Accelerometry in persons with multiple sclerosis: Measurement of physical activity or walking mobility?

https://doi.org/10.1016/j.jns.2009.12.021Get rights and content

Abstract

Objective

Motion sensors such as accelerometers have been recognized as an ideal measure of physical activity in persons with MS. This study examined the hypothesis that accelerometer movement counts represent a measure of both physical activity and walking mobility in individuals with MS.

Methods

The sample included 269 individuals with a definite diagnosis of relapsing–remitting MS who completed the Godin Leisure-Time Exercise Questionnaire (GLTEQ), International Physical Activity Questionnaire (IPAQ), Multiple Sclerosis Walking Scale-12 (MSWS-12), Patient Determined Disease Steps (PDDS), and then wore an ActiGraph accelerometer for 7 days. The data were analyzed using bivariate correlation and confirmatory factor analysis.

Results

The results indicated that (a) the GLTEQ and IPAQ scores were strongly correlated and loaded significantly on a physical activity latent variable, (b) the MSWS-12 and PDDS scores strongly correlated and loaded significantly on a walking mobility latent variable, and (c) the accelerometer movement counts correlated similarly with the scores from the four self-report questionnaires and cross-loaded on both physical activity and walking mobility latent variables.

Conclusion

Our data suggest that accelerometers are measuring both physical activity and walking mobility in persons with MS, whereas self-report instruments are measuring either physical activity or walking mobility in this population.

Introduction

Multiple sclerosis (MS) is a chronic, neurodegenerative disease characterized by inflammation in the central nervous system (CNS) [1] and demyelination and destruction of the axons [2]. This prevalent disease is associated with worsening of symptoms, function, disability, and quality of life across time [3]. There is considerable evidence for positive effects of physical activity on the aforementioned undesirable consequences of MS [4], [5], [6], but persons with MS are largely inactive and sedentary [7], [8].

The veracity of conclusions on the benefits and rates of physical activity in persons with MS relies upon the psychometric integrity of the measures. Physical activity, defined as bodily movement produced by skeletal muscle contraction resulting in increased energy expenditure [9], is most often measured by motion sensors (e.g., accelerometers) and self-report questionnaires (e.g., the Godin Leisure-Time Exercise Questionnaire) [10], [11]. There is evidence that scores from both motion sensors and self-report questionnaires provide valid [12] and reliable [13] measures of physical activity in persons with MS. We further note that accelerometers are seemingly the preferred method of measuring physical activity in persons with MS because this device provides an objective marker of free-living physical activity that is not susceptible to recall bias [7], [14], [15].

Accelerometers, such as the ActiGraph model 7164, typically contain a single, vertical axis piezoelectric bender element that generates an electrical signal proportionate to the force acting on it during movement. The resulting electrical signal is digitized by an analog-to-digital converter and numerically integrated over a pre-programmed epoch interval into movement counts. The movement counts are a summation of accelerations measured during a cycle period and represent a quantitative measure that is linearly related to the intensity of a participant's movement. Variation in the accelerometer signal has two primary contributors, physical activity and walking mobility. Indeed, the accelerometer movement counts have been associated with energy expenditure during ambulatory physical activity [16] and walking mobility [17]. We further note that researchers have considered an accelerometer as a referent standard of both physical activity [7] and walking mobility [18] in MS. This joint contribution of physical activity and walking mobility into the raw accelerometer signal might present a measurement problem in cohorts of persons with MS (i.e., there might not be a straight-forward, logical interpretation of movement counts from an accelerometer). One possible illustration of this problem would be that an individual can be very physically active, but gait abnormalities (e.g., shuffling) or reliance upon an assistive device during ambulation (e.g., cane or walker) can result in minimal vertical acceleration along the sensitive axis of an accelerometer, thereby yielding a reduction in movement counts. Such a situation might result in the erroneous interpretation that an individual is physically inactive when really the accelerometer signal is reflecting mobility impairment.

Preliminary evidence of this measurement problem was provided in a study of relationships among scores from an ActiGraph accelerometer and self-report measures of walking mobility and physical activity in individuals with MS (N = 82) [19]. The primary results of the study were (a) strong correlations between scores from two self-report physical activity measures (r = .71) and between scores from two self-report walking mobility measures (r = .82), (b) moderate correlations across scores from self-report physical activity and walking mobility measures (e.g., r = −.45), and (c) similarly sized correlations between accelerometer counts and scores from self-report physical activity (e.g., r = .64) and walking mobility (e.g., r = −.70) measures. Such findings suggest that ActiGraph accelerometers provide a measure of both physical activity and walking mobility in individuals with MS [19]. The two noteworthy limitations of that study were a relatively small sample of individuals with MS and data-driven statistical analyses (i.e., Pearson product-moment correlations).

One statistical platform for directly testing the measurement problem with accelerometry involves confirmatory factor analysis. Confirmatory factor analysis is part of a family of statistical methodologies, termed structural equation modeling, that take a hypothesis-driven approach for analyzing data [20], [21]. This technique is appropriate for testing and comparing the fit of measurement models that describe the pattern of associations among a set of observed variables or indicators. The measurement models are derived in advance and represent the latent variables (i.e., theoretical constructs that cannot be measured directly) that underlie the pattern of covariance among a set of observations. Overall, confirmatory factor analysis focuses on how, and the degree to which, observed variables (e.g., accelerometer counts) are linked with latent variables (e.g., physical activity or walking mobility) [22].

This study involves a further examination of the possibility that accelerometry provides a measure of both physical activity and walking mobility in people with MS. To that end, we included a sample of individuals with MS over three times larger than previous research [19] and tested a series of a priori models for the associations among scores using confirmatory factor analysis. The primary hypothesis we tested with confirmatory factor analysis was that accelerometers measure both physical activity and walking mobility in individuals with MS. To verify this hypothesis, we tested a baseline measurement model of two correlated latent variables (i.e., physical activity and walking mobility) that described the associations between scores from two self-report measures of physical activity and two self-report measures of walking mobility. This analysis confirmed the convergent and divergent validity aspects of the self-report measures of physical activity and walking mobility. We then tested a series of nested measurement models whereby accelerometer movement counts either cross-loaded on both physical activity and walking mobility latent variables or independently loaded on either physical activity or walking mobility latent variables. Cross-loading of accelerometer movement counts on both physical activity and walking mobility latent variables would support the hypothesis of a measurement problem, whereas independent loading would support accelerometer movement counts as either a measure of physical activity or walking mobility.

Section snippets

Participants

The data presented in this paper are from the baseline portion of a longitudinal investigation of symptoms and physical activity behavior in persons with RRMS that was funded by the National Multiple Sclerosis Society (NMSS). The protocol was approved by a University Institutional Review Board and all of the participants were required to read and sign an informed consent document before beginning the study. The sample was recruited through a research advertisement that was posted on the NMSS

Sample characteristics

The sample consisted of 223 women and 46 men who had a mean age of 45.9 years (SD = 9.6). The sample was mostly Caucasian (91%), married (70%), employed (59%), and educated (25% had some college education and 58% were college graduates) with a median annual household income of greater than $40,000 (68%). The mean time since definite diagnosis was 8.8 years (SD = 7.0) and the mean time since onset of first symptoms was 13.3 years (SD = 9.2). The median PDDS score was 2 (range = 0–6) and corresponds with

Discussion

The findings of this study indicated that (a) GLTEQ and IPAQ scores were strongly correlated and loaded significantly on a physical activity latent variable, (b) MSWS-12 and PDDS scores were strongly correlated and loaded significantly on a walking mobility latent variable, and (c) accelerometer movement counts correlated similarly with the scores from all four self-report questionnaires and cross-loaded on both physical activity and walking mobility latent variables. Such results confirm our

Acknowledgement

This investigation was supported by a grant from the National Multiple Sclerosis Society (RG 3926A2/1).

References (29)

  • R.W. Motl et al.

    Physical activity and multiple sclerosis: a meta-analysis

    Mult Scler

    (2005)
  • C. Bouchard et al.
  • R.K. Dishman et al.

    Measurement of physical activity

    Quest

    (2001)
  • C.E. Tudor-Locke et al.

    Challenges and opportunities for measuring physical activity in sedentary adults

    Sports Med

    (2001)
  • Cited by (83)

    • Accelerometer measured physical activity and sedentary time in individuals with multiple sclerosis versus age matched controls: A systematic review and meta-analysis

      2023, Multiple Sclerosis and Related Disorders
      Citation Excerpt :

      Mean step count per day for PwMS is 5896.39 ± 2876.36 and controls 8778.46 ± 2955.56, a difference of 2882 steps per day (SMD 0.959 p < 0.001). Thirteen of the 21 studies reported activity counts, including 645 people, 330 with MS and 315 controls (Blikman et al., 2015; Chung et al., 2016; Fjeldstad et al., 2015; Hale et al., 2008; Ickmans et al., 2014; Kent-Braun et al., 1997; Klassen et al., 2008; Kos et al., 2007; Ranadive et al., 2012; Sandroff et al., 2012; Sandroff and Motl, 2013; Scott et al., 2011; Weikert et al., 2010). Data pooling demonstrates a significant difference between groups with greater activity counts in controls, p < 0.001.

    View all citing articles on Scopus
    View full text