We will calculate the proportion of children who wear the accelerometer and provide usable data. We will assume periods of sixty minutes or more with zero readings as “non:wear” time. Participant’s data will be included if they provide two or more weekdays of data with at least 500 minutes of data between 6am and 11pm.
The authors are introducing a rule on how to interpret accelerometer data that can only increase, never decrease, apparent activity levels.
It's possible that their assumption is valid. One would have to check how they justify this.
It's also possible that "periods of sixty minutes or more with zero readings" might just be periods where the patient is sleeping during the day due to exhaustion. In my experience, periods where I attempted to be more active and walk every day or most days, resulted in much more daytime sleep. If the device is not highly sensitive or the patient simply doesn't move, then these daytime naps could be erroneously counted as no wear time. The kind of exhaustion that makes me sleep or nap during the day tends to be so profound that I don't even moving a limb or turn around in bed.
In addition to questionnaire measures, participants in both trial arms will be asked to wear an accelerometer (GT3X+) to measure physical activity for seven days within one month of randomisation and at 3 and 6 months follow:up.
Does this indicate that a patient could have received treatment for up to 3 weeks before the first measurement of activity levels are made? Both treatment arms are described as having an initial activity reduction component. If the patient reduces their activity as told, and only then the first measurement of activity levels is performed, it could lead to the illusion of improvement.
This could easily happen if the patient, during the first contact with the service, receives both the accelerometer and the advice to reduce activity levels.
The accelerometer data will be processed to identify mean minutes of sedentary, light and moderate to vigorous intensity physical activity per day using established accelerometer cut off points and protocols.
Such subgrouping could also give an unscrupulous reeseacher more opportunities for spin and p-hacking. For example it's possible that light activity increases at 3 and 6 months, but that moderate and vigorous activity decrease. If the researcher only reported that light activity levels increased it would be misleading. What is wrong with just reporting accelerometer counts?
Accelerometers are small, match box sized devices that measure physical activity. They have been shown to provide reliable indicators of physical activity among children and adults.
I recall reading that people tend to temporarily modify their activity levels when they know that it's being recorded with accelerometers. So any measurements need to take this effect into account and be performed for long enough periods for this to not be a significant factor. Whether one week is enough needs to be checked.
Edited for clarity.
