Risk Factors for Late-Stage HIV Disease Presentation at Initial HIV Diagnosis in Durban, South Africa

Background After observing persistently low CD4 counts at initial HIV diagnosis in South Africa, we sought to determine risk factors for late-stage HIV disease presentation among adults. Methods We surveyed adults prior to HIV testing at four outpatient clinics in Durban from August 2010 to November 2011. All HIV-infected adults were offered CD4 testing, and late-stage HIV disease was defined as a CD4 count <100 cells/mm3. We used multivariate regression models to determine the effects of sex, emotional health, social support, distance from clinic, employment, perceived barriers to receiving healthcare, and foregoing healthcare to use money for food, clothing, or housing (“competing needs to healthcare”) on presentation with late-stage HIV disease. Results Among 3,669 adults screened, 830 were enrolled, newly-diagnosed with HIV and obtained a CD4 result. Among those, 279 (33.6%) presented with late-stage HIV disease. In multivariate analyses, participants who lived ≥5 kilometers from the test site [adjusted odds ratio (AOR) 2.8, 95% CI 1.7–4.7], reported competing needs to healthcare (AOR 1.7, 95% CI 1.2–2.4), were male (AOR 1.7, 95% CI 1.2–2.3), worked outside the home (AOR 1.5, 95% CI 1.1–2.1), perceived health service delivery barriers (AOR 1.5, 95% CI 1.1–2.1), and/or had poor emotional health (AOR 1.4, 95% CI 1.0–1.9) had higher odds of late-stage HIV disease presentation. Conclusions Independent risk factors for late-stage HIV disease presentation were from diverse domains, including geographic, economic, demographic, social, and psychosocial. These findings can inform various interventions, such as mobile testing or financial assistance, to reduce the risk of presentation with late-stage HIV disease.


Introduction
South Africa has more HIV-infected people than any other country, and over 1.4 million South Africans are now receiving antiretroviral therapy (ART) [1,2]. However, despite this progress, 45-51% of ART-eligible people are still not receiving treatment [1,3]. Not only do these HIV-infected people constitute nearly one-quarter of all AIDS-related deaths in sub-Saharan Africa, but they also transmit HIV to others [1]. Furthermore, late-stage HIV disease at initial HIV diagnosis has been associated with poor treatment response rates and higher mortality [4][5][6][7][8][9][10][11]. In August 2011, the South African Department of Health increased the ART initiation threshold from CD4,200 to ,350 cells/mm 3 to help reduce AIDS-related deaths [12]. However, the median CD4 count at the time of ART initiation was 111 cells/mm 3 in a recent large South African cohort [4]. Increasing the CD4 count treatment threshold will have little benefit if the majority of people continue to present with dangerously low CD4 counts and late-stage HIV disease. Strategies to reduce AIDS-related mortality and decrease HIV transmission must include earlier diagnosis of HIV [13], which has been shown to improve survival [14]. Intensified efforts to promote early diagnosis of HIV-infected people in resource-limited settings are needed, but little is known about how to best target people in HIV testing campaigns [15,16]. We conducted a large, prospective study to assess both the real and perceived barriers to presenting for HIV care in South Africa.

Data collection
We asked participants about personal demographics, proximity to the HIV clinic, and prior healthcare usage and HIV testing. We recorded responses to 12 questions related to perceived personal barriers for seeking HIV testing and medical care during the prior 6 months [17]. We asked 5 questions about emotional health over the previous month with each response rated on a 6-point Likert scale (ranging from 1 being ''all of the time'' or ''always'' to 6 being ''none of the time'' or ''never''). These questions, which were adapted from the 5-item Mental Health Inventory (MHI-5) screening test, were used to calculate a mental health composite (MHC) score [18]. We asked 13 questions about availability of personal social support with each response rated on a 5-point Likert scale (ranging from 1 being ''none of the time'' to 5 being ''all of the time''). These questions incorporate four social support scales (emotional/informational, tangible, positive interaction, affectionate), and were used to calculate the Social Support Index (SSI), from the Medical Outcomes Study [19]. Both the MHC and SSI scores were independently averaged and then transformed to scores ranging from 1 to 100, with higher numbers signifying better emotional health and more social support. We asked participants if they had gone without food, clothing, or housing (''basic necessities'') during the prior 6 months because they needed money for healthcare, or if they had foregone healthcare during the prior 6 months because they needed money for food, clothing, or housing [20].
After completing the survey, participants were offered free HIV counseling and testing, and HIV-infected participants were offered free CD4 count testing. Those who tested positive for HIV were referred for appropriate care and treatment. All HIV testing, care, and treatment was provided in accordance with current South African Department of Health HIV testing and treatment guidelines [3].

Statistical analyses
Late-stage HIV disease presentation was defined as a CD4 count ,100 cells/mm 3 at the time of initial HIV diagnosis. To allow incorporation of the perceived barriers into multivariate models, while also minimizing possible collinearity, we categorized the 12 barriers into 5 groups (service delivery, financial, personal health perception, logistical, and structural). Service delivery barriers included ''have to wait too long to see the nurse/doctor,'' ''the nurse/doctor does not speak my language,'' and ''not treated with respect by the nurse/doctor.'' Financial barriers included ''could not afford medications'' and ''could not afford the cost of transportation.'' Personal health perception barriers included ''didn't think it was necessary, because didn't feel sick'' and ''felt too sick.'' Logistical barriers included ''could not get time off work'' and ''had to take care of someone else.'' Structural barriers included ''could not get to the clinic during the hours it was open,'' ''could not arrange transport to the clinic,'' and ''did not know where to find care.'' Poor emotional health and poor social support were defined as an MHC and SSI score below the median value, respectively.
We used Chi-squared and Fisher's Exact tests to compare potential risk factors between those presenting with and without late-stage HIV disease. We used an iterative model building approach to construct a series of logistic regression models to identify factors associated with late-stage disease presentation. First, we used bivariate logistic regression models to determine odds ratios (OR) of presenting with late-stage HIV disease. To

Cohort characteristics
Among 3,669 people screened for the study, 2,694 met eligibility criteria and enrolled in the study. Among those enrolled, 1,026 (38.1%) tested positive for HIV, of which 830 adults (80.9% of HIV-infected participants) completed both CD4 testing and the study survey. Among those, the median CD4 count was 186 cells/ mm 3 (interquartile range 70-345 cells/mm 3 ), and 279 (33.6%) participants had late-stage HIV disease at the time of their initial HIV diagnosis.
Within the cohort, 249 (30.0%) were .40 years of age, 415 (50.0%) were male, and 449 (54.1%) had not completed high school (Table 1). Over half (56.5%) worked outside the home, and 688 (82.9%) participants reported living $5 kilometers from the clinic. Most participants (79.7%) had never previously been tested for HIV, including 54 of the 60 participants (90.0%) who had spent an overnight in a hospital during the prior year. Reported HIV testing of participants hospitalized during the prior year (6/60 or 10.0%) was significantly lower than the reported HIV testing among participants who had not been hospitalized during the prior year (162/769 or 21.1%) (p = 0.04).

Competing needs to healthcare
Overall, 227 (27.3%) participants had ever gone without healthcare because they needed money for basic necessities (food, clothing, or housing). In bivariate analysis, this was more common among those presenting with late-stage HIV disease (p = 0.02). Participants who presented with late-stage disease were also more likely to have gone without healthcare to pay for food (23.3% vs. 16.3%, p = 0.02), housing (22.5% vs. 15.6%, p = 0.04), and food and housing (13.3% vs. 7.6%, p = 0.009) (Figure 1, top). Similarly, 196 participants (23.6%) had ever foregone basic necessities because they needed money for healthcare. Participants who presented with late-stage disease were more likely to have foregone housing (18.6% vs. 13.2%, p = 0.04), and food and housing (10.8% vs. 6.4%, p = 0.03) to pay for healthcare (Figure 1, bottom).

Perceived barriers to medical care
The most commonly reported perceived barriers to medical care were ''have to wait too long to see the nurse or doctor'' (31.4%), ''could not afford medications'' (26.0%), and ''didn't think it was necessary, because didn't feel sick'' (24.7%) (

Factors associated with late-stage HIV disease presentation
In the multivariate logistic regression model (Table 3), factors associated with presentation to care with late-stage HIV disease were living $5 kilometers from the clinic, having gone without healthcare because money was needed for basic necessities, being male, working outside the home, having a perception of health service delivery barriers, and poor emotional health. Living $5 kilometers from the clinic conferred a 2.80-fold (95% CI 1.68-4.67) higher odds of presenting with late-stage HIV disease. Having gone without healthcare because money was needed for basic necessities (AOR 1.67, 95% CI 1.17-2.37), being male (AOR 1.66, 95% CI 1.22-2.26), and working outside the home (AOR 1.48, 95% CI 1.07-2.05) had higher odds of late-stage HIV disease presentation. Among the categories of perceived barriers to medical care, only a perception of barriers related to health service delivery (AOR 1.48, 95% CI 1.07-2.05), which included ''have to wait too long to see the nurse/doctor'', was significantly associated with late-stage HIV disease presentation in multivariate analyses. Finally, while both poor emotional health and poor social support were significant in bivariate analyses, only poor emotional health (AOR 1.41, 95% CI 1.03-1.94) was significant in the multivariate model.

Discussion
As both guidelines and data increasingly support earlier HIV treatment, it is imperative to understand why patients continue to present with advanced HIV. In a large cohort of outpatient clinic attendees newly diagnosed with HIV in South Africa, the main risk factors for presenting with late-stage HIV disease were living further from the clinic, being male, and having gone without healthcare to pay for basic living necessities. Other variables associated with late-stage disease presentation were working outside the home, having a perception of barriers to health service delivery, such as long wait times, and having poor emotional health. These findings provide focused targets for improving HIV testing programs in order to diagnose people earlier and reduce the number of adults presenting to care with late-stage HIV disease.
Several studies have examined risk factors for late-stage disease presentation in sub-Saharan Africa. In Uganda, studies by Kigozi et al. and Wanyenze et al. found significant risk factors were being male, older, and having no secondary education, similar to our findings [21,22]. They additionally found that people receiving healthcare in a non-medical setting (pharmacy, home, or by traditional healer) or having many sexual partners, both of which were variables we did not obtain, were more likely to present with late-stage disease. In Ethiopia, which has a very different cultural population than South Africa, non-pregnant women, frequent alcohol users, those in a long-term relationship, and people who perceived ART to have many side effects or HIV as a stigmatizing disease were more likely to present with late-stage disease [23]. In our cohort, relationships longer than 6 months had no impact on late-stage presentation, and we did not assess alcohol use or perception of HIV as a stigmatizing disease. The current study is unique from these previous studies by assessing distance to the clinic, perceived barriers to healthcare, competing needs to healthcare, and emotional health and social support structures, before participants were aware of their HIV-infected status.
In our cohort, several structural barriers, such as longer distance to the clinic, a perception of poor service delivery, and working outside the home, were among the strongest risk factors for latestage HIV disease. Other studies in Africa have shown similar structural barriers, such as longer distances and higher transportation costs, to be related to loss-to-care before ART initiation [24][25][26][27][28]. One potential approach to ensuring better service delivery and earlier HIV diagnosis could be the use of active, mobile HIV testing strategies [29][30][31].
Despite the frequent occurrence of late-stage HIV disease presentation in sub-Saharan Africa, there has been an incomplete and inconsistent understanding of the perceived personal barriers to HIV testing [32]. In Botswana, perceived barriers to HIV testing included fear of learning one's status, lack of perceived HIV risk, and fear of having to change sexual practices if positive [33]. In Ethiopia, a perception of HIV as a highly stigmatizing disease was common among people who presented with late-stage disease [23]. Although we did not assess HIV-related stigma, our findings indicate that a perception of service delivery, structural, or financial barriers are obstacles that prevented people from learning their HIV-infected status.
Over one-quarter of those in our cohort reported having gone without healthcare because they needed money for basic necessities, and they were more likely to present with late-stage disease. In several sub-Saharan African studies, food insecurity has been associated with poor ART adherence, more opportunistic infections, missed clinic visits, and increased hospitalizations [34][35][36]. Food insecurity is more common among older, unmarried, HIV-infected adults [34], and thus food insecurity should be addressed as part of comprehensive HIV treatment programs in resource-limited settings [37]. While our findings support the observed negative effects of food insecurity, our results suggest that housing insecurity is also a common problem and associated with late-stage HIV disease presentation.
We found that poor emotional health and social support, both of which were assessed before HIV testing, were associated with a higher likelihood of presentation with late-stage HIV. Depression is common among HIV-infected adults [38,39], and mental health problems are vastly undertreated in resource-poor settings [40,41]. Rates of depression are also higher in symptomatic HIV-infected people [42]. While little data exists on outcomes among HIVinfected adults with mental health issues in developing countries, one U.S.-based study found higher rates of mortality among HIVinfected adults with depression [43]. Studies from Uganda and Ethiopia have reported conflicting results about whether poor mental health influences ART adherence [44,45]. While evidence has been lacking on the association between mental health disorders and uptake of diagnostic and treatment services for HIV [40], our findings suggest that mental health issues and poor social support structures may be important risk factors for delayed HIV testing.
In our study, which excluded known HIV-infected adults, the 60 participants who had been hospitalized during the prior year represented important missed opportunities for HIV testing. Surprisingly, self-reported prior HIV testing was less common among those who had been hospitalized, since South African guidelines recommend ''offering HIV testing to all clients attending health-care facilities as a standard component of medical care, unless the client actively refuses'' [46]. Our finding suggests that either providers were not adherent to national guidelines or hospitalized patients refused HIV testing at very high rates (90%). Regardless, those missed opportunities represented 7.2% of our cohort, suggesting that HIV testing of hospitalized patients remains suboptimal.
Our findings were primarily limited by the accuracy of selfreported survey data. In this cross-sectional survey, we were not able to determine causality with associations and the analysis was not designed to fully examine the social stigma of HIV as a barrier to testing. We used a definition of late-stage HIV disease of CD4,100 cells/mm 3 , which portends a major risk for cryptococcal meningitis, a leading cause of AIDS-related deaths in sub-Saharan Africa [47]. This definition allowed us to identify risk factors and perceived barriers among the people with the greatest risk for severe opportunistic infections and mortality. While we did not assess some variables shown to be associated with late-stage presentation in other African studies (alcohol consumption, number of sexual partners, or ownership of material goods), we examined many important measures not studied in other studies of late-stage HIV disease presentation in sub-Saharan Africa. Finally, a small number of HIV-infected participants either did not fully complete our survey or agree to CD4 count testing, so we cannot assess their influence on our findings.
In conclusion, we found independent risk factors for presentation with late-stage HIV disease from diverse domains, including geographic, economic, demographic, social, and psychosocial. These can directly inform efforts to improve HIV testing. Such efforts should focus on various interventions, such as the use of active mobile testing strategies, financial assistance, or providing food supplementation, to reduce late-stage disease presentation in resource-poor settings (Table 4). Simply expanding HIV testing will not ensure ART-eligible people are enrolled in care or initiated on ART; innovative approaches are also needed to improve subsequent linkage to treatment [48,49]. Providing increased and targeted HIV testing to those at greatest risk for late-stage HIV disease, and subsequently linking them to HIV care should reduce AIDS-related morbidity and mortality.