Mind the analytics when studying the benefits of physical activity with accelerometers

A consensus on analytical approaches to investigate the associations of physical activity with health outcomes

The benefits of physical activity have been recently described in the World Health Organization guidelines [1]. Movement sensors have enhanced the research of the physical activity benefits by allowing a detailed analysis of the physical activity patterns throughout the day. Furthermore, they also allow the assessment of other related behaviours, such as sedentary activities and sleep-related outcomes. These behaviours (i.e., physical activity, sedentary time, and sleep) are inter-dependent as they share the 24 hours of every day and are related to health outcomes. Dedicating more time to any of these behaviours would reduce the time in the others as the daily time is constrained to 24 hours. As such, some challenges still require action to get the most from these devices. The challenges appear mainly at two levels:

  • How to reduce the high amount of data that is recorded by an accelerometer to obtain meaningful information on physical activity.
  • How to statistically analyse the data without losing relevant information and considering the inter-relationships between physical activity, sedentary time and sleep behaviours.

The “International Workshop: A focus on statistical methods to analyse accelerometer-measured PA” was held in Granada on October 21st-22nd 2019. This event brought together a panel of researchers to discuss, reach consensus, and provide recommendations about the most-frequently used analytical approaches in the field, and about future research directions in physical behaviour epidemiology. The focus was on modelling physical behaviour constructs (mainly related to PA and SB, although we also included sleep to cover the 24-h continuum) as exposure variables and health indicators as outcomes. In addition to the consensus authors from 9 European countries, the document was reviewed by an international panel composed of 10 experts on the topic from Europe, USA and Australia, which improved and strengthened the present consensus article [2].

Consensus points

After exploring the different point of views of the panel invited to this International Workshop, the discussions led to a five consensus points that were agreed to ensure good practices when investigating the associations of physical activity with health outcomes in epidemiological studies. The consensus points were (see Infographic):

  1. The study of the association between physical behaviours (i.e., PA, SB and sleep) and health should move to a more thorough investigation of the interactions and co-dependencies between different behaviours (or physical activity intensities) and health. Several analytical approaches are provided in this consensus document, although none of them is free from limitations.
  2. We recommend investigating more detailed physical activity intensities than the typically studied (i.e., SB and MVPA). Examples include light physical activity of different intensities or the more fine-grained intensity bands as described in this document.
  3. Public health guidelines on physical behaviours should acknowledge that behaviours are co-dependent and this may affect the guidelines as traditionally understood.
  4. Further investigation in functional data analysis and machine learning is needed concerning the associations of physical behaviours with health.
  5. There is not a gold-standard able to test which analytical approach is the best for a given research question. Thus, we cannot make a strong recommendation on a single analytical approach. Instead, we provide some practical recommendations to select analytical approaches well-suited for a given research question. Triangulation across findings from different analytical approaches is currently the best solution.

Decision tree and future perspectives

Apart from the consensus points, the authors also developed a decision tree to assist researchers with the decision making to analyse accelerometer-measured physical activity data in relation to health outcomes [2]. Furthermore, a number of future perspectives to reach in the field were listed related to the appropriate report of the methods used and interpretation of the findings reached by these methods, and to the motivation to investigate new methods (mainly machine learning approaches).

Authors & Affiliations:

Jairo H. Migueles1,2, Eivind Aadland3, Lars B. Andersen3, Jan Christian Brønd4, Sébastien F. Chastin5,6, Bjørge H. Hansen7,8, Kenn Konstabel9,10,11, Olav M. Kvalheim12, Duncan E. McGregor5,13, Alex V. Rowlands14,15,16, Séverine Sabia17,18, Vincent T. van Hees19,20, Rosemary Walmsley21, Francisco B. Ortega1,22 

Note: Except for the first and last author, contributing authors are listed in alphabetic order.

1 PROFITH “PROmoting FITness and Health through physical activity” Research Group, Sport and Health University Research Institute (iMUDS), Department of Physical Education and Sports, Faculty of Sport Sciences, University of Granada, Granada, Spain.

2 Department of Health, Medicine and Caring Sciences, Linköping University, 581 83, Linköping, Sweden.

3 Faculty of Education, Arts and Sports, Western Norway University of Applied Sciences, Sogndal, NORWAY.

4 Department of Sport Science and Biomechanics, University of Southern Denmark, Odense, Denmark.

5 School of Health and Life Science, Glasgow Caledonian University, Glasgow, UK.

6 Department of Movement and Sport Science, Ghent University, Belgium.

7 Department of Sports Medicine, Norwegian School of Sport Sciences, PO Box 4014, Ullevål Stadion, 0806 Oslo, Norway.

8 Departement of Sport Science and Physical Education, University of Agder, Norway

9 Department of Chronic Diseases, National Institute for Health Development, Hiiu 42, Tallinn, Estonia.

10 School of Natural Sciences and Health, Tallinn University, Tallinn, Estonia.

11 Institute of Psychology, University of Tartu, Tartu, Estonia.

12 Department of Chemistry, University of Bergen, Bergen, Norway.

13 Biomathematics and Statistics Scotland, Edinburgh, UK.

14 Diabetes Research Centre, University of Leicester, Leicester General Hospital, Gwendolen Road, Leicester LE5 4PW, UK.

15 NIHR Leicester Biomedical Research Centre, Leicester General Hospital, Gwendolen Road, Leicester LE5 4PW, UK.

16 Alliance for research in Exercise, Nutrition and Activity (ARENA), Sansom Institute for Health Research, Division of Health Sciences, University of South Australia, Adelaide SA 5001, Australia.

17 Université de Paris, Inserm U1153, Epidemiology of Ageing and Neurodegenerative diseases, 75010 Paris, France.

18 Department of Epidemiology and Public Health, University College London, London, UK.

19 Accelting, Almere, The Netherlands.

20 Amsterdam UMC, Vrije Universiteit Amsterdam, Department of Public and Occupational Health, Amsterdam Public Health research institute, The Netherlands.

21 Nuffield Department of Population Health, University of Oxford, Oxford, OX3 7LF, UK.

22 Department of Biosciences and Nutrition, Karolinska Institutet, Huddinge, Sweden.

Competing interests: None.


  1. Bull F, Saad Al-Ansari S, Biddle S, et al. World Health Organization 2020 Guidelines on Physical Activity and Sedentary Behaviour. Br J Sports Med 2020;:1451–62. doi:10.1136/bjsports-2020-102955
  2. Migueles JH, Aadland E, Andersen LB, et al. GRANADA consensus on analytical approaches to assess associations with accelerometer-determined physical behaviours (physical activity, sedentary behaviour and sleep) in epidemiological studies. Br J Sports Med 2021;:1–9. doi:10.1136/bjsports-2020-103604

(Visited 548 times, 4 visits today)