Phew, that was an extremely long month of data analysis! In August, our team decided to conduct a randomized controlled trial in one of our ungraduate public health classes. The title of our study was “Healthy Eating Using a Mobile App Among University Students in Singapore”. In order to gather the data that we needed, we designed questionnaires. The students were given a participant information sheet, some dietary guidelines (Health Promotion Board’s “My Healthy Plate”) and instructions on how to use the intervention (for those part of the intervention group). As the intervention, we used MyFitnessPal, a calorie counter app that helps you to keep a diary of your food intake and gives information on the nutritional value of foods you eat. We were interested to know whether the app could improve students’ dietary habits.
We asked students to complete a baseline questionnaire at the beginning of the semester, to gather information about their dietary and exercise habits. We then randomly allocated half the class to the intervention group and the other half to the control group. The intervention group was assigned to used the calorie counter app with the dietary guidelines, whereas the control group was just provided with the dietary guidelines. The task of analyzing the trial data fell on this newbie’s plate. Here I am, a novice at Stata, tasked to complete the data analysis within 5 weeks. It was a steep learning curve for me, I had to learn to clean, manage and analyze the data appropriately. Renaming and labeling the variables alone took me a week. 117 observations and 80 variables later I had a passable dataset that I could present at our team meeting. After 8 weeks from baseline, it was time to administer the final follow up questionnaire. Now I had 2 datasets to clean, manage and analyze within 2 weeks… I was overwhelmed. However, with some solid support from chocolates, chips and coconut water I was able to complete the analysis in time.
Stay tuned for the results of our trial!
-Nivedita
Did you also pull data from the app, for people the group that used it?
Thanks for the comment. No, we did not. We did consider this, though this would require a third-party app to interface with the MyFitnessPal API, or for the students to download and submit their own data. In the end, we decided not to do it, partly because we didn’t have the time to develop the interface and because it’s more intrusive. But it also introduces potential bias, in that information on food intake in those who use the app is collected differently (and we’d expect more accurately) from information in the control group. In fact, that’s a general problem with these types of studies, even without pulling the data from the app – those in the intervention group are likely to provide more accurate data, simply because they log their intake and/or physical activity routinely.