Linking Human Health and Neighborhood Using Big Data: A Case Study in Torino, Italy. (June 2017)

• Record Type: Journal Article
• Authors: Rybarczyk, Greg; Melis, Giulia
• Is Part Of: Journal of transport & health. Volume 5(2017)Supplement
• Page Start: S74
• Page End: S75
• Publication Date: 2017-06
• Subjects: Transportation -- Health aspects -- Periodicals; Transportation -- Periodicals; Public Health -- Periodicals; Noise, Transportation -- Periodicals; Air Pollutants -- Periodicals; 388
• DOI: 10.1016/j.jth.2017.05.224
• Languages: English
• ISSNs: 2214-1405

Abstract

Abstract : Background: Assessing and mapping public health within neighborhoods has long been a focus area for researchers. The major weaknesses of previous research have been the use of untimely and aggregated data. The advent of Big Data, however, has opened up new level of analysis and type of information which can be advantageous for assessing mental health. Geosocial media (GSM), such as Twitter, contains three attributes which sets it apart from other data: a timestamp, location, and human discourse. These three pieces of information collectively reveal previously unknown psychosocial relations to one's environment. The purpose of the current research was to use GSM to examine where and why human sentiment among neighborhoods in Torino, Italy fluctuates spatially and temporally. Methods: Twitter data was the source of all GSM data and collected through an open-source API. The database was filtered and used to derive a valance (i.e. sentiment) index. Compositional and contextual information for each statistical zone were then collected from government sources and then linked to the Twitter database using GIS. The zonal indicators included in this research consisted of: public health, infrastructure, travel safety, incivilities, accessibility, socioeconomic standing, and density. Exploratory spatial data analysis (ESDA), GIS, and regression modeling were implemented to investigate the spatial patterning of sentiments, and highlight important explanatory factors associated with increased or decreased values. Results: The ESDA results showed that the spatial patterning of Tweets clustered near the central portion of the study area, a spatial clustering index supported this finding. Not surprisingly, it was also found that sentiment values differed depending on time-indicating that human sentiment varies temporally. A global model, although week (R 2 = .03) pointed to strong and statistically significant relations (p-value < .05) to sentiment values. Most notably, urban density, mortality rates, and disturbances negatively affected sentiments, while pedestrian infrastructure had the opposite influence. The results validate a growing body of work on how urban stressors affect mental health spatially and temporally. Conclusions: We expect the results of this research to be valuable for practitioners and researchers on two fronts. First, though the models' explanatory powers were weak, several noteworthy coefficients were observed. Second, we have shown that GSM data is a valid dataset for understanding the nuanced psychosocial relationships among mental health, time, and space.