ENABLING A HEALTHIER FUTURE FOR ALL THROUGH PRECISION MEDICINE

ABSTRACT

The National Institutes of Health’s All of Us Research Program is building one of the most comprehensive research cohorts in the United States. Launched in 2018, the program has enrolled more than 849,000 individuals to empower precision medicine relevant to all populations. Participants in All of Us share complete surveys, share electronic health records, donate biospecimens, and contribute other data, such as wearable device data. To date, All of Us has generated whole genome sequences for more than 414,000 participants and returned actionable genetic health information to more than 100,000. By examining the technological, engagement, and data integration challenges encountered, this paper highlights how All of Us builds on pioneering efforts in biomedical research and points to future directions, including expanded data sharing, international cooperation, and the potential to drive transformative discoveries in personalized health care. Current successes also highlight necessary future directions: even larger and more comprehensive cohorts, better data sharing, and more international cooperation.

INTRODUCTION

Precision medicine seeks to tailor health care treatments and preventive strategies to the unique genetic, environmental, and lifestyle characteristics of each individual. This approach builds on the legacy of many large observational cohort studies that transformed public health through discovery of intervenable risk factors. One notable example is the Framingham Heart Study that was launched in 1948. It became a landmark in epidemiology by identifying lifestyle and biological risk factors for cardiovascular disease, all starting with 5,209 individuals, which is modest by today’s standard (1). Insights from Framingham revealed crucial connections between smoking, cholesterol, diet, and hypertension and the risk of heart disease. Treatments and public health interventions based on such intervenable risk factors have resulted in a more than 60% reduction in cardiovascular mortality from the 1950s to now (2).

Other major cohort studies, such as the Nurses’ Health Study (3) and the Women’s Health Initiative (4), expanded this model by focusing on additional populations and health outcomes. These studies revealed links between hormone replacement therapy and cancer risk, dietary factors and chronic disease, and other relationships between lifestyle factors and long-term health outcomes. Over decades, these cohorts highlighted the power of observational research to reveal preventable risk factors, contributing to medical guidelines and shifting health care practices to prioritize preventive care.

While observational cohort studies have long been foundational to population health, their effectiveness has been limited by a lack of participant variability. Other limiting factors include numbers of participants enrolled, breadth and depth of data collection, ability to recontact and resample participants, and length of follow-up. In particular, more than 90% of genome-wide association studies (GWAS) draw data from individuals of primarily European genetic ancestry (5). This skewed composition can lead to inaccurate polygenic risk scores when applied to individuals predominantly from other genetic ancestries and incorrect interpretations of genetic variant pathogenicity, reducing the effectiveness of precision medicine for all populations (6,7).

To address these opportunities, the National Institutes of Health (NIH) launched the All of Us Research Program (All of Us) in 2016 as part of the Precision Medicine Initiative. The program aims to recruit over 1 million participants from across the Unites States, including individuals from a broad range of age groups and backgrounds (8). By collecting comprehensive data, including genetic information, electronic health records (EHRs), biospecimens, and health-related surveys, All of Us seeks to advance scientific understanding to drive advances in diagnosis, therapeutics, prevention, and maintenance of health. The program’s large cohort will provide the foundation for research that can drive more objective, higher-quality health care by ensuring discoveries and treatments are based on data from all populations. All of Us is designed as a platform to enable research through broad access to program-generated data as well as partnered research studies that generate additional data from the participants. Approved researchers from around the world access the All of Us Researcher Workbench to study disease etiologies and risk factors, treatment effects, and health care gaps.

All of Us aspires to extend the legacy of observational cohort studies, harnessing today’s technological advancements to create a platform that will enable rapid growth of personalized health care. This paper provides an overview of All of Us’ design and data collection methods, its platform model for research, and the importance of returning results to participants. This paper concludes with a discussion of future opportunities.

All of Us Program Design and Data Collection

A primary goal of All of Us is to include and engage populations through strategies including partnerships with community organizations, regional and academic medical centers, and federally qualified health centers. The program launched with the only enrollment requirements of living within the United States and being over age 18. Inclusion criteria are broad, encompassing communities with different health needs and unique health experiences. Over 80% of current participants have had fewer opportunities to contribute to and benefit from medical research, including individuals living in rural areas, older adults, and those with less access to care (9).

The program uses a multifaceted data collection approach. Participants provide baseline health data through digital surveys that cover topics such as lifestyle, habits, personal and family medical history, and mental health. Additional data, including physical measurements (e.g., blood pressure, body mass index) and biospecimens (e.g., blood, urine), are collected at enrollment sites across medical centers throughout the United States, including some in U.S. territories. Participants not located near an enrollment site can send in saliva samples for DNA. Participants are also encouraged to contribute data from wearable devices, such as Fitbits, to track activity metrics over time. This robust, longitudinal data collection framework enables an in-depth analysis of health patterns and trends across the population (Figure 1).

Fig. 1.

Distribution of data types in the All of Us Research Program Curated Data Repository (CDR) version 8. The timeline (1980–2023) illustrates the breadth of data collected from EHRs, surveys, and wearable devices. The gradient shading indicates increasing density of data over time, with a significant increase in data density over time.

Data are brought together in a raw data repository and then harmonized with removal of personal identifiers within the Data and Research Center (10). The data repository encompasses EHRs, genetic data, lifestyle surveys, and environmental factors, creating a multidimensional dataset accessible through the Researcher Workbench. This platform allows registered researchers to access de-identified participant data within a secure cloud-based environment. With tiered access levels (Public, Registered, and Controlled), the Researcher Workbench enables data-driven studies while safeguarding participant privacy.

Key Areas of Research Enabled by the Program

The All of Us Research Program enables a wide range of research across basic, clinical, and population health science. Researchers can study complex interactions between genetic, environmental, and lifestyle factors. Through its inclusion of structured EHR data going back up to 40 years, both common and rare diseases research is supported. Disease conditions tend to be overrepresented within All of Us compared to population estimates, and pairwise associations of conditions have similar effect sizes to well-established measures (11). Links to survey data, the American Community Survey, and other resources support investigation into health outcomes, such as exploring how social determinants influence disease prevalence and health status. Clinical research benefits from the program’s pharmacogenomic data, enabling studies on drug efficacy and safety tailored to different genetic ancestries. Additionally, links to environmental data and future generation and incorporation of exposomic data may facilitate research on how pollution and other environmental and occupational exposures influence health. Machine learning and artificial intelligence are also applied to predict disease risk and progression, as demonstrated in studies like those on glaucoma (12) and risk of stroke in atrial fibrillation (13).

All of Us aims to improve health by supporting research that evaluates genetic risk factors across a wide range of populations. Gaps in health outcomes are multifactorial. For example, studies are exploring the variability of polygenic risk scores and testing their effectiveness among different genetic ancestries and incorporating other measures that influence health outcomes, such as social determinants of health, lifestyle, and environmental exposures. By providing broad access to these data, researchers can refine disease risk models, which may lead to more accurate predictions and interventions for a wider range of populations.

Environmental exposures, such as air pollution, play a significant role in health outcomes. Participants in All of Us provide residential and work locations, which includes geospatial linkage to environmental data. Currently, researchers have access to three-digit zip code information in the Controlled Tier, which some have used to facilitate studies on environmental risk factors for diseases. For example, recent findings have demonstrated links between fine particulate matter (PM2.5) exposure and cancer risk (14). We are exploring approaches that would allow more granular geospatial linkages.

Return of Health Information to Participants

Genetic information from the program is transforming approaches to preventive health care. One of the foundational goals of All of Us is not only to collect health data from populations for research but also to empower participants by returning actionable health information directly to them. The program has implemented several initiatives to provide participants with personalized reports that can have significant implications for their health and well-being. This return of information includes detailed genetic insights related to hereditary disease risk and medication responses, making All of Us one of the largest health research programs globally to prioritize participant engagement through actionable health data.

Hereditary Disease Risk

A central aspect of the All of Us information return strategy involves reports on hereditary disease risks. As of December 2024, over 128,200 participants had accessed personalized reports that identify genetic variants associated with 59 serious health conditions, including hereditary cancers and cardiovascular diseases. These reports enable participants to understand their potential genetic risks, empowering them to make informed decisions about their health in collaboration with their health care providers. To date, approximately 3% of participants who received their hereditary risk reports have discovered a potentially life-changing genetic variant, which may prompt further medical evaluation, preventive measures, or family screening.

Medication and Pharmacogenomic Insights

The program also provides a “Medicine and Your DNA” report, which focuses on pharmacogenomics—how genetic factors influence a person’s response to medications. Over 121,600 participants have accessed their personalized reports, which analyze seven genes that impact the body’s processing of more than 50 different medications, including those for pain management, cardiovascular conditions, and mental health. These pharmacogenomic insights are critical, as they can influence which medications participants may respond to best or help avoid drugs that could cause adverse reactions. More than 90% of participants who have received this report have had an actionable drug-gene interaction that could directly affect their medication choices, demonstrating the value of personalized insights in daily health care​.

Broader Impact and Future Directions

The return of health information aligns with the program’s mission of participant-centered research and illustrates a shift toward more interactive, reciprocal research models. By actively involving participants in their health journeys, the program seeks to strengthen trust, improve participant retention, and encourage individuals to engage more fully with their health data. Future expansions of the information return may include additional genetic return of results, environmental data, and other health reports. We also seek to make available information about clinical trials for which they may be eligible.

Through these efforts, All of Us not only serves as a data resource for researchers but also as a valuable tool for participants, turning research findings into actionable health information that may lead to improved health outcomes and a greater sense of agency in personal health care.

Current Achievements and Milestones

Since the launch of national enrollment in 2018, All of Us has made significant strides in participant enrollment, with over 849,000 individuals enrolled as of December 2024 (Figure 2). The program released its eighth public data set that includes more than 633,000 individuals with some type of data, the minimum of which includes survey data. The data set includes more than 414,000 whole genome sequences and EHRs on 393,000 individuals. Importantly, following three formal consultations with tribal nations, this data set is the first to include data from individuals who self-identify as American Indian/Alaska Native (AI/AN). Individuals who identify as AI/AN have been able to join All of Us since its inception, and we have included them in return of genetic result information if they wished.

Fig. 2.

Status of participant enrollment and research utilization in the All of Us Research Program. This figure highlights key metrics of the All of Us Research Program as of December 2024.

To date, more than 14,300 researchers from over 1,000 organizations have used All of Us data to publish more than 800 studies. In addition to typical academic researchers and well-funded institutions, the All of Us researcher population includes high school students and students from community colleges and other nontraditional institutions. This wide range of backgrounds fosters additional research perspectives and drives innovative studies across health, basic science, and clinical conditions, making the program’s data widely accessible for varied research needs. Research findings have included novel insights into gaps in health outcomes, genetic risks for chronic diseases, and disease risk factors. Nearly all research studies using All of Us data include populations that typically have had limited representation in research in the past, and closing gaps in health outcomes is a common theme. It has been an important resource for artificial intelligence and machine learning as well. For example, a study using machine learning approaches to predict who would need glaucoma surgery demonstrated how the comprehensive, multisite data derived from many different health systems in All of Us dramatically improved predictive models for surgery needs over models trained on single-site data (12). These studies highlight the potential of All of Us to contribute to scientific knowledge across many health conditions.

Looking to the Future: More Collaboration and Exponentially More Data

All of Us exemplifies a transformative approach to precision medicine, but its success represents only a part of a broader movement toward global and integrated health research. International collaboration across larger cohorts, standardized data practices, routine clinical use of advanced genomic and other -omic (e.g., proteomic, metabolomic, exposomic, and microbiomic) technologies, and sensor and wearable devices will play critical roles in deepening our understanding of health and disease (Figure 3). Together, these efforts aim to create a more complete and precise picture of human health across populations and environments.

Fig. 3.

Envisioning how precision medicine will affect clinical medicine and research in the next decade. This figure illustrates seven foundational components essential for the transformation of health care through precision medicine by 2030 [adapted and used with permission from Denny and Collins (15)].

The Rise of International Cohorts and Global Collaboration

A key advancement in the field has been the growth of global research consortia, such as the International Hundred Thousand Cohort Consortium (IHCC), which connects over 100 cohorts representing nearly 50 million participants across 43 countries (16). This network and others promote large-scale data sharing, enabling researchers to investigate health trends and outcomes on a global scale, addressing diseases that transcend borders, and studying genetic differences across populations. Supporting these efforts, initiatives like the Global Alliance for Genomics and Health (GA4GH) provide frameworks and tools to ensure data interoperability, ethical sharing, and collaboration across nations. Together, these programs promote the scalability and integration of precision medicine efforts, ensuring that discoveries can benefit people regardless of geography.

Clinical Sequencing and Multidimensional Omics

As sequencing costs decline, genomic sequencing has become more accessible in both research and clinical settings. The integration of genetic data into more routine clinical care represents a shift in health care and will allow providers to identify disease risks earlier and tailor treatments more effectively. Beyond genomics, advancements in other -omic fields—such as exposomics, metabolomics, and the microbiome—offer new ways to understand the complex interplay between genetics, environmental exposures, and lifestyle factors. Research into the microbiome further reveals how microbial populations within the human body influence various health outcomes, from immune function to metabolic health.

Expanding the Role of Wearable Devices and Sensors

Wearable technology, such as Fitbit, Apple Watch, and other health sensors, has introduced the potential for real-time health monitoring, offering insights into health metrics outside the clinical setting. These devices capture continuous data on physical activity, sleep patterns, heart rate, and more, providing a dynamic view of health. This type of data could redefine our understanding of health by integrating daily fluctuations and responses to environmental changes. Such measures allow for early detection of deviations from baseline health, potentially leading to preventive interventions before issues escalate into chronic conditions.

Integrating Data to Advance Precision Medicine

The future of precision medicine hinges on harmonizing and integrating data from different sources, including EHRs, genomic data, sensor data, and multiple -omic fields. By creating integrated data ecosystems, researchers can uncover complex patterns and relationships that may be invisible within isolated, episodic data sets. Efforts to standardize data collection, as championed by GA4GH and IHCC, will facilitate these advancements, allowing global research networks to analyze and share insights seamlessly. These integrated databases can support predictive models and personalized interventions, potentially improving health outcomes across populations.

The rapid expansion of these technologies, along with the collaboration of research communities, represents a future where health research is not limited by geographical, technological, or disciplinary boundaries. Together with programs like All of Us, these efforts aim to establish a new era of health care that is as complex as the populations it seeks to serve, bringing precision medicine closer to a global reality.

DISCUSSION

All of Us represents a milestone in making precision medicine accessible to populations from a wide range of age groups, backgrounds, and life experiences. By addressing the gaps in genomic data representation, the program enables a more comprehensive approach to health risk prediction and treatment. This is critical for realizing precision medicine’s promise to provide targeted health care for all.

Several challenges persist in implementing the program’s objectives, including harmonizing EHR data from various sources, filling in missing data, and maintaining participant engagement over time. EHRs are a valuable and necessary source of disease and treatment phenotypic data in All of Us but are subject to data fragmentation. Additionally, we still are missing EHR data for many participants who have volunteered but are not part of an existing All of Us health care provider organization. Finally, there is a need for ongoing evaluation of data generalizability, as differences in survey completion rates and data completeness could affect research outcomes.

Looking ahead, the program seeks to expand its data repository by incorporating new data types, such as imaging and exposomic data. Most of these types of additional data will be developed through partnered research studies. These projects will involve partnerships with other parts of the NIH or external organizations. Several such studies are already underway, bringing together expertise and resources from across the NIH. Examples include Nutrition for Precision Health, a study of environmental and genetic causes of type 2 diabetes in partnership with the National Institute of Environmental Health Sciences, and a study of optical coherence tomography data in partnership with the National Eye Institute. We believe there will be incredible power of multimodal data collections with participants who may be recontacted for additional studies and data, such as in All of Us, to accelerate new research while simultaneously reducing costs of recruitment and data collection and generation.

Enhancing collaboration with other U.S.-based and international cohorts will further enhance the utility of the data. Indeed, studies already routinely include data from multiple cohorts. Attention paid to phenotype and genotype data standards, facilitated by organizations such as GA4GH and IHCC, will aid in multicohort analyses.

Finally, we recognize the importance of new methods, such as artificial intelligence, applied to large multimodal health data sets. The rise of artificial intelligence has yielded incredible insights into complex data. It is essential that broad, deep, and inclusive data sets like All of Us are available for such methods to make sure their insights are applicable to the broad U.S. population.

ACKNOWLEDGMENTS AND FINANCIAL SUPPORT

The All of Us Research Program would not be possible without the partnership of its participants. I acknowledge the contributions of NIH, program funders, researchers, participants, and partnering organizations whose efforts make the All of Us Research Program possible. The All of Us Research Program is supported by the National Institutes of Health, Office of the Director: Regional Medical Centers (1 OT2 OD026549; 1 OT2 OD026554; 1 OT2 OD026557; 1 OT2 OD026556; 1 OT2 OD026550; 1 OT2 OD 026552; 1 OT2 OD026553; 1 OT2 OD026548; 1 OT2 OD026551; 1 OT2 OD026555; IAA: AOD21037, AOD22003, AOD16037, AOD21041), Federally Qualified Health Centers (HHSN 263201600085U), Data and Research Center (5 U2C OD023196), Biobank (1 U24 OD023121), The Participant Center (U24 OD023176), Participant Technology Systems Center (1 U24 OD023163), Communications and Engagement (3 OT2 OD023205; 3 OT2 OD023206), and Community Partners (1 OT2 OD025277; 3 OT2 OD025315; 1 OT2 OD025337; 1 OT2 OD025276).

Finally, I am grateful for the assistance of Hannah Coleman and Rachele Peterson for their assistance in reviewing and editing this manuscript. In addition, parts of this paper were drafted with the assistance of large language models, primarily ChatGPT.

DISCUSSION

Hochberg, Baltimore: As a former Veterans Affairs (VA) investigator, are you partnering with the Department of Veterans Affairs and the Million Veteran Program (MVP)?

Denny, Bethesda: Yes, we are partnering with the VA. The VA is one of our enrollment partners, and it has enrolled about 25,000 veterans in the program. In addition, other veterans have enrolled outside of the VA enrollment efforts. The MVP is a great program, and some people have enrolled in both MVP and All of Us. MVP is not as open to outside researchers as All of Us, but we certainly partner with MVP researchers and do joint research projects with them.

Wilson, Durham: I have a question that’s not in reference to your data, which is a really fantastic endeavor. I am actually somewhat concerned about the security of the data, having recently received a notice that all of our records were hacked, signed up for identity protection, and so forth. How secure is your data and the association with specific identifiers?

Denny, Bethesda: Great question. It’s actually the most common question we get from participants, as you might imagine. I used to be at Vanderbilt and led the data and research center for All of Us, which housed the majority of the data at that point. Maintaining privacy and security of people’s data is Job One. I think we’re really best in class in what we do with this and how we test it. We also do things like having “white hat” hackers try to break into the system with rewards if they are successful. Basically, we have the good people trying to help us find bugs and make the system better. We also have all sorts of automated third-party tools and people testing the system. We maintain common best practices, like having data encrypted at rest, separating identifiers from the identified data. It’s also why we don’t allow individuals to download data. Instead, we require them to use our clouds so that we can audit and monitor what happens. If there are worrisome flags of activity, we can shut off data access instantly.

Wilson, Durham: You know you need to speak to the private sector, thank you.

Denny, Bethesda: Thank you.

Feldman, Philadelphia: This is a little bit more of a comment than a question, but my center was a site for All of Us. I have tried repeatedly to get data on myself and my family. I have also tried to get data on patients as a group, and it’s been impossible. As investigators, we don’t really know how to get through whatever walls necessary to get that data, and I think that could be much simpler and more straightforward for us. The other side of this that what we’re seeing in our field, which is heart failure, is that industry is now jumping into that empty cauldron. A huge paper came out just a short while ago in which the industry representatives were the first authors and the last authors of the paper. The paper that was published in a society journal stated implicitly that none of the data would ever be available to anybody who was outside the study. We can’t check their data or see if it’s right or wrong. Is that something you can address through federal opportunities because I think it’s a real danger out there?

Denny, Bethesda: That’s a really great question and feel free to ask a follow-up. In terms of data to individuals, we, in response to our engagement studios and input from participant and community advisory boards, chose to have participants be the arbiter of who gets their identified genetic results. At this point, a participant can’t download their raw genetic files, but I’d like to get there. Participants can get clinically actionable genetic results as a PDF, and genetic counselors are available to talk with them. That has actually led to cascade screening at points in families, which has been really exciting in terms of downstream health effects.

We also don’t allow external researchers to come in and identify people—that’s part of our privacy protections. We do have ways to work on recontact potentially, and that will grow over time.

In terms of data reproducibility, I think we’re actually very good at it. That’s really a foundation of all we want to do because everything is done within our data clouds. Researchers come into that, the work they do is there, and those data are available and can be shared to any researcher approved in the system. You could reproduce the study and get all the raw data that someone else had produced.

Feldman, Philadelphia: Okay, thank you.

Denny, Bethesda: Thank you.