Recent national data (2014–2017) show annual increases in all 3 reportable bacterial sexually transmitted infections (STIs) (chlamydia, gonorrhea, and syphilis) in the United States (US).1 These increases occurred in all regions and among both sexes.1 The increases in STI burden imply increases in the associated economic burden on the US health care system (over US $700 million annually in direct medical costs for chlamydia, gonorrhea, and syphilis infections).2 The increases in the burden of these STIs suggest the need for upgrading––and in some cases changing––prevention infrastructure to effectively meet the existing (and growing) needs. For planning, designing, and implementing effective (and/or cost effective) STI prevention and control interventions, it is important to understand the relative burden of the infection(s) across race (racial disparity). Additionally, because race and geography (or location) are closely related when considering STI burden,3,4 it is equally important to examine the disproportionate burden across geography (spatial disparity). Therefore, studies that can improve our understanding of spatial disparities (or patterns) in STI burden are needed.
A number of studies have used spatial analytical methods and geographic information systems to identify high concentrations of STI burden in different communities using different geographic units.3–12 In 2017, Chang et al.13 used a combination of hot spot analyses (HSA) and spatial logistic regression techniques to identify clusters of relatively high reported bacterial STIs (chlamydia, gonorrhea, and syphilis) at the county-level in the US, and examined the correlates of the 3 STI hot spots. Their analyses focused on the 3 STIs individually. As a result, detail spatial information on the extent of the intersection/overlaps between the hot spots among the 3 STIs (multiple-STI hot spots) was not reported. In this study, we add relevant information to Chang et al.’s study by analyzing data from the same years to:
- Identify and map overlapping county STI hot spots (multiple-STI hot spots), and
- Examine the correlates of the multiple-STI hot spot classes.
Following Chang et al.,13 we assembled 5 years (2008–2012) data on the reported total county-level cases of chlamydia, gonorrhea, and primary and secondary (P&S) syphilis on all the counties in the contiguous states in the US from the AtlasPlus web tool.14 Next, we computed temporally smoothed rates as the sum of all cases divided by the sum of the resident population across the years, and then multiplied the result by 100,000 for each county.15,16
We obtained 5-year estimates (2008–2012) of county-level economic and sociodemographic data from American Community Surveys.17 Violent crime data (2008–2012) were obtained from the Uniform Crime Reporting Program Data webpage18 and converted to temporarily smoothed rate (number/1000 residents) using the method described above. We also obtained the 2013 rural-urban continuum codes to categorize counties as “urban” (i.e., rural-urban continuum codes 1–3) and “rural” (i.e., rural-urban continuum codes 4–9) counties.19 See the footnotes for Table 1 for more details.
Hot Spot Analyses
The HSAs were performed using the Getis-Ord-Gi statistic, which provided z scores.20 The spatial relationships between the counties was operationalized using the first-order queen contiguity weights.21 We also applied the false discovery rate correction to account for multiple testing and spatial dependency as well as increase the power of the analyses.22 For each STI, counties with high z score statistic (> + 1.64) were categorized as “hot spot” and all the other counties with low (negative) and non-significant z scores were categorized as “cold spot/not-significant” for the purpose of this study. Next, we mapped the individual STI hot spots. We then determined and mapped 4 multi-STI hot spot classes based on the number of individual STI hot spots identified for each county (0 through 3) by superimposing all the 3 individual STI hot spot county maps.
Spatial Regression Analyses
The selection of the correlates (i.e., economic and sociodemographic variables) for the spatial regression was based on numerous previously published works on bacterial STIs.3–6,10,11,13,15,16,23–26 The economic and sociodemographic variables included percent black (non-Hispanic), percent white (non-Hispanic, referent race category), percent Hispanic/Latino, percent Asian, percent American Indian and Alaska Native (AIAN), percent Native Hawaiian and Other Pacific Islander (NHOPI), income inequality (i.e., Gini coefficient), birth rate, population density (i.e., residents/square mile), sex ratio (i.e., male-to-female ratio), commute index (i.e., percent with >1 hour commute time), percent of residents aged 15 to 24 years and 25 to 44 years.
Because our likelihood-ratio test for proportionality of odds showed that the proportional odds assumption was not violated, we applied an ordered spatial logistic regression technique with a spatial error model (SEM) specification25,27 in which the multi-STI hot spot classes variable generated above (0 for no hot spot counties; 1 for counties determined to have 1-STI hot spot; 2 for counties determined to have 2-STI hot spots, and 3 for counties determined to have all 3-STI hot spots) was the dependent variable. Because almost all the hot spots were in the south, we created and included a binary variable for the south––equal to 1 if the county was from a southern state, 0 otherwise (included the counties from the West, Midwest, and Northeast regions) to compare the counties in the south to those in the other regions. Finally, to account for the rural-urban disparities in STI morbidity in the south, we included 2 categorical variables––“south–urban” and “south-rural” based on the rural-urban continuum categories described above.
Following previous studies and results from our preliminary regression analyses, we excluded Florida and Illinois from our final analyses because of the lack of reliable data on crime,18 which was included to improve our regression results.24,26 Consequently, the data for the HSA included information on counties from 46 contiguous states and the District of Columbia (for the purpose of this study, the District of Columbia was categorized as a county), bringing the total number of counties (including county equivalents) to 2935 (>93% of the counties in the US). Because of incomplete data from one of the counties in Arkansas, it was dropped from the regression analyses. As a result, the number of observations/counties used in the ordered spatial logistic regression analyses was 2934 (>93% of the counties in the US).
We used ArcMap version 10.3 (ESRI Redland, CA) for mapping and conducting the HSA for all the counties in the 46 contiguous states and the District of Columbia. We conducted the spatial regression analyses with GeoDa version 0.9.5-I28 and STATA version 14.2 (StataCorp LP, College Station, Texas). We checked the potential for multicollinearity issues using the calculated variance inflation factors (VIF).
STI Hot Spots
Based on our methods and assumptions, the hot spot counties that were identified for each STI are depicted in Figures 1A–C. Of the 2935 counties that were used, there were 263 chlamydia, 324 gonorrhea, and 155 P&S syphilis hot spot counties (Fig. 1). When we overlaid the individual STI county hot spot maps to identify counties with multiple-STI hot spot, we found that 85 were hot spots for all 3 STIs (3-STI hot spot counties), 177 were hot spots for 2 STIs (2 STI-hot spot counties); 145 were hot spots for only 1 STI (1-STI hotspot county), and 2528 were not hot spots for any of the STIs (0 STI-hotspot counties). Figure 2 depicts the resulting overlaid map generated from the individual STI county hot spot maps displayed in Figure 1. We found that approximately 93% (79 of 85) of the counties that were determined to be 3-STI hot spots were found in 4 southern states with Mississippi having the largest number followed by Arkansas––Mississippi (n = 25), Arkansas (n = 22), Louisiana (n = 19), and Alabama (n = 13). The remaining six were in Texas, Georgia, and North and South Carolina. Counties determined to be 2-STI hot spots were found in seven southern states––Arkansas, Louisiana, Mississippi, Alabama, Georgia, and North and South Carolina had at least ten 2-STI hot spot counties each (Fig. 2).
Ordered Spatial Logistic Regression
Table 1 depicts the summary statistics of the select economic and sociodemographic variables used in the ordered spatial logistic regression analyses. The summary results of the ordered spatial logistic regression are presented in Table 2. Model 1 included 1 regional categorical variable (south), while model 2 included 2 regional-rural/urban categorical variables (south-urban and south-rural). The quantitative and qualitative results (estimated coefficients of the sociodemographic variables) were consistent across models 1 and 2 (Table 2). The estimated VIFs were less than the suggested limit of 1029 (average = 1.31; highest = 1.85).
The results indicated that a 1 percentage point increase in the percentage of Black non-Hispanic residents was associated with a 3.2 (P < 0.01) percent increase (95% confidence interval [CI], 2.5–4.0) in the odds of being in a one-step higher multi-STI hot spot class. A 1 percentage point increase in the percentage of Hispanic residents was associated with a 1.6 (P < 0.01) percent increase (95% CI, 0.8–2.4) in the odds of being in a one-step higher multi-STI hot spot class. A 1 percentage point increase in the percentage of AIAN residents was associated with a 3.3 (P < 0.01) percent increase (95% CI, 1.8–4.9) in the odds of being in a one-step higher multi-STI hot spot class. A 10-unit increase in the population density was associated with 0.9 (P < 0.01) percent decrease (95% CI, −1.4 to −0.4) in the odds of being in a one-step higher multi-STI hot spot class. A 10-unit increase in the sex ratio (i.e., the number of males per 100 females) was associated with 11.2 (P < 0.05) percent decrease (95% CI, −20.5 to −0.8) in the odds of being in a one-step higher multi-STI hot spot class. A 1 percentage point increase in the percent of residents aged 25 to 44 years was associated with 4.5 (P < 0.05) percent increase (95% CI, 0.4–8.7) in the odds of being in a one-step higher multi-STI hot spot class. A unit increase in the violent crime rate was associated with 4.7 (P < 0.05) percent increase (95% CI, 0.3–9.3) in the odds of being in a one-step higher multi-STI hot spot class.
The odds of being in a one-step higher multi-STI hot spot class was 4 (P < 0.01) times higher (95% CI, 2.8–5.9) for the counties in the south when compared with the counties in the other regions (West, Midwest, and Northeast). Furthermore, the odds of being in a one-step higher multi-STI hot spot was 4.5 (P < 0.01) times higher (95% CI, 3.0–6.7) for urban counties in the south and 3.7 (P < 0.01) times higher (95% CI, 2.5–5.5) for rural counties in the south when compared with the counties in the other regions. However, we did not find any significant difference between the estimated odds ratios for the south-urban versus south-rural.
On the other hand, our analyses indicated that percent Asian, percent NHOPI, income inequality (i.e., Gini coefficient), birth rate, commute index (i.e., percent with >1 hour commute time), and percent of residents aged 15–24 years were not significant at the 5% significance level.
In this study, we analyzed 5 years (2008–2012) of cross-sectional data and used a combination of HSA and ordered multivariate spatial logistic regression analyses to identify and examine the correlates of multi-STI hot spot counties for the 3 most commonly reported STIs (chlamydia, gonorrhea, and syphilis) within 46 contiguous states (and the District of Columbia) in the US. Consistent with Chang et al.’s study and reported STI morbidity data,1,13 a large majority of the hot spot counties were found in the south. Additionally, the counties that were determined to be 3-STI hot spots were largely concentrated (>90%) in 4 southern states (Mississippi, Arkansas, Louisiana, and Alabama). Seven states (Arkansas, Louisiana, Mississippi, Alabama, Georgia, and North and South Carolina) had at least ten 1-STI hot spot counties each. These findings are consistent with previous reports/studies of high-burden clusters of county bacterial STIs in the southern states.1,6,8,12,15,16,25
Our results also showed that the proportion of the resident minority populations (black [non-Hispanic], Hispanic, and American Indian) were positively associated with the odds of being in a higher multi-STI hot spot county. This positive association between the proportion of the resident minority populations and the odds of being in a higher multi-STI hot spot county is likely the result of racial segregation, because segregated minority populations suffer from poor education, economic hardship, increased drug/alcohol use, and decreased access to health care, all of which can lead to increased risk of STI transmission.13,23,30
Consistent with Chang et al.'s recent study,13 we found that population density was negatively associated with the odds of being in a higher multi-STI hot spot class, which could be expected as most of the counties with the highest population density (i.e., New York, King, or Bronx counties) were not identified as hot spots.13 We also found that counties in the southern states were at least 3 times more likely to be in a higher multi-STI hot spot class than counties in other regions. The male-female ratio was negatively associated with the odds of being in a higher multi-STI hot spot class, likely because most of the STIs were among females given the increased testing in this population. The proportion of the population aged 25 to 44 years was positively associated with the odds of being in a higher multi-STI hot spot class. In part, this finding is likely because the age group with the highest risk for syphilis infection is older.1 Violent crime rate was positively associated with the odds of being in a higher multi-STI hot spot class. Violent crime has been shown to be positively associated with STIs, and has consequently been suggested as a proxy for hard-to-measure social determinants.24 In fact, it has been demonstrated that including violent crime rate in ecological STI regression analyses can improve model quality.26
Just as in Chang et al.’s study,13 several limitations in this study are worth mentioning. Because the STI morbidity data used in this study were from the national STI surveillance data, all the limitations associated with STI surveillance data (missing/unknown data, inconsistent testing, and reporting practices) are applicable.1 Additionally, similar to the Chang et al.’s study,13 these analyses only provide a national-level insight into the distribution of county multiple-STI hot spots. Finer, small-scale analyses (such as census tracts/blocks or cities) would provide information that is more detailed. We excluded data from Florida and Illinois from our analyses due to lack of consistent data on violent crime rate. However, based on Chang et al.'s13 study and results from our preliminary analyses, we did not lose much information because there were very few STI hot spot counties in both states––Florida had less than 3% of the hot spots (3 hot spot counties for chlamydia, 10 for gonorrhea, and 6 for syphilis); Illinois had no (zero) hot spot counties.
This study adds to the literature on bacterial STI hot spot counties in the US. Although Chang et al.,13 determined and mapped individual STI hot spots for all 3 bacterial STIs, there was no spatial analyses on the overlapping hot spots. This study used HSA to identify clusters of relatively high STI rates, determined multi-STI hot spot counties by overlaying all 3-STI hot spot maps, and then used ordered spatial logistic regression techniques to examine population-level associations with the multi-STI hot spots using county-level data on over 93% of the counties in the US (from 46 contiguous states and the District of Columbia). This study used publicly available data; therefore, our results can be replicated. We used data around 2010 (a decennial census year), which increased the reliability of the denominators (population estimates) because of population counts in the 2010 census. Additionally, we used temporally smoothed rates for each STI to increase the robustness of the measure of morbidity of all the STIs for each county. Correcting for multiple testing and spatial dependence when conducting the HSA increased confidence in the hot spot counties we identified.22 Finally, we used the appropriate and validated technique (i.e., spatial regression), which reduced potential problems of spatial dependence/autocorrelation.27
To determine and examine the correlates of multiple bacterial STI (chlamydia, gonorrhea, and syphilis) hot spot counties in the US, we analyzed five-year (2008–2012) cross-sectional county-level data from 46 contiguous states (including the District of Columbia) using HSAs and ordered spatial logistic regression. The disproportionate concentration of 3-STI hot spots (over 90%) in 4 southern states (Mississippi, Arkansas, Louisiana, and Alabama) further illustrates the geographic disparities related to bacterial STIs in the US. Our regression analyses show that the proportion of black (non-Hispanic), Hispanics, and American Indians were positively associated with multi-STI hot spot class. Additionally, population density and male-female sex ratio were negatively associated with multi-STI hot spot class, while the proportion of the population aged 25 to 44 years and violent crime rate were positively associated with multi-STI hot spot class. The results from our analyses can help to identify counties with multiple-STI hot spots, which can be used to assist with planning, designing, and implementing effective (and/or cost effective) integrated STI prevention and control interventions.31 It may also assist in deciding where multiple STI tests (such as dual chlamydia-gonorrhea tests) can be most effective (and/or cost effective), recognizing that within these multi-STI hot spot counties, the high-risk populations may be different for each STI (e.g., men who have sex with men are the high-risk population for syphilis, while young males and females are the high-risk population for chlamydia and gonorrhea8). Future studies may consider using more recent data for smaller geographic units such as census tracts/blocks or cities.
1. Centers for Disease Control and Prevention. Sexually transmitted disease surveillance, 2017. Atlanta, GA: U.S. Department of Health and Human Services, 2018.
2. Owusu-Edusei K Jr., Chesson HW, Gift TL, et al. The estimated direct medical cost of selected sexually transmitted infections in the United States, 2008. Sex Transm Dis 2013; 40:197–201.
3. Jennings JM, Curriero FC, Celentano D, et al. Geographic identification of high gonorrhea transmission areas in Baltimore, Maryland. Am J Epidemiol 2005; 161:73–80.
4. Kerani RP, Handcock MS, Handsfield HH, et al. Comparative geographic concentrations of 4 sexually transmitted infections. Am J Public Health 2005; 95:324–330.
5. Owusu-Edusei K Jr., Owens CJ. Monitoring county-level chlamydia incidence in Texas, 2004–2005: Application of empirical Bayesian smoothing and exploratory spatial data analysis (ESDA) methods. Int J Health Geogr 2009; 8:12.
6. Owusu-Edusei K Jr., Doshi SR. Assessing spatial gaps in sexually transmissible infection services and morbidity: An illustration with Texas county-level data from 2007. Sex Health 2012; 9:334–340.
7. Patel S, Tuchman L, Hayes K, et al. Spatial distribution of adolescents with sexually transmitted infections diagnosed in the pediatric emergency departments of Washington, DC. International Society for Disease Surveillance 2015.
8. Escamilla V, Hampton KH, Gesink DC, et al. Influence of detection method and study area scale on syphilis cluster identification in North Carolina. Sex Transm Dis 2016; 43:216–221.
9. Gesink D, Li Y. What can we infer about incarceration and sexually transmitted diseases? Sex Transm Dis 2015; 42:329–330.
10. Gesink D, Wang S, Norwood T, et al. Spatial epidemiology of the syphilis epidemic in Toronto, Canada. Sex Transm Dis 2014; 41:637–648.
11. Gesink DC, Sullivan AB, Norwood TA, et al. Does core area theory apply to sexually transmitted diseases in rural environments? Sex Transm Dis 2013; 40:32–40.
12. Chesson HW, Sternberg M, Leichliter JS, et al. Changes in the state-level distribution of primary and secondary syphilis in the USA, 1985–2007. Sex Transm Infect 2010; 86(Suppl 3):iii58–iii62.
13. Chang BA, Pearson WS, Owusu-Edusei K Jr. Correlates of county-level nonviral sexually transmitted infection hot spots s: Application of hot spot analysis and spatial logistic regression. Ann Epidemiol 2017; 27:231–237.
14. NCHHSTP AtlasPlus. Updated 2017. Available at: https://www.cdc.gov/nchhstp/atlas/index.htm
. Accessed November 15, 2018.
15. Owusu-Edusei K, Doshi SR. County-level sexually transmitted disease detection and control in Texas: Do sexually transmitted diseases and family planning clinics matter? Sex Transm Dis 2011; 38:970–975.
16. Owusu-Edusei K, Chesson HW, Leichliter JS, et al. The association between racial disparity in income and reported sexually transmitted infections. Am J Public Health 2013; 103:910–916.
17. American FactFinder: 2012–2016 American Community Survey. United States Census Bureau's American Community Survey Office, 2018. Available at: http://factfinder.census.gov/faces/nav/jsf/pages/download_center.xhtml#none
. Accessed February 20, 2019.
18. United States Department of Justice, Office of Justice Programs, Federal Bureau of Investigation. Uniform Crime Reporting Program Data: County-Level Detailed Arrest and Offense Data, United States 2008–2012. Inter-university Consortium for Political and Social Research [distributor], 2014.
19. Rural Urban Continuum. 2013. Available at: http://www.ers.usda.gov/data-products/rural-urban-continuum-codes.aspx
. Accessed January 10, 2019.
20. Getis A, Ord JK. The analysis of spatial association by use of distance statistics. Geographical Analysis 1992; 24:189–206.
21. Modeling spatial relationships. Environmental Systems Research Institute (ESRI) Inc., 2016. Available at: http://resources.arcgis.com/en/help/main/10.1/index.html#//005p00000005000000
. Accessed February 1, 2019.
22. Hot spot analysis (Getis-Ord Gi*). Environmental Systems Research Institute (ESRI) Inc., 2015. Available at: http://pro.arcgis.com/en/pro-app/tool-reference/spatial-statistics/hot-spot-analysis.htm
. Accessed February 1, 2019.
23. Farley TA. Sexually transmitted diseases in the southeastern United States: Location, race, and social context. Sex Transm Dis 2006; 33:S58–S64.
24. Chesson HW, Owusu-Edusei K Jr., Leichliter JS, et al. Violent crime rates as a proxy for the social determinants of sexually transmissible infection rates: The consistent state-level correlation between violent crime and reported sexually transmissible infections in the United States, 1981–2010. Sex Health 2013; 10:419–423.
25. Owusu-Edusei K, Chesson HW. Using spatial regression methods to examine the association between county-level racial/ethnic composition and reported cases of chlamydia and gonorrhea: An illustration with data from the state of Texas. Sex Transm Dis 2009; 36:657–664.
26. Owusu-Edusei K Jr., Chang BA, Aslam MV, et al. Does including violent crime rates in ecological regression models of sexually transmissible infection rates improve model quality? Insights from spatial regression analyses. Sex Health 2019; 16:148–157.
27. LeSage JP. Econometric toolbox. 2010. Available at: www.spatial-econometrics.com
. Accessed January 22, 2019.
28. Anselin L, Syabri I, Kho Y. GeoDa: An introduction to spatial data analysis. Geographical Analysis 2006; 38:5–22.
29. Belsley DA, Kuh E, Welsch RE. Regression Diagnostics. New York: John Wiley & Sons, 1980.
30. Aral SO. Social and behavioral determinants of sexually transmitted disease: Scientific and technologic advances, demography, and the global political economy. Sex Transm Dis 2006; 33:698–702.
31. Law DC, Serre ML, Christakos G, et al. Spatial analysis and mapping of sexually transmitted diseases to optimise intervention and prevention strategies. Sex Transm Infect 2004; 80:294–299.