Motivation

The neighborhoods in which people live can impact their health and wellbeing. The health effects of living in socioeconomically disadvantaged neighborhood effects are well-established, with significant associations having been found for childhood asthma [1], diabetes [2], mortality [3], heart attack [4], cancer [5], drug use, physical activity, tobacco use, alcohol consumption, and dietary fat intake. Some of these health effects may be explained by environmental variables such as poor housing conditions, air pollution [1], proximity to hazardous waste sites [6], food deserts [7], easy access to alcohol retailers [8], and low density of recreation facilities [9,10]. Despite this knowledge, there are gaps in information concerning neighborhood-level health behaviors, attitudes, and resources.

Significance

To reduce neighborhood-based health disparities, we require access to meaningful, timely, and actionable information regarding place-based health determinants. Yet, health-related organizations, public health practitioners, and the policymakers lack sufficient access to up-to-date, spatially granular, and actionable data concerning neighborhood-level health behavior and its determinants. As our prior work has shown, social media sites may be potential sources of such data [11]. However, there are challenges in leveraging the volumes and variety of social media data, along with the uncertainty and biases that can arise from them. Modeling and accounting for bias will address a key obstacle in the use of this new stream of data for neighborhood health-related research, programs and policies.

Innovation

Social media sites have been widely adopted as venues for conversation, information sharing, and recommendations. Social media can provide a real-time signal about public health, and our previous work shows that this is possible at the neighborhood (census tract) level [11,12]. Our previous analyses of Twitter posts showed that social media provides can be used to characterize of dietary behaviors and their determinants at the census tract level. Specifically, we found significant associations between our novel tweet-level variables (aggregated to the census tract level) and tract-level variables known to correlate with dietary behavior such as demographics and mortality from obesity-related conditions [13]. However, a weakness of this work was the lack of direct measures of health behaviors and attitudes with which to evaluate the association between individual behavior and attitudes and the signals in individual’s tweets. Furthermore, there is unknown bias in the content of tweets both based on the demographics of Twitter users, and due to the social meanings accorded to tweeting certain types of content (e.g. positivity bias) [14].

Project Aims

We hypothesize that social media-based measures can accurately characterize health behaviors, and attitudes at the census tract level. We also posit that these measures can be effectively generalized to populations in census tracts in which social media is less widely adopted, such as rural and low-income areas. Study aims are:

Aim 1: To validate social media-derived measures of health behavior, and related attitudes, at both individual- and census-tract levels using self-reported survey data.

Aim 2: To validate algorithmic approaches to assigning social media posters to census tracts using self-reported survey data.

Aim 3: Model bias in social media-based measures of health behavior and attitudes based on: (a) tweeter demographics compared to population size, population density, rural and urban geographic location, and socioeconomic conditions, and (b) social desirability and selection biases affecting social media posting content concerning health behaviors.

References

1. Corburn J, Osleeb J, Porter M. Urban asthma and the neighbourhood environment in New York City. Health Place [Internet]. 2006 Jun;12(2):167–79. Available from: https://linkinghub.elsevier.com/retrieve/pii/S1353829204001029

2. Auchincloss AH. Neighborhood Resources for Physical Activity and Healthy Foods and Incidence of Type 2 Diabetes Mellitus. Arch Intern Med [Internet]. 2009 Oct 12;169(18):1698. Available from: http://archinte.jamanetwork.com/article.aspx?doi=10.1001/archinternmed.2009.302

3. Subramanian SV, Chen JT, Rehkopf DH, Waterman PD, Krieger N. Racial Disparities in Context: A Multilevel Analysis of Neighborhood Variations in Poverty and Excess Mortality Among Black Populations in Massachusetts. Am J Public Health [Internet]. 2005 Feb;95(2):260–5. Available from: http://ajph.aphapublications.org/doi/10.2105/AJPH.2003.034132

4. Rose KM, Suchindran CM, Foraker RE, Whitsel EA, Rosamond WD, Heiss G, et al. Neighborhood Disparities in Incident Hospitalized Myocardial Infarction in Four U.S. Communities: The ARIC Surveillance Study. Ann Epidemiol [Internet]. 2009 Dec;19(12):867–74. Available from: https://linkinghub.elsevier.com/retrieve/pii/S1047279709002865

5. Gomez SL, Shariff-Marco S, DeRouen M, Keegan THM, Yen IH, Mujahid M, et al. The impact of neighborhood social and built environment factors across the cancer continuum: Current research, methodological considerations, and future directions. Cancer [Internet]. 2015 Jul 15;121(14):2314–30. Available from: http://doi.wiley.com/10.1002/cncr.29345

6. Smith CL. Economic deprivation and racial segregation: Comparing Superfund sites in Portland, Oregon and Detroit, Michigan. Soc Sci Res [Internet]. 2009 Sep;38(3):681–92. Available from: https://linkinghub.elsevier.com/retrieve/pii/S0049089X09000210

7. Beaulac J, Kristjansson E, Cummins S. A Systematic Review of Food Deserts, 1966-2007. Prev Chronic Dis. 2009;6(3).

8. Berke EM, Tanski SE, Demidenko E, Alford-Teaster J, Shi X, Sargent JD. Alcohol Retail Density and Demographic Predictors of Health Disparities: A Geographic Analysis. Am J Public Health [Internet]. 2010 Oct;100(10):1967–71. Available from: http://ajph.aphapublications.org/doi/10.2105/AJPH.2009.170464

9. Dahmann N, Wolch J, Joassart-Marcelli P, Reynolds K, Jerrett M. The active city? Disparities in provision of urban public recreation resources. Health Place [Internet]. 2010 May;16(3):431–45. Available from: https://linkinghub.elsevier.com/retrieve/pii/S135382920900135X

10. Smiley MJ, Diez Roux A V., Brines SJ, Brown DG, Evenson KR, Rodriguez DA. A spatial analysis of health-related resources in three diverse metropolitan areas. Health Place [Internet]. 2010 Sep;16(5):885–92. Available from: https://linkinghub.elsevier.com/retrieve/pii/S1353829210000614

11. V. G. Vinod Vydiswaran, Daniel M. Romero, Xinyan Zhao, Deahan Yu, Iris Gomez-Lopez, Jin Xiu Lu, Bradley Iott, Ana Baylin, Philippa Clarke, Veronica Berrocal, Robert Goodspeed TV. “Bacon Bacon Bacon”: Food-Related Tweets and Sentiment in Metro Detroit. In: AAAI Conference on Web and Social Media. 2018.

12. Shen Y, Clarke P, Gomez-Lopez IN, Hill AB, Romero DM, Goodspeed R, et al. Using social media to assess the consumer nutrition environment: comparing Yelp reviews with a direct observation audit instrument for grocery stores. Public Health Nutr [Internet]. 2019 Feb 8;22(2):257–64. Available from: https://www.cambridge.org/core/product/identifier/S1368980018002872/type/journal_article

13. Vydiswaran VGV, Romero DM, Zhao X, Yu D, Gomez-Lopez I, Lu JX, et al. Uncovering the relationship between food-related discussion on Twitter and neighborhood characteristics. J Am Med Informatics Assoc [Internet]. 2020 Feb 1;27(2):254–64. Available from: https://academic.oup.com/jamia/article/27/2/254/5601669

14. Spottswood EL, Hancock JT. The positivity bias and prosocial deception on facebook. Comput Human Behav [Internet]. 2016 Dec 1 [cited 2020 Apr 29];65:252–9. Available from: https://www.sciencedirect.com/science/article/pii/S0747563216305842?via%3Dihub