The national launch of the National Institutes of Health’s (NIH) All of Us Research Program is well underway, aiming toward the goal of enrolling one million plus participants. This large platform of diverse participants presents an unprecedented opportunity to explore scientific questions that will advance the understanding of precision prevention and management of heart, lung, blood, and sleep (HLBS) conditions. Engaging the National Heart, Lung, and Blood Institute (NHLBI) external and internal community on how best to utilize this historic opportunity to address unanswered questions related to HLBS conditions in ways that cannot be done with NHLBI’s current cohort studies and clinical trials is imperative.
All of Us has developed a scientific framework to guide NIH Institutes and Centers (ICs) and the extramural community in identifying research priorities for the program. It conducted a crowdsourcing initiative (IdeaScale) and hosted a scientific priorities workshop March 21-23, 2018. NHLBI hosted a workshop to follow this initial effort on July 11-12, 2018 in Bethesda, Maryland. This trans-NHLBI workshop, with additional funding support from the NIH Office of Research on Women’s Health (ORWH) extends these initial efforts from All of Us to further elucidate the areas of synergy and alignment with NHLBI research priorities. Workshop participants included experts in HLBS science, with about half having some prior involvement with the All of Us, either as a current grantee and/or participant in the March All of Us Scientific Priorities workshop.
The overarching goal of this workshop was to discuss the potential of All of Us to address scientific priorities in HLBS conditions. Key questions addressed during the workshop were:
- How can All of Us be leveraged to fill research gaps relevant to precision medicine that cannot be addressed with the NHLBI’s existing investments in cohorts, clinical trials, registries, and repositories?
- What could be the NHLBI’s priorities in terms of influencing the future versions of the All of Us protocol?
- What are ways the NHLBI can synergize with All of Us given current NHLBI investments in precision medicine, such as Trans-Omics for Precision Medicine (TOPMed)?
- How can All of Us inform/stimulate future NHLBI studies?
- What are some of the issues related to participant engagement and access to data by participants and researchers that are of particular interest to the NHLBI stakeholder community that may inform how All of Us is leveraged to advance HLBS science?
A detailed description of the content and format of workshop can be found in Appendix I.
The following are the main recommendations that arose from combined group discussions.
Immediately available research opportunities and Influencing future protocol versions
Workshop participants acknowledged that enrollment and the current version (v.1) of the All of Us protocol were very early stage and therefore research opportunities are limited with the current available data. However, many potential protocol elements (PPEs) were discussed in the breakout groups and identified as high priorities to include in the next version of the protocol to enhance utility for HLBS relevant research questions.
- A list of the issues and themes that emerged from the breakout groups, including the many PPEs identified as high priority, can be found on Appendix II.
- Participants recommended that All of Us leverage the content expertise of the NHLBI research community to inform protocol development, particularly participant-provided information (PPI) content by including NHLBI researchers in the working groups and/or an NHLBI expert group consisting of academic and industry researchers specifically developed to advise the study leadership. Participants also requested clearer communication about how researchers can provide input throughout the protocol development and vetting process prior to final adoption.
Investing in All of Us
Because the majority of All of Us Research Program’s resources will be on creating and maintaining the cohort and supporting infrastructure, NHLBI may wish to support the following HLBS-focused priorities [or “needs”]:
- Support development of research core groups to validate the algorithm development of outcomes measures, as captured by electronic health records (EHR), by conducting manual outcome ascertainment and adjudication on a sample of events.
- Identify subgroup(s) of All of Us participants for more in-depth phenotyping with measures such as, but not limited to: repeated biospecimen measures, imaging, spirometry, overnight oximetry, and ecological momentary assessment. Subgroups of particular interest may be younger adults age 20-50 and childbearing/pregnant women who are having increasing events but are not captured by traditional risk assessment.
- Augment recruitment and engagement of specific populations of interest to the NHLBI.
- Augment education and engagement of primary care and specialty providers of HLBS patient populations to enhance recruitment of individuals traditionally under-represented in biomedical research.
- Engage and leverage Small Business Innovation Research (SBIR) programs and the National Center for Advancing Translational Sciences (NCATS) for development, validation, and implementation of HLBS specific mobile applications and sensors.
- Initiate partnership with the Center for Medicaid and Medicare Services (CMS) to ensure All of Us participants insured by CMS programs receive screening imaging for which they are already eligible (i.e., low dose CT imaging for lung cancer screening).
Informing future research
Participants identified a number of ways All of Us can help inform future of HLBS research:
- All of Us has the potential to make clinical trials more efficient by (a) improving case finding through improved subject characterization/phenotyping; (b) pioneering eConsent and integration of mobile technology on a scale that has not previously been done and may be of benefit for future NHLBI studies, particularly pragmatic clinical trials; (c) identifying current clinical treatment practices for various disease conditions to identify the presence or absence of clinical equipoise (for clinical trials planning); (d) collecting large scale passive data of patient outcomes (such as hospitalizations, emergency room or urgent care visits, hospital mortality, and other measures of utilization); and (e) facilitating multi-center studies through the use of the shared All of Us research platform and common data fields and procedures.
- There is the potential for All of Us to improve understanding of not just appropriate treatments, but also appropriate timing of treatments in the “acute” healthcare setting.
- The large scale of All of Us has great potential for helping assess safety and efficacy of device or drug therapies, such as pharmacovigilance or post market surveillance.
- All of Us can inform implementation science by identifying current clinical practices for various disease conditions and thereby identify implementation gaps (NHLBI Strategic Objective #6). Embedding interventions as part of ongoing communication with members of the cohort will allow us to use principles of implementation science and assess the impact of improved dissemination and implementation of evidence-based interventions on health disparities.
- The large number and diversity of healthcare provider organizations (HPOs) in All of Us will contribute to understanding of geographic variability in healthcare delivery and how healthcare delivery influences biological systems, health, and well-being.
- All of Us may inform the future of observational cohorts in HLBS conditions by shedding light on: (a) feasibility and reliability of collecting participant provided information using the various mobile modules; (b) ability to gain HLBS relevant measures from novel wearables, and (c) representation by difficult to access populations.
Engaging the research community and accessing data
- Encourage All of Us to clarify to its investigators and on its public websites how it defines underrepresented in biomedical research (UBR), as there is confusion with how it matches with other entities, such as that defined by the NIMHD Minority Health and Health Disparities Research Framework
- Avoid the incorrect interpretation/assumption by the research community or the public at large that the All of Us cohort can provide “representative” population estimates in the same way as a study like the National Health and Nutrition Examination Survey (NHANES). Consider:
- Making the limitations known on its data-request website and insertion of a disclaimer for use in publications resulting from the All of Us data.
- Creating a ‘rapid reaction team’ to address potential misinterpretation of data.
- Address equitable access to research, particularly in light of differing capacities for research in terms of access to technological and statistical expertise.
- Leverage opportunity to advance data science:
- Enable researchers to develop and share code and utilize their own computational capabilities in the anticipated cloud-based research workspace, in addition to using analytic tools provided by the All of Us workspace.
- Provide synthetic datasets for data scientists to help innovate analytic methods.
- NHLBI should encourage trainees and early stage investigators to participate in All of Us-related HLBS research through new and/or existing mechanisms for training and career development and loan repayment.
- Seek early and adequate engagement of NHLBI stakeholders (which includes patients, healthcare systems, payors, product manufacturers, informaticists, and policymakers) for solicitation of input on prioritization of future research, particularly trials or observational subcohorts that emanate from All of Us.
- Recommend data harmonization/cleaning and version control be resourced/clearly documented throughout the study’s lifetime. This would minimize confusion over conflicting results/findings due to different versioning used in the analyses by different researchers at different times.
With the exception of the breakout groups, all presentations and group discussions were broadcast live and archived on NIH videocast:
NHLBI Staff Contact
Nicole Redmond, MD, PhD, MPH
NHLBI Workshop Planning Team
Nicole Redmond, MD, PhD, MPH
NHLBI representative (alternate), All of Us Research Program Liaison Coordinating Team
Medical Officer, Division of Cardiovascular Sciences
Gina S. Wei, MD, MPH
NHLBI representative, All of Us Research Program Liaison Coordinating Team
Associate Director, Division of Cardiovascular Sciences
Whitney Barfield, PhD
Program Director, Center for Translation and Implementation Science
Aaron D. Laposky, PhD
Program Director, National Center On Sleep Disorders Research
Lisa Postow, PhD
Program Director, Division of Lung Diseases
Sharon Smith, PhD
Program Director, Division of Blood Disorders and Resources
Christine M. Albert, MD, MPH
Professor, Harvard Medical School
Physician, Cardiovascular Medicine and Preventive Medicine, Brigham and Women's Hospital
Russell Bowler, MD, PhD
Professor, Department of Medicine, Division of Pulmonary, Critical Care & Sleep Medicine
Philip M. Alberti, PhD
Senior Director for Health Equity Research and Policy, Association of American Medical Colleges (AAMC)
Eric Boerwinkle, PhD
Dean M. David Low Chair in Public Health, University of Texas Health Science Center at Houston (UTHealth) School of Public Health
Harold Collard, MD
Professor, Department of Medicine, University of California San Francisco
J. Michael Gaziano, MD
Principal Investigator, Million Veteran Program and Scientific Director of the Massachusetts Veterans Epidemiology Research and Information Center (MAVERIC), VA Boston Heathcare System; Chief, Division of Aging, Brigham and Women's Hospital, Professor of Medicine, Harvard Medical School
Philip Greenland, MD
Director, Institute for Public Health and Medicine (IPHAM) - Center for Population Health Sciences
Harry W. Dingman Professor of Cardiology; Professor of Preventive Medicine (Epidemiology) and Medicine (Cardiology)
Nadia Hansel, MD, MPH
Associate Professor of Medicine and Associate Dean for Research, Johns Hopkins Medicine
Elliot Israel, MD
Physician, Brigham and Women's Hospital and Professor of Medicine, Harvard Medical School
Girardin Jean-Louis, PhD
Professor, Departments of Population Health and Psychiatry, New York University, Langone Health
Jill M. Johnsen, MD
Associate Member, Bloodworks Northwest Research Institute and Associate Professor, Department of Medicine, University of Washington
Muin J. Khoury, MD, PhD
Director, Office of Public Health Genomics, Division of Public Health Information Dissemination
Center for Surveillance, Epidemiology, and Laboratory Services
Office of Public Health Scientific Services
Latrice Landry MS, PhD, MSC
Clinical Molecular Genetics Fellow, Department of Biomedical Informatics, Harvard Medical School
Cora E. “Beth” Lewis, MD, MSPH, FACP, FAHA
Professor, Department of Medicine, UAB School of Medicine
Chair, Department of Epidemiology, UAB School of Public Health
Elizabeth Ofili, MD, MPH, FACC
Morehouse School of Medicine, Professor of Medicine and Director and Senior Associate Dean of the Clinical Research Center & Clinical and Translational Research
Olugbenga “Gbenga” Ogedegbe, MD
Professor of Population Health & Medicine, Chief Division of Health & Behavior and Director Center for Healthful Behavior Change in the Department of Population Health, School of Medicine, New York University
Pilar Ossorio, PhD, JD
Professor of Law and Bioethics, University of Wisconsin Law School
Susan Redline, MD, MPH
Peter C. Farrell Professor of Sleep Medicine, Harvard Medical School
Senior Physician, Division of Sleep and Circadian Disorders, Departments of Medicine and Neurology, Brigham and Women's Hospital
Physician, Division of Pulmonary Medicine, Department of Medicine, Beth Israel Deaconess Medical Center
George Saade, MD
Professor & Director, Division of Maternal-Fetal Medicine, Obstetrics and Gynecology, University of Texas Medical Branch
Christopher W. Seymour, MD, MSc
Associate Professor, Departments of Critical Care and Emergency Medicine, University of Pittsburgh
Svati H. Shah, MD, MHS
Associate Professor, Medicine, Nursing and Global Health, School of Medicine, Duke Global Health Institute
Nicolas L. Smith, PhD
Director, Seattle Epidemiologic Research and Information Center (ERIC)
Affiliate Investigator at Group Health Research Institute, Kaiser Permanente Washington
Associate Graduate Program Director, Department of Epidemiology, University of Washington
Wally Smith, MD
Professor of Internal Medicine and Scientific Director of the Virginia Commonwealth University, Center on Health Disparities
Kim Smith-Whitley, MD
Elias Schwartz, MD, Endowed Chair in Hematology and Clinical Director, Division of Hematology; Director of the Comprehensive Sickle Cell Center; Associate Professor of Pediatrics, Children's Hospital of Philadelphia (CHOP)
Scott Tillman Weiss, MD
Director of Respiratory, Environmental, and Genetic Epidemiology at the Channing Laboratory and Professor of Medicine, Harvard Medical School
Physician, Brigham and Women's Hospital
Consuelo H. Wilkins, MD, MSCI
Executive Director, Meharry-Vanderbilt Alliance
Associate Professor of Medicine, Vanderbilt University Medical Center and Meharry Medical College
Pamela K. Woodard, MD
Professor, Radiology; Professor, Biomedical Engineering
Head, Advanced Cardiac Imaging (Cardiac CT/MRI), Division of Diagnostic Radiology, Cardiothoracic Imaging Section and Director, Center for Clinical Imaging Research, Mallinckrodt Institute of Radiology, Washington University
Phyllis Zee, MD, PhD
Chief of Sleep Medicine in the Department of Neurology and Benjamin and Virginia T. Boshes Professor of Neurology, Northwestern Memorial Hospital
NIH All of Us Research Program Leadership
Eric Dishman, PhD
Director, All of Us Research Program
Stephanie Devaney, PhD
Deputy Director, All of Us Research Program
Stephen C. Mockrin, PhD
Senior Advisor, All of Us Research Program
National Heart, Lung, and Blood Institute (NHLBI) Presenters
Gary H. Gibbons, MD
Rebecca Beer, PhD
Program Official (TOPMed), Epidemiology Branch, Division of Cardiovascular Sciences
Jared Reis, PhD
Acting Deputy Branch Chief, Epidemiology Branch, Division of Cardiovascular Sciences
Appendix I: Content and Format of the Workshop
Dr. Gary H. Gibbons, Director, National Heart, Lung, and Blood Institute, gave the opening remarks, describing NHLBI’s commitment to advancing research in precision medicine, including its role in standing up the All of Us Research program. Mr. Eric Dishman, Director of All of Us, provided its current status, including protocol development and participant enrollment. This was followed by presentations from NHLBI Epidemiology Branch staff, Dr. Rebecca Beer and Dr. Jared Reis, on NHLBI’s precision medicine portfolio, longitudinal cohorts, and biorepository.
During the March use cases (hypothetical studies designed to answer important research questions-Figure 1) previously submitted via IdeaScale, identified gaps, and created new use cases based on a scientific framework of health conditions and cross-cutting themes. In addition, datatypes that were common across use cases were identified as “potential protocol elements” (PPEs).participants reviewed
Following this model, NHLBI workshop participants were divided into breakout groups by health condition (heart, lung, blood, sleep) or cross-cutting themes (Risk Factors, Prevention and Wellness/Therapeutic and Preventive Interventions; Health Disparities/Health Care Quality and Access; Genomics and Other Omics/Informatic Methodologic, Ethical, Legal, and Statistical Research; Mobile Health/Environmental and Other Contextual Effects) to further prioritize previously generated use cases and PPEs relevant to heart, lung, blood, and sleep science in accordance with NHLBI’s scientific research priorities as articulated in the Institute’s Strategic Vision. Each breakout group identified specific measures that should be obtained from participant provided information (PPI), biospecimens, physical exam, electronic health records (EHR), and mobile apps/sensors that were relevant for each health condition and/or cross-cutting theme with an emphasis on measures that had not previously been identified or needed to be highlighted as high priority for the NHLBI. The larger group discussion focused on justifications for advocating the inclusion of such measures in future versions of the All of Us protocol and/or for investment by the NHLBI to add such measures to All of Us via pilots and ancillary/sub-studies. Considerations for prioritizing measures were based on criteria below, derived from All of Us’ prioritization framework.
- IMPACT: To what extent does the use case offer the potential for a return of value for society (e.g., to maintain health, assess disease risk, detect disease, treat/cure disease, or reduce disease impact)?
- SCOPE: To what extent does the use case offer broad applicability that can advance many areas of knowledge? Does this PPE have cross-cutting capabilities that can advance many areas of knowledge and/or show synergies and system dynamics across conditions?
- FIT/STATE OF THE FIELD: To what extent is All of Us the right platform to address this research question/topic?
- VALUE TO PARTICIPANTS: To what extent is the science and potential impact meaningful and valuable to diverse participants, including those from communities traditionally underrepresented in biomedical research?
- FEASIBILITY/AFFORDABILITY/VALIDITY: To what extent are the proposed tools, measures, and analyses feasible, valid/reliable (i.e., accurately and reliably measuring the risk factors or outcomes of interest), and affordable at scale?
- PARTICIPANT BURDEN: To what extent will the additional elements place a significant burden or safety risk upon the participants?
- RESEARCHER ECOSYSTEM: Are there investigators who are able to pursue the research question or topic and use the resulting data?
Following the breakout groups, additional discussion was focused on the following topics.
- Immediately address: Are there research questions/priorities that can be addressed NOW with the current version of the protocol?
- Influence: Define requirements for All of Us platform/future protocol versions that are critical for NHLBI research priorities. What should NHLBI push for in the version 2 protocol and beyond?
- Invest: Which use cases/PPE seem more appropriate for ancillary studies, pilots, etc. requiring NHLBI support? How can All of Us contribute to NHLBI clinical trials (e.g., recruitment pool for trials, integrating trial data with All of Us data, embedding pragmatic trials in All of Us)?
- Inform: How can All of Us inform current/future NHLBI cohorts and trials (e.g., data sharing, mobile tech integration, participant engagement, EHR follow-up, linkage with other datasets such as CMS data), and vice versa?
- Engage: How do we ensure diversity and enhance engagement of participants and researchers?
Appendix II: Issues/Themes that Emerged from the Breakout Groups
Breakout groups were tasked with identifying measures that are of high interest for studying HLBS conditions, but were not clearly elucidated or prioritized in the prior All of Us Scientific priorities workshop.
Participant Provided Information
Participants identified several HLBS topics and in some cases, existing survey instruments, that should be considered for future protocols as listed below.
- Respiratory symptoms may be measured with instruments such as the modified Medical Research Council [mMRC] dyspnea scale and/or the American Thoracic Society-NHLBI Division of Lung Diseases [ATS-DLD] questionnaire;
- Bleeding symptoms may be measured with an instrument such as the International Society for Thrombosis and Haemostasis - Bleeding Assessment Tool [ISTH-BAT]; bleeding phenotypes should also be assessed but additional input is needed to determine the best measures;
- Sleep measurement should be informed by the multi-dimensional Sleep Health framework SATED (Sleep Satisfaction, Alertness, Timing, Efficiency, Duration). This framework has been shown to predict mortality and risk for cardiometabolic diseases and other health outcomes. Several specific sleep-related measures were suggested including the PROMIS short form Sleep Disturbance Scale and Sleep Impairment Scale, Munich Chronotype Micro Questionnaire, snoring/stop breathing self-report, National Health Interview Survey [NHIS] shift work history, self-reported use of hypnotics and/or CPAP, and family history of sleep apnea.
- Diet-related questionnaires should be sensitive to cultural and regional diversity in food intake. NHLBI’s experience with diverse cohorts, particularly the Jackson Heart Study, can contribute to identifying such measures
- Multi-level social determinants of health should be assessed using a biopsychosocial model, such as the one that informs The NIMHD Minority Health and Health Disparities Research Framework.
In addition, participants discussed the potential of developing a machine learning algorithm to create a customized experience skip/branch patterns in order to offer survey questions to diverse populations. Development and validation of such algorithms will require a formal process for engaging the scientific community. Lastly, participants noted that some data elements that are assumed to come from EHR may be better assessed via PPI modules (e.g., circumstances related to sudden death, some demographic classifications).
Mobile app/sensor data
Sensors may monitor physiologic processes (e.g. blood pressure, heart rate and rhythm, respiratory rate, snoring activity, electrodermal/sweat for sympathetic activity, oximetry, movement/accelerometry) or environmental conditions (e.g., indoor or outdoor air quality, location, light/dark). Integrated data on timing of movement, heart rate, light, and position with participant reported data such as dietary intake, stress, and mood could allow contextual assessments of integrated behaviors relevant for HLBS health, inform multiple health dimensions, and provide robust participant feedback. Although sensors and/or apps may exist for some measures, they may not necessarily be ready for use in All of Us due to questionable accuracy, lack of validation in diverse cohorts, or cost. Therefore, All of Us may be an opportunity for further development and validation of new technologies (e.g., mobile enabled spirometry, sleep app that integrates patient-reported behaviors with sensor data). Proprietary algorithms of sensors may present a challenge to data sharing; a formal process for engaging the scientific community for validating algorithms in diverse participants is recommended. More data science research is needed around these issues which may require more “open source” code and algorithms to facilitate.
Electronic Health Record (EHR)
Many EHRs link to external data systems that have data of high interest to HLBS research; therefore, systems will need to be developed to integrate this information into the All of Us data repository. Some key examples include:
- Newborn screening for sickle cell trait
- Pharmacy data for medication adherence
- Time stamping of medication administration
- Sleep lab results and sleep apnea treatments
- Genomic testing/referrals
- Electrocardiogram (ECG) images/reports
- Raw image data (not just image interpretation reports) for common imaging modalities (e.g., computed tomography [CT], magnetic resonance imaging [MRI], echocardiogram)
- Blood bank records
- Specialty (“send-out”) labs
- Pulmonary function tests/spirometry results and interpretation reports
It will be important to involve participants to provide additional self-report measures to provide context to EHR inputs. In addition, machine learning may be employed to generate disease phenotypes and predictive models using EHR data; however, there is some concern that confounding by indication may bias the results. A formal process for engaging the scientific community to validate algorithms should be encouraged. Lessons from analysis of big cohorts from TOPMed and other large cohort studies with phenotypes and -omics may be applied to All of Us and influence future versions of the protocol.
Biomarkers and Genomics
With regard to specific biomarkers of interest for HLBS conditions, consider analyzing all participants with respect to complete blood count (CBC), lipids, glucose, and creatine using blood collected at baseline exam. Important, but perhaps lower priority measures include Hemoglobin A1C, troponin, and high sensitivity C-reactive protein (hs-CRP). Prospectively, the added value of repeated blood measures should be considered, as limiting to specific disease or at-risk subgroups of interest may be more feasible.
Fidelity of the collection and handling of biological samples by way of the standardization of sample collection and transport is critically important to ensure the ability to obtain high quality, reproducible samples in order to enable -omics analyses. Timing of blood collection relative to meal intake may be important for some measures. Processing of “-omic” (i.e., genome, transcriptome) data potentially introduces bias; therefore, storing raw sequence data for All of Us participants would be preferable in order to enable studies in new and emerging spaces.
Participants discussed return of results for actionable genes according to the American College of Medical Genetics recommendations, including the implications of enrollment and return of results in children (while a child) and the need to re-consent at the age of majority.
Prospective Non-Blood Exam Measures
Currently, All of Us plans to conduct physical exams on only a subset of the program’s participants. Workshop participants felt that exam measures should be customized based on age group and life-stage (e.g., pregnancy). Workshop participants discussed the potential importance of repeated exam measurements, specifically body mass index (BMI), given the high prevalence of obesity and the importance of longitudinal monitoring of this measure. This may provide some consistency over BMI measures gleaned from the EHR.
Imaging and other diagnostics
Workshop participants discussed the need to develop new technologies or more portable/affordable technologies that could be used by participants or integrated into the All of Us exam. For example, pulmonary function tests/spirometry are crucial for the diagnosis of most lung diseases and have implications for many non-pulmonary diseases, but current technologies are too expensive and require an office visit. Development and validation of more affordable and portable technologies for measuring lung function would be of high interest. All of Us should consider other innovative imaging technology, such as participant provided video uploads of behaviors such as gait or sleep.
A cardiovascular imaging repository could be generated for the All of Us by leveraging clinically indicated imaging data. For example, low-dose computed tomography (CT) is recommended for lung-cancer screening for individuals with current smoking. In addition to assessing for presence of lung cancer/nodule (its intended clinical use), it can also be used to assess for presence/extent of chronic obstructive pulmonary disease (COPD), vertebral body bone density for osteoporosis, coronary artery calcium scoring, and presence of sarcopenia (loss of muscle mass). Imaging findings could also be correlated with genetic information and smoking history to determine genetic factors influencing which smokers develop emphysema/lung cancer/coronary artery disease vs. those who do not; and whether smoking is associated with earlier age of osteoporosis development or loss of muscle mass, especially in women. Workshop participants cautioned about potential privacy concerns related to imaging; e.g., brain MRI could potentially be used to identify participants.
There was significant discussion regarding the current enrollment/recruitment strategy of All of Us. Because there is no pre-defined sampling frame for All of Us enrollment, this raised several concerns.
- Without a clearly defined sampling frame, there is the threat of the “atomistic fallacy”, which is an erroneous inference about causal relationships in groups made on the basis of relationships observed in individuals. Because of this, significant efforts to communicate the potential limitations of the current enrollment and data collections in order to ensure responsible interpretation and dissemination of data should be made. It is important to avoid the incorrect interpretation/assumption that the All of Us cohort can provide “representative” population estimates in the same way as a study like the National Health and Nutrition Examination Survey (NHANES). One recommendation was to create a ‘rapid reaction team’ to address potential misinterpretation of data.
- Because enrollment in All of Us has only recently begun, with just under ten percent of the targeted one million participants enrolled, there is still the possibility to consider this current process a “pilot for enrollment” and potentially impose a more NHANES type sampling frame to future participant enrollment in order to avoid the misinterpretation.
- The risk of inaccurate and misleading inferences is particularly concerning with regard to enrollment targets for certain subgroups, as “underrepresented in biomedical research” does not necessarily mean that these populations also experience health disparities.
- However, All of Us is likely to have more “real-world” data in comparison to traditional cohorts; by accessing individuals within healthcare systems All of Us can circumvent “healthy volunteer bias” that leads to sampling bias in cohorts, clinical trials, registries, and repositories.
Furthermore, how underrepresented in biomedical research (UBR) is defined may be specific to All of Us and needs to be aligned with how other entities define UBR. The All of Us enrollment targets are not clear as it relates to the four populations that have health disparities as defined by the NIMHD Minority Health and Health Disparities Research Framework: a) racial and ethnic minorities, b) sexual and gender minorities, c) underserved rural residents, and d) low socioeconomic status (SES). Being underrepresented in research does not necessarily mean that population is experiencing health disparities. Furthermore, because there are no defined enrollment targets, it is plausible that some potential participants may decide not to enroll or discontinue participation because they feel their population group is already well represented (e.g., if we advertise 70% “UBR”, some groups may not feel compelled to enroll).
Income alone is insufficient to determine SES. A composite measure that includes, but is not necessarily limited to income, employment, education and housing may be more robust. Specific engagement efforts may be needed in order to access difficult to reach (under-represented) populations who traditionally have greater representation within healthcare systems and are less likely to volunteer from the community. Such difficult to reach populations include older adults (> 65 years), African Americans, Latinos, sexual and gender minorities, rural residents, individuals with disabilities (e.g., hearing or visually impaired, etc.), and lack of internet access or with limited technical proficiency. In addition, it will be important to understand if there is differential response rates by underrepresented in biomedical research (UBR) groups for various data collections. Furthermore, adding participants who have completed prior NHLBI studies (vs. in ongoing NHLBI studies) may be particularly relevant for minority health given NHLBI’s portfolio of diverse cohorts.
All of Us participants within single provider/small clinical practices are likely to enroll through the direct volunteer pathway. Thus, a provider outreach strategy may be needed to educate providers who are not a part of the healthcare provider organizations (HPOs) affiliated with All of Us in order to empower and/or incentivize them to encourage their patients to enroll, and to facilitate biospecimen collection from enrolled participants. Recruitment of “specific” populations (e.g., populations with disease conditions of interest or other specific characteristics) was distinguished from recruitment of “special” populations as defined by IRB (e.g. children, incarcerated individuals, etc.). Recruitment of specific populations may be challenged due to diagnostic bias as some populations may not yet be diagnosed with conditions of interest and thus may not be included; particularly for queries of self-reported diagnoses phrased as, “Did a PROVIDER tell you had [diagnosis]…”
Workshop participants also recommend leveraging enrollment of the family unit (i.e., adult and their child/children). For mother/newborn pairs, this could enable an “antenatal cohort” to explore the health status of women pre, during, and post pregnancy (e.g., prior pregnancies, hypertension during pregnancy, etc.), and newborns to predict HLBS conditions. This may require addressing consent issues in order to obtain data that may not be in the EHR such as birth certificates and newborn screening information. Lastly, workshop participants recommended that All of Us consider expanding recontact consent in order to allow participants who would potentially be eligible for HLBS focused studies outside of All of Us to be invited to participate.
Adjudication of outcomes is critically important, particularly for HLBS conditions where ICD codes or self-report are poor (e.g., heart failure, myocardial infarction, atrial fibrillation, sudden cardiac death, stroke, peripheral artery disease, venous thromboembolism). There may be an opportunity to use machine learning to find new, standardized ways to assess outcomes using EHR, and All of Us should learn best practices for outcome ascertainment from other groups (e.g. registries, the Million Vets Program, Patient Centered Outcomes Research Institute [PCORI]). Adjudication of a certain percentage of self-reported or ICD- code based HLBS cases as gold standard cases should be considered.
There was also discussion about the need for more clarity about how often and for what purposes participants may be re-contacted. Re-contact is important for longitudinal following of exposures and endpoints, particularly for patient-reported outcomes (PRO) and endpoints that occur outside of the healthcare system. Discussion also focused on the critical role of participant proxies to assist with participant follow-up, and the need for continuity of proxies for longitudinal assessment of some outcomes such as cognitive decline and dementia. Ascertainment of the occurrence and cause of death is a high priority outcome for NHLBI. Workshop participants recommended obtaining consent to contact proxies to look into cause of death, beyond the National Death Index and/or ICD codes.