Improving research methods for the study of geography and mental health: Utilization of social networking data and the ESRI GeoEvent Processor

Abstract

The purpose of this article is to review the literature on geography and mental health, report on a case example using new methods for studying this topic, and provide recommendations for future research. Over 25 years ago, Holley (1988) conducted a review of the literature on geography and mental health and astutely stated, ‘… it is surprising that geographic study designs … have not received greater attention as an important and viable method of assessing population mental health’ (p. 535). In 1997, Parr echoed a similar statement, indicating that little has been researched on this topic and suggested this area be termed ‘geography of mental health’. Decades later, these statements still hold true – researchers have given little attention to the intersection of geography and mental health. Yet, there is great potential for this research to expand in a way that may be of great benefit to those studying mental health as well as the many suffering with mental health problems. In this aricle, a case example is provided to demonstrate the possibilities of utilizing current technologies, Twitter and the ESRI GeoEvent Processor, to expand the methods for research on this topic.

Keywords

geography mental health real time data social networking technology

For decades, there has been a lack of literature on geography and mental health. Over 25 years ago, Holley (1988) conducted a review of the literature on geography and mental health and astutely stated,… it is surprising that geographic study designs … have not received greater attention as an important and viable method of assessing population mental health’ (p. 535). In 1997, Parr echoed a similar statement, indicating that little has been researched on this topic and suggesting this area be termed ‘geography of mental health’. Decades later, these statements still hold true – researchers have given little attention to the intersection of geography and mental health. Yet, there is great potential for this research to expand in a way that may be of great benefit to those studying mental health as well as the many suffering with mental health problems. Parr concurred with this assertion in 1997 by recognizing that this research, with the right methods, could significantly impact social, political, economic, and cultural aspects. With current technology, researchers can now improve the methods for studying this topic by utilizing geolocation and social networking data to better understand group trends and patterns regarding mental health.

An individual’s mental health can be impacted in a matter of seconds from family events to regional or national news events. Some events may even impact more than just individuals and may include larger regions of people in which hundreds or even thousands of peoples’ mental health is impacted. Yet, professionals in the mental health field typically quantitatively and qualitatively measure the emotional states of individuals rather than collecting data about larger regions. Further, the individual data gathered are collected using surveys, rating scales, and interviews that are completed on one day, at one period of time, and often in clinical settings. Moreover, due to the methodological and legal constraints of collecting and managing individual data, this individual data is not considered within the context of other individuals’ data. Therefore, group mental health data are difficult to obtain and, when they are obtained, it is often no longer reflective on one’s mental health. Thus, the problem is professionals are left with a lot of individual data, which is often stagnant and is not contextualized with geography.

Background

The literature on geography and mental health dates back to the 1970s (Giggs, 1973; Wolch & Philo, 2000). During the 1970s through the early 1990s, researchers were plotting incidences of individual mental health problems or diagnoses to examine geography and mental health. Methodologies focused on spatially tracking people with mental health problems. For example, Giggs and Cooper (1997) published the findings from a study that incorporated geography and mental health and utilized a method other than survey data. More specifically, they plotted on a map (by district clusters) the addresses of 68 patients with each patient’s ICD-9 schizophrenia diagnosis. Later, they plotted the residences for 132 patients with ‘case-register diagnoses’ of affective psychosis. They found an increased concentration for schizophrenia in areas of low socio-economic status and an increased concentration for affective psychosis in areas of middle and low socio-economic status. The methodological approach used by Giggs and Cooper is unique because it is very different from what has been used to examine this topic in more recent years (i.e., 1990s to 2000s).

During the 1990s and early 2000s, researchers primarily used methodology that depended on self-report data via surveys to explore the relationship between geography and accessibility to mental health services. The focus had been on the location of mental health services. In this type of research, participants’ geography was usually categorized as urban, suburban, or rural settings. For example, Foley and Platzer (2007) examined mental health services offered in London via a questionnaire and confirmatory focus groups. They wanted to know if there were sufficient mental health services within the city. As a result of their study, they discovered trends and patterns in the geospatial data that assisted with future planning and decision-making. More specifically, they began to target ‘unconsidered’ geographic areas within the city. Arcury et al. (2005) also used a survey to explore geography and accessing health care. Participants were asked to report on locations they utilized for health care services in relation to their home. According to the findings, concerns regarding rural health care access were discovered. Therefore, the authors suggested the continuing need for public policy on rural health care access. Additionally, the authors stressed the importance of researchers continuing to study health care, especially in urban areas, due to the increased symptoms associated with mental health difficulties when compared to adolescents in suburban and rural settings. Finally, they discovered geographic patterns regarding paying for mental health services, knowledge of mental health services, and availability of mental health services. As an extension of this methodology, some researchers collected qualitative data in an effort to move beyond the survey-level data to examine the differences between individuals with mental health problems and those without mental health problems within the context of geography (i.e., Pinfold, 1998).

While these studies have contributed to the literature-base on this topic, little methodologically has changed in the past decade. Wolch and Philo (1988) expressed concerns in 2000 that the research on this topic was reaching a stalemate and that it was becoming ‘increasingly problematic to conduct large-scale or longitudinal research associated with mental (ill) health, and there is indeed a risk that some researchers become paralytically self-reflexive …’ (p. 149).

Mental health and social networking sites

Previous research on geography and mental health utilized spatially tracking individuals and self-report data on accessing services and/or self-report data on mental health status. The type of data collected as a part of these research methods is usually time-sensitive, meaning that after one reports the information, the information becomes old and may or may not accurately reflect information in the present. For example, when an adolescent struggling with feelings associated with depression completes a questionnaire about her emotional state for the school psychologist as a part of a screening tool or evaluation, this information is reflective only of the moment in which she completed the self-report measure. Although the reliability of many self-report measures is acceptable, information on these self-report measures present one opportunity during a 30-minute session for an adolescent to report on his/her emotional state. The data gathered may be time limited and have the potential to quickly become stagnant. Further, the data do not capture any significant life changes that may impact a person’s emotional state.

Another resource for collecting self-report data that is ongoing and potentially more reflective of one’s current state is social networking sites (SNS). SNS, which are self-report, provide users opportunities to emote via words and pictures whenever, wherever, and as often as they want. SNS, such as Twitter and Facebook, provide a real time pulse on the emotional state of individual users. For example, someone struggling with depression may write ‘feel so depressed, I don’t know what to do’, or someone struggling with anxiety may write ‘so sick of worrying about everything all the time’. By harvesting this publically available data, researchers can begin to understand more about groups of people and their emotional states. Of course, these data also have limitations and care needs to be taken to interpret group data versus individual data to assist in compensating for some of the limitations associated with individual users.

Utilization of social media data for research has increased in the past several years (i.e., Yang et al., 2016). As social media continues to expand and become a predominant form of communication in society, it is likely more researchers will look to this data source. Researchers studying geography have also looked to social media as a data source. For example, a group of researchers at San Diego State University have conducted several studies using a technology they developed to explore social media as an analytic tool (i.e., Yang et al., 2016). More specifically, this research group utilized Twitter in some of their research (i.e., Han, Tsou, & Clarke, 2015). Social media data also present an opportunity for researchers in the field of mental health and, more specifically, those looking to expand the literature on geography and mental health.

Mental health and the ESRI GeoEvent Processor

The purpose of SNSs, such as Twitter, is to share information about what is happening in an individual’s life at that moment in time. Therefore, it typically contains current information about a user’s thoughts. Utilizing these data for groups of users may provide the most accurate and up-to-date information about SNS users’ mental health. By further linking these data to geography, it could provide a lens into the mental state of a group of people in a specific region or even nation. This is achievable through use of the Environmental Systems Research Institute GeoEvent Processor (ESRI, 2016)

The ESRI GeoEvent Processor (2016) was created to monitor moving assets in real-time. When coded, the ESRI GeoEvent Processor also allows for professionals to connect streaming data feeds into ESRI ArcGIS, a software program used to map data. By utilizing publically available data from SNS and importing it into the ESRI GeoEvent Processor, posts that are geotagged utilize geolocation to map the data, which gives professionals the opportunity to examine the impact real-time events have on individuals associated with specific regions. This has been used in many fields to improve work flow with utility companies (i.e., efficiency of vehicles traveling), emergency responder vehicles (i.e., efficiency of vehicles traveling), and self-reports on social media for drug/alcohol use (i.e., Yang et al., 2016). For the purposes of examining geography and mental health, an example relevant question may be: How would Twitter posts (with geotags) be impacted during the Boston Marathon bombings one day after, one week after, one month after, and one year after? Would there be differences in posts made by the age of the Twitter user or by his or her geographical proximity to the bombing? What types of emotions would there be more of anger, aggression, sadness, or anxiety?

When utilizing SNS that are in written form, such as Twitter, the data being imported into ESRI GeoEvent Processor (2016) depend on the linguistics associated with the SNS message. For example, if SNS messages associated with depression were being mapped, words or word phrases such as ‘sad’, ‘blue’, ‘isolated’, ‘feeling depressed’ might be programed into the processor. In addition to the linguistics, tweets with geotags can be sorted by demographic information, including age of the user. However, the accuracy of this is dependent on the profile created by the user.

A case example utilizing Twitter data with the ESRI GeoEvent Processor (2016) was conducted to demonstrate the capabilities of these two technologies to determine the viability of this methodology.

Case example

Participants

Participants included 1% of Twitter users across the world, who were using Twitter during a 36-hour period of time with a publically accessible profile who tweeted in English or Spanish ‘depressed’ or ‘feeling depressed’. Therefore, Twitter users who did not have publically available profiles were excluded from the sample. In addition, Twitter has restrictions on accessing data. Thus, 99% of tweets that met the above criteria were excluded. The 1% that is permitted to be accessed is randomly selected.

Method

In a 36-hour period, Twitter was used to monitor 1% of publically accessible tweets across the world for key words ‘depressed’ and ‘feeling depressed’. The ESRI GeoEvent Processor (2016) was coded to input the Twitter data and to map each tweet using longitude and latitude in real time. The ESRI GeoEvent Processor was coded to digitally place a ‘pin’ on the map of the world within a several hundred-foot circumference of the original tweet. This random placement of the pin was used to assist in maintaining confidentiality of the Twitter user, even though the user chose to make this information publically accessible.

Results

In the 36-hour period of time, exactly 136,385 tweets with ‘depressed’ and ‘feeling depressed’ were retrieved with geotags that were used for geolocation in the ESRI GeoEvent Processor (2016). This amounts to approximately one tweet every four seconds containing either ‘depressed or ‘feeling depressed’. Again, it is important to note that this represents 1% of the data (randomly selected by Twitter), per Twitter’s research data restrictions. By utilizing Twitter and the ESRI GeoEvent Processor, a real-time map was produced in ESRI ArcGIS in which the researcher could watch the map of the world as users tweeted in real time with messages containing ‘depressed’ and ‘feeling depressed’. Concentrations of these tweets were in the Pennsylvania and New York regions of the United States and in Europe (see Figure 1). See Table 1 for examples of specific tweets that were randomly selected during this period of data collection.

Figure 1.

Twitter feed map. This figure illustrates the frequency of tweets containing the words “depressed” and “feeling depressed” during the data collection period.

Table 1.

Examples of twitter content.

Tweet Content
‘I really just want to feel content again. I hate being so stressed out and depressed every day’
‘Okay but is there a difference between “I feel depressed” and “I have depression”? Or is it incorrect to use it like the first saying’
‘What’s wrong with me that I’m so unlikable to my family #depressed’
‘Never in my life have I been this depressed about going to school’
‘I’m so depressed and stressed rn and it’s my birthday tomorrow gosh someone help me idk what’s wrong with me’
‘Being depressed is like being dead in life…’
‘16 yeah, maybe I’m depressed’
‘Being at school makes me depressed’

Limitations and ethical factors

It is important to note that Twitter only permits users to search 1% of tweets. Therefore, these data are a drastic under-representation of the actual frequency of these words in tweets. In addition, the ESRI GeoEvent Processor (2016) was coded to map the data within several hundred feet of the actual latitude and longitude to protect users’ confidentiality, even though they chose for their data to be publically accessible.

Caution needs be taken when accessing the potentially sensitive data utilized in this case study, which included information about individuals’ mental health statuses. More specifically, when the ESRI GeoEvent Processor (2016) is coded to collect Twitter data, the software can identify tweets with geotags within a very precise longitude and latitude (depending on the GPS satellites and triangulation using cell towers, etc.). Using the software with a high degree of precision (i.e., within feet of the original tweet) for the purposes outlined in this case example was not necessary and may raise ethical issues if used with that level of accuracy. In contrast, there are researchers studying personal sensing – using smartphones to study an individual’s movements that may signal mental health problems (Pisani, Wyman, & Mohr, 2016; Weir, 2017). Another limitation of this case example is that it represents a specific group of users. These users likely live in environments in which they have access to technology and they have been provided enough education to support them in utilizing and engaging on a SNS site, such as Twitter. Therefore, the results of this case example have limited generalization.

Implications

Future research is needed to expand the literature on geography and mental health with innovative methods that will extend the current literature base. With the technology available to researchers, it is imperative that professionals studying this topic begin to utilize tools like the ESRI GeoEvent Processor (2016) and SNS (i.e., Twitter) to collect real time data about mental health. Self-report scales have been used to collect information about a person’s thoughts and feelings, which captures one moment in time for one person. Although these measures are often reliable, they do not capture changes in groups of people and utilizing this technology would allow for an analysis at a more global level.

Future research is also needed to improve the linguistic aspect of utilizing SNS with the ESRI GeoEvent Processor (2016) to explore the validity of phrases selected (i.e., ‘depressed’ and ‘feeling depressed’). Eventually, these phrases should be examined within the context of mental health disorders and the relationship of these disorders on geography. Additionally, it is imperative to control for events such as Suicide Prevention Week when SNS is populated with messages such as, ‘Have you thought about killing yourself?’ which are followed by, ‘Get help by calling that National Suicide Prevention Hotline’. Another linguistic concern is SNS messages that include ‘depressed’, but the context is, ‘Have you heard that new song called “Depressed”?’. Note that this case study was conducted during a 36-hour window with no known major events. Finally, coding for the ESRI GeoEvent Process needs to expand to account for languages of all users across the world.

Clearly, utilizing this type of a methodological approach involves the management of big data. As linguistic chains are improved, these data will be used to identify trends and patterns to begin to improve mental health supports in areas with users who identify needs. Researchers need to give consideration to the ways in which data are consumed. There is a lot of information coming in, and managing it, understanding it, and disseminating it in a meaningful way is critical to making a real impact on the field of geography and mental health.

Conclusion

Researchers have documented the importance of the study of geography and mental health for both individuals and groups of people since the 1970s. Yet, the literature on geography and mental health has been slow to progress and has reached a standstill for a little over a decade. This pause in the literature is largely due to the lack of evolution for methods to study this topic. Use of SNS and technology like the ESRI GeoEvent Processor (2016) appear to have the potential for an immense application to the mental health field, especially in terms of the literature on geography and mental health. Pairing big ideas with innovative technologies, such as the ESRI GeoEvent Processor, may hold the key to making real-time, meaningful changes in the way we analyse and provide mental health supports across regions and nations, in providing interventions, and how we understanding the way the world responds on an emotional level to real-time events.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Author Biography

Courtney L. McLaughlin PhD, NCSP is an Associate Professor in the Educational and School Psychology Department at Indiana University of Pennsylvania, USA, where she directs the School Psychology PhD Program. Her research interests focus on school-based mental health, social media and mental health, geography and mental health, and training future school psychologists.

References

Arcury

T. A.

Gesler

W. M.

Presisser

J. S.

Sherman

Spencer

Perin

(2005) The effects of geography and spatial behavior on health care utilization among the residents of a rural region. Health Services Research 40(1): 135–155. doi: 10.1111/j.1475-6773.2005.00346.x.

Environmental Systems Research Institute (ESRI) [Computer software]. (2016). Retrieved from http://www.esri.com/.

Foley

Platzer

(2007) Place and provision: Mapping mental health advocacy services in London. Social Science and Medicine 64(3): 617–632. doi: 10.1016/j.socscimed.2006.09.014.

Giggs

J. A.

(1973) The distribution of schizophrenics in Nottingham. Transactions of the Institute of British Geographers 59: 55–76.

Giggs

J. A.

Cooper

J. E.

(1987) Ecological structure and the distribution of schizophrenia and affective psychoses in Nottingham. British Journal of Psychiatry 151: 627–633.

Han

S. Y.

Tsou

Clarke

(2015) Do global cities enable global views? Using Twitter to quantify the level of geographical awareness of US cities. PLoS ONE 10(7): 1–23. doi:10.1371/journal.pone.0132464.

Holley

H. L.

(1998) Geography and mental health: A review. Social Psychiatry and Psychiatric Epidemiology 33(11): 535–542.

Parr

(1997) Mental health, public space and the city: Questions of individual and collective access. Environment and Planning D: Society and Space 15: 435–454. doi: abs/10.1068/d150435.

Pinfold

(1998) Moving away from the psychiatric asylum: Nottingham’s rehabilitation landscape. East Midland Geographer 21: 23–34.

10.

Pisani

A. R.

Wyman

P. A.

Mohr

D. C.

et al. (2016) Human subjects protection and technology in prevention science: Selected opportunities and challenges. Prevention Science 17(6): 765–778. doi:10.1007/s11121-016-0664-1.

11.

Weir

(2017) Exploring how location affects mental health. Monitor on Psychology 48(1): 18–20.

12.

Wolch

Philo

(2000) From distributions of deviance to definitions of difference: Past and future mental health geographies. Health and Place 6(3): 137–158. doi: 10.1016/S1353-8292(00)00019-8.

13.

Yang, J., Tsou, M., J, C., Allen, C., Spitzber, B., Gawron, J., & Han, S. (2016). Social media analytics and research testbed (SMART): Exploring spatiotemporal patterns of human dynamics with geo-targeted social media messages. Big Data and Society, 3, 1–19. doi: pdf/10.1177/2053951716652914.