Actical accelerometer users rejoice! Rachel Colley of the Healthy Active Living and Obesity Research Group at the Children’s Hospital of Eastern Ontario Research Institute has posted annotated SAS files to use for cleaning and combining accelerometry data from multiple participants.
The Accel+ program is free to download:
- User’s Guide (Download)
- SAS Code
Contrary to the title of this post, the files aren’t foolproof (the HALO site includes a disclaimer to that effect), but they are extremely helpful for folks like myself who know a little bit about SAS, but aren’t experts. These Accel+ files can also be viewed as essentially companion files to the National Cancer Institute’s publicly available files for cleaning Actigraph accelerometer data. I really can’t say enough about how great it is that Rachel and her colleagues (Didier Garriguet, Glenn Glover, Elyse Labonte and Janine Clarke)made these files available for anyone who wants to use them.
The casual reader may wonder why I could possibly be so excited about this.
Well, when researchers want to measure physical activity and/or sedentary behaviour, they usually choose one of two approaches. They can ask people how much time they spend being active and/or sedentary (e.g. self-report), or they can measure it directly using an accelerometer (there are other approaches you can use, but these are by far the most common).
Anyone who has ever used accelerometry data will know that it can be tricky to analyse.
This has to do with 2 basic issues:
1. Accelerometers provide an unbelievable amount of data
The two major accelerometers on the market (Actigraph and Actical) give a movement “count” for each minute of the day (1440 datapoints/day), and researchers usually collect data for at least 7 days. If you have a study with 1000 participants (which isn’t that large as far as epidemiological studies go), you wind up with a file that has 10080000 (1440*7*1000) data points. That’s for a smallish epidemiological study. To put this in perspective, you can only fit the data of 4-5 participants into an Excel spreadsheet before you get an error message telling you that you’ve run out of cells.
Let’s say that one of the accelerometers malfunctions – you do not want to sift through 1 million datapoints by hand in order to find it. As a result of this giant tsunami of data, we tend to use code-based programs like SAS to clean and analyse accelerometer data. And since these programs are code-based, they require that you literally learn a new language to use them. So analysing accelerometer data can be tricky, especially for physiologists who don’t have a biostatistics/epidemiology background.
2. It can be tricky to be consistent across studies
While accelerometer data is “objective”, there are a lot of somewhat arbitrary decisions that need to be made when cleaning the data (they are usually based on evidence, but still somewhat arbitrary). For example, should you include participants who only wore the accelerometers for 6 hours/day, instead of the requested 14? Should you include participants who only wore the accelerometer on weekends, but not weekdays? How should you decide what values constitute various intensitites of physical activity (light, moderate, vigorous), and how do you determine whether a participant is being sedentary, or whether they took off the accelerometer altogether?
If I decide to include all the data, and you decide to only include data from participants who wore the accelerometer for 10+ hours/day over at least 4 separate days, then you and I are going to wind up with a very different values based on the same dataset. For the true nerds out there, check out this recent paper from MSSE which illustrates just how these minor changes can affect the relationship between movement and health.
Accelerometers come with their own software for downloading and analyzing data, but it doesn’t really work for the purposes of research. Hence why programs like Accel+ are so incredibly useful. Playing with these sorts of pre-made files can also be a very good way to figure out how to use SAS, if you’re into that sort of thing.
If you are a researcher who will soon be undertaking an analysis of Actical accelerometry data, download the above files and get analyzing!
Accelerometer analysis for dummies by PLOS Blogs Network, unless otherwise expressly stated, is licensed under a Creative Commons Attribution 4.0 International License.