To maximize the likelihood of health benefits, physical activity should be performed regularly over many years.1 Physical activity interventions have demonstrated moderate success at promoting initial improvements in physical activity, but there is limited evidence that interventions can support sustained behavior change.2 In fact, sustained behavior change has received strikingly less attention as an intervention outcome.3 A reason for this state of affairs is the absence of a clear criterion for physical activity maintenance.4 The development of common indicators of physical activity maintenance is critical to evaluate the success of interventions designed to promote long-term behavior change.

Although physical activity maintenance is conceptually defined as sustained behavioral engagement above a threshold for a duration of time sufficient to improve health or well-being,5 operational definitions vary widely across studies. The World Health Organization Guidelines on Physical Activity recommend a threshold of at least 150 minutes per week of moderate or 75 minutes per week of vigorous-intensity aerobic activity6; however, studies have also used person-specific thresholds based on personal goals or relative change in physical activity volume.4 Likewise, the duration of time used to define physical activity maintenance has ranged from 3 to 24 months.4 Furthermore, there is not clear agreement regarding the duration of a temporary deviation below the threshold that should be allowed during maintenance. Deviations ranging from a single session to 3 months have been used.4 These inconsistencies make it difficult to compare findings across studies and establish benchmarks to evaluate interventions to promote physical activity maintenance.

A significant challenge to standardizing indicators of physical activity maintenance is the discontinuous manner by which long-term intervention outcomes are assessed. Typically, follow-up assessments use self-report questionnaires of current physical activity level at single time points occurring 6, 12, or 24 months after the conclusion of an intervention. This strategy fails to provide any information about behavior patterns leading up to, and after, the follow-up assessment. A noteworthy improvement upon this approach is the use of intensive longitudinal assessments that capture continuous physical activity patterns across months or years using passive wearable motion sensors, such as accelerometers. Intensive longitudinal accelerometer assessment after the conclusion of interventions may serve as an improved criterion for evaluating long-term behavior maintenance success.

To help the field move toward consensus, an evidence base testing predictive, construct, and external validity of longitudinal accelerometer indicators of physical activity maintenance needs to be built. Focusing on threshold (eg, level or volume of physical activity required), duration (eg, days required above the threshold), and deviation allowance (eg, amount and days allowed below the threshold) will enable faster progress toward consensus. Following, we propose a framework for evaluating the validity of these indicators.

Step 1: Predictive Validity

Compare how different values of longitudinal accelerometer indicators predict future behavior engagement and health benefits. The first step is evaluating whether accelerometer indicators predict the intended outcomes, which are sustained behavior engagement and health benefits in the future.5 Predictive validity can be tested by comparing metrics such as receiver operator characteristic curves7 that evaluate the predictive performance of intended outcomes (eg, sustained behavior engagement for the next month, attaining a personal physical activity goal) across different values of indicators (eg, after an initial period of engagement of 1–30 d). Receiver operator characteristic curves can generate values of sensitivity (ie, true positive rate) and specificity (ie, true negative rate) for the likelihood of values over (or under) an indicator threshold predicting an intended outcome (or not). This type of analysis, for example, could indicate the minimum number of days required to be initially above the threshold to maximize the chance of sustained physical activity in the future. A similar approach could identify the maximum number of days a person’s physical activity can fall below the threshold before it undermines their chance of resuming their level of physical activity. Different combinations of indicators and outcomes can be tested, and outcomes can be behavioral (as described earlier) or health related (eg, lowering cholesterol or hbA1c levels).

Step 2: Construct Validity

Evaluate the degree to which longitudinal accelerometer indicators are associated with theory-based psychosocial mechanisms of physical activity maintenance. The next step involves testing whether proposed psychosocial mechanisms underlying physical activity maintenance (eg, habits,8 automatic processes,9 and satisfaction with outcomes10) are associated with the accelerometer indicators identified in step 1. Construct validity can be tested by classifying periods of activity in a study as maintenance (vs not maintenance) based on values of longitudinal accelerometer indicators. The likelihood of psychosocial mechanisms measured during (if captured through ecological momentary assessment) or prior to (if captured through retrospective questionnaires) being associated with maintenance or not maintenance classifications can also be evaluated through receiver operator characteristic curves as described earlier.

Step 3: External Validity

Determine whether values of accelerometer indicators are generalizable across different populations, devices, and algorithms. The last step consists of testing whether the values obtained in the previous steps operate similarly across a range of populations varying in sex, gender, age, race/ethnicity, socioeconomic status, weight status, fitness level, able-bodied status, and country. It will also be important to determine whether values of accelerometer indicators are comparable across accelerometer devices and data processing algorithms.

Conclusion

Widespread application of a common framework to evaluate longitudinal accelerometer-based indicators is necessary to develop standardized operational definitions of physical activity maintenance. Once a sufficient evidence base on longitudinal accelerometer indicators of physical activity maintenance is available across diverse samples and populations, the research community can convene to decide whether universal indicators can be recommended. This process may be cyclical because growing an evidence base informs consensus definitions that can guide future research, which, in turn, generates evidence that either affirms or guides modifications to the consensus definitions. After a consensus definition is reached, studies will be able to determine who will reach maintenance and when maintenance will occur as well as who will temporarily or permanently fall out of maintenance and when these slips and lapses will occur. Answers to these questions will form the basis of interventions, policies, and guidelines to promote long-term physical activity maintenance in the general population.

Acknowledgments

This work was supported by a grant (U01HL146327) from the National Heart, Lung, and Blood Institute, National Institutes of Health.

References

  • 1.

    Shortreed SM, Peeters A, Forbes AB. Estimating the effect of long-term physical activity on cardiovascular disease and mortality: evidence from the Framingham Heart Study. Heart. 2013;99(9):649654. PubMed ID: 23474622 doi:

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 2.

    Conn VS, Hafdahl AR, Mehr DR. Interventions to increase physical activity among healthy adults: meta-analysis of outcomes. Am J Public Health. 2011;101(4):751758. PubMed ID: 21330590 doi:

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 3.

    Sheeran P, Wright CE, Listrom O, Klein WM, Rothman AJ. Which intervention strategies promote the adoption and maintenance of physical activity? Evidence from behavioral trials with cancer survivors. Ann Behav Med. 2023;57(9):708721. PubMed ID: 37078961 doi:

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 4.

    Dunton GF, Leventhal AM, Rebar AL, Gardner B, Intille SS, Rothman AJ. Towards consensus in conceptualizing and operationalizing physical activity maintenance. Psychol Sport Exerc. 2022;61:102214. doi:

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 5.

    Seymour RB, Hughes SL, Ory MG, et al. A lexicon for measuring maintenance of behavior change. Am J Health Behav. 2010;34(6):660668. PubMed ID: 20604692 doi:

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 6.

    World Health Organization. WHO Guidelines on Physical Activity and Sedentary Behaviour: At a Glance. 2020.

  • 7.

    Mandrekar JN. Receiver operating characteristic curve in diagnostic test assessment. J Thorac Oncol. 2010;5(9):13151316. PubMed ID: 20736804 doi:

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 8.

    Hagger MS. Habit and physical activity: Theoretical advances, practical implications, and agenda for future research. Psychol Sport Exerc. 2019;42:118129. doi:

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 9.

    Rebar AL, Dimmock JA, Jackson B, et al. A systematic review of the effects of non-conscious regulatory processes in physical activity. Health Psychol Rev. 2016;10(4):395407. PubMed ID: 27118430 doi:

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 10.

    Williams DM, Lewis BA, Dunsiger S, et al. Comparing psychosocial predictors of physical activity adoption and maintenance. Ann Behav Med. 2008;36(2):186194. PubMed ID: 18777124 doi:

    • Crossref
    • Search Google Scholar
    • Export Citation