We collected information on the use of 3 different MSM GSN apps using a geographically systematic sample of points in Fayette County, Kentucky [43]. Fayette County encompasses the city of Lexington, Kentucky, and has a land area of 284 square miles and a population of 295,805 people who are predominately white (75.7%) and non-Hispanic (93.1%), with 30.7% between the ages of 18 and 34 years [43].

The road network in the county resembles a spoke pattern in which the city is bisected by a series of main roads that extend to the county’s perimeter. The bisecting routes that offered the best county coverage were selected using Google Maps [44]. Data collection points (n=62) were spaced in 2-mile increments along 9 routes with a total driving distance of 112 miles (Figure 1).

We visited each point twice: once during weekday daytime (Monday through Friday, 9:00 am to 3:02 pm) and once during weekend nighttime (Friday or Saturday, 8:02 pm to 11:50 pm) hours. While stopped at each point (approximately 5 minutes per stop), we logged into a blank profile created for study purposes on each of the 3 GSN apps. Once we were logged into the profile, the app displayed the number of users within varying distances from the collection point. We recorded the number of users within 1 mile on each app, the time of collection, and the latitude and longitude at that collection point on a paper form and via the Fulcrum data collection app [45]. All apps used in this study have been self-described as providing a space for gay men to look for dates, friends, fun, relationships, and hookups. Apps 1 and 3 are nongroup/niche specific apps that have been described in previous research [46-49], with the latter attracting gay, bisexual, and curious men [50]. App 2 targets a subgroup identified as “bears” [47,48], who have a more traditional masculine style and acceptance of diverse body shapes [49].

According to preliminary data from interviewer-administered surveys from 57 MSM aged 18 to 35 years from Lexington (data collection ongoing), apps 1, 2, and 3 were in the top 5 most frequently named MSM GSN apps used by respondents to “meet people for sex.”

Variables included in the regression model were derived from the 2011-2015 American Community Survey 5-year estimates [51], the 2010 Census data from the US Census Bureau [43], and the Lexington-Fayette Urban County Government Open Data Web Portal [52]. Independent variables in the model included block group population density (population count per square mile; range 3.3-15,737.5), median age (range 7.5-64.5 years), median annual household income (range $7,905-$201,429), percent white (range 10.8%-100.0%), percent Hispanic (range 0.0%-64.2%), and business zoning amounting to greater (or equal) versus less than 1% of the area of the block group (binary: 40.4% had greater or equal to 1% of the area business zoned). These variables were included in the analyses because past research on GSN app use has found that demographic variables such as age [10,15], race/ethnicity [10,15], and income [22] were associated with app use. Business zoning, the areas specified for business use by the local government, was included to capture the presence of gathering spaces (eg, restaurants, bars, shopping venues, employers).

The number of GSN app users within a 1-mile radius of the data collection points ranged from 0 to 36 during weekday daytime hours and 0 to 39 during weekend nighttime hours. The median number of estimated GSN app users in each Census block group in Fayette County during weekday daytime and weekend nighttime varied by app (10.0, 2.0, 10.5, and 10.0, 1.0, 9.0, respectively), but within-app comparisons revealed no significant temporal differences for any of the 3 apps (P=.95, .39, and .65; see Table 1).

We created choropleth maps to display differences in the spatial distribution and number of GSN app-using MSM by time of day (Figure 3). From weekday daytime to weekend nighttime, the spatial distribution of app-using MSM varied for 2 of the 3 apps. The use of app 1 was concentrated in downtown Lexington (ie, the center of the county) during weekday daytime and weekend nighttime; however, at weekend nighttime, a second area south of the city emerged as an area of high app usage. For app 3, use was concentrated in the downtown area both during weekday daytime and weekend nighttime but use was more intensely concentrated in the downtown area at weekend nighttime.

Fayette County comprises 208 Census block groups. Based on the dependent variable of estimated number of GSN app users in each Census block group, unadjusted negative binomial regression with each app stratified by time of collection was used to estimate crude risk ratios (Table 2). In all combinations of apps and collection times, the number of GSN app-using MSM was positively associated with population density (P<.001) and presence of business zoning of the block group area (P<.001), and negatively associated with age (P<.001 to P=.03) and median income (P<.001). For every 100 persons per square mile increase in population density, the number of GSN app users increased by 1.1% to 1.5%. In the presence of business zoning, the number of GSN app users increased by 106% to 188%. For every year increase in median age, the number of GSN app users decreased by 1.5% to 3%. For every $5,000 increase in median income, the number of GSN app users decreased by 5.7% to 8.9%.

Percent Hispanic ethnicity was not significantly associated with the number of GSN app-using MSM in any app or time combinations (P=.20 to P=.99). The statistical significance of the association between percent white and number of GSN app-using MSM varied between app and collection time (P<.001to P=.13); percent white was negatively associated with the number of GSN app users for app 1 and app 3 during the daytime and app 1 and app 2 during the weekend nighttime, but not for app 2 during the weekday daytime or app 3 during weekend nighttime.

The multivariable models are shown in Table 3. In all combinations of app and collection times, number of GSN app-using MSM was positively associated with population density (P<.001 to P=.045) and business zoning (P<.001) and negatively associated with median income (P<.001) and percent Hispanic (P<.001 to P=.045), adjusting for all other variables. The statistical significance of the association of median age (P<.001 to P=.46) and percent white (P=.02 to P=.77) with the outcome of interest varied between app and collection time in the final model; median age was positively associated with the number of GSN app users for app 1 during the weekday daytime but not for any other time or app. Percent white was negatively associated with the number of GSN app users for app 1 during the weekday daytime and app 2 during the weekend nighttime but not for any other time or app. In multivariable models with income, age, and population density log-transformed (data not shown in table), the results were similar except that in both of the models for app 2 and app 3, median age became statistically significant and in app 1 day and app 2 night, percent white lost statistical significance.

This study revealed that the presence of business zoning and population density were positively associated with the number of GSN app-using MSM during both weekday daytime and weekend nighttime for all GSN apps. We also found that median income and percent of the population who were Hispanic were negatively associated with the number of GSN app-using MSM during both weekday daytime and weekend nighttime for all GSN apps, adjusting for other variables in the model. Increased app use in areas with the presence of business zoning could imply that app users may be using these apps in areas of economic activity (eg, bars, restaurants, stores) including areas that cater to a predominantly gay clientele. This could highlight an important overlap between virtual and in-person partner-seeking spaces.

By using Wilcoxon signed rank sum tests and comparing choropleth maps of the differences in the spatial density of GSN app-using MSM, we determined that the total number of users between weekday daytime and weekend nighttime was not significantly different but that specific areas within the county could be experiencing changes in the number of partner-seeking MSM between these 2 time periods. This could imply either that the same users are migrating to different areas over time, different users at different locations are logging in at different times, or a combination of these. Previous research with GSN app-using MSM indicates that more than 50% log on 5 or more times per day with significantly greater percentages logging on during evening hours [55]; however, no studies to our knowledge have tracked use by time and geographic position. This information could be informative to local health departments because instead of using mobile HIV testing units only during weekend nighttime hours at nightlife venues, these units could also be used at specified hotspots during weekday daytime hours. These data may also be able to inform more cost-efficient geotargeted and temporally targeted recruitment strategies for research. However, more research is needed to explore attitudes about the presence of HIV outreach activities such as mobile HIV testing units or research recruitment efforts near daytime hotspots such as neighborhoods and places of business.

The novel methods of data collection used in this study addressed some of the limitations of past GSN app research. For example, prior research using GSN apps for data collection largely relied on a single app [12,14,23,26,41,42,56]. Our study used 3 apps, and the variance in spatial distribution of use and number of users across apps highlights the importance of using multiple apps in future research. Popularity of GSN apps shifts over time, and some apps are more popular among specific subgroups of MSM [57]. Previous research has also aggregated geographic data across time points rather than examining daytime and nighttime use separately [23]. Our study, which examined weekday daytime and weekend nighttime use separately, revealed that spatial patterns of use may vary by time of day.

Finally, data on the use of GSN apps is also novel in the context of midsize cities as most studies to date have focused solely on larger cities [13,15-17,20,22-24,26]. Collecting data on GSN apps for MSM in midsize cities could be of increased importance due to differences in social context [29-33,38,39] and the way in which users interact with the app in contrast with urban GSN app users [37]. This is especially important given the disparities in resources for sexual and gender minority people in rural areas compared to large urban centers. Prior research has demonstrated the importance of technology in the lives of rural sexual and gender minority people in combating social isolation and homophobia in public social spaces [37-41]. In general, more research on the role of technology in fostering resilience among MSM in rural areas is needed. However, given the literature linking GSN app use and sexual risk behavior, a closer examination of how these dynamics play out in rural settings is crucial to developing targeted HIV prevention efforts for rural MSM.

Although the maps demonstrated substantial differences in estimated GSN app usage throughout the study area, more data collection points could have improved the accuracy of the EBK estimation. Also, our study compared differences between weekday daytime and weekend nighttime but did not include weekday nighttime hours as a sampling window. The sampling time window was large, more so for weekday daytime hours (6-hour window) than weekend nighttime hours (4-hour window); future studies may benefit from narrowing the time window of data collection; although this may require data collection occurring over more calendar time depending on personnel. Additionally, this study used a multiapp approach for data collection, but there were other GSN apps that were not used for this study. At the outset of the study, we contacted 6 app companies to notify them of our intent to do the study and dropped 3 apps due to the fact that it violated their user agreements.

In respect of app users’ privacy, we did not record any identifying information about GSN app users, making it impossible to compare demographic characteristics of users with ecological variables based on census data. This also prevented us from determining the extent to which GSN app-using MSM were using more than 1 app and whether the differences in spatial distribution were due to the temporal migration of the same users to different locations or if other app users in different locations were logging on at different times. Although the choice not to collect identifying data resulted in analytic limitations, we believe that this was an important tradeoff to make in respect of user privacy and app company policies. Ethical issues surrounding the collection and use of GSN profile data for research are complex, rapidly evolving, and warrant increased attention. In the interim, we believe it is important for researchers to communicate with GSN app companies and take steps to prevent violation of user privacy.

The number of GSN app users within a 1-mile radius of the data collection points ranged from 0 to 36 during weekday daytime hours to 0 to 39 during weekend nighttime hours. In the multivariate analysis, the number of GSN app-using MSM was positively associated with business zoning and population density and negatively associated with median income and percent Hispanic residents. Methodology using geospatial data and area demographic data can lead to insights for tailored and targeted interventions to support better health outcomes in underserved populations.