oa Tracking health awareness in MENA through Facebook Advertising audience estimates

Diabetes epidemic is claiming more lives than ever, with the number of people with diabetes rising from 108 million in 1980 to 422 million in 2014. World Health Organization estimates that in 2015, diabetes was the direct cause of 1.6 million deaths and in 2012 high blood glucose was the cause of another 2.2 million deaths (http://www.who.int/mediacentre/factsheets/fs312/en/). The obesity and diabetes are especially prevalent in GCC, and Qatar in particular where 70% of the population either overweight or obese (http://www.qatar-tribune.com/news-details/id/54297). The most common form of diabetes – Type II – is caused by a combination of lifestyle factors and genes, thus numerous campaigns are underway to encourage change in daily behaviors which may lead to obesity and diabetes. Recently, Facebook and other social media has become a platform for patient interaction, outreach and education.The scale and reach of Facebook and other social media provide a unique opportunity to track the awareness and interest in health-related topics of large slices of population. This is imperative, as the awareness of an issue is the first step to lifestyle behavior change. In this work, we use the Facebook Advertising platform to track the interests of millions of people across the MENA regions on health-related topics. In particular, before ordering an advertising campaign, the potential advertiser is free to query Facebook about the potential reach, or estimated audience, of a particular selection of location, gender, age, language, interests, and a variety of other attributes. For example, when we select Arabic-speaking women living in Qatar between the ages of 30 and 50 who are interested in dieting, Facebook gives an estimated audience of 4,300 users who would see our hypothetical ad. As such reach estimates do not divulge information on any particular user, this information provides a window into otherwise private behaviors of Facebook users, in aggregate. In the present study, we examine the interest Facebook users from the countries in the MENA region have concerning health-related topics such as Obesity and Diabetes awareness, as well as general Fitness and Wellness and Healthcare. These interests are then compared to baselines such as Luxury Goods and Shopping, as well as health-related ones like Fast Food. We designed a visual analytic interface to enable exploration of these health awareness data. Once the interests are selected (Fig. 1, top), demographic slices of the data are presented in tree maps (Fig. 1, left) and on a choropleth map (Fig. 1, right), with each segment colored with a Health Awareness Score. The user can select a particular demographic (such as male gender) or country, automatically updating all other demographic segments to correspond with the current selection.Fig. 1. Comparison of interest in Physical Activity vs Fast Food for young women who are local to each of the MENA countries. Tool is available online at http://sha.qcri.org/ Using Facebook Advertising, or any other social media, as a source for health awareness monitoring has several drawbacks, the most prominent of which is the biased sampling of the population. However, with the continued adaptation of these platforms across the world and segments of population, they become increasingly representative of the general population. This can be mediated by careful cross-correlation with authoritative data sources such as statistics provided by the World Health Organization. Another drawback comes from the black box nature of the platform, as we are not purview to the internal processing and definitions the platforms use to come up with the metrics they reveal. Our ongoing efforts are addressing this by comparing the metrics provided by Facebook to the known disease prevalence as collected. In particular, if we look at the rates of diabetes (as published by the Institute for Health Metrics and Evaluation (IHME) (http://www.healthdata.org/), we can compare them to the estimates of interest Facebook provides. In Qatar, the Pearson correlation between diabetes rates and Facebook awareness of Diabetes Mellitus Type II across different age groups is r = 0.963 (p < 0.001), suggesting the awareness of disease rises as the incidence of the disease increases in the population. Interestingly, the trend reverses when we correlate interest in Diabetes Mellitus Type I (a much rarer form of disease, usually diagnosed in children) at r = –0.844 (p < 0.002), indicating the focus of the population remains with the Type II. In summary, this demographically rich data source has a great potential for improving awareness campaign tracking and dynamic opinion monitoring. Our study develops new methodologies and tools to collect, verify, visualize, and make this data available to the policy makers, campaign organizers, and many other potential actors interested in monitoring public opinion.