The Pre-Pandemic Research Informatics Environment at CTSAs

The CTSA consortium has been working together for years to build an informatics infrastructure to support collaborative clinical and translational research. By way of the survey, we reviewed the “CTS environment” in the academic medical centers before the pandemic and the extent to which they were prepared for the challenges as the pandemic unfolded. In terms of informatics, we highlight seven areas:

• Governance: We observed varying levels of maturity of the formal processes for (i) allocation of scarce resources and (ii) data sharing. Generally, these processes were not designed to and did not work rapidly and in times of peak demand.

• Tools: The CTSAs have built shared informatics tools with varying degrees of adoption. The positive example of REDCap [Reference Harris, Taylor, Thielke, Payne, Gonzalez and Conde3] shows the value of widespread adoption of a general purpose, versatile database design, and management tool [Reference Harris, Taylor and Minor4].

• CDMs: CDMs are needed to centralize or federate multiple sources of data. For EHR data, many have emerged over the last 15 years – and five now have a prominent place [links embedded] – Observational Medical Outcomes Partnership (OMOP), FDA Sentinel, Patient-Centered Outcomes Research Network (PCORNet), Informatics for Integrating Biology and the Bedside and Accrual to Clinical Trials network (i2b2/ACT) [Reference Visweswaran, Becich and D’Itri5], and TriNetX [Reference Topaloglu and Palchuk6]. While CDMs enable multisite, data-driven research, the implementation and maintenance of even one CDM consume human and other resources. Our survey results show that CTSAs are supporting an average of three CDMs.

• Data terminology standards: Heavy reliance on standards is common from ICD-10 to LOINC, SNOMED, HL7, etc. [Reference Schulz, Stegwee, Chronaki, Kubben, Dumontier and Dekker7,Reference Awaysheh, Wilcke, Elvinger, Rees, Fan and Zimmerman8] However, any standard terminology is needed for a new disease, whether for CDMs or phenotyping takes time to be formalized, disseminated, adopted, and implemented.

• Training: Informatics training has been a focus for CTSAs from the start. To the extent that a CTSA has reached a certain level of informatics self-reliance among its users, the ability of the informatics core team to support the surge in requests is made easier, thereby contributing significantly to institutional agility.

• Participant protections: All CTSAs are migrating to the use of a central IRB for multisite studies. Local IRBs may lack experience in reviewing and approving remote consenting, remote research visits, and other applications of telehealth and mHealth technologies whose use in the pandemic has grown.

• Research use of the EHR: There is significant variation across CTSAs with respect to EHR integration to serve research needs. Experience with cohort identification, composing and validating computable phenotypes, and populating patient registries is valuable [Reference Reeves, Hollandsworth and Torriani9] User-friendly tools to evaluate cohorts unburden informatics teams and encourage investigator-level familiarity with and exploration of EHR data.

Informatics During COVID-19: Challenges and Solutions

Despite this solid background of accomplishments, the pandemic precipitously and profoundly affected the conduct of clinical research across the CTSAs and the world. By requiring physical distance between clinicians and patients, investigators, and research participants, research was forced overnight into the virtual space, urgently requiring innovation and expanded support from informatics teams within each institution and across the CTSA consortium. One major theme that we detect running across many of these actions is the speed with which, under pandemic pressure, decisions were made. At our institutions, risk tolerance, particularly with respect to telemedicine, was recalibrated. Decision-making timelines were shortened.

Public health directives to minimize in-person interactions to reduce the risk of COVID-19 transmission created a surge in demand for innovations to support remote interactions with patients. With rapid coverage changes made by CMS and insurance providers that allowed for reimbursement of telehealth visits, many concerns about reimbursement rates were minimized. The need for social distancing and for protection of patient and healthcare worker safety required the rapid uptake of telehealth services at a new scale for both urgent and routine care visits as well as for research visits. This was also essential to support ongoing, non-COVID-19-related studies. Remote clinical and research visits, and the electronic conduct and capture of informed consent (eConsent) moved from being a futuristic option weighted by privacy concerns to a state of urgent and high demand followed by immediate adoption and scale-up. Importantly, many pre-COVID-19 barriers to the implementation of these activities rapidly fell away as institutional leadership, administrators, regulators, and researchers fell into unprecedented alignment of purpose. Technical projects that previously were predicted to require many months or longer were completed in weeks. Operational challenges of access to technology, personnel, bandwidth, and training were matched by informatics challenges to integrate platforms (Clinical Trials Management Systems (CTMSs), EHR, video teleconferencing), create biorepositories, conduct cohort identification, etc.

Universally, we observed that investigators requested platforms and technical solutions for not only virtual research visits, but remote conduct of informed consent and for electronic capture of the interaction. There are limited available platforms for eConsent, some of which are expensive to license or complex to configure. We observed that the majority of CTSA organizations use DocuSign (https://www.docusign.com/), REDCap Cloud, (https://redcapcloud.com/), and REDCap for nonprofit organizations (https://projectredcap.org/). The FDA released its own eConsent platform, (https://www.fda.gov/drugs/science-and-research-drugs/covid-mystudies-application-app) which is limited to FDA-regulated COVID-19 studies, generating both inconvenience and complexity for informatics teams supporting the operation and integration of multiple platforms for eConsenting. None of the available eConsent platforms was interoperable with common CTMSs or EHR platforms. As there was limited use of eConsent pre-pandemic, there is a limited evidence base regarding the potential impact of the shift to remote consent platforms on human research protections, equitable access to research, or the potential impact on the demographics of cohort identification, enrollment, and representative research [Reference Doerr, Maguire Truong, Bot, Wilbanks, Suver and Mangravite10]. As eConsent is conducted through individual secure login accounts, there may be insufficient tracking of virtual versus in-person consent and data capture to support robust comparisons. Vanderbilt’s eConsent (REDCap eConsent framework) is the dominant source for a low-cost eConsent platform within the consortium [Reference Lawrence, Dunkel and McEver11]. Challenges and innovations are summarized in Table 1.

Table 1.

Challenges, response, and innovations in informatics during COVID-19

The quality, depth, and breadth of data collected by healthcare organizations have provided increasing opportunities to advance knowledge and population health, especially during the COVID-19 pandemic. To be able to make sense of the information contained in EHRs and data warehouses, the scale is an absolute necessity. That scale makes it possible to ask and answer important questions, provide evidence for policy changes in the standard of care, care delivery, coverage structure, and emerging population health trends. Further, the required scale is, in the USA, rarely achieved by any single health system. Thus, scale requires that data must be aggregated, and this requires consistent use of standard definitions and terminology.

The National COVID-19 Cohort Collaborative (N3C)

Prior to the COVID-19 pandemic, the proliferation of CDMs for EHR data across the CTSA consortium was already changing the landscape of multisite, data-driven clinical research. Benefits of CDMs include the ability to share code for phenotyping, analysis, and predictive analytics; consistently defined, EHR vendor- and institution-agnostic clinical variables; and the availability of well-documented data quality assurance scripts.

Despite the advantages of CDM infrastructure, the five major CDMs (OHDSI/OMOP, i2b2/ACT, Sentinel, PCORnet, TriNetX) do not readily interoperate with each other. If institutions that wish to share data do not support the same CDM – perhaps one uses i2b2/ACT, while another uses OMOP – there are very few options available to bridge that semantic and structural gap. Moreover, due to the fact that mapping between CDMs is resource- and personnel-intensive, there has historically been little institutional will to launch full-scale efforts to do so. Thus, when NCATS devised a plan for a centralized, nationwide EHR data repository for COVID-19 research, it was necessary to overcome this long-standing barrier. When designing its repository, N3C ultimately chose to harmonize four CDMs to OMOP as shown in Fig. 1.

Fig. 1.

N3C sites can submit EHR data for their COVID-19 population in any one of the four data models. Once transmitted to NCATS, a transformation pipeline maps fields and value sets from the source data models to the OMOP data model. In the near future, privacy-preserving hashing methods will allow for some deduplication of patients as part of the pipeline. Harmonized data in the OMOP model are made available to researchers in a secure analytics enclave. N3C, National COVID-19 Cohort Collaborative; OMOP, Observational Medical Outcomes Partnership; i2b2, Informatics for Integrating Biology and the Bedside; ACT, Accrual to Clinical Trials network; TriNext, company named TriNext.

N3C enables broad access to harmonized EHR data to support community-driven, reproducible, and transparent analyses of COVID-19 data (Fig. 1). Key to N3C’s architecture and success is the capability to map and transform between CDMs. Clear prioritization from NIH/NCATS and the weight of the pandemic both contributed to institutions’ willingness to expend resources on mapping data models, working out governance and regulatory challenges, and extracting data for contribution. At the time of this writing, the N3C repository contains EHR data from 41 sites and over 3 million patients, representing large-scale harmonization of 4 CDMs into a single dataset. After regulatory and governance approval, investigators can access and analyze these data in a secure enclave for COVID-19-related research questions. As of this writing, 164 institutions have signed the Data Use Agreement (DUA) that allows their researchers to access the enclave for approved projects [Reference Haendel, Chute and Gersing12].

CTSA Informatics Survey

Necessary Innovation, Lessons Learned, and Recommendations for Future Study

The USA and our academic medical centers were inadequately prepared for the COVID-19 pandemic, despite >$35B spent on electronic medical records, decades of warnings of the threat of pandemics, and significant public investments in emergency preparedness for chemical, biological, radiological, and cyberattacks. In numerous cities and regions, the healthcare delivery system was or is severely strained. The current crisis is far from over: as this is being written, we are in the third wave in COVID-19 cases in the USA and elsewhere. In December, the first two highly effective vaccines received FDA Emergency Use Authorization and their administration began to those at highest risk. By mid-February, more than 40 million doses have been administered in the USA. Adequate suppression of the pandemic to return to some normalcy, given the widespread use of effective vaccination coupled with public health measures, is thought to be still half a year or more away. The pandemic has been a stress test to the system, and it’s useful to consider where strengths and weaknesses were identified so that our institutions are better prepared to respond to the next challenges of the current pandemic and to manage better when the next one strikes. In our increasingly digital world, the centrality of data to health has never been more clear or important. Our data performance has shown numerous gaps in data, data quality, and data flow.

Pre-pandemic, the informatics organizations of the CTSA network were focused on conducting and supporting diverse areas of biomedical research. As shown above, most groups developed processes and tools to handle hundreds of research projects. During this time, almost everyone had to support multiple standards for health and biomedical data and had to develop distinct ways of sharing those data with each of the respective networks to which they belonged. Furthermore, many of the teams were participating in and/or supporting projects with tools and analytics that were designed for interventional clinical and community outreach projects. In March2020, nearly all research studies halted and nearly all research and operations shifted to support COVID-19 patients and COVID-19-related projects in real time. Due to the nature of the pandemic, the speed of change is unprecedented at every level.

We can start illustrating this in the area of governance. Typically, DUAs and Memoranda of Understanding (MOUs) require months of back and forth refinement. Now, driven by clinical needs, these have been arranged in a matter of weeks. What exactly was the value of all those rounds of back and forth? Confronted by a life-threatening pandemic, some of those concerns gladly receded into the background. Datasets are now being shared with multiple stakeholders within and outside our institutions while still complying with regulations and protecting patient privacy. How can we learn from this and extend the agility of data sharing under COVID-19 to be a routine feature of our post-pandemic life? Privacy protections are and will remain important, but we recommend including some extended protections to data custodians under-declared public health emergencies.

Rapid data-sharing agreements are clearly a feature of operating in this pandemic. But, such flexibility only increased the speed with which organizations needed to consider issues of both technical and semantic interoperability, as well as data quality. New datasets from a number of different sources, such as the state-level immunization registries, Health Information Exchanges (HIEs), claims databases such as CMS (Medicare, Medicaid) and all payers’ databases need to be linked, evaluated, and studied. There has been a significant challenge with the collection of the data, and modeling and analysis of the data attributed to different individuals collected under difficult situations. The pressure of the pandemic does not allow much in-depth communication to understand the overall objectives.

COVID-19 testing databases and COVID-19 immunization registries capture numerators while EHRs integrated across multiple healthcare systems, e.g., through HIEs, capture one key denominator that reflects persons in care. The true population denominator, necessary for accurate population-based estimates, might be census-based, or some combination of social security, passport/visa, and immigration registration-based data sources, or an eventually, a national patient ID.

Since many patients have or will receive COVID-19 testing data and COVID-19 vaccination from providers that are not their usual sources of care, ensuring that the test and immunization data are transferred into consolidated medical records is important for calculating overall population and population subgroup rates. This requires the de-duplication and integration of these multiple data streams at the individual patient level, as well as careful capture not only of the event data, but also the metadata (type of test, vaccine manufacturer, lot number, etc.). Data integration efforts that combine such numerator and denominator information are critical for several purposes, including:

1. 1) Public health – establishing vaccine coverage rates with the granularity (e.g., demographic and geographic data) to examine subgroup reach and to target underserved/underrepresented groups (especially those individuals and groups who may be excluded systematically or who avoid healthcare settings).

2. 2) Pharmacovigilance – to capture vaccine adverse events, some potentially rare, subtle, and unrecognized and to have demographic and clinical information to examine the heterogeneity of treatment effects (HTE) in demographic and clinical subgroups.

3. 3) Clinical – to evaluate up-to-date status for clinical preventive services so clinicians at the point of care can help patients become and stay up to date for indicated preventive services (e.g., get tested, receive second dose, and probable booster doses, etc.) and to capture longitudinal information on infection incidence, severity, and outcome, especially to characterize, phenotype, and treat long-haul COVID-19.

New natural history and clinical outcomes studies had to be set up in record time, while data and data standards were often changing as IRB protocols were being written. An example is laboratory data. New COVID-19 tests are being approved (or authorized under a EUA) frequently, producing new definitions, new LOINC codes, SNOMED terms, values, etc. Studies to evaluate those tests, in terms of sensitivity, specificity, etc., are still in the early stages. The US Department of Health and Human Services offered a solution. SHIELD (https://aspe.hhs.gov/shield-standardization-lab-data-enhance-patient-centered-outcomes-research-and-value-based-care) offers a standard for reporting lab data but many organizations are still not aware of this standard, let alone implementing it. The N3C effort, by tackling the CDM interoperability challenge as described above, has helped substantially. Now, there is better coordination and many are working toward common goals. This is an effort we expect to continue through the pandemic and beyond since it is an essential part of health informatics preparedness.

Faced with an unknown new disease with no preexisting coding, we relied on standard-setting bodies (for ICD10, LOINC, SNOMED, etc.) to rapidly provide new codes and terms. However, the variables that investigators want to work with are often not the same variables that are coded and standardized – e.g., visit types, ventilator settings, ICU admissions and length of stay, oxygen saturation, and symptoms. This speaks to the need for more agile standardization for certain use cases – a “good enough” approach that may lack in the robustness of HL7 or SNOMED, but does not rely on lengthy deliberation and approval processes. Some of the CDM communities took steps in this direction in response to COVID-19; PCORnet quickly developed standard ways of representing certain COVID-19-specific variables, such as ventilator and ICU flags, and the ACT ontology devised its own system of flagging positive COVID-19 lab tests (see https://github.com/shyamvis/ACT-COVID-Ontology/blob/master/ontology/README.md).

Informatics groups are in the middle of all of these rapid changes. Multiple demands and resource constraints are very real. Most groups had to innovate and adapt fast. As indicated in our survey, many groups developed dashboards, COVID-19 data marts, standardized extracts, and deployed eConsent digital tools for clinical trials. They also developed new data models and participated in research to evaluate the process and clinical outcomes of telehealth encounters as compared to face-to-face visits. Such dramatic changes to the delivery of health care are only now being rigorously evaluated. The effectiveness of these changes and innovations is yet to be ascertained, and the research has already started. Informatics groups have to be ready to respond to this challenge, including developing new, agreed-upon metrics for the process of care that can be captured passively or with minimal interference with care.

As mentioned earlier, investigators needed to conduct the informed consent process remotely and to capture the interaction electronically. A majority of institutions expanded or developed new eConsent infrastructure and most will be keeping it. Management of eConsent for FDA-regulated studies frequently requires the support of a separate platform adding complexity for both users and informatics teams. Due to the lack of interoperability with common CTMS or EHR platforms, continued innovation and harmonization will be needed in this area. We also want to emphasize that the potential positive and negative impacts of remote consent procedures on human subject protections, engagement of hard-to-reach populations, and the role of the digital divide inequitable and representative research will be important topics for near-term research inquiry and debate.

Finally, we are still in the middle of the pandemic and many additional challenges are yet to come. Some are just around the corner and will require massive informatics efforts that go beyond our individual institutions. The evolving N3C resource should prove to be a valuable asset to a diverse set of investigators. But, the more “typical” research project might now involve state and national immunization registries, real-time immunization information, record linkage, data fusion, data curation, and machine-to-machine understanding of adverse effects. We, therefore, anticipate new challenges in the next phase of battling the pandemic. This leads us to make some recommendations.

1. 1) There is a fundamental and perhaps somewhat underappreciated difference between clinical informatics and public health informatics. It seems fair to say that the CTSAs, by and large, have been evolving in the ecosystem of clinical and translational research, and not public health research. As a result, many – perhaps most – were and are unprepared to quickly integrate with preexisting public health data infrastructure and respond to public health research needs. Further, as we have explored our local or state public health data infrastructure, we are concerned that significant investment is needed to raise the level of maturity and agility of many public health information systems. These gaps need immediate attention as many COVID-19 vaccines will become available over the next year or two. Collaborating with existing organizations, many of them at the state level will be both mutually beneficial and necessary. Focusing on the data sharing and DUAs with state immunization registries and clinical laboratory testing data systems, perhaps at the national level with the CDC to avoid 64-fold repetition, can help to implement a safe, equitable, and effective mass vaccination campaign.

2. 2) We see real value in examining “data preparedness” and “data logistics” in declaring public health emergencies [Reference Yaraghi13]. Imagine if there were an equivalent to the Federal Emergency Management Agency (FEMA) for health data that could pre-position assets at the earliest phase of a pandemic.

3. 3) We urge the NIH to support research to refine culturally appropriate best practices and privacy guidelines as healthcare research moves to a more virtual/distanced world. From eConsent to home device monitoring, both care and research will be more of a noncontact affair in the future. This shift is worthy of study in and of itself so that we have the evidence to refine our distanced approaches to both care and research in a patient-centric way.

4. 4) This rethinking of “place” also has interesting implications for talent development and staffing. As informatics skills, particularly in areas like artificial intelligence (AI) and machine learning, are in such high demand, academic institutions feel at a disadvantage. HR policies should be reevaluated to seek ways to broaden applicant pools by supporting WfA – Work from Anywhere – telecommuting and remote work policies, while maintaining the deep collaboration required by academic work.

5. 5) The COVID-19 pandemic has put all the CTSAs on high alert status, working as fast as possible to be of service. It is fair to say that “pandemic preparedness” is not a standard part of CTSA evaluation. We encourage NCATS to begin a series of discussions with CTSA leadership on this topic. A workshop in 2022 to layout the plans for the next pandemic would help the CTSAs consolidate and share lessons from this one and highlight to NIH leadership potential areas for further infrastructure investment. It will be important to engage other DHHS entities, such as CDC, FDA, ONC, CMS, and HRSA in these discussions. “Dual-use” informatics can help both clinical research and serve public health needs.

6. 6) In the spirit of never wasting a good crisis, it is worth reviewing those processes that previously required months and hours of meetings. What is the real value of those steps? Let’s not slip back?

7. 7) Regarding disparities and equity, That COVID-19 does not impact all communities equally in the USA was a fairly early observation and soon became a substantiated finding. It’s time to ask ourselves – “How did we do?” – and be prepared to spend the time to answer it rigorously and then commit resources, including in informatics, to do better for all phases of pandemic response.