2014.ISDS.Abstracts.Final.pdf


ISDS Annual Conference Proceedings 2014. This is an Open Access article distributed under the terms of the Creative Commons Attribution-
Noncommercial 3.0 Unported License (http://creativecommons.org/licenses/by-nc/3.0/), permitting all non-commercial use, distribution, 
and reproduction in any medium, provided the original work is properly cited.

42

ISDS 2014 Conference Abstracts

Digital Sources of Food Purchasing Data for the 
Surveillance of Dietary Patterns
Katia Charland*, Hiroshi Mamiya and David Buckeridge

McGill University, Montreal, QC, Canada

Objective
To demonstrate the utility of automatically captured store-level 

(i.e. point-of-sale) food purchasing data for the surveillance of dietary 
patterns before and after interventions. We assessed the effects of two 
interventions in Montreal, Canada that were intended to reduce the 
consumption of sugary drinks.

Introduction
In Canada, the economic impact of unhealthy eating is estimated at 

$6.3 billion annually and in the US the estimated cost is $87 billion. 
Despite the critical need to identify effective diet-related interventions 
through empirical evaluation, public health practitioners and 
researchers lack timely access to representative data sources collected 
at a fine spatial and temporal resolution. Food surveys, for example, 
are costly, infrequent, delayed, and subject to biases [1].

The Nielsen Corporation collects data on food purchasing directly 
from scanners in grocery and convenience stores around the world. 
These data hold great potential for public health practice. We were 
interested in using these data to analyze purchases of regular (sugary) 
soda and water, before and after two interventions aimed at reducing 
sugary drink consumption. The first intervention, ‘Gobes-tu ça’, was a 
counter-advertising campaign targeting the age group with the highest 
consumption of soda, 12-17 year olds [2]. The second intervention, 
‘Sois-futé, bois santé’, targeted elementary school students. Both 
began in the Fall of 2011 and ramped up over time.

Methods
Using Nielsen data, we determined the monthly number of (240ml) 

servings of soda and water purchased at 180 sampled convenience 
and grocery stores in Montreal, Canada, from 2008 to 2013. The 
outcomes were the number of soda and water servings purchased per 
month and the proportion of servings (out of water and soda) that 
were soda servings. We first identified significant breakpoints in both 
the soda and water series [3]. Then, to assess the relationship between 
the timing of the intervention and the purchasing outcomes, we used 
log-linear and logistic time series regression. Heteroskedasticity 
and autocorrelation consistent estimators were used to account for 
temporal correlation [3]. The independent variable representing the 
timing of the intervention, had a value of zero before the intervention 
(before September 2011). Afterwards, to account for the intervention 
ramping up, the variable had a value equal to the number of months 
since the intervention was implemented. Covariates were the month, 
the average monthly maximum temperature, and the ratio of the 
average price of soda to water.

Results
RESULTS The number of soda servings sold and the proportion 

that were soda dropped in late 2011 (Figure). There was one 
significant breakpoint for the soda series in Dec 2011 (breakpoint 
month: 12/2011; 95% confidence interval 3/2011, 12/2012) and none 
in the water series. After accounting for the average price ratio of soda 
to water, the month and the average maximum temperature, from the 
time of the intervention to the end of the study period there was an 
average 26% increase in water sales (1.26, 95% CI 1.09, 1.46) and 

12% decrease in soda sales (0.88; 95% CI 0.78, 0.97), resulting in an 
average 30% decrease in the proportion of sales of soda (odds ratio 
0.70; 95% CI 0.61, 0.95).

Conclusions
Using digital food purchasing data we identified a significant 

drop in the sales of soda relative to water following the public health 
interventions. Future work will involve examining geographical 
variations in the effect of the intervention, possibly due to socio-
economic status or demographic factors.

Figure: Monthly sales (/10 million) of soda, water, and the proportion of soda 
sales. The shaded region indicates the intervention period.

Keywords
Evaluation; Diet; Intervention

References

1. Diane McAmmond and Associates. Food and Nutrition Surveillance in 
Canada: An Environmental Scan. 2000 Mar.

2. Langlois K, Garriguet D. Sugar consumption among Canadians of all
ages. Health Rep. 2011;22.

3. Zeileis A. Econometric Computing with HC and HAC Covariance
Matrix Estimators. J Stat Softw. 2004 Nov; 11(10).

*Katia Charland
E-mail: charlandk@yahoo.com    

Online Journal of Public Health Informatics * ISSN 1947-2579 * http://ojphi.org * 7(1):e12, 2015