Article Text

Original article
Condom deserts: geographical disparities in condom availability and their relationship with rates of sexually transmitted infections
  1. Enbal Shacham,
  2. Erik J Nelson,
  3. Lauren Schulte,
  4. Mark Bloomfield,
  5. Ryan Murphy
  1. College for Public Health and Social Justice, Saint Louis University, St Louis, Missouri, USA
  1. Correspondence to Professor Enbal Shacham, Department of Behavioural Science and Health Education, College for Public Health and Social Justice, Saint Louis University, 3545 Lafayette Avenue, Office 315, St Louis, MO 63104, USA; eshacham{at}slu.edu

Abstract

Background Identifying predictors that contribute to geographical disparities in sexually transmitted infections (STIs) is necessary. This study assesses the spatial relationship between condom availability to locations of STIs in order to better understand these geographical disparities.

Objectives We conducted a condom availability audit among potential condom-selling establishments. New gonorrhoea and chlamydia cases in 2011 (n=6034) and HIV infection cases from 2006 to 2011 (n=565) were collected by census tract in St Louis, Missouri. 829 potential condom-selling establishments participated in the condom availability audit in St Louis City; 242 of which sold condoms.

Results A negative linear relationship exists between condom vendors and cases of gonorrhoea and chlamydia, after adjusting for concentrated disadvantage and free condom locations. Higher concentrated disadvantage, higher proportions of convenience vendors and free locations were associated with higher rates of HIV.

Conclusions This study was conducted to provide evidence that lack of condom availability is associated with STI rates, and likely is an integral component to influencing the subjective norms surrounding condom use and STI rates. Condom distribution interventions may be addressing availability needs and social norms, yet are more likely to be effective when placed in locations with the highest STI rates.

  • CONDOMS
  • EPIDEMIOLOGY (CLINICAL)
  • GONORRHOEA
  • CHLAMYDIA INFECTION
  • HIV

Statistics from Altmetric.com

Introduction

Sexually transmitted infections (STIs), including HIV infection continue to be a public health concern. There are up to 3 million new infections of chlamydia and gonorrhoea and an estimated 50 000 new cases of HIV each year in the USA.1 ,2 STIs and HIV infection occur at higher rates among young individuals, African Americans, in urban settings and in impoverished areas.3 In addition to the behavioural risk factors of having an STI, individuals with STIs are at higher risk for HIV infection, due to ulcerative lesions and viral susceptibility.4 In addition, these infections tend to cluster geographically due to partnership formation within sexual networks, by sociopolitical factors and access to specialised STI healthcare.5 ,6

Consistent and correct use of condoms and monogamous relationships can prevent incident STIs among populations that are sexually active.7 Behavioural interventions to prevent STIs/HIV have not been particularly effective, as incident STIs/HIV infections persist. Biomedical interventions have been found to be effective in preventing HIV transmission, adherence to medication among individuals with HIV8 and pre-exposure prophylaxis among individuals at risk for HIV infection.9–11 As biomedical prevention efforts are being disseminated, efforts to improve behavioural interventions continue.

There is a need to better understand the real-world challenges related to sexual health that occur outside of research settings, which limit the effectiveness of behavioural STI/HIV evidence-based interventions.12 This is partially due to the lack of consideration of the environmental context, specifically where condom use decisions are made. Individual factors associated with unprotected sex such as younger age,13 heavy alcohol and illicit drug use,14 ,15 increased psychological distress,16 ,17 and racial discrimination18 have been well documented. Similarly, higher rates of STIs/HIV infections in individuals have been observed among those with older sex partners,19 ,20 the unemployed21 and living in neighbourhoods with higher concentrations of STIs/HIV infections.22 However, there are few studies that have examined the association between environmental determinants of condom use, such as condom availability and accessibility, and STIs. Identifying environmental determinants that potentially hinder condom use is critical in order develop appropriate interventions to reduce STIs/HIV infections.23

Geographical variation occurs in STI/HIV infection rates, with higher rates occurring in areas with higher concentrated disadvantage.22 ,24–26 Similarly, the geographical patterns of risk factors by location have been associated with STI/HIV infections. Studies have found that sexual partner selection within the same geographical region occurs more often among high HIV prevalence neighbourhoods,22 higher presence of gay communities were associated with lower rates of unprotected sex,27 higher proportions of monogamous relationships vary by region28 and earlier sexual initiation also varies across geographical regions.29 These studies highlight the variation of sexual health norms in neighbourhood contexts, suggesting that more research needs to be done to understand the pattern of risk factors and environmental determinants of sexual health within these geographical areas.

The purpose of this study was to assess the geographical relationship between condom availability and STI/HIV infection rates. Previously, we identified patterns of condom vendors (ie, condom deserts) which highlighted wide geographical variation in condom availability and accessibility in St Louis City, Missouri (E Shacham, EJ Nelson, L Schulte, et al. Condom Deserts: Evidence for Geographic Disparities in Condom Availability and Accessibility. 2015; under review). We hypothesised that the lack of available condoms within geographical regions are associated with higher rates of STIs.w1

Materials and methods

Data sources

Sexually transmitted disease rates

Diagnosed cases of gonorrhoea, chlamydia and HIV were counted and aggregated to the census tract level in St Louis City by the Missouri Department of Health and Senior Services.w2 Rates of diagnosed gonorrhoea and chlamydia cases reported in 2011 were calculated separately by summing the total number of cases in each census tract and dividing by the total population of the corresponding census tract. Of the 9989 gonorrhoea cases reported in St Louis City, only 4053 (40.6%) occurred within our a priori study area of St Louis City County. Similarly, 1981 out of the 3736 chlamydia cases (53.0%) occurred within the study area. Notably, 740 (7.4%) gonorrhoea cases and 234 (6.3%) chlamydia cases did not have a geographical identifier and were not included in the analysis. The diagnosed rate of reported HIV cases from 2006 to 2011 was also calculated for each census tract by dividing the number of HIV cases by the total census tract population. Of the 1365 cases of HIV that were reported in St Louis City, 565 (41.4%) occurred within the study area. Of all HIV cases reported in St Louis City, 374 (27.4%) did not have a geographical identifier and were not included in the analysis. Finally, we used indirect standardisation to calculate the standardised incidence ratio (SIR) of gonorrhoea, chlamydia and HIV for each census tract, which represents the observed disease rate relative to the expected disease rate.w3

Condom vendor locations

The location of 242 businesses that distribute condoms in St Louis City were identified during an extensive business audit and are described elsewhere (E Shacham et al, 2015; under review). Briefly, businesses were identified through an online audit and contacted via telephone to ascertain information regarding condom availability, pricing and where condoms were sold within the store. This audit included a total of 1262 potential locations that sold condoms; 829 participated in the audit which identified a total of 242 businesses that sold condoms. Businesses were categorised as (1) free locations, where condoms were free of charge; (2) convenience vendors (defined as gas stations, convenience stores, liquor stores, bars and pharmacies) or (3) other (defined as beauty salons, barbershops, grocery stores and retail stores). This resulted in 49 free locations, 170 convenience locations and 61 other locations. These categories were mutually exclusive. For example, if a vendor provided free condoms, they were not included in the convenience or other locations categories. Free condom locations were part of a condom distribution campaign sponsored by the City of St Louis’ Department of Communicable Diseases, and included health clinics, barbershops and bars where condoms could be obtained free of charge. All locations were geocoded and mapped using ArcGIS V.10.2.2 (ESRI, Redlands, California, USA). As has been done in other studies of the built environment, we counted the number of condom vendors and free condom locations within a 0.4 km buffer of each census tract.w4 The mean number of vendors in each census tract was 4.96 (SD=2.8; min=0; max=19) for condom vendors and 0.93 (SD=1.4; min=0; max=15) for free condom locations. The number of condom vendors and free locations were then divided by the total number of businesses that distributed condoms to estimate the proportion of condom vendors and free locations within each census tract. The initial analyses suggested that there are areas within the study region that have limited or no available condoms, and those with availability have low accessibility, condoms being sold behind the counter, selling only single condoms, and no choice in brand. These results identified a ‘condom desert’ effect, where there is limited availability and accessibility of condoms.

Concentrated disadvantage

We constructed an index of concentrated disadvantage using principal component common factor analysis of eight variables using 5-year estimates from the 2008 to 2012 American Community Survey.w5 Based on the literature and prior research,w6–w8 the census variables included were the proportion of the African-American population, proportion of female-headed households, proportion of households receiving food stamps, proportion of individuals using public health insurance programmes (all ages and races), the proportion of households with children under age 18 years, percentage of households without employment during the past 12 months, proportion of households below the federal poverty line and the median household income (mean-centred). This factor explained 74.9% of the total variance of the eight variables. This variable was standardised and weighed by a factor loading coefficient for use in the regression models described below.

Statistical analysis

We conducted analyses separately using cases of gonorrhoea, chlamydia and HIV, respectively, as the dependent variable. Given the nature of our spatially dependent data, traditional regression based approaches (eg, Poisson or negative binomial regression) would likely violate the independence assumption required for generalised linear models. Ignoring the dependence of spatial data can underestimate SEs, resulting in overly narrow CI estimates and, consequently, incorrect statistical inference.w9 Thus, we first examined the nature of the relationship between condom availability and disease rates by testing for the presence of spatial autocorrelation using Moran's I.w10 Values of I greater than (less than) the expected value indicate positive (negative) spatial autocorrelation: nearby census tracts tend to exhibit similar (or dissimilar) rates of diseases.

Due to the presence of spatial autocorrelation, we fit sparse spatial generalised linear mixed models,w11 which greatly improves regression inference relative to the traditional areal mixed modelw12 because it is not spatially confounded.w13 Under a Bayesian setting, the current analyses involved the use of the sparse spatial generalised linear mixed models to perform spatial Poisson regressions with offset, where the offset for the ith census tract was the population of the census tract. Using Bayesian modelling approaches, the prior distributions recommended by Hughes and Haranw11 were included, and we chose the first q=39 Moran eigenvectors, which virtually exhaust the possible patterns of spatial attraction for the St Louis City County adjacency structure. We drew 100 000 samples from the posterior distribution. The resulting Monte Carlo SEs were of the order of 0.0001, which indicates convergence of the Markov chain.w14 We used the package ngspatialw15 in V.3.0.1 of R (R Development Core Team, 2013) for all analyses. We first modelled the bivariate relationship of the outcomes with convenience vendors, then fit multivariate models adjusting for concentrated disadvantage and free condom locations. We then exponentiated the model coefficients (ie, exp(β)) in order to present relative risks (RRs) with their corresponding 95% credible intervals. We also report choropleth maps of the adjusted SIR (using the predicted values from the multivariate model adjusted for all confounding variables and spatial dependence) to show the census tracts with extreme high (low) disease rates.

Results

A total of 4053 gonorrhoea cases, 1981 chlamydia cases and 565 HIV cases occurred within our a priori study area of St Louis City County. Condom availability was assessed in 280 businesses throughout the study region, including 49 locations where condoms were freely available.

Figure 1 presents a graphical display of diagnosed gonorrhoea, chlamydia and HIV rates, as well as the distribution of condom vendors across census tracts in the city. Table 1 suggests the presence of significant autocorrelation for cases of gonorrhoea, chlamydia and HIV, and therefore requires the use of spatially dependent modelling approaches. A negative linear relationship exists between condom vendors and cases of gonorrhoea (RR=0.75; 95% CI 0.61 to 0.92) and chlamydia (RR=0.60; 95% CI 0.43 to 0.83), after adjusting for concentrated disadvantage and free condom locations (table 2). However, this association was not significant for HIV cases (RR=1.04; 95% CI 0.66 to 1.63).

Table 1

Global test of spatial autocorrelation in sexually transmitted infections: Moran's I

Table 2

Examining the intersection of sexually transmitted infections and condom vendors using spatial Poisson regression

Figure 1

Crude rates of diagnosed gonorrhoea, chlamydia, and HIV infections overlaid with the distribution of condom vendors across the study area.

Higher concentrated disadvantage was associated with higher rates of gonorrhoea (RR=2.01; 95% CI 1.92 to 2.09), chlamydia (RR=2.09; 95% CI 1.96 to 2.23) and HIV (RR=1.16; 95% CI 1.06 to 1.27). Higher proportions of free condom locations was associated with higher risk of gonorrhoea, chlamydia and HIV, however the association was only statistically significant for HIV cases (RR=3.52; 95% CI 1.84 to 6.81). Figure 2 presents choropleth maps of the SIRs from the findings contained in table 2 (ie, the fitted values of the fully adjusted spatial Poisson models). Qualitatively, higher diagnosed rates of gonorrhoea, chlamydia and HIV infection tend to occur in the northern region of the city, with lower disease rates tending to occur in the southern part of the city. Similarly, free condom locations tend to be located in the central and northern regions of the city, while convenience and other condom vendors were more evenly distributed throughout the study area.

Figure 2

Choropleth maps of the SIRs (Susceptible, Infected, or Recovered) within the study area by STI.

Discussion

This study found limited availability of condoms to be significantly associated with higher STI rates. This finding supports the notion that ‘condom deserts’ do indeed exist and may partially explain disparities in STI rates (E Shacham et al, 2015; under review). The observed geographical patterns indicate that condoms are less likely to be available in disadvantaged areas, which is likely to fuel the known disparity in STIs in these neighbourhoods.22 To our knowledge, condom availability and its relationship with STI/HIV rates have not previously been studied in this manner. Thus, understanding the role of condom access points within neighbourhoods may serve to develop more effective interventions within high-risk communities. Furthermore, increasing our understanding of the potential influence of condom deserts on STI/HIV infections could provide insight into social norms and neighbourhood contexts that influence condom use decision making. More research is needed to understand how condom availability and importantly, condom access within stores, influences the social norms and perceptions of condom use within communities.

Higher gonorrhoea and chlamydia infection rates occurred more often in areas with less condom availability. The bivariate relationship between HIV and condom vendors also suggested that the risk of HIV infection decreased as the proportion of condom vendors increased. However, the relationship was not statistically significant after adjusting for concentrated disadvantage and free condom locations. This may be partially explained by the efforts of St Louis City to position free condom locations in areas with historically higher risks of HIV infection. If so, then the association between HIV infection and condom vendors may have been biased towards a null finding. Increasing the sale of condoms in traditional outlets (eg, convenience stores) where they have not previously been available is likely to increase their use and prevent the spread of HIV (and other STIs).

Much of the previous condom use research conducted has focused on individual and interpersonal factors, which claim that condom use is primarily influenced by planning and types of sex partners.w16 Accessibility issues related to condom purchasing serve as barriers to use as well,w17 yet availability has not been examined in this manner. This study aimed to add to that body of literature in order to explain more of the gap between using condoms and condom use intent. By examining neighbourhood-level condom availability and accessibility, we were able to determine that condom availability varies greatly across census tracts. Condom distribution campaigns have been found to increase use, delay sexual initiation among youth and reduce incident STIs.w18 In this study, locations of free condoms were associated with lower STI/HIV infection rates. It is worth noting that the City of St Louis’ Department of Communicable Diseases initiated a condom distribution campaign during the business audit. The main focus of the campaign was to reduce HIV incidence rates through condom distribution. Findings from the current study suggest that condom distribution was appropriately targeted in geographical areas with the highest rates of HIV infection. As these programmes continue to improve condom availability, our findings may serve to inform where these programmes should position new condom locations.

Opportunities also exist for the development of sustainable partnerships between private and public sectors to improve access to condoms in order to reduce STI/HIV rates. For example, encouraging private businesses to (1) sell condoms and (2) to offer condoms in more accessible (open) locations within stores may increase purchasing patterns. In addition, perceived social norms regarding condom use and sexual health may potentially be altered as businesses participate in making condoms more available and accessible.w17 w19 It is important to note that condoms need to be available if they are to be used, yet that alone is not a comprehensive solution to prevent infections. We hypothesise that sexual norms are influenced by the availability of these resources and that making these resources available will alter sexual norms. Future research is needed to understand how sexual norms can be influenced through location-based condom accessibility across geographical areas.

This study is not without limitations. The cross-sectional ecological design does not permit causal inference nor does it permit a temporal sequencing of events. Furthermore, relying upon an audit to determine condom availability comes with limitations; specifically, stores in low income areas tend to close down in shorter time frames. Thus, condom availability may be in consistent over time; although during the inperson validity check we found that most new businesses that had opened in place of old stores that offered condoms continued to do so. In addition, a substantial number of diagnosed gonorrhoea, chlamydia and HIV cases did not have geographical identifiers and were not included in this study. Although, many of the HIV cases were not actually missing location information, but rather the infections did not occur within the study area. However, it is not surprising that many men and women did not disclose their address due to the sensitive nature of STI testing. HIV infection rates from correctional facilities were not specifically identified, which may pose a limitation, although there is clear lack of condom availability in these facilities. Additionally, the data were not differentiated for individuals who were co-infected with gonorrhoea and chlamydia. Finally, data were available for only one time period and we were not able to examine annual trends of these geographical relationships.

Conclusions

In conclusion, this study highlights the relationship between condom deserts and gonorrhoea, chlamydia and HIV infection across census tracts in St Louis, Missouri. Prevention efforts designed to increase condom availability and promote condom use should consider geographical disparities to design innovative approaches to addressing inequities in STI/HIV infection rates and access to health resources.

Key messages

  • Geographical regions with less available condoms are associated with higher gonorrhoea/chlamydia and HIV infection rates.

  • Limited visibility of condoms likely impacts norms related to condom use, and is thus related to higher rates of sexually transmitted infections (STIs).

  • This initial study identified condom availability as a determinant of geographical disparities of sexual health outcomes, specifically STI/HIV infections.

Acknowledgments

The authors acknowledge the research participation of The Missouri State Department of Health and Human Services; Courtney Brightharp, MPH; Max Holtz, MPH; Lauren Ho, MPH; and Elizabeth Baney, MPH and the Department of Behaviour Science and Health Education for support for this study.

References

Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

Footnotes

  • Handling editor Jackie A Cassell

  • Contributors ES conceived of the research question, developed the research design, managed the data collection and analysis, and wrote the manuscript. EJN conducted the geospatial analyses, wrote sections of the manuscript and revised the manuscript. LS collected the condom availability data, wrote sections of the manuscript and revised the manuscript. MB assisted the geospatial analyses, wrote sections of the manuscript and revised the manuscript. RM assisted the geospatial analyses, wrote sections of the manuscript and revised the manuscript.

  • Competing interests None declared.

  • Ethics approval The Institutional Review Board at Saint Louis University, State of Missouri Department of Health and Human Services (#201209019).

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data sharing statement The STI/HIV infection data were collected from the State of Missouri Health and Human Services Department. The condom availability and accessibility audit data are available from the author.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.