Physical activity assessment – Pedometry
Pedometers are low-cost motion sensors which are typically worn on a belt or waistband and respond to vertical accelerations of the hip during gait cycles (Welk et al, 2000). Pedometers provide data on steps taken, and therefore, only really measure walking activity. Due to this, they will not capture activities such as cycling, swimming, walking on an incline or weight lifting. Walking however is one of the most common forms of physical activity and pedometers readily measure this.
There are wide variations between pedometer models and this is reflected in their cost. Generally the cost of a pedometer is proportional to its accuracy. Several papers have rigorously reviewed and compared commonly available pedometers (Crouter et al, 2003; Schneider et al, 2003; Tudor-Locke et al, 2006). In one study, eight out of 10 electronic pedometers were considered accurate when recorded steps were compared to actual steps, but three models (Kenz-Lifecorder, New Lifestyles NL-2000 and Yamax Digiwalker SW-701) were noted to be superior over a 400-m track walking test, as they were accurate within ±3% (Schneider et al, 2003). Comparison of pedometers over fixed distances or at a variety of treadmill speeds is not reflective of their performance in free-living conditions. The Yamax SW-200 (YX200) has been used widely in the literature and as a criterion pedometer in a comparison study in free-living conditions (Schneider et al, 2004).
Pedometers used in research must be of research grade and the following specifications have been recommended (Tudor-Locke et al, 2006):
- Sensitivity threshold of 0.35Gs;
- Detection of ± 1 step error on the 20 step test (i.e. within 5%);
- Detection of ± 1% error most of the time during treadmill walking at 80m.min-1 (3 miles.hr-1);
- Detect steps/day within 10% of the Actigraph at least 60% of the time, or be within 10% of the Yamax under free-living conditions.
The variation in cost and accuracy is due to different internal mechanisms. There are at least three basic mechanisms, including the spring-suspended lever arm with metal-on-metal contact, a magnetic reed proximity switch and horizontal beam, and piezo-electric crystal i.e. an accelerometer-like mechanism (Crouter et al, 2003; Schneider et al, 2003).
Some models allow for a low level of calibration – subjects walk a set distance and count the number of strides taken. This is entered as normal stride length and from this, distance covered can be estimated. Pedometer steps are influenced by body size and speed of locomotion so should be used with caution in children or groups of children of vastly different ages and heights (Trost, 2007).
A systematic review of the validity of pedometers compared to accelerometers, observation, energy expenditure and self-report concluded that they are a valid method for assessing physical activity when compared to different accelerometers (specifically uniaxial accelerometers), with a reported median correlation of r=0.86, depending on the specific instruments used, monitoring frame and conditions implemented, and the manner in which the outputs are expressed (Tudor-Locke et al, 2002). With regards to ambulatory activities only, better accuracy has been reported at faster walking speeds (i.e. 4.8-6.4 km/hr) compared to slower walking speeds but not at running speeds (Crouter et al, 2003). More sensitive (e.g. piezoelectric) pedometers have been recommended in individuals who naturally ambulate at a slower pace e.g. the elderly (Melanson et al, 2004). For the estimation of energy expenditure, one study in children showed good correlations between step counts and oxygen uptake expressed as body mass scaled (O2/kg-0.75min); r=0.806 for all activities (Eston et al, 1998). However, Crouter et al (2003) concluded that pedometers are most accurate for assessing steps, less accurate for assessing distance, and even less accurate for assessing kilocalories, in their study validating ten electronic pedometers. Therefore, as pedometers provide a total daily step count, the resulting data should ideally be kept in this form rather than estimating energy expenditure of distance travelled to prevent the addition of unnecessary error (Corder et al, 2007).
The number of days of monitoring required for a reliable estimate of physical activity has been investigated in a number of studies. High intra-individual variability has been demonstrated with a coefficient of variation of ≈32%; 3 days wearing was recommended in a study of 90 adult males and females (Tudor-Locke et al, 2005). Two days was found to be sufficient in a study of diabetic women over 40 years (Stryker, et al, 2007). Ideally seven days of monitoring would be undertaken but this may not be feasible (Corder et al, 2007) or may compromise compliance.