This commentary is based on a keynote presentation at the 2022 International Conference on Ambulatory Monitoring of Physical Activity and Movement (ICAMPAM). My wish is to provide historical context for newer researchers and to take the opportunity to propose suggestions to move forward. While I have no financial conflicts to disclose, the views are my own, so in an epidemiological context, this commentary is subject to being selective and potentially biased. I will describe several aspects of the first fielding of accelerometer-based devices in the National Health and Nutrition Examination Survey (NHANES), highlight issues raised in past meetings related to device-based measures, and provide comments on the current situation for population measures with devices. However, I will start with some personal background.
I began my professional career at the Centers for Disease Control and Prevention as an Epidemic Intelligence Service Officer with the National Center for Health Statistics (NCHS). NCHS had a plastic ruler give-away for survey participants with the phrase, “When We Measure, We Know Better.” That slogan has been a guiding principle for me throughout my career. After the Epidemic Intelligence Service postdoc, I was employed at NCHS to support NHANES. Among my first tasks were redesigning the physical activity questionnaire and developing a treadmill cardiovascular fitness test protocol for the upcoming “continuous NHANES” scheduled to begin in 1999 when the survey would be conducted with ongoing sequential 2-year cross-sectional samples.
To help determine best approaches to accomplish these tasks, I reached out to experts like Carl Caspersen, known to many as the source for the definition of physical activity in early paragraphs of your papers (Caspersen et al., 1985), Steve Blair, Russ Pate, and Bill Haskell. I especially want to encourage early stage researchers to not be shy and to reach out to colleagues outside your lab group, even “famous” ones. I cannot guarantee that others will be receptive to sharing ideas, but I am confident that you will not get their ideas without asking.
The Accelerometer Component in NHANES 2003–2006
The genesis of using accelerometers in NHANES is another example of the importance of dialog. After moving to the National Cancer Institute (NCI), colleagues from NCHS contacted me to ask if I knew of a better way to measure physical activity among youth, as they had been hearing concerns about the existing NHANES questions. I had attended a conference at the Cooper Institute where Jim Sallis presented work with Brian Saelens to summarize the status, limitations, and future directions for assessment of physical activity by self-report (Sallis & Saelens, 2000), so it seemed that Jim would be a good resource.
When Jim and I spoke, he told me, “It’s time to put devices in the Survey.” Jim had been using Actigraph accelerometers in relatively large deployments with the Neighborhood Quality of Life Study, so he was convinced of the feasibility. The timing was right. NCI was able and willing to financially support the effort. My supervisors at multiple levels, from the Risk Factor Monitoring and Methods Branch to the Applied Research Program to the Division of Cancer Control and Population Sciences, all agreed that implementing device measures of physical activity in the NHANES was consistent with the branch mission to improve assessment methods and to support population surveillance of cancer risk factors.
As the planning and data collection proceeded, we needed to make some decisions, such as what cut points to use to define physical activity intensity. The challenge was that many cut points had been proposed and published that varied depending upon the study samples and activities included in the validation studies. A small group at NCI decided that a reasonable approach would be to take a weighted average of the existing ambulatory cut points because the device was waist worn, which primarily captured locomotor activity. Those became the Troiano cut points used in the first publication of the data, Physical Activity in the United States Measured by Accelerometer (Troiano et al., 2008), and fortuitously adopted and named by many subsequent studies.
The NHANES accelerometer data were a rich resource for the research community. By December 2011, about 3 years after the data were released, more than 50 papers had been published using the publicly available data with a variety of associated end points (Tudor-Locke et al., 2012). The initial 2008 paper was highly influential, which was particularly rewarding for an extramural government researcher. Citation estimates vary by database, but as of September 2022, Scopus, Web of Science, and Google Scholar all currently show the paper with more than 5,000 citations. Varela et al. (2018) published a structured literature review and citation network analysis that stated the 2008 paper was among the five most highly cited articles in field of physical activity and health research.
Key Workshops and Relevant Outcomes
I want to highlight three workshops that influenced how accelerometers came to be used in large studies. These were also workshops in which I was involved, influencing my selection among the many accelerometer-focused meetings and workshops over the years. Somewhat surprisingly, several speakers at ICAMPAM 2022 referenced these same workshops. These were the workshops: Objective Monitoring of Physical Activity: Closing the Gaps in the Science of Accelerometry, held at the University of North Carolina in 2004 (Ward, Evenson, et al., 2005); Objective Measurement of Physical Activity: Best Practices and Future Directions, held at NIH in 2009 (Freedson, Bowles, Troiano, & Haskell, 2012); and Measurement of Active and Sedentary Behaviors: Closing the Gaps in Self-Report Methods, held at NIH in 2010 (Bowles, 2012).
The prevailing perspective regarding device use in large studies around the time of the 2004 workshop is reflected in a quote from a 1999 Cooper Institute Conference on physical activity measurement: “… objective motion sensors were not practical for large scale studies because of high cost, uncertain reliability, and difficulties in the interpretation of data” (Troiano, 2005). However, by 2004, much progress had occurred and enthusiasm for devices was high. This momentum has only continued. Figure 1a is from an overview article in the supplement that summarized the 2004 meeting (Troiano, 2005). The plot shows the number of publications identified on Scopus.com with titles or abstracts that contained the terms “physical activity” or “exercise” along with any terms that included “acceleromet*.” At that time, the peak was just less than 90 articles in 2003 and 2004. Figure 1b is an updated version from 2021. Note the location of the peak article count highlighted in 2005. The negative assessment from 1999 was apparently overcome.
—(a) Trends in accelerometer articles (2005) from “A Timely Meeting: Objective Measurement of Physical Activity,” by R.P. Troiano, 2005, Medicine & Science in Sports & Exercise, 37(Suppl. 11), S487–S489 (https://doi.org/10.1249/01.mss.0000185473.32846.c3). Copyright 2005 by American College of Sports Medicine. (b) Trends in accelerometer articles (2021).
Citation: Journal for the Measurement of Physical Behaviour 6, 1; 10.1123/jmpb.2022-0038
—(a) Trends in accelerometer articles (2005) from “A Timely Meeting: Objective Measurement of Physical Activity,” by R.P. Troiano, 2005, Medicine & Science in Sports & Exercise, 37(Suppl. 11), S487–S489 (https://doi.org/10.1249/01.mss.0000185473.32846.c3). Copyright 2005 by American College of Sports Medicine. (b) Trends in accelerometer articles (2021).
Citation: Journal for the Measurement of Physical Behaviour 6, 1; 10.1123/jmpb.2022-0038
—(a) Trends in accelerometer articles (2005) from “A Timely Meeting: Objective Measurement of Physical Activity,” by R.P. Troiano, 2005, Medicine & Science in Sports & Exercise, 37(Suppl. 11), S487–S489 (https://doi.org/10.1249/01.mss.0000185473.32846.c3). Copyright 2005 by American College of Sports Medicine. (b) Trends in accelerometer articles (2021).
Citation: Journal for the Measurement of Physical Behaviour 6, 1; 10.1123/jmpb.2022-0038
It is interesting to look back at prior recommendations considering current practices. Here are some recommendations from the 2004 workshop (Ward, Evenson, et al., 2005):
- •Encourage open-source technology to enhance comparability.
- •Tap into cell phone technology.
- •Develop a core set of activities to employ in calibration studies.
These recommendations have clearly been accepted and have led to evolution in physical activity research. However, there is one more recommendation to highlight from the 2004 meeting:
- •Avoid wrist placement.
Not everything goes according to plan. In this case, the explosion of wrist-worn consumer devices and greater acceptability of the wrist over waist placement prevailed. The accelerometer component in NHANES 2011–2014 used wrist placement with 24-hr wear, which improved wear compliance and provided measures related to sleep (Belcher et al., 2021).
We find a somewhat more detailed set of recommendations from the 2009 workshop, including these (Freedson, Bowles, Troiano, & Haskell, 2012):
- •“Monitor data should be collected and saved as raw signals with post-processing used for data transformation.”
- •“Organiz[e] multi-disciplinary teams … to develop tools, process data, and perform calibration/validation studies ….”
- •“… discontinue development and use of cutpoint methods to define intensity categories … .”
Device manufacturers were present at the 2009 workshop and responded to the messages they heard. All research-focused accelerometer devices now provide access to unprocessed signal data. The call for multidisciplinary teams was repeated several times at ICAMPAM 2022. The need for collaboration has only increased as algorithm development becomes more sophisticated, open access data processing tools are developed, and physical activity researchers become interested in the 24-hr activity cycle, which includes sleep.
The third recommendation bears special mention and additional comment. The recommendation to abandon cut point methods was voiced by Patty Freedson, the first person to have accelerometer cut points that bore her name, stemming from calibration studies done in her lab at University of Massachusetts, Amherst, in the late 1990s (Freedson et al., 1998). Patty had an impressive impact on our field by being the academic mentor to many influential researchers, including several presenters at ICAMPAM 2022, Ed Melanson and Kate Lyden, the host cochairs of the meeting, and Sarah Kozey Keadle, the scientific planning committee colead.
The final meeting to highlight is the 2010 NCI workshop on self-report methods. You may wonder why discussion of self-report methods would be relevant to a commentary about applications of accelerometer measures. While not focusing on devices, this workshop contributed important advances that continue to affect interpretation of device-based measures. The workshop provided a framework for physical activity as a complex and multidimensional behavior that clarified how movement that is detected by devices relates to the behaviors that generate it (Figure 2, Pettee Gabriel et al., 2012). It also included a call for standardized, precise, consistent terminology, and definitions. One specific example was to replace the judgment-laden terms of objective and subjective measures with the more neutral terms of device-based and report-based measures (Troiano et al., 2012). And perhaps most importantly, discussions at the workshop and the resulting publications recognized how the distinct measurements obtained by accelerometer-based devices and self-report were not better or worse representations of a single construct but provided measures of complementary aspects of the underlying physical activity behavior (Troiano et al., 2012). Importantly, it is easy to see why these complementary measures are not very likely to be quantitatively identical.
—Conceptual framework from “Framework for Physical Activity as a Complex and Multidimensional Behavior,” by K.K. Pettee Gabriel and J.R. Morrow Jr., 2012, Journal of Physical Activity & Health, 9(Suppl. 1), S11–S18 (https://doi.org/10.1123/jpah.9.s1.s11). Copyright 2012 by Human Kinetics.
Citation: Journal for the Measurement of Physical Behaviour 6, 1; 10.1123/jmpb.2022-0038
—Conceptual framework from “Framework for Physical Activity as a Complex and Multidimensional Behavior,” by K.K. Pettee Gabriel and J.R. Morrow Jr., 2012, Journal of Physical Activity & Health, 9(Suppl. 1), S11–S18 (https://doi.org/10.1123/jpah.9.s1.s11). Copyright 2012 by Human Kinetics.
Citation: Journal for the Measurement of Physical Behaviour 6, 1; 10.1123/jmpb.2022-0038
—Conceptual framework from “Framework for Physical Activity as a Complex and Multidimensional Behavior,” by K.K. Pettee Gabriel and J.R. Morrow Jr., 2012, Journal of Physical Activity & Health, 9(Suppl. 1), S11–S18 (https://doi.org/10.1123/jpah.9.s1.s11). Copyright 2012 by Human Kinetics.
Citation: Journal for the Measurement of Physical Behaviour 6, 1; 10.1123/jmpb.2022-0038
I-Min Lee presented the opening keynote at ICAMPAM 2022, the inaugural Hans Bussman Lecture, “Maximizing the utility and comparability of accelerometer data from large-scale observational epidemiologic studies.” During her presentation, I-Min provided a personal example of her likelihood of reporting that she played tennis for 2 hr and noted that it was highly unlikely that a device would register 2 hr of movement during that period. Neither measure would be wrong. I-Min and other questionnaire respondents report behaviors as they perceive and understand them while a device detects moment-by-moment body movement that occurs during that reported behavior. This example also highlights the unique contribution of a report measure. While a device may provide an accurate measure of body movement across shorter or longer periods, a researcher is unlikely to know that I-Min played tennis without a report. Reports can provide information about context and purpose that are not available from device measures alone.
To transition from a historical look back to thoughts about moving forward, I highlight a couple of good ideas from previous ICAMPAM conferences that relate to noted workshop recommendations. At ICAMPAM 2013 in Amherst, James McClain and I proposed a Sensor Methods Collaboratory. The intent was to provide an online discussion space that would foster cross-disciplinary collaboration toward algorithm development to process the newly available unprocessed signal data (2009 workshop recommendations). We were encouraged that more than 100 researchers signed up to be part of the effort. At ICAMPAM 2017 in Bethesda, Sebastien Chastin presented a proposal for a behavioral taxonomy called AlphaBET, consistent with the call for common terminology from the 2010 workshop. Unfortunately, neither of these efforts was fruitful. Without active encouragement or incentives, little engagement occurred with the Sensor Methods Collaboratory. Research groups developed independent approaches to data processing. As far as I can tell, no further progress has occurred with AlphaBET. However, to provide an optimistic note on the goals of these efforts, I note two successes. Based on the successful International Children’s Accelerometer Database that pooled data from ActiGraph devices for youth (Sherar et al., 2011), Katrien Wijndaele and Genevieve Healy led an effort using a Delphi process to address harmonizing accelerometer data from adults. The project documented data from 76 studies with more than 275,000 adults and provided consensus recommendations on how to facilitate data harmonization (Wijndaele et al., 2015). More recently, a team of researchers from multiple universities collaborated toward a consensus approach to labeling behaviors from direct video observation. This is an example of fostering collaboration to achieve common terminology that is required to pool data for potential device-based behavior identification algorithm development. A publication will hopefully soon be forthcoming.
Current Situation
In looking back over approximately the past two decades, research with accelerometer devices has achieved tremendous progress, but much work is still required. Because device manufacturers heeded the call, we now have raw data readily available, at least from research devices. Raw data provided the potential to move away from the plethora of proprietary cut point thresholds that limited comparability of results across studies. Interestingly, at ICAMPAM 2022, the availability of nonproprietary count metric derivation and disclosure of previous proprietary methods led to reconsideration of cut point metrics. I hope that if the pendulum swings back toward cut points, a consensus approach will arise rather than a variety of competing approaches.
With increasing computational power and engagement with researchers from other fields, rapid advances have occurred toward sophisticated high-resolution signal processing algorithms to identify physical activity-related behaviors. However, consensus on how to process the signals and interpret the resulting metrics is lacking. One issue is the desired or required degree of resolution in identifying behaviors. Is it sufficient to know if someone is inactive or perhaps specifically sitting, or do we need to know whether they are sitting at a computer working rather than sitting at a table eating? A means to translate the rich signal data from devices into behavioral metrics that the public can understand and that will support population surveillance remains elusive. We also still have the challenge of classifying activity intensity.
One area where the issue of activity intensity arises is in applying device measures to population surveillance to monitor compliance with physical activity guidelines (Troiano et al., 2020). Current U.S. and World Health Organization guidelines (Bull et al., 2020; Piercy et al., 2018) mostly rely on epidemiology based on reports with associated qualifiers like moderate or vigorous intensity. Recent contributions from device data have allowed current guidelines to evolve by recognizing that a primary health driver is total physical activity volume. These observations led to the elimination of bout criteria in guidelines and the increasing recognition that light-intensity activity can contribute to health enhancing physical activity. The challenge is to transition from quantitative targets based on questionnaires with bout requirements and subjective intensity descriptors to targets that incorporate more precise device data, while retaining the ability to communicate the targets meaningfully to the public and monitor compliance practically in populations around the world.
The evolution of device-based measures of physical activity and the potential to apply them to cohort studies and even population surveillance can be described as an area where amazing advances and frustrating inertia coexist. Advances are so rapid that technology and processes that are current at the beginning of a study have changed by the end of the study. On the other hand, challenges that were discussed in workshops more than 15 years ago remain unresolved issues. A contributor to the lack of progress is the incentive structure of grant funding and professional promotion within the academic research culture that promotes entrepreneurship over collaboration.
Collaboration Is Key
Evidence of collaboration is increasing. Publication author lists demonstrate greater engagement across disciplines as well as across institutions. Such collaborations need to be fostered and expanded to achieve the consensus required to truly advance the application of devices to public health. Challenges where collaboration is needed include the ability to pool data for enhanced metric and algorithm development and testing, which requires accepted common terminology, protocol harmonization, and perhaps even standards development. Ideally for public health applications, our research field would be able to provide a consensus recommendation on physical activity metrics and how to translate those metrics for population recommendations. It is important to avoid the “cut point conundrum” of some years ago as new approaches are developed. Multiple sets of competing proposed terminology or standards are less valuable than consensus approaches.
There is likely little argument that multidisciplinary teams and collaboration are desirable to achieve consensus goals, but the means to implement mechanisms and incentives are less clear. The major levers of publication requirements (e.g., clinical trial registration, required review process evaluations) and incentives for professional advancement are resistant to change. We know from experience that gaining wide acceptance among competing approaches from different individuals or institutions is challenging. Prescription of standards from funding organizations also has only limited success. However, the International Society for the Measurement of Physical Behaviour (ISMPB) could be well situated to support consensus building. The Society has broad member engagement within ISMPB and across partner societies. ISMPB also has growing means to foster communication. The Society could potentially serve the role of convener to promote efforts such as consensus guidance for device signals or metrics and associated data to be pooled with relevant standards for protocol description and data labeling. If accomplished, these consensus standards could provide a foundation to partner with organizations that have resources (financial or in-kind) to provide an accessible data repository and incentives for opportunities such as a data challenge.
My retrospective view along with multiple presentations and discussions at ICAMPAM 2022 (and previous meetings) highlight the value of multidisciplinary collaboration for physical activity device-based research. I hope that early career researchers are motivated to reach out to investigators outside their lab group or academic institution to expand ideas and collaborations. Meetings like ICAMPAM provide excellent opportunities to foster those expanded networks. While the path forward may not be obvious, I hope that we as a research field can utilize ISMPB to provide opportunities to achieve consensus approaches to device-based metrics and their applications for population surveillance and health improvement.
Acknowledgments
I had the privilege of participating in influential aspects of the evolution of device-based measures of physical activity and the application to public health. Many colleagues contributed to these opportunities, and only a small proportion is represented as coauthors on publications. Others were supervisors at multiple levels who supported or facilitated efforts, academic and government researchers in the United States and abroad who participated in formal discussions or informal conversations, and administrative staff who helped me negotiate the many bureaucratic aspects of travel, contracts, and interagency agreements. They are too many to name, but their contributions are greatly appreciated.
References
Belcher, B.R., Wolff-Hughes, D.L., Dooley, E.E., Staudenmayer, J., Berrigan, D., Eberhardt, M.S., & Troiano, R.P. (2021). US population-referenced percentiles for wrist-worn accelerometer-derived activity. Medicine & Science in Sports & Exercise, 53(11), 2455–2464. https://doi.org/10.1249/mss.0000000000002726
Bowles, H.R. (2012). Measurement of active and sedentary behaviors: Closing the gaps in self-report methods. Journal of Physical Activity & Health, 9(Suppl. 1), S1–S4.
Bull, F.C., Al-Ansari, S.S., Biddle, S., Borodulin, K., Buman, M.P., Cardon, G., Carty, C., Chaput, J.-P., Chastin, S., Chou, R., Dempsey, P.C., DiPietro, L., Ekelund, U., Firth, J., Friedenreich, C.M., Garcia, L., Gichu, M., Jago, R., Katzmarzyk, P.T., … Willumsen, J.F. (2020). WHO 2020 Guidelines on physical activity and sedentary behaviour. British Journal of Sports Medicine, 54(24), 1451–1462. https://doi.org/10.1136/bjsports-2020-102955
Caspersen, C.J., Powell, K.E., & Christenson, G.M. (1985). Physical activity, exercise, and physical fitness: Definitions and distinctions for health-related research. Public Health Report, 100(2), 126–131.
Freedson, P., Bowles, H.R., Troiano, R. (2012). Objective measurement of physical activity: Best practices and future directions. Medicine & Science in Sports & Exercise, 44, S1–S89.
Freedson, P., Bowles, H.R., Troiano, R., & Haskell, W. (2012). Assessment of physical activity using wearable monitors: Recommendations for monitor calibration and use in the field. Medicine & Science in Sports & Exercise, 44(Suppl. 1), S1–S4. https://doi.org/10.1249/MSS.0b013e3182399b7e
Freedson, P.S., Melanson, E., & Sirard, J. (1998). Calibration of the computer science and applications, Inc. accelerometer. Medicine & Science in Sports & Exercise, 30(5), 777–781. https://doi.org/10.1097/00005768-199805000-00021
Pettee Gabriel, K.K., Morrow, J.R., Jr., & Woolsey, A.L. (2012). Framework for physical activity as a complex and multidimensional behavior. Journal of Physical Activity & Health, 9 (Suppl. 1), S11–S18. https://doi.org/10.1123/jpah.9.s1.s11
Piercy, K.L., Troiano, R.P., Ballard, R.M., Carlson, S.A., Fulton, J.E., Galuska, D.A., George, S.M., & Olson, R.D. (2018). The physical activity guidelines for Americans. JAMA, 320(19), 2020–2028. https://doi.org/10.1001/jama.2018.14854
Sallis, J.F., & Saelens, B.E. (2000). Assessment of physical activity by self-report: Status, limitations, and future directions. Research Quarterly for Exercise Sport, 71(Suppl. 2), S1–S14. https://doi.org/10.1080/02701367.2000.11082780
Sherar, L.B., Griew, P., Esliger, D.W., Cooper, A.R., Ekelund, U., Judge, K., & Riddoch, C. (2011). International children’s accelerometry database (ICAD): Design and methods. BMC Public Health, 11, 485. https://doi.org/10.1186/1471-2458-11-485
Troiano, R.P. (2005). A timely meeting: Objective measurement of physical activity. Medicine & Science in Sports & Exercise, 37(Suppl. 11), S487–S489. https://doi.org/10.1249/01.mss.0000185473.32846.c3
Troiano, R.P., Berrigan, D., Dodd, K.W., Masse, L.C., Tilert, T., & McDowell, M. (2008). Physical activity in the United States measured by accelerometer. Medicine & Science in Sports & Exercise, 40(1), 181–188. https://doi.org/10.1249/mss.0b013e31815a51b3
Troiano, R.P., Pettee Gabriel, K.K., Welk, G.J., Owen, N., & Sternfeld, B. (2012). Reported physical activity and sedentary behavior: Why do you ask? Journal of Physical Activity & Health, 9(Suppl. 1), S68–S75. https://doi.org/10.1123/jpah.9.s1.s68
Troiano, R.P., Stamatakis, E., & Bull, F.C. (2020). How can global physical activity surveillance adapt to evolving physical activity guidelines? Needs, challenges and future directions. British Journal of Sports Medicine, 54(24), 1468–1473. https://doi.org/10.1136/bjsports-2020-102621
Tudor-Locke, C., Camhi, S.M., & Troiano, R.P. (2012). A catalog of rules, variables, and definitions applied to accelerometer data in the National Health and Nutrition Examination Survey, 2003–2006. Preventing Chronic Disease, 9, E113. https://doi.org/10.5888/pcd9.110332
Varela, A.R., Pratt, M., Harris, J., Lecy, J., Salvo, D., Brownson, R.C., & Hallal, P.C. (2018). Mapping the historical development of physical activity and health research: A structured literature review and citation network analysis. Preventive Medicine, 111, 466–472. https://doi.org/10.1016/j.ypmed.2017.10.020
Ward, D.S., Evenson, K.R., Vaughn, A., Rodgers, A.B., & Troiano, R.P. (2005). Accelerometer use in physical activity: Best practices and research recommendations. Medicine & Science in Sports & Exercise, 37(Suppl. 11), S582–S588. https://doi.org/10.1249/01.mss.0000185292.71933.91
Wijndaele, K., Westgate, K., Stephens, S.K., Blair, S.N., Bull, F.C., Chastin, S.F., Dunstan, D.W., Ekelund, U., Esliger, D.W., Freedson, P.S., Granat, M.H., Matthews, C.E., Owen, N., Rowlands, A.V., Sherar, L.B., Tremblay, M.S., Troiano, R.P., Brage, S., & Healy, G.N. (2015). Utilization and harmonization of adult accelerometry data: Review and expert consensus. Medicine & Science in Sports & Exercise, 47(10), 2129–2139. https://doi.org/10.1249/MSS.0000000000000661
