Optimal Instruments to Measure Diet, Physical Activity, and Sleep

The National Institutes of Health’s (NIH) National Heart, Lung, and Blood Institute (NHLBI), in partnership with the National Cancer Institute (NCI) and the National Institute on Diabetes and Digestive and Kidney Diseases (NIDDK), convened a workshop on “Optimal instruments for measurement of dietary intake, physical activity, and sleep among adults”. This workshop assembled national and international experts from multiple disciplines to review the state of the science regarding the validity, reliability, and sensitivity of instruments used to assess dietary intake, physical activity, and sleep in large observational studies of adults.

Background

Research has shown that healthy behaviors are associated with greater longevity and disease-free survival. [1] Mounting evidence tells us that eating healthy, engaging in regular physical activity, and having adequate sleep of good quality can improve overall health. These behaviors have been identified as key components of the American Heart Association’s Life Essential 8 construct that is used to guide cardiovascular health. [2] Understanding the relationship between diet, physical activity, and sleep with health outcomes is crucial for advancing our understanding of diseases. The recently unveiled national strategy on Hunger, Nutrition, and Health emphasizes the importance of advancing nutrition metrics, increasing data collection, and providing access to safe venues for physical activity as a means to enhance health and overall well-being. [3] However, measuring these dynamic behaviors accurately can be difficult, leading to variations in the observed short- and long-term associations with specific health outcomes, and making it hard to interpret the relative importance of these behaviors when viewed across an individual's lifespan.

Workshop Purpose and Objectives

The objectives of the workshop were to:

• Review the state of the science regarding the validity, reliability, and sensitivity of instruments used to assess dietary intake, physical activity and sleep in large observational studies among adults.
• Evaluate the pros and cons, including the feasibility of instruments to measure dietary intake, physical activity, and sleep in large population studies.
• Discuss best practices to harmonize data from differing instruments across studies within each measurement domain to facilitate pooling studies as well as the possible incorporation of new instruments within each domain into ongoing studies while retaining the ability to perform valid longitudinal comparisons from previous instruments.
• Identify gaps and opportunities for future research to develop improved and objective measurement tools for large studies to assess these key health behaviors including the use of biomarkers (“omics” such as metabolomics), mobile technology, and sensors).

The recurring themes from discussions across domains were:

Measuring human behaviors consistently across different population groups can be a challenging task. Factors such as variations in eating habits, ethnic-specific diets, access to recreational spaces, and the use of different measurement tools have hindered the inclusion of diverse population groups and reduced the applicability of such tools in large population studies. However, many aspects of chronic diseases are closely related to diet, physical activity, and sleep. Research efforts on these behaviors in large pooled samples have been challenging due to the lack of standardization and harmonization of measured variables. By leveraging objective measures, we may be able to improve our understanding of the effects that these behaviors have on health and provide a platform for comparisons across studies. The participation of different communities has expanded the applicability of diet measurement tools by broadening dietary patterns and ethnic-specific foods. Additionally, new technologies, such as wearables, allow for greater participation and objective data collection in physical activity and sleep. While much more research is needed, the improved digitization of measurements and improved data gathering tools provide a unique opportunity to strengthen the evidence observed within and across the three behaviors.

Assessing behavioral patterns throughout the lifespan has been deemed essential in light of the dynamic changes that occur over time. Measurement of behaviors at a single time-point may result in variations in the observed health outcomes and the proposed recommendations for a healthy lifestyle. Furthermore, it is imperative to recognize the complexity associated with the measurement of multiple behaviors and the limitations encountered when attempting to integrate them. The recent advancements in technology, such as the utilization of time-use compositional data analysis and machine learning algorithms, have facilitated the analysis of 24-hour cycles, and have been instrumental in the interpretation of large-scale, real-world data into reproducible and quantifiable behavioral patterns.

Below is a short list of discussed research opportunities and strategies to improve the measurement of diet, physical activity, and sleep in population-based studies.

1. Improvement of measurement tools for diet, physical activity, and sleep.
   
   Integrate different sources of data to assess diet. Factors that influence diet at the neighborhood and work environment level can be studied with Geographic Information System (GIS) data (e.g., distance from work/school to food settings), along with developing tools to assess community nutrition environments (e.g., Nutrition Environment Measures Survey - NEMS), degree of food processing (e.g., The NOVA classification system of processed foods) and sales data to evaluate the effects of different policies on changes in population-level diet metrics.
   
   Expand the collection of data from physical monitors such as consumer-based wearables that provide the measurement of physical activity and sleep during 24-hour cycles and longer periods. An identified limitation is the lack of access to and use of the raw data, including those from accelerometers and health measurement devices (e.g., heart rate, pulse oximetry). Validation of measurement tools from devices with standardized raw values should be encouraged.
   
   Define clear protocols to harmonize exposures, covariates, and outcomes. Define the methods for analyzing and pooling data from diverse cohorts.
   
   Develop “OMICs” biomarkers that are specific to each of the behaviors (diet, physical activity, sleep), as well as standardized and developmentally translatable biomarkers that can be reproduced in population studies.
   
   Data-driven approaches should be derived from survey, accelerometry, and wearable data, as some behaviors are still broadly categorized (e.g., sedentary, light, and moderate to vigorous). This also emphasizes the need to gather diet, physical activity, and sleep measurements in larger datasets to support the development of machine learning inference methods and apply such approaches to epidemiological cohorts.
    
2. Standardization and Harmonizing of data across population studies
   
   Challenges in standardization and harmonization of diet, physical activity, and sleep data include a lack of nationally representative datasets with well-validated measurements (for objective and subjective measurements). In some cases, each one of these behaviors measures only one aspect such as one time point or survey tool, while assessments should be made through multiple dimensions, time points, and multiple measurement tools. Furthermore, improved standardization is needed across different surveys and measurement tools.
   
   Develop and promote consistent standards (e.g., standard file types for device recording, standard variables, reporting, and wording) and, in the implementation of measuring behaviors, a combination of tools should be considered (surveys are the easiest, wearables can be variable).
   
   Leverage validated population-specific surveys (e.g., Food Frequency Questionnaire - FFQ, Automated Self-Administered Dietary Assessment Tool - ASA24) that have been used in large population cohorts. Surveys will continue to be the primary instrument in terms of dietary assessment, which when done repeatedly, greatly improves their accuracy and reliability. While wearables continue to show promise for accurate and continuous measurement of physical activity and sleep, efforts to obtain raw wearable data are needed to improve the validity across devices.
   
   The creation of standardized databases for use in pooled analyses of multiple cohorts is needed, to maximize the comparability of data by reducing inconsistencies, improve the quality and interpretation of data, and enhance scientific rigor.
   
   Harmonization of biomarkers (e.g., metabolomics) is required while taking into consideration differences in the platform that are very similar but not completely overlapping. Metabolite data should be harmonized across different cohorts before being combined.
   
   Metadata and data should follow the FAIR guidelines which aim to improve the Findability, Accessibility, Interoperability, and Reuse of digital assets. Achieving the FAIR principle require skilled personnel, such as data engineers and coders.
    
3. Additional Future Aims
   
   Integration of the different health behaviors can be evaluated with data science tools and advancement in OMICs science for nutritional profiles, 24-hour cycles, type of physical activity (e.g., standing, driving, sleeping, etc.) and through the different dimensions of sleep. Understanding individuals’ performances in free-living conditions is crucial, as clinical research usually collect the best performances but not habitual ones.
   
   A combination of methods and statistical modeling is needed for the correction of measurement errors. Biomarkers should be integrated and used to improve the estimates of association with outcomes. Measurement precision should be improved, particularly with regard to differences in behavior interrelation.
   
   Approach large cohorts using questionnaire-based measurements, and in sub-cohorts, wearable devices and smartphone-based apps could be used for a more detailed assessment of diet, physical activity, and sleep, while incorporating biomarkers and metabolomics for validation and calibration.

Publication Plans
The meeting organizers and participants will develop a workshop report for submission for publication in a peer-reviewed journal. The report will describe in more depth the discussions and research opportunities, including those under each of the five domain areas.

References:

1. Lloyd-Jones, D.M., et al., Defining and setting national goals for cardiovascular health promotion and disease reduction: the American Heart Association's strategic Impact Goal through 2020 and beyond. Circulation, 2010. 121(4): p. 586-613.
2. Lloyd-Jones, D.M., et al., Life's Essential 8: Updating and Enhancing the American Heart Association's Construct of Cardiovascular Health: A Presidential Advisory From the American Heart Association. Circulation, 2022. 146(5): p. e18-e43.
3. Biden-Harris Administration, National Strategy on Hunger, Nutrition, and Health. 2022.

Workshop Chairs:
Kelley Pettee Gabriel, The University of Alabama at Birmingham
Linda Van Horn, Northwestern University
Marie-Pierre St-Onge, Columbia University Irving Medical Center

Workshop Members:
Maya Vadiveloo, The University of Rhode Island, USA
Michael LaMonte, State University of New York at Buffalo, USA
Michael Grandner, The University of Arizona College of Medicine, USA
Amanda Paluch, University of Massachusetts Amherst, USA
Frank Hu, Harvard University, USA
Julio Fernandez-Mendoza, Penn State University College of Medicine, USA
Dorothea Dumuid, University of South Australia, Australia
Sylvia Badon, Kaiser Permanente Northern California, USA
Kristen Knutson, Northwestern University Feinberg School of Medicine, USA
Carol Boushey, University of Hawai’i at Mānoa, USA
Christopher Depner, University of Utah, USA
Jennifer Schrack, John Hopkins University, USA
Katherine Tucker, University of Massachusetts - Lowell, USA
Keith Diaz, Columbia University Medical Center, USA
Erin Dooley, University of Alabama at Birmingham, USA
Diana Thomas, United States Military Academy at West Point, USA
Aiden Doherty, University of Oxford, UK

NIH Staff:

National Heart, Lung, and Blood Institute
Division of Cardiovascular Sciences
Gabriel Anaya, Prevention and Population Sciences Program, Epidemiology Branch
Yuling Hong, Prevention and Population Sciences Program, Epidemiology Branch
Jared P. Reis, Prevention and Population Sciences Program, Epidemiology Branch
Alison Brown, Prevention and Population Sciences Program, Clinical Applications and Prevention Branch
Charlotte A. Pratt, Prevention and Population Sciences Program, Clinical Applications and Prevention Branch

Division of Lung Disease
Alfonso Alfini, National Center on Sleep Disorders Research

National Cancer Institute
Division of Cancer Control and Population Sciences
Jill Reedy, Epidemiology and Genomics Research Program, Risk Factor Assessment Branch
Kirsten Herrick, Epidemiology and Genomics Research Program, Risk Factor Assessment Branch
Marissa Shams-White, Epidemiology and Genomics Research Program, Risk Factor Assessment Branch
Dana L. Wolff-Hughes, Epidemiology and Genomics Research Program, Risk Factor Assessment Branch

National Institute of Diabetes and Digestive and Kidney Diseases
Division of Digestive Diseases and Nutrition
Mary E. Evans, Special Projects in Nutrition, Obesity, and Digestive Diseases