Practical considerations
At the outset, decide the primary outcome of the study; total volume of activity, time spent at different intensity levels, or estimates of physical activity energy expenditure.
It is not possible to recommend one device over another. The following should be considered when deciding on which model of accelerometer to use:
- Cost and available budget
- Burden to respondents, e.g. size, attachment
- Uniaxal, biaxial or triaxial
- Data storage capacity
- Sampling frequency, higher sampling frequency will more accurately detect and differentiate activities
- Epoch setting - newer accelerometers can collect data on a 1-second basis
- Number of days sampling required
- Robustness and reliability in the field
- Feasibility
There are important considerations for any researcher wishing to use accelerometry that must be thought through carefully at the outset and these are discussed below. The data produced from accelerometers is somewhat lengthy, detailed and complex compared to questionnaires. Analysis programs have become more sophisticated and more user friendly, but collaboration with experienced researchers is advisable for new users. See resources required for information on data management.
Number of days monitoring
Accelerometers are often used to capture habitual activity and should therefore capture day-to-day variation in activity including the likely differences between weekdays and weekends. The number of required days of wearing is controversial (Reilly et al, 2008). Day-to-day variation tends to be greater in children and 4-9 days of monitoring are common, in adults wearing time tends to be shorter, e.g. 4-5 days (Trost et al, 2005). Researchers should consider if there is likely to be a seasonal effect in activity levels, e.g. different patterns of activity during holiday times, and then consider the likely implications of this. This may be more feasibly addressed by repeated monitoring, rather than attempting continuous monitoring across seasons.
Epoch time
Generally the shortest epoch time possible is desirable and newer accelerometers have sufficient memory capacity to capture data in at least 15 second epoch resolution over the desired monitoring period – these two are of course inversely related, so a decision should be made that optimises this balance. A 60 second epoch has often been used in adults but shorter epochs increase the sensitivity of the measurement in children (Treuth et al, 2004). Studies have shown the effect of different epoch lengths on the description of children’s physical activity patterns but this was confined to vigorous activity (Nilsson et al, 2002). Baquet and colleagues (2007) showed that 80% of moderate physical activity and 93% of vigorous activity lasted less than 10 seconds in children aged 9-10 years. These results suggest that a 5-second epoch may be appropriate in children. However, to date evidence for the optimum epoch length does not exist; combining moderate and vigorous activity may be appropriate in some studies and overcome limitations of longer epochs (Reilly et al, 2008).
Wearing position
The most appropriate wearing position of an accelerometer depends on the study question but the most common place is the hip or lower back (Trost et al, 2005). This location allows for tracking the movement of the largest and most central part of the body, the trunk. Attachment to the arm or leg, however, may be more feasible locations for some individuals.
Nilsson et al (2002) compared the MTI/CSA accelerometer worn at the hip and the lower back in 16 children and did not find significant differences in counts or time spent in moderate, vigorous or very vigorous activity. Similarly, an adult study compared this monitor worn in these two positions, and no significant differences in time spent in moderate or vigorous activity were observedd (Yngve et al, 2003).
Number of monitors
Multiple monitors reduce the feasibility via both increased participant burden and by increasing the amount of data to be processed. Multiple monitors may, however, provide a more detailed picture of the individual level of physical activity, for example the types of physical activity undertaken. Using simple regession analysis, some marginal improvement in the estimation of energy expenditure from information provided by two monitors has been found (Swartz et al, 2000) but this is negated by the increased time needed for data analysis and the anticipated reduction in compliance. However, different analytical techniques, using conditional modelling may shift this balance (Rothey et al, 2008).