the responsibility of vendors and institutional IT staffs. Recommendations Recommendation D1: Define policies to allow data sharing with clinical researchers in proprietary EHRs while managing protected health information and metadata. Recommendation D2: Provide physicians and other clinicians with incentives to enter research data comprehensively and accurately into EHRs. Recommendation D3: Identify common data elements frequently used in clinical research, request vendors to structure these data using nationally standard code sets such as SNOMED CT, LOINC and RxNorm, and educate clinicians to improve the completeness of these data. Recommendation D4: Ensure that research needs are integrated into meaningful use requirements and that these requirements call for interoperability of EHRs containing research data. Recommendation D5: Develop better tools to enrich data in EHRs and improve their reusability, including: Define shared data dictionaries Develop interface with patients and other facilities such as pharmacy to incorporate data generated by personal devices or reported by patients Provide patients with access to their own data. 7

10 Libraries of EHR data collection forms which can be shared and reutilized by clinicians using the same EHR (e.g., Epic Community Library) 3 Recommendation D6: Standardize data quality assessment methods and improve the transparency of EHR data quality. Recommendation D7: Involve clinical researchers in configuring EHR systems. E. Management of Data Access Task: Obtain, manage, and secure EHR data for research. Background Decisions must be made regarding whether to develop a federated or centralized model and how to provide patients with access to their own data. Managing data access will also require interinstitutional agreements. Key Stakeholders In order to streamline the use of clinical data for research, institutions should clarify the regulatory environment under which they hold data (HIPPA vs. IRB) and their policies regarding patient consent. Requests for research use of clinical data should be standardized. Policy makers should address the false dichotomy between clinical and research use of data in learning health care systems by evolving the regulatory environment to encourage the advancement of medical science while providing robust protection of patients. Protection of sensitive information should be of the highest priority, as a breach of confidentiality is harmful to patients, institutions and the research endeavor in general. This must be balanced by the need for reasonable access to data for legitimate purposes. Techniques that should be considered to assist in navigating this balance include audit trails, deidentification, security metadata and the provision of data in a tailored fashion ( data marts ). Adequate security is accomplished by a combination of institutional policies (including educational activities) and thoughtful design of EHRs by vendors. Commented [CW3]: Related references are: * Weiskopf NG, Bakken S, Weng C, Are EHR Data Suitable for Secondary Use? Researcher Views AMIA Joint Summits on Translational Science, in press. Fort D, Weng C, Bakken S, Wilcox AB, Considerations for Using Research Data to Verify Clinical Data Accuracy, 2014 AMIA Joint Summits on Translational Science, in press. * Weiskopf NG, Rusanov A, Weng C, Sick Patients Have More Data: The Non-Random Completeness of Electronic Health Records Proc of AMIA 2013 Fall Symp, , November * Weiskopf NG, Hripcsak G, Sushmita S, Weng C, Defining and measuring completeness for electronic health records for secondary use. J Biomed Inform, 2013, Jun 29. doi:pii: S (13) /j.jbi [Epub ahead of print]. (Best Talk for Day 2 of the 2013 NLM Informatics Training Conference ) (PMID: ) * Weiskopf NG, Weng C, Methods and Dimensions of EHR Data Quality Assessment: Enabling Reuse for Clinical Research, J Am Med Inform Assoc Jan 1;20(1):144-51, PMID: * Botsisa T, Hartvigsena G, Chen F, Weng C, Secondary Use of EHR: Data Quality Issues and Informatics Opportunities, Proc of Clinical Research Informatics Summit 2010, 1-5. PMID: Data access management will require: Tracking memoranda of agreement Data stewardship Tools that provide access to the data for researchers Maps of data within EHRs Maps of local codes to national codes Honest brokers 3 Wright A, Bates DW, Middleton B, Hongsermeier T, Kashyan V, Thomas SM, Sittig DF. Creating and sharing clinical decision support content with Web 2.0: Issues and examples. Journal of Biomedical Informatics 2009; April;42(2):

11 Rules governing how the data may be used Role-based security systems Protocol-specific access control Access to data for study monitors and auditors Viewable catalogs Tracking from data input to publication to ensure reproducible science Business analytic tools Needed security elements: Firewalls with management so to allow for appropriate access by authorized researchers, etc. Rules about transmission of protected health information Active monitoring Strong policies and procedures Role-based, multidimensional security procedures Federated authentication Recommendations Recommendation E1: Standardize consents, IRB approvals and MOUs to enable machinetracking of allowable data access. Recommendation E2: Enable end-to-end audit trails of data access. Recommendation E3: Provide standard de-identification technologies including Honest Broker. Recommendation E4: Create research-specific and general-purpose data marts. Recommendation E5: Harmonize HIPAA and IRB regulations, especially regarding use of PHI for QA/QI and research. Recommendation E6: Enable researchers to establish different levels of security for different data elements in EHRs. F. Integrate Data from Multiple Sources Task: Integrate data from multiple internal and external sources, including EHRs, pathology databases, and registries, for clinical research. Background Integrating data from multiple sources requires: Understanding of the data elements, metadata, and how the data might be used Standardized data aggregation 9

12 Determination of who owns the data Regulatory clarity Approaches for managing exceptions Information on the location and timeframe for each data system, including how far back its data go, when the system went live, and how often it is refined An interface could be created that integrates data from EHRs and research systems, but data from the two systems must be merged carefully. A problem might arise, for example, if information on a research participant s risky sexual behaviors is transferred to the patient s EHR and released to employers or insurers. Some vendors offer a retrieve form for data capture that can prepopulate CSRs with clinical data and store the CSRs in a study database. Vendors could also develop plug-in applications that researchers can use to transfer EHR data into their research system. However, such applications are difficult to develop in a way that protects data quality and safety. Data for observational studies often come primarily from EHRs and could be stored in a clinical enterprise-wide data warehouse. However, clinical trial data do not come primarily from EHRs. Should these data be entered into EHRs to support clinical decision making? Or is this not possible because the data are proprietary and sharing them with EHRs would be too complex? A substantial amount of data needed for certain types of clinical research, such as comparative effectiveness research, comes from sources other than EHRs, such as billing data. Other data sources, such as public health and education data, could also be valuable for clinical research. Vendors of clinical research systems, not EHRs, might need to be approached to integrate data from these systems for clinical research. Other issues to address are the ethics of secondary data uses, sustainability, responsibility for data quality, and unintended customers. In addition, the elements of an integrated data system might depend on the uses of its data. Key Stakeholders Lack of interoperability of EHRs is a major obstacle to the aggregation of data for research. Improving interoperability is the responsibility of vendors and the standards community. Ideally, all EHR data should be standardized and interoperable, but data entry should remain flexible so as to meet the particular needs of the wide range of clinical environments. Ownership/control of data is an area of controversy. It is probable that transparency is the best mechanism for resolving concerns about the use of data. To accomplish this, institutions should set up governance mechanisms to monitor data use and vendors should provide security that can be based on multiple parameters. The current state of low interoperability means that a great deal of source-specific information must be available to maintain ongoing aggregation. Institutions should staff data repositories with this in mind. 10

13 Recommendations Recommendation F1: Provide substantial flexibility in EHR data entry but standardize mechanisms for integrating EHR data with data from other sources. Recommendation F2: Create standards for integrating data from several sources. Recommendation F3: Develop site-based, project-based, and source-specific security mechanisms, including audit trails, encryption, and de-identification, and enable nonemployees (such as representatives of sponsors and regulatory agencies) to monitor these systems. Recommendation F4: Store data from multicenter clinical trials locally so that investigators have the ability to analyze their own data. Recommendation F5: Encourage vendors to support common data models that allow for data integration. Recommendation F6: Identify and train research informatics experts who understand data capture, storage and retrieval in relevant health/research information systems (e.g. EHRs, CRMS, etc.) to support logical and meaningful integration of data from disparate sources for research. G. Create Governance Models Task: Create a centralized governance structure at each institution for overseeing clinical research and bioinformatics. Background Traditional models for IT governances at academic health centers tend to include leaders who oversee the technological (CIOs) and clinical (CMIOs) portions of the enterprise mission. While such leaders typically are concerned about also supporting the needs of researchers; research IT needs and issues are usually not their top priority. More recently, some AHCs have found it useful to establish IT leadership roles specifically focused on advancing the research mission, thereby centralizing and empowering the governance of research informatics/it alongside clinical and operational IT. Such models have many potential benefits. First, they facilitate responsiveness and strategic planning that is focused on the unique and often complex needs of the research community. Second, they can yield economies of scale both by enabling investments in IT infrastructure that can often serve clinical, operational and research mission areas, and they can also reduce the need to respond, often in an ad hoc and non-strategic manner, to the needs of individual researchers. As such, research focused governance models can be important to ensuring that IT investments and infrastructure decisions take the needs and opportunities of research into account and proceed in a strategic fashion for maximal benefit. 11

14 When operationalized, essential elements of research IT governance include: Process for determining and enabling research-specific data access, data stewardship, and data reuse policies Leadership for research IT with clear priorities and authority relative to clinical and operational IT leaders Understanding the specific IT needs of researchers and opportunities for collaboration Research-focused resources, policies and processes are resourced proportionately with clinical care within IT Collaboration among clinical, financial, and administration chief information officers Multiple advisory committees to coordinate efforts Recommendations Recommendation G1: Operationalize research IT governance structures that enable an evidence-generated medicine paradigm alongside one that values evidence-based medicine 4 Recommendation G2 Create best practices for governance structures that bring together relevant stakeholders to enable appropriate and efficient policies for data access, stewardship and security, in order to facilitate rather than hinder data-driven research activities. Recommendation G3: Assign a senior staff member (e.g., research IT director, chief research information officer, etc.) as a counterpart to the chief information officer or chief medical information officer to be responsible for all research IT strategy at the institution. Recommendation G4: Dedicate resources to clinical research IT that can be leveraged along an efficient timeline required by research efforts as compared to some operational or even clinical services. Recommendation G5: Educate those responsible for regulatory requirements at each institution, including IRBs, about the nuances of research governance, informed by the designated leader of research IT. H. Make the Case for the Value of EHRs to Support Clinical Research Task: Make the case for allocating more resources to electronic clinical research resources. Background Clinical research benefits patients, populations, health systems, and the academic community through, for example, quality improvements and associated cost savings and by drawing patients to the institution. In addition, clinical research is key to the learning health system and 4 Embi P, Payne PRO. Evidence Generating Medicine: Redefining the research-practice relationship to complete the evidence cycle. Medical Care 2013; August; 51:S87-S91. 12

15 accountable care. Clinical research IT can help institutions comply with meaningful use requirements, serve as a resource on compliance and policies, and jointly fund infrastructures. Documenting these benefits requires built-in measures of value, such as measures of deferred resource utilization. In addition, steps must be taken to avoid negative effects by, for example, billing correctly for dual clinical and research visits by patients to health centers. Recommendations Recommendation H1: Enable measurement and analytics of relevant clinical outcomes that stakeholders value alongside research-related measures, to demonstrate dual-benefit of such health/research IT infrastructure. Recommendation H2: Develop approaches to measure the value to healthcare organizations of hosting a robust clinical research enterprise as it relates to clinical and operational goals (e.g. improved reputation, referral patterns, etc.). Recommendation H3: Encourage patients, clinicians, and others to become engaged in clinical research by showing the value of this research to them, their families, and the general public. Recommendation H4: Incorporate the concept of enabling a learning health system into communications and marketing efforts and its relationship to improvements in healthcare as well as into the future. Recommendation H5: Offer incentives for front-line staff whose main focus is not on research to participate in the learning health system and make this participation easy. 5 I. Facilitate Registration of Research Participant Consent and Preferences Task: Create structured forms to collect patient preferences regarding being contacted about clinical research and consents (assent for pediatric patients) to participate in research that are computable, modular, and available to relevant stakeholders. Background Confidential disclosure agreements could provide an architecture for the structured preference and consent forms needed for clinical research. Injecting these agreements into the hospital system, where many of these questions are asked anyway, could reduce risks to institutions. Consent forms in EHRs need to be adaptable, with enough variability so that people will use them. However, they must be inflexible within a single system. Securing the information in 5 Embi PJ, Tsevat J. Commentary: the relative research unit: providing incentives for clinician participation in research activities. Academic Medicine: Journal of the Association of American Medical Colleges Feb;87(1):

16 consent forms requires role-based access, with rules regarding who may look at information and tracking of who has looked at information. The authenticity of the information in these forms must be guaranteed. Moreover, all authorization and consent documentation must comply with local and state laws. Joint Commission regulations require health systems to store consent forms for all treatments in medical records. Therefore, consent forms for treatment, diagnostic, and other services required for clinical research must be stored in EHRs. However, the Joint Commission does not require storage of consent forms regarding participating in questionnaire-type studies in EHRs. Recommendations Recommendation I1: Develop structured, computable ways (with limited flexibility to address institutional needs) to document: Patients authorization to be contacted about clinical research and to use their specimens and data for research Blanket consent to screen their records to determine whether they are eligible for studies and use their biospecimens for research Active consent to participate in specific trials, observational studies, or biorepositories Consent to participate in long-term followup activities for studies in which they have participated Changes in permissions and consents over time Duration of consents and authorizations Withdrawals from studies Recommendation I2: Provide verifiable capability in EHRs to collect, store, and track digital signatures to prevent the storage of duplicate authorizations or consents. Recommendation I3: Determine how data can be used for research from patients who have not consented to participate in research and those who have died as well as how anonymized or de-identified data may be used. Recommendation I4: Educate patients by providing information at appropriate literacy levels on the content of the documents they are asked to sign. Recommendation I5: Use EHR systems to manage all documents and materials (including biospecimens) related to patient authorizations and consents. Recommendation I6: Provide different levels of access to patient consents and authorizations based on the user s role. Recommendation I7: Develop standardized consent language that can be used across institutions. 14

17 Interfacing with other data and systems should include linkages to authoritative sources, audit processes, other regulatory adherence, and disambiguation. Existing ontologies should be leveraged. J. Create Metadata Surrounding Research Data Task: Develop and leverage standardized metadata (data describing the elements and attributes of primary clinical/research data) to enhance the functionality of clinical and research information systems for research purposes. Background While metadata standards exist for a variety of health information (e.g. laboratory data 6, radiology data 7, etc.), research data and even many clinical data elements that have the potential to be repurposed as research data are typically not associated with metadata. Examples of metadata include information such as: when data were captured, by whom, in what setting/context, and units of measure (including normal or expected ranges), to name a few. Metadata are critical to enabling the re-use of data collected for clinical purposes and the primary use of data collected for research purposes. Metadata can also enhance researchers ability to understand and analyze data including adjusting for potential biases. On the contrary, the absence of such metadata tagging makes it challenging to leverage data for research purposes with the needed degree of understanding and certainty that rigorous research activities require. While recognized as important and there are some good examples of metadata and standards development efforts for research data 8, attempts to attach metadata to vast amounts of data in EHRs that could be useful for research has met with limited success to date. Recommendations Recommendation J1: Create a community repository of the top 100 (most common) pieces of metadata that clinical researchers need in EHRs, including who entered the data and where, when the data were entered, the context of the data entry (e.g., during a scheduled visit, an acute visit, or in between visits), and whether the patient or a proxy provided the information. Recommendation J2: Create flexible yet standardized ways to collect and store a minimum set of metadata in EHRs. Recommendation J3: Provide researchers with access to metadata in EHRs. Recommendation J4: Create a centralized governance structure to store metadata and determine how metadata can be used when repurposing and using clinical data from EHRs. 6 LOINC article 7 DICOM article 8 BRIDG model, CDISC efforts, etc. 15

18 Recommendation J5: Harmonize efforts to collect metadata and make them accessible to researchers with existing standards, such as those of the Clinical Data Interchange Standards Consortium. Vendor Perspective The following comments and recommendations are based on a panel discussion with several EHR and CTMS vendors during the October 14 16, 2013, IT Roundtable meeting. Vendors are already addressing some of the issues discussed at this meeting, including recruiting participants to clinical studies using EHRs, facilitating communications between clinicians and researchers, and facilitating data capture for research while minimizing the need for duplicate data entry. However, CTMSs are designed to organize and track protocols and EHRs are designed to organize and track patients, so merging the two systems is challenging. Investigators with NIH funding are required to share their data collection forms, and NIH has developed open-source forms. These forms can be used to identify data elements from EHRs that are needed to populate CRFs. The vendor community does not always observe that NIH investigators are interested in easily sharing their data collection and sharing forms. Recommendations from vendors for the clinical research community were to: Conduct pilot tests using EHRs for clinical research and share lessons learned with vendors. Invest time in understanding institutional EHRs and how to maximize their use for research. Recognize that some of the problems with using EHRs for clinical research are due to cultural barriers within the research community and not available technology. Develop o Open standard data forms, including common data elements, so that vendors do not need to request permission from investigators and institutions whenever they want to make a form available o A standard consent form o A core set of EHR data that are relevant to most clinical studies o Standardized expression of recruitment rules that can be imported into a patient portal or EHR system, patient-reported outcomes, and study calendars Advise the research community to communicate with vendors: o What problems they need to address without dictating the solutions o Whether EHR systems should interact directly with patients to manage their informed consent o What tools vendors could develop to ensure that research participants are compliant Recommendations for vendors were: Let clients know about the clinical research capabilities that EHRs already have Provide greater, more flexible access to data that have already been collected 16

19 Create systems that can be customized to meet the needs of individual institutions and changing use cases over time Create online forums for users to share experiences with software Make it easier for EHR users to share their enhancements with one another without requiring the vendor s permission Partner with one another to make their products interoperable, create interfaces between systems from different companies, and fill other gaps AHC Perspective The following recommendations are based on a panel discussion with AHC leaders. Recommendations for the clinical research community were to: Develop use cases that extend beyond cohort identification and data access to identifying adverse events Collect examples of efforts to use EHRs for clinical research at different AHCs Create a Clinical and Translational Science Award (CTSA) health informatics initiative to address the issues raised at this meeting in partnership with the Clinical Research Forum Make research informatics experts available to consult with clinical researchers Involve clinical research informatics experts early in the development of clinical research projects Incorporate the patient perspective in defining health IT needs for clinical research A recommendation for vendors was to: Make key data elements readily available (such as left ventricular ejection fraction) in discrete fields Next Steps Key questions that must be answered before sharing recommendations with vendors are: Should institutions collect and store clinical data needed for research in EHRs and data used purely for research in research systems? How should data from these two systems be integrated for research? Participants suggested the following next steps: Refine the list in this report of major clinical research IT barriers and potential solutions, including determining which recommendations can be implemented in the short term and which ones will require more time Create multidisciplinary working groups (with sponsorship from the Clinical Research Forum and others) that include researchers, vendors, IT experts, and financial officers to identify solutions to clinical research IT issues Form regional working groups to bring together interested parties from within a region 17

20 Identify stakeholders to work with to implement solutions, including AHC leaders, patients, and government agencies Collect examples of experiences (both positive and negative) of using EHRs for clinical research and share them with stakeholders Initiate pilot studies in AHCs to test solutions to clinical research IT problems Organize a meeting for participants in this meeting and/or working groups in conjunction with upcoming CTSA or PCORI meetings Determine whether PCORI could issue grants to implement some of the recommendations Establish an online discussion board or LinkedIn group to share questions and ideas Potential partners to implement the recommendations in this report include: Academy Health American Medical Informatics Association Association of American Medical Colleges CTSA Informatics Key Function Committee Electronic Data Methods Forum Health Data Consortium Effort by NIH and the Society for Behavioral Medicine to evaluate and recommend common data elements for patient-reported measures of health behaviors and psychosocial factors for use in EHRs PCORI, which plans to issue grants for projects using patient-reported outcomes in EHRs Commented [EP4]: Unfortunately the CTSAs have eliminated the Key function committees Recommendations for the Clinical Research Forum were to: Avoid duplicating the activities of the Clinical Data Interchange Standards Consortium, Home Location Register, ASTM, and International Organization for Standardization Avoid creating yet another set of terminology standards Make clinical researchers aware of ongoing standardization efforts 18