HIV Testing Deserts and Vulnerability Among Adolescents and Young Adults in Tampa Bay, Florida

Abstract

Purpose

HIV burden among Florida adolescents and young adults (AYA, ages 13–24), particularly in Tampa Bay, is among the highest in the nation. We sought to determine the association between zip code-level test site accessibility and AYA HIV burden, compare this association with adult (ages 25–44) HIV burden, and identify local AYA HIV testing deserts. We further aimed to identify the association between test site accessibility and population-level markers of social disadvantage.

Methods

We geocoded HIV test sites and determined the percent surface area per zip code within 15-minute walking distance to ≥1 test sites (PSA15) in Pinellas and Hillsborough counties. We calculated Pearson’s correlation coefficients for the association of PSA15 and HIV burden by age group and, separately, the association of PSA15 and population-level characteristics.

Results

Of the 96 zip codes analyzed, 36.5% had a PSA15 of 0%. The association between PSA15 and HIV burden was substantially higher for adults (r=0.51, p<0.001) than AYA (r=0.09, p=0.38). Overall, we identified 4 potential AYA testing deserts. We also found that greater PSA15 was correlated with greater %Black/African-American residents (r=0.32, p=0.002), greater %residents living in poverty (r=0.27, p=0.008), and lower child opportunity index scores (r=−0.29, p=0.004).

Discussion

Walking-accessible HIV test sites in Tampa Bay were limited and geographically distributed largely based on adult HIV burden, population-level markers of social disadvantage, and among areas with higher percentages of Black/African-American residents. Test site distribution was less correlated with AYA HIV burden, leaving this population vulnerable across multiple testing deserts.

Introduction

Despite enormous strides made over the last three decades in reducing HIV transmission across the United States, adolescents and young adults (AYA) remain a vulnerable population. Of all age groups, AYA are least likely to know their HIV status, with only 56% of AYA aware of their HIV status compared to 87% among all ages [1]. Furthermore, estimates show that only 63% of AYA living with HIV have achieved viral suppression, a percentage that, while comparable to other age groups in the U.S., remains well below the Joint United Nations Programme on HIV/AIDS goal of achieving 90% viral suppression worldwide by 2020 [1, 2].

Across the U.S., trends in AYA HIV transmission are further complicated by state- and county-level geographic disparities. In Florida in particular, AYA HIV prevalence is nearly double the national average for their age range [3]. Florida adolescents (ages 13–19) and young adults (20–24) also face the 3rd and 5th highest rates of progression to Stage 3 disease (AIDS), respectively [3]. Youth from minoritized racial and ethnic groups in Florida are disproportionately impacted by these disparities. In 2020, the prevalence rate of HIV among Black/African American AYA in Florida was 278 per 100,000 compared with 69 per 100,000 for Hispanic AYA, 29 per 100,000 for Asian AYA, and 27 per 100,000 for white AYA [4]. Within Florida, Pinellas and Hillsborough counties—collectively known as the Tampa Bay region—are among the 48 counties and territories identified by the national Ending the HIV Epidemic campaign as accounting for greater than 50% of all new HIV diagnoses [5]. Together, these epidemiological findings point to an urgent need to understand the geo-social factors that contribute to such trends in HIV transmission among AYA living in Tampa Bay, Florida.

The concept of healthcare “deserts” has previously been employed to understand how geographic access to healthcare resources, such as infectious disease testing or pharmacies, impacts disease burden [6–9]. Drawing on this framework, the overall goal of our study was to characterize HIV testing deserts for Tampa Bay AYA. We began by determining the association between geographic access to HIV testing and zip code-level measures of HIV burden for Tampa Bay AYA. As part of this analysis, we proposed a novel method for quantifying geographic access to testing at the zip code level that can be applied to other urban regions. We next sought to compare the AYA association to that of adults ages 25–44, a demographic that has historically been prioritized in HIV outreach due to having the greatest HIV prevalence among all age groups. In doing so, we also identified local AYA HIV testing deserts. Finally, we investigated the association between zip code-level population characteristics (e.g., percent living in poverty) and HIV burden, as well as the association of population characteristics and geographic access to testing. To our knowledge, this study represents one of the first published efforts to describe AYA HIV testing deserts.

Methods

Data Sources

Several publicly available data sources were used to evaluate access to HIV testing relative to AYA HIV burden in Tampa Bay, Florida. Specifically, HIV test site addresses and measures of HIV burden by zip code were obtained from aidsvu.org [4], while population characteristics by zip code—including race, ethnicity, poverty, and population density data—were obtained from the 2020 American Community Survey [10]. Additional population characteristics included the child opportunity index (COI), a quantitative measure of neighborhood access to resources that was developed by researchers in the Institute for Child, Youth, and Policy at Brandeis University. We obtained 2020 COI scores by zip code for the education, social/economic, health, and overall domains from diversitydatakids.org [11].

HIV Burden

Measures of HIV burden included the number of new HIV diagnoses (aggregated 2016–2020) and the number of prevalent cases (2020) per zip code for AYA ages 13–24 and, separately, for adults ages 25–44. HIV new diagnoses per 100,000 and prevalence per 100,000 by zip code were calculated for each age group by dividing the number of new or prevalent cases by the number of residents in that age range, then multiplying by 100,000. If the number of new or prevalent cases for an age group within a given zip code is between 1 and 4 cases, aidsvu.org suppresses the data to maintain patient privacy. To maximize the utility of this data in the analysis, the new or prevalent case count for suppressed data was set to the midpoint of 2.5 cases, as described in Stephenson et al 2021 [12]. Because the American Community Survey does not provide population estimates for ages 13–24 as a discrete category, population density for AYA was estimated by summing two-fifths the number of individuals ages 10–14, the number of individuals ages 15–19, and the number of individuals ages 20–24, as also described in Stephenson et al 2021 [12]. Population density for ages 25–44 was calculated directly as the sum of the number of individuals in the 25–29, 30–34, 35–39, and 40–44 age categories. Zip codes were excluded from analysis of both age groups if either age group had 125 or fewer residents.

Geographic Accessibility of HIV Test Sites

Using the geocoding software ArcGIS Pro 3.0.0, we created maps of HIV new diagnoses per 100,000 and prevalence per 100,000 by zip code for each age group. The maps were overlayed with HIV test site locations, and a 15-minute walking distance radius was generated around each test site. Both 15-minute walking distance and 30-minute driving have previously been used in the testing desert literature as reasonable distances to access healthcare services [8]. Given that AYA may be less likely to own a car and may be reluctant to ask family for a ride to a sexual health appointment, we chose 15-minute walking distance. To quantify geographic access to testing, we then calculated the percent surface area per zip code that was within 15-minute walking distance to one or more HIV test sites (PSA15). As a secondary measure of geographic testing accessibility, we also calculated the number of test sites per 100,000 total residents in each zip code. This test site density metric has also been used in the literature [12]; unlike the percent surface area calculation, it does not account for inter-zip code travel for testing.

Population Characteristics by Zip Code

Separate maps depicting population characteristics by zip code (percent of residents who are Black, percent of residents who are Hispanic, percent of residents living in poverty, and COI score) were created and overlayed with HIV test site locations surrounded by 15-minute walking distance radii to visualize the relationship between social determinants of health and test site distribution. Of note, the American Community Survey does not delineate population characteristic data by age group, so data on race, ethnicity, and poverty account for residents of all ages in each zip code. Given that race and income are often shared within families, it is reasonable to estimate that the percent of residents per zip code who have a given characteristic is approximately equal to the percent of AYA within that zip code who have the characteristic.

Statistical Analysis

Maps of HIV burden were used to identify HIV testing deserts. These testing deserts were defined qualitatively as regions with non-zero AYA HIV prevalence (high need) and limited or no surface area within 15-minute walking distance to one or more test sites (low supply). Prevalence per 100,000 was chosen for this definition because it captures more long-term trends in HIV burden than does new diagnoses per 100,000. Pearson’s correlation coefficients and simple linear regression models were calculated in R (version 4.2.0) for the association of geographic access to testing (PSA15 or test site density) and HIV burden (HIV new diagnoses per 100,000 or prevalence per 100,000) for each age group. Simple linear regression analysis was also performed for the association of population characteristics with HIV burden and the association of population characteristics with geographic test site accessibility.

All analyses described were carried out under a Johns Hopkins Medicine IRB-approved protocol, IRB00239645.

Results

Test Site Accessibility

Of the 96 Tampa Bay zip codes analyzed, greater than a third (36.5%) had a PSA15 of zero percent (Figure 1). The median PSA15 was 1.9%. As a means of validating our measure of test site geographic accessibility, we also calculated the number of test sites per 100,000 residents (“test site density”) for each zip code and found that more than half (56.3%) of zip codes do not have any HIV test sites (Figure 1).

Figure 1. Geographic accessibility of HIV test sites in Tampa Bay, FL.

a. There is a paucity of walking-accessible HIV test sites in Tampa Bay, FL, as demonstrated by the finding that nearly one third of zip codes have a PSA15 of 0%.

b. When using number of test sites per 100,000 residents (“test site density”) as the measure of geographic access to HIV testing, there is a similarly right-skewed distribution to that of PSA15. The median test site density per zip code of zero further corroborates the lack of HIV test sites in the region.

Association of Testing Access and HIV Burden

We next sought to understand how geographic access to testing is associated with measures of HIV burden for AYA and adults ages 25–44 (Table 1 and Figure 2). The correlation between PSA15 and HIV prevalence per 100,000 was substantially greater for adults ages 25–44 (r = 0.53, p < 0.001) than for AYA (r = 0.10, p = 0.34). Similarly, the correlation between PSA15 and new diagnoses per 100,000 was 0.53 (p < 0.001) for adults ages 25–44 but only 0.27 (p = 0.01) for AYA. These trends held true when using test site density as the measure of geographic access to testing (Table 1). Based on these disparities, we ultimately identified 4 AYA HIV testing deserts in Tampa Bay (Figure 3).

Table 1.

Pearson’s correlation coefficients with 95% confidence interval for the association of geographic test site accessibility (PSA15 or test site density) and HIV burden (prevalence per 100,000 or new diagnoses per 100,000) for AYA and adults ages 25–44.

		Measure of HIV Burden
		New Diagnoses Per 100,000 2016–2020		Prevalence Per 100,000 2020
		AYA	Adult	AYA	Adult
Measure of Geographic Access to Testing	PSA15	0.27 (0.07, 0.45) **	0.50 (0.33, 0.64) ***	0.10 (−0.10, 0.29)	0.53 (0.36, 0.66) ***
	Test Site Density	0.22 (0.02, 0.41) *	0.46 (0.28, 0.60) ***	0.05 (−0.15, 0.25)	0.50 (0.34, 0.64) ***

^*p ≤ 0.05

^**p ≤ 0.01

^***p ≤ 0.001

Figure 2. HIV test site distribution follows trends in adult—but not necessarily AYA—HIV burden.

Inner: Maps of HIV prevalence per 100,000 (top, panel a) or new diagnoses per 100,000 (bottom, panel b) for AYA (left) and adults (right) overlayed with radii representing 15-minute walking distance to test sites (yellow). Outer: Scatterplots depicting the association of HIV prevalence per 100,000 (top, panel a) or new diagnoses per 100,000 (bottom, panel b) for AYA (left) and adults ages 25–44 (right) with PSA15.

a. There is a statistically significant correlation between prevalence per 100,000 and PSA15 for adults but not for AYA, suggesting that test site distribution more closely follows trends in adult HIV prevalence.

b. There is a statistically significant correlation between new diagnoses per 100,000 and PSA15 for both AYA and adults, but this correlation is stronger for adults.

Figure 3. AYA HIV testing deserts in Tampa Bay.

We identified 4 AYA HIV testing deserts: central Pinellas County (red box/A), Thonotosassa city (orange/B), south Tampa city (blue/C), and south Hillsborough County (purple/D).

Notably, one of the testing deserts contains a military base that has its own local healthcare system. Given the unique nature of this zip code, correlation coefficients for the association between geographic test site accessibility and HIV burden were recalculated with the zip code removed from the dataset. This sensitivity analysis yielded similar results to the original analysis.

Association of Population Characteristics and Testing Access

We further wanted to understand whether national and state trends regarding racial and socioeconomic disparities in HIV burden are present in the Tampa Bay region as well (Table 2). We found that the number of new HIV new diagnoses per 100,000 was positively correlated with the percent of residents per zip code who identify as Black/African American (r = 0.37, p < 0.001) and the percent of residents per zip code who live in poverty (r = 0.25, p = 0.01). In contrast, the number of new diagnoses per 100,000 was negatively correlated with all 3 domains of the COI, as well as the overall COI score (r = −0.33, p = 0.001). There was no statistically significant correlation between the number of new diagnoses per 100,000 and the percent of residents who identify as Hispanic or white, non-Hispanic. Similar trends were identified for the association between HIV prevalence per 100,000 and demographic variables, though the percent of residents living in poverty and the health domain of the COI did not have statistically significant correlations with AYA HIV prevalence.

Table 2.

Pearson’s correlation coefficients with 95% confidence intervals for the association of population characteristics variables with HIV burden or geographic access to testing.

	HIV Burden		Testing Access
	New Diagnoses Per 100,000 (2016–2020)	Prevalence Per 100,000 (2020)	PSA15	Test Site Density
% Black	0.37 (0.18, 0.52) ***	0.22 (0.02, 0.41) *	0.31 (0.12, 0.48) **	0.33 (0.14, 0.49) **
% Hispanic	−0.02 (−0.22, 0.17)	−0.10 (−0.30, 0.10)	0.14 (−0.06, 0.33)	0.16 (−0.03, 0.36)
% White Non-Hispanic	−0.19 (−0.38, 0.01)	−0.03 (−0.23, 0.17)	−0.27 (−0.44, −0.07) **	−0.30 (−0.47, −0.10) **
% Living in Poverty	0.25 (0.06, 0.43) *	0.15 (−0.05, 0.34)	0.28 (0.08, 0.45) **	0.35 (0.17, 0.52) ***
COI: Education	−0.29 (−0.47, −0.10) **	−0.24 (−0.42, −0.04) *	−0.23 (−0.41, −0.03) *	−0.28 (−0.46, −0.09) **
COI: Social & Economic	−0.33 (−0.50, −0.14) **	−0.22 (−0.40, −0.02) *	−0.28 (−0.45, −0.08) **	−0.32 (−0.49, −0.13) **
COI: Health	−0.21 (−0.4, −0.01) *	−0.11 (−0.30, 0.10)	−0.35 (−0.51, −0.16) ***	−0.38 (−0.54, −0.19) ***
COI: Overall	−0.33 (−0.50, −0.14) **	−0.23 (−0.41, −0.03) *	−0.29 (−0.46, −0.09) **	−0.34 (−0.51, −0.15) ***

^*p ≤ 0.05

^**p ≤ 0.01

^***p ≤ 0.001

Lastly, we investigated the association between geographic accessibility of HIV test sites and population characteristics (Table 2). We found that PSA15 was positively correlated with the percent of residents who are Black/African American (r = 0.31, p = 0.002) and the percent of residents living in poverty (r = 0.28, p = 0.006), but negatively correlated with the percent of residents who are white, non-Hispanic (r = −0.27, p = 0.008), all three domains of the COI, and the overall COI score (r = −0.29, p = 0.004). There was no statistically significant correlation between geographic accessibility of testing and the percent of residents who identify as Hispanic. Similar trends were noted when test site density was used as the measure of testing accessibility instead of PSA15 and in a sensitivity analysis with data from the military base zip code removed.

Discussion

Measuring Testing Accessibility

While the concept of deserts has previously been applied across a wide range of public health domains, including research on food insecurity and pharmacy access, we are among the first to utilize this concept to understand HIV test site accessibility. In doing so, we have introduced a new definition of testing deserts using an integrated approach between epidemiologic data, geocoding, and calculated walking distance to test sites. We propose that identification of testing deserts will provide new perspectives to better understand potential barriers to accessing HIV testing in a priority region and among a priority population.

AYA HIV Testing Deserts

Our analysis indicates that there are at least 4 unique AYA HIV testing deserts in Tampa Bay, Florida and highlights a significant overall lack of walking-accessible HIV test sites in the region. This issue impacts individuals of all ages but may be especially detrimental to AYA, who have limited access to cars and few options for public transportation in the Tampa Bay region. Where present, HIV test sites are generally distributed around regions with the greatest adult, but not necessarily AYA, HIV burden, resulting in multiple AYA HIV testing deserts across Tampa Bay. These testing deserts may contribute to adverse trends in HIV transmission and acquisition among local AYA and represent targeted opportunities to enhance access to testing for this priority population.

In addition to adverse trends in HIV transmission, Hillsborough and Pinellas counties have seen significant recent increases in bacterial sexually transmitted infections (STIs)—including syphilis, gonorrhea and chlamydia—among AYA. These rates are higher among AYA compared to other age groups in Tampa Bay, as well as higher than rates in many other Florida counties and at the state level [13]. This greater and increasing STI burden among Tampa Bay AYA highlights a lack of access to sexual health services more broadly. Given that AYA are the least likely of all age groups to know their HIV status and carry the highest STI burden, enhancing access to HIV testing is a critical step towards linking AYA to sexual health services and reducing HIV/STI transmission. In Tampa Bay, increasing geographic access to HIV testing for AYA will likely require a multifactorial approach that includes increasing the number of brick-and-mortar walk-in test sites, as well as incorporating greater mobile and school-based testing in low-access zip codes. For AYA over 18 years old, rideshare programs to test sites may be another valuable approach to enhance access [14]. Such interventions are especially urgent as sexual and gender-diverse youth, who are disproportionately affected by HIV/STIs, are facing increasing geopolitical barriers to accessing health services [15, 16].

Population Characteristics Associated with Testing Access

Initially, we hypothesized that disparities in HIV transmission among socially disadvantaged AYA may be driven in part by decreased geographic access to HIV testing for these youth. This hypothesis is consistent with a recent study that found racial and socioeconomic disparities in geographic access to social resources for sexual and gender diverse residents of Chicago [17]. However, our data demonstrates that, in Tampa Bay, geographic access to testing is instead greater among zip codes with a higher proportion of residents who identify as Black/African American or who live in poverty. This finding suggests that, while enhancing geographic access to testing is necessary, it is not a sufficient strategy on its own to address racial and socioeconomic disparities in HIV transmission.

Limitations and Future Directions

There were several notable limitations to our study. First, our analysis was limited to data at the zip code level. Typically, geospatial analyses are conducted with data at the census tract or blocking group level, both of which are more precisely defined and represent smaller geographic units than zip codes. However, to maintain the privacy of protected health information, aidsvu.org restricts publicly available data on HIV transmission to the zip code level, which limited our analysis to this geographic unit.

Second, we were not able to access local surveillance data on other STIs, such as syphilis, gonorrhea or chlamydia, which are important predictors of HIV risk and could further inform geographic gaps in test site accessibility.

Third, our measure of geographic access does not account for HIV testing through methods other than brick-and-mortar clinics, such as mobile outreach programs or online mail-order test kits. To address this limitation, future analyses will utilize outreach data from local mobile testing organizations to understand the extent to which their services penetrate the AYA testing deserts we have identified. Moreover, in-depth, qualitative interviews with local AYA—especially those from minoritized racial and ethnic backgrounds—will be invaluable to understanding where and how Tampa Bay youth access HIV testing.

Lastly, this study focused on geographic access to HIV testing, but we recognize that other structural barriers to testing also contribute to HIV transmission. Focusing on AYA, one such barrier may be the youth friendliness of test sites. While a single validated measure of youth friendliness does not currently exist in the literature, studies suggest that factors such as confidentiality and cost play a substantial role in making healthcare sites more welcoming to AYA [18, 19]. At a more basic level, age restrictions also contribute to youth friendliness (e.g., some of the commercial test sites included in our analysis only serve AYA older than 18 years). Therefore, in addition to qualitative interviews, future research will also focus on analyzing the youth friendliness of digital tools (e.g., websites and social media) employed by test sites to understand how these facilities communicate information related to cost, confidentiality, and other factors that are important to AYA.

Conclusion

Despite these limitations, our findings have important public health implications for AYA HIV control and prevention. We have identified novel AYA HIV testing deserts in an increasingly challenging geo-political region as it relates to health care access for sexual and gender diverse youth. Our novel approach can be applied to other regions across the United States, including other Ending the HIV Epidemic priority regions, to discover similar AYA deserts. Identification of these testing deserts represents a critical first step toward understanding the nature of geographic testing disparities, which in turn can help inform the design of targeted, AYA-focused interventions to halt HIV acquisition and transmission in this priority population

Implications and Contributions.

The authors propose a novel approach to quantify geographic access to HIV testing that can be applied to metropolitan regions and use it to identify HIV testing deserts for Tampa Bay AYA. This study is among the first descriptions of AYA-specific HIV testing deserts in the literature.

List of Abbreviations

AYA

adolescents and young adults

COI

child opportunity index

PSA15

percent surface area within 15-minute walking distance to at least one HIV test site

STI

sexually transmitted infection