Are neighborhood conditions associated with HIV management?

Authors


Abstract

Objectives

HIV infection has become a manageable chronic disease as a result of treatment advances. Secondary prevention efforts have proved inadequate to reduce the estimated incidence of new HIV infections. Epidemiological data suggest that geographical clustering of new HIV infections is a common phenomenon, particularly in urban areas among populations of low socioeconomic status. This study aimed to assess the relationship between neighbourhood conditions and HIV management and engagement in high-risk behaviours.

Methods

During routine out-patient HIV clinic visits, 762 individuals from the St Louis metropolitan area completed behavioural assessments in 2008. Biomedical markers were abstracted from their medical records. Multi-level analyses were conducted based on individuals' census tracts.

Results

The majority of the sample were male and African American. In the adjusted models, individuals residing in neighbourhoods with higher poverty rates were more likely to have lower CD4 cell counts and be current smokers. In neighbourhoods with higher rates of unemployment, individuals were less likely to have a current antiretroviral prescription. In more racially segregated neighbourhoods, individuals reported more depressive symptoms.

Conclusions

Despite the advances in HIV disease management, neighbourhood characteristics contribute to disparities in HIV care. Interventions that address neighbourhood conditions as barriers to HIV management may provide improved health outcomes.

Introduction

With treatment advances over the past decade, HIV infection has become a manageable chronic disease [1]. With this encouraging success in terms of reduced mortality [2], new challenges have emerged. Specifically, secondary prevention efforts have been inadequate to prevent new HIV infections. Test and treat strategies are now being promoted as an effective method of HIV prevention [3]. Epidemiological data suggest that geographical clustering of new HIV infections is a common phenomenon, particularly in urban areas with populations of low socioeconomic status [4]. Unfortunately, the development of evidence-based interventions to reduce HIV transmission has been limited in impact. Notably, few interventions have incorporated the neighbourhood environment in which individuals live as a determinant of behaviours [5].

Previous research has examined how social disorganization theoretical factors are associated with HIV infection and health outcomes [6]. Some measures of the social disorganization theory are routinely operationalized by census tract-level data, including per cent in poverty, the racial make-up of a census tract, and unemployment rates. Early in the HIV epidemic, patterns of neighbourhood-level poverty were identified and this poverty was demonstrated to adversely affect infection and survival rates [7-9]. Neighbourhood-level poverty is associated with poorer HIV-related outcomes even within universal health care settings [10-12]. More recent research has demonstrated that neighbourhood poverty factors are associated with higher AIDS-related mortality [13]. Additionally, higher AIDS prevalence occurred in the lowest income neighbourhoods and African Americans were more likely to live in those neighbourhoods [13]. Other studies identified that higher rates of HIV testing occurred among neighbourhoods with more African American residents [14, 15]. This increased HIV testing rate may be a positive public health outcome; yet as it relates to residential segregation, it was also noted to be associated with stronger feelings of perceived racism [14, 16]. Thoughtful consideration of these factors and how they impact health behaviours and outcomes is imperative for furthering intervention development.

Previous research has suggested that there are neighbourhood factors that impact health behaviours, and thus health outcomes. Some of that work has been carried out in relation to risk behaviour engagement. Specifically, lower income neighbourhoods have been noted to have higher rates of alcohol use, attributable to a higher density of alcohol-selling establishments and the marketing of alcohol using mass media such as billboards [17-25]. In other studies, HIV infection rates were predicted by the density of alcohol-serving establishments among men who have sex with men (MSM) [26]. These factors matter to HIV management, as alcohol use is associated with increased engagement in unplanned and unprotected sex [27-31], as well as medication nonadherence [32-34].

Reducing high-risk sexual behaviour also remains germane to HIV prevention efforts. Sexually transmitted infection (STI) incidence rates remain high in populations with HIV despite the decline in self-reported high-risk behaviours after HIV diagnosis [35, 36]. Individual-level risk reduction efforts have not been broadly successful. Evidence-based community-level interventions have rarely addressed neighbourhood factors that may be associated with HIV transmission.

This study aimed to assess the constructs associated with the social disorganization theory (poverty, unemployment and residential segregation) related to HIV management, specifically behaviours that facilitate the transmission of HIV and affect HIV-related health outcomes [6]. HIV-related spatial analyses most often have been conducted to examine HIV infection rates by location and region to for service provision needs [37-40], and have been used to assist in the massive roll-out of combination antiretroviral therapy (cART). We aimed to study specific characteristics of adverse socioeconomic conditions to assess their relationship to poor HIV management and engagement in high-risk behaviours in this care context.

Methods

Study population

This cross-sectional study assessed the associations between neighbourhood conditions and HIV-related factors and risk factors (medication nonadherence, depression and current smoking) among an urban out-patient HIV clinic population. As part of standard of care, all patients who attended the Washington University HIV Clinic (WU HIV Clinic) in St Louis, MO in 2008 completed a behavioural assessment during routine clinic visits. Interviews were conducted by trained medical assistants as individuals were waiting to be seen by their health care providers. Participant residential addresses were also collected during this timeframe. Their addresses were then geocoded using Geographic Information Systems (ArcGIS, version 9.3; ESRI, Redland, CA), to allow for neighbourhood level analyses. Census tract-level socioeconomic indicators were obtained from US Census 2000 and linked to individual-level data. This study was approved by Washington University School of Medicine Human Research Protection Office.

Behavioural and psychological distress measures

The assessments included measures of demographic characteristics (gender, race/ethnicity, employment, education and annual income) and depressive symptomatology as measured by the Patient Health Questionnaire (PHQ-9) [41]. The PHQ-9 is used to calculate severity and symptom counts that signify major depressive disorder (MDD) and other depressive disorders (ODDs); this measure requires a count of the number of severity symptoms endorsed as well as other calculations, as explained elsewhere [41]. In these analyses, we dichotomized the PHQ-9 results into mild symptoms and MDD/ODDs. Variables were categorized and dichotomized. Education level was dichotomized as ≤ high school graduate/equivalent or > high school degree. Employment status was dichotomized into unemployed (including receiving disability benefits) and employed (part- or full-time). Duration of HIV infection (in years) was analysed as a continuous variable. Depression severity was dichotomized into those who expressed symptoms of MDD/ODDs within the past 2 weeks and those who did not.

HIV management measures

Data on current CD4 cell count, plasma HIV RNA viral load, and use and types of prescribed antiretroviral therapies (ARTs) were collected at the time of the visit. ART was defined as the use of at least three drugs from two different antiretroviral drug classes or the use of at least three nucleoside reverse transcriptase inhibitors (NRTIs). Virological suppression was considered < 400 HIV-1 RNA copies/mL at the time of these analyses. Presence of clinical AIDS was determined by a CD4 cell count of <200 cells/μL. Viral loads were used as proxy measures of medication adherence based on published work [42]. We identified factors of HIV management, using the Gardner's Cascade [3], in efforts to seek a method to quantify chronic disease management. If individuals are receiving cART and have suppressed viral loads and high CD4 cell counts, they are likely to be more consistently engaged in care [3].

Neighbourhood measures

Measures of the social disorganization theory were operationalized in this study by census tract-level data, including per cent in poverty, racial make-up and unemployment rates. Three census tract-level factors were dichotomized into lower or more than median based on their distributions among respondents. To characterize neighbourhoods, we analysed three variables for this initial analysis of neighbourhood characteristics: percentage poverty, percentage of the census tract that was majority African American, and percentage of unemployment, as has previously been done in other studies [6, 13, 43]. We chose to analyze these separately, as our previous research suggested that race, unemployment and low income were independent factors associated with poor virological outcomes. Thus, we hypothesized that these census tract-level factors may identify similar relationships, yet differences would be important to identify in order to highlight specific intervention opportunities. We dichotomized the tract-level factors by the median, primarily because of a small sample size. If we had analysed by quintiles or quartiles, the sample size for each category would have been much smaller and decreased the statistical efficiency. In addition, based on our previous experience, tract-level socioeconomic status indicators might indicate a threshold effect without dose−response. We aimed to include these structures of socioeconomic status using a quantitative approach, while recognizing that there are other factors that may influence HIV management in neighbourhood contexts.

Statistical analyses

Using a generalized linear mixed model, we performed a series of multilevel logistic regressions to examine the effect of adverse neighbourhood conditions. Individual covariates were examined for the potential differences by census tract-level characteristics, including the percentage of the population below the federal poverty line, the percentage of the population who were unemployed, and the percentage of the population who were African American. Depression was controlled for in the multilevel modelling regarding its differences between neighbourhoods. All tests were two-sided and P < 0.05 was considered significant. Data analyses were performed using sas software (version 9.2; SAS Institute Inc., Cary, NC).

Results

A total of 762 individuals completed the assessment and had addresses in the St Louis metropolitan area. The majority of the sample were male (n = 512; 67.2%) and African American (n = 566; 74.6%). Nearly 53% of the sample were unemployed or receiving disability benefits. A large proportion endorsed symptoms of MDD/ODDs (n = 278; 36.5%). Approximately one-quarter of the sample were not currently in receipt of ART and 66% (n = 501) of the sample had suppressed HIV viral loads, while 81% (n = 614) had CD4 cell counts > 200 cells/μL. Individuals had been living with HIV for a median of 8.6 years [interquartile range (IQR) 3.0, 13.0 years]. This sample was similar to the larger clinic population. Additional characteristics of the sample are included in Table 1. A total of 273 census tracts were used in the analysis (Table 1).

Table 1. Sociodemographic and HIV-related parameters, and neighbourhood characteristics (n = 762)
Agen%
< 30 years12916.9
30–39 years17222.6
40–49 years29138.2
≥  50 years17022.3
Gender  
Male51267.2
Female25032.8
Race  
Caucasian19325.3
African American/other56674.3
Education  
≤ High school diploma39151.3
> High school diploma37148.7
Employment  
Unemployed39952.4
Employed28637.5
Annual income  
≤ $10,00040152.6
> $10,00034645.4
Sexual partners  
None41654.6
One31841.7
Multiple222.9
Sexual activity in last 3 months  
No43256.7
Yes32142.1
CD4 count  
< 200 cells/μL14418.9
≥ 200 cells/μL61480.6
Viral load  
< 400 copies/mL50165.7
≥ 400 copies/mL26134.3
Receipt of ART  
No19826.0
Yes56474.0
Current smoker37248.8
ODD/MDD symptoms27836.5
Census tract variableMinimum25th percentileMedian75th percentileMaximumRange
  1. ART, antiretroviral therapy; MDD, major depressive disorder; ODD, other depressive disorder.
% persons below poverty line1.306.1413.0525.8665.2163.91
% non-Hispanic African Americans0.002.6123.6477.43100.00100.00
% civilian labour force unemployed1.013.776.4012.3458.0557.04

In an unadjusted model, the following parameters were associated with neighbourhoods that had higher poverty rates: male sex, African American race, lower education, current smoking, and the presence of clinical AIDS (see Table 2 for greater detail). Neighbourhoods with higher unemployment rates were associated with male sex, older age, African American race, smoking, lower income, clinical AIDS, and lower rates of ART prescription. Racially segregated neighbourhoods were associated with male sex, lower education, unemployment, and the presence of depressive symptoms (Table 2).

Table 2. Associations among individual- and census-level sociodemographic, behavioural and HIV-related parameters
 Below poverty lineAfrican AmericanUnemployed
< median≥ medianP< median≥ medianP< median≥ medianP
n%n%n%n%n%n%
  1. ART, antiretroviral therapy; MDD, major depressive disorder; ODD, other depressive disorder.
Age               
< 30 years597.7709.20.358678.8628.10.083719.3587.60.003
30–39 years84118811.6 9412.37810.2 9913739.6 
40–49 years15820.713317.5 15119.814018.4 15019.714118.5 
≥ 50 years8310.98711.4 719.39913 668.710413.7 
Gender               
Male24031.527235.70.00524231.827035.40.01823931.427335.80.002
Female14418.910613.9 14118.510914.3 14719.310313.5 
Race               
Caucasian466.114719.40.000304.016321.50.001374.915628.90.001
African American/other33644.323030.3 35146.321528.3 34745.721928.9 
Education               
≤ High school diploma23230.515920.90.00122829.916321.40.00123230.515920.90.001
> High school1522021928.7 15520.321628.4 15420.221728.5 
Employment               
No20529.919428.30.79120229.519729.90.03220429.819528.50.983
Yes1442114220.7 14220.814421 14621.314020.4 
Annual income               
≤ $10,00024432.7157210.00122630.317523.40.00123331.216822.50.001
> $10,00013417.921228.4 15020.119626.2 14619.520026.8 
Sexual partners               
None22029.119625.90.24621127.920527.10.82921528.420126.60.823
One14919.716922.4 15620.616221.4 15720.816121.3 
Multiple121.6101.3 121.6101.3      
Sexual activity in last 3 months               
No23030.520226.80.07722229.521027.90.44922129.4211280.786
Yes15019.917122.7 15620.716521.9 16121.416021.3 
CD4 count               
< 200 cells/μL8611.4587.70.0127610.0689.00.5258411.1607.90.041
≥ 200 cells/μL29538.931942.1 30640.430840.6 30039.631441.4 
Viral load               
< 400 copies/mL25633.624532.20.592593424231.80.27225934.024231.80.426
≥ 400 copies/mL12816.813317.5 12416.313718.0 12716.713417.6 
Receipt of ART               
No11014.48811.60.09111014.48811.60.08311515.18310.90.015
Yes27436.029038.1 27335.829138.2 27135.629338.5 
Current smoker               
No17122.621528.40.00018424.320226.70.14518224.020426.90.041
Yes21328.115921 1972617523.1 20326.816922.3 
Depression               
Mild or no symptoms19729.120229.80.36721631.918327.00.00821131.218827.80.167
ODD/MDD symptoms14721.713119.4 12218.015623.0 13219.514621.6 

In the adjusted models, individuals residing in neighbourhoods with higher poverty rates were more likely to have CD4 cell counts < 200 cells/μL [odds ratio (OR) 1.56; 95% confidence interval (CI) 1.05–2.44] and more likely to be current smokers (OR 1.45; 95% CI 1.04–2.02). In neighbourhoods with higher rates of unemployment, individuals were less likely to have a current ART prescription (OR 1.47; 95% CI 1.05–2.04). In neighbourhoods that were more racially segregated, individuals more often reported depression symptoms (OR 1.51; 95% CI 1.04–2.22) (Table 3). These additional factors failed to demonstrate an independent association after controlling for other factors: age, gender, race, education, income, depression (except depression outcome), number of sex partners and events of unprotected sex.

Table 3. Significant sociodemographic and HIV-related characteristics and their relationship with neighbourhood characteristics [odds ratio (95% confidence interval)]
VariableCD4 cell countHIV viral loadReceipt of cART PrescriptionDepressionSmoking
UnadjustedAdjustedUnadjustedAdjustedUnadjustedAdjustedUnadjustedAdjustedUnadjustedAdjusted
  1. Adjusted models were adjusted for variables that were significant in unadjusted models: age, gender, race, education, income, depression (except depression outcome), number of sex partners and unprotected sex events.
  2. cART, combination antiretroviral therapy.
% below poverty line (vs. < median)0.62 (0.43–0.90)1.56 (1.05–2.44)0.91 (0.66–1.25)0.78 (0.55–1.12)0.75 (0.53–1.06)0.70 (0.47–1.05)1.13 (0.81–1.57)1.34 (0.94–1.93)1.68 (1.25–2.25)1.45 (1.04–2.02)
% African American (vs. < median)0.86 (0.59–1.25)1.07 (0.69–1.66)0.88 (0.64–1.22)0.79 (0.55–1.14)0.75 (0.53–1.05)0.79 (0.52–1.18)0.68 (0.49–0.94)1.51 (1.04–2.22)1.25 (0.92–1.70)1.08 (0.77–1.53)
% unemployed (vs. < median)0.66 (0.45–0.96)0.73 (0.47–1.13)0.91 (0.66–1.25)0.84 (0.59–1.21)0.68 (0.49–0.95)1.47 (1.05–2.04)0.82 (0.59–1.15)0.90 (0.63–1.30)1.34 (0.99–1.82)1.13 (0.81–1.59)

Discussion

This is one of the first examinations of the relationship between neighbourhood characteristics and HIV management. We identified that neighbourhoods with higher rates of poverty and unemployment, and those with higher residential segregation were associated with poorer health behaviours and HIV management. These findings suggest that adverse neighbourhood conditions have a negative impact on HIV-related health outcomes (unsuppressed viral loads, low CD4 cell counts, and lack of cART prescription). Further studies are needed to conduct more in-depth analyses of other neighbourhood factors and their influence on behaviour over time and across a larger geographical region.

Overall, neighbourhoods with higher poverty levels and those that had higher rates of unemployment were associated with more advanced HIV disease, as manifested by lower CD4 cell counts. Previous studies have identified that higher rates of HIV incidence and AIDS mortality occur in poorer neighbourhoods [7, 8, 13, 44]. Thus, our finding adds to the evidence that poverty is fuelling this epidemic. We failed to identify a relationship between neighbourhood conditions and HIV viral loads. This may reflect an inherent bias given that all subjects in our sample were engaged in care. We were unable to include individuals with HIV infection from these neighbourhoods who were not engaged in care. Nevertheless, in neighbourhoods with higher unemployment, fewer individuals were receiving ART, suggesting disparities based on neighbourhood. These data confirm similar reports from locations with universal health care systems [10, 11]. These data are particularly relevant because all participants in the current study had access to medications through the Ryan White programme, yet disparities persisted.

We found higher rates of smoking in poorer and more unemployed neighbourhoods, confirming previous findings that neighbourhoods with higher rates of poverty were also associated with higher rates of smoking [45]. While populations with HIV infection have been documented to have higher rates of smoking, these neighborhood-level findings suggest that being engaged in care is not protective to manage the perceived stress that is experienced in neighborhoods with high unemployment rates. We hypothesize that higher rates of smoking may be independently associated with neighborhoods with higher unemployment rates, as a consequence of the stress of limited financial support. This suggests that barriers to practising positive self-care behaviours persist regardless of engagement in medical care in neighbourhoods with more adversity. The daily challenges of living in impoverished communities may limit the positive impact that access to medical care provides.

We also identified that, in neighbourhoods that were more residentially segregated, individuals more often expressed symptoms of depression. Areas where few non-African Americans live offer residents limited choices in employment and education, and visions of upward mobility [46]. While the correlated factors of poverty and unemployment are also relevant in this discussion, these findings suggested that these residents, and patients, were more depressed than those living in neighbourhoods that were more racially diverse. Previous findings suggested that residents in neighbourhoods such as these more often reported perceived racism, experienced challenges to accessing medical care, and had sexual networks where higher rates of STIs existed [14]. Individuals experiencing these daily complexities may face challenges in managing their HIV infection. We previously reported that individuals with greater symptoms of depression were less likely to adhere to their medication or attain successful virological suppression [47-49].

Individual-level factors such as race, gender and employment have been shown to affect engagement in medical care, and specifically medication adherence [42, 50]. Thus, analysing the neighbourhood characteristics that align with these individual-level factors offered additional insight into how these sociodemographic characteristics may influence health behaviours and outcomes. Neighbourhood-level stress persisted and had an adverse impact on HIV-related health outcomes. These inherent inequalities probably contribute to negative health behaviours and outcomes in differing ways as a consequence of unemployment, racial segregation or low income. Thus, the findings suggest that, even though individuals may be engaged in medical care and understand how and where to access it, overcoming the challenges in managing their health conditions in the context of neighbourhood stressors may be too difficult without additional support. The granularity of these census-level factors was utilized to offer insights into how to best develop interventions to overcome apparent barriers that exist in specific communities.

This study was limited in the geographical region of the sample, as all individuals came from the St Louis metropolitan area. Longitudinal examinations of these factors would be important to better understand the causal relationships between neighbourhood conditions and health outcomes (e.g. CD4 cell counts, depression and smoking). Furthermore, as this was a cross-sectional assessment, we did not identify the intensity of engagement in medical services among the sample. Additionally, these neighbourhood factors are complicated characteristics and are highly correlated. We aimed to identify independent relationships between these individual-level factors and neighbourhood characteristics, hypothesizing that there are neighbourhood factors that are likely to be related to individual-level behaviours or outcomes. This study identified that individuals living in neighbourhoods with higher unemployment rates were more likely to engage in stress-related behaviours and less likely to receive cART. Similarly, depression was associated with residential segregation, perhaps as a consequence of the limited exposure to diverse living conditions.

Neighbourhoods are complicated networks that influence individual behaviours. These findings suggest that neighbourhoods and specifically the context in which individuals live serve as a determinant of health. This study highlighted the significant disparities in HIV-related health outcomes, even among individuals who were engaged in care, by neighbourhood context. In future studies, we will explore details of the neighbourhood context and the characteristics of individuals within their neighbourhood context. Further development of an in-depth assessment of neighbourhood context beyond census tract data (specifically, community norms, and the availability and accessibility of health protective and non-protective resources such as licensed alcohol outlets, condoms and clinics) is needed to better contextualize these findings. Such research will lead to the development of location-based interventions.

Ancillary