Highly active antiretroviral therapy (HAART) became the standard of care for the treatment of HIV/AIDS in 1996. HAART decreases plasma viral load and improves CD4 cell counts, resulting in decreased HIV-related morbidity and mortality.1-3 The effectiveness of HAART is well established, yet ensuring optimal access to HAART by all eligible persons remains an elusive goal. Each year, a significant number of eligible persons living with HIV/AIDS die never having accessed treatment. The rationale behind an individual's decision to access treatment varies from person to person and is of great importance, because inequalities in access hinder our ability to care for the HIV-positive population.4
Health care and HAART are provided free of charge in British Columbia to all eligible persons living with HIV/AIDS. Accessing treatment is therefore theoretically universal, and the risk of mortality should be equal across the socioeconomic gradient. Previous research in British Columbia has indicated that there is significant variation in the incidence of HIV/AIDS and HIV-related mortality across specific subpopulations and neighborhoods, however.5-7 Building on previous research, the objectives of this study were to examine neighborhood measures of socioeconomic status and their effect on the risk of mortality among HIV-positive persons accessing and not accessing treatment, the effects of late access to treatment by CD4 cell count, and survival among those who accessed treatment.
HIV/AIDS Drug Treatment Program
The British Columbia Center for Excellence in HIV/AIDS (the Center) is responsible for the distribution of antiretroviral medications through the Drug Treatment Program (DTP). Details of the distribution and dispensation of medications through the DTP have been described in detail elsewhere.8,9 The DTP provides antiretroviral therapy free of charge to all eligible HIV-infected patients in British Columbia and has received ethical approval from the University of British Columbia/Providence Health Care Research Ethics Board.
The Center's guidelines for the use of HAART have remained consistent with those recommended by the International AIDS Society-USA Panel since 1996.11-15 HIV-positive men and women are entered into the Center's DTP when they are first prescribed antiretroviral agents. Physicians enrolling an HIV-positive individual must complete a drug request form. This form acts as a legal prescription and captures the patient's postal code and basic demographic and clinical information. All patient data are entered into the Center's monitoring system, and patients are asked to provide informed consent for accessing electronic records from linkages to other sources of data in the province.
Eligible participants were those who were 18 years of age and older. In the first analysis, patients included those who had at least 1 prescription for treatment (thus were enrolled in the DTP) and those who had never accessed treatment. All patients in this analysis died between September 1, 1997 and November 30, 2005. In subsequent analyses, all patients were antiretroviral naive when first dispensed HAART and entered into the DTP between September 1, 1997 and November 30, 2004 and were right-censored as of November 30, 2005.
Outcome Measures and Explanatory Variables
The primary outcomes were HIV-related death and late access to antiretroviral treatment, as defined by a CD4 count of <50 cells/mm3. Mortality data were provided by the British Columbia Vital Statistics Agency. This file includes HIV/AIDS-related deaths of individuals enrolled in the DTP and persons who are not in the DTP but died of HIV/AIDS-related causes in the province. The advantage of this linkage is that it allows us to identify HIV-positive persons not accessing treatment. All deaths related to HIV/AIDS are classified as HIV underlying or HIV associated, as identified by the International Classification of Diseases (ICD), ninth revision (ICD-9; 1997 to 1999), and ICD, 10th revision (ICD-10; 2000 to 2005), codes. Vital statistics define a death as HIV underlying if the attending physician and/or coroner determines that the death was directly attributable to HIV infection, whereas HIV associated defines a death when HIV infection is a secondary condition contributing to an individual's underlying cause of death.
Patient characteristics derived from the DTP database were considered in each model. These included gender, age, AIDS-defining illnesses, history of injection drug use, HIV-1 plasma viral load, CD4 cell counts, and treatment adherence. HIV plasma viral load was measured using the Roche Amplicor Monitor assay (Roche Diagnostics, Laval, Quebec, Canada) using either the standard method or the ultrasensitive adaptation. CD4 cell counts are recorded using flow cytometry and fluorescent monoclonal antibody analysis (Beckman Coulter, Inc., Mississauga, Ontario, Canada). Our measure of adherence was defined as the number of days medication was dispensed divided by the number of days medication was prescribed in the first year of treatment. For the purposes of this study, adherence measures were dichotomized (<95% and ≥95% for nonadherence and adherence, respectively). This measure of adherence has previously been associated with survival among HIV-positive patients in the DTP.16
Patient characteristics were linked to neighborhood socioeconomic characteristics by means of their residential postal code using the 2001 Canadian Census database. This database provides data by neighborhood (known as a census tract or census subdivision) on socioeconomic status. This method of linking census-based data as a surrogate when the characteristics of each individual are unknown has been used successfully in other studies.7,17 Six variables to be used as measures of socioeconomic status were obtained from the census data: percent of residents with postsecondary education, percent of unemployed residents, percent of aboriginal residents, percent of residents living below the Canadian poverty line, median neighborhood income, and whether the neighborhood was urban or rural. In the first analysis, residential postal codes were linked to census data at the time of death because postal codes were only available at death for those who never accessed treatment. In all subsequent analyses, we were able to link postal codes to census data at the time of DTP enrollment, thus minimizing the downward drift effect that can occur as patients become more ill over time.6,18
Three analyses were conducted in this study. The first analysis was undertaken to compare HIV/AIDS-related mortality in persons who had received at least 1 prescription for treatment and died with that in persons who had never accessed treatment and died. Differences in the socioeconomic status measures and patient characteristics between persons who did or did not access treatment were compared using a Pearson χ2 test for categoric variables or a Wilcoxon sign rank test for continuous variables. Because clinical data (eg, CD4 cell count, HIV-1 plasma viral load) were not available for the group who never sought treatment, these variables were not included in the model.
In the second analysis, we compared patients who accessed treatment early with those who accessed treatment late. Logistic regression was used to evaluate variables associated with late versus early access to treatment. Patient characteristics (eg, gender, age, history of injection drug use), including clinical characteristics (eg, year of therapy initiation, prior AIDS-defining illness, baseline plasma viral load), and all socioeconomic status measures were included in the analyses. All variables were entered into a multivariate logistic model, and a backward Akaike information criterion (AIC)-based selection procedure determined the final model. We evaluated model fit using the Hosmer-Lemeshow statistic for goodness-of-fit.19
In the third analysis, comparing factors associated with HIV-related mortality among persons first prescribed HAART, all deaths from non-HIV-related causes were censored and classified as nonevents in this analysis. The variables of interest were gender, age, history of injection drug use, medication adherence, timing of access to treatment, year of treatment initiation (1997 to 1999 vs. 2000 to 2004), previous AIDS-defining illness, baseline CD4 cell count, and baseline plasma viral load. All socioeconomic measures were also included in the analysis. Weibull survival models were used to calculate unadjusted and adjusted relative risks of mortality and 95% confidence intervals (CIs) for all explanatory variables.20 Variables independently associated with increased risk of mortality were identified using the same backward AIC-based selection procedure as previously explained. We assessed the appropriateness of the Weibull model and the assumption of proportional hazards by plotting the log-negative log of the survival function against the log of time. We visually assessed the graph for lines that were linear and parallel to ensure that these assumptions were met.
All analyses were conducted using SAS version 8 (SAS Institute, Cary, NC). All tests of significance were 2-sided, with a P value less than 0.05 indicating that an association was statistically significant.
There were 1436 HIV-related deaths between September 1, 1997 and November 30, 2005. Of these, 1406 (98%) were successfully matched by means of residential postal code to the 2001 census. Those not linkable to the census were more likely to be female (P = 0.006). Table 1 shows that no differences were found between persons who accessed treatment and those who did not for age, gender, or any of the socioeconomic status variables. In total, 567 (40%) persons died without ever accessing treatment and 839 (60%) persons died having accessed at least some treatment. Figure 1 illustrates that access to treatment before death was lowest in 1997 through 1999 and increased significantly during the period 2000 through 2005 (P < 0.001).
There were 2168 treatment-naive patients who accessed HAART between September 1, 1997 and November 30, 2004. Of these, 2080 (96%) had identifiable postal codes at DTP enrollment and were successfully matched with data from the 2001 census database. Those not linkable to the census database were more likely to be injection drug users (P = 0.020) and to have started therapy early in the study period (P = 0.003).
Table 2 presents the variables associated with accessing treatment late. In total, 341 (16%) patients accessed treatment late. These patients were more likely to reside in neighborhoods characterized by high levels of unemployment (odds ratio [OR] = 1.41, 95% CI: 1.14 to 1.74) when controlling for age, previous AIDS-defining illness, and baseline plasma viral load. Using the Hosmer-Lemeshow statistic, we obtained a nonsignificant P value of 0.517, indicating a good model fit. Figure 2 illustrates that timing of access to treatment for patients with CD4 counts between 50 and 200 cells/mm3 improved over time, whereas that for patients with CD4 counts <50 cells/mm3 showed no improvement over time.
Table 3 summarizes the univariate and multivariate analyses of the baseline factors associated with survival time to HIV-related mortality. Of the 2080 patients included in the analysis, there were 349 HIV-related deaths identified during the study period, with a crude mortality rate of 16.8%. The median length of follow-up was 4.0 years (interquartile range [IQR]: 2.1 to 6.3 years). The risk of mortality in the univariate analyses was significantly higher in neighborhoods characterized by low levels of postsecondary education, high unemployment, a high percentage of aboriginal residents, a high percentage of residents living below the poverty line, and low median income. In addition, mortality was associated with older age, nonadherence, late treatment access, low baseline CD4 cell count, and high baseline plasma viral load. In multivariate analysis, mortality was associated with a low level of postsecondary education (hazard ratio [HR] = 0.80, 95% CI: 0.71 to 0.91) and a high percentage of residents living below the poverty line (HR = 1.07, 95% CI: 1.01 to 1.13) when controlling for age, baseline CD4 cell count, plasma viral load, adherence, and late access to treatment. The appropriateness of the Weibull model and the assumption of proportional hazards were validated by ensuring that the log-negative log of the survival functions were linear and parallel across the log of time.
We found no differences in neighborhood-level socioeconomic status between persons who did not access treatment and died and persons who did access treatment and died. Our findings draw special attention to the number of eligible persons living with HIV/AIDS who do not access treatment. On average, 40% of people who died of HIV/AIDS-related causes during the study period had never accessed treatment. This finding is of particular concern, given that treatment is universal and provided free of charge in British Columbia. Barriers preventing access to care need to be studied in greater detail among subgroups of the population that may be at high risk of HIV infection (eg, injection drug users, men who have sex with men, refugees). Additionally, we must consider that transportation problems, illegal activity, or language barriers, for example, play a role in an individual's ability to access treatment. Understanding such issues is necessary before we can develop a successful approach to reach these individuals.
Our study also found that 16% of patients delayed accessing treatment until their CD4 counts were low (<50 cells/mm3). Accessing treatment late has been associated with an increased risk of AIDS-related opportunistic infection and mortality.21,22 Several studies have demonstrated that neighborhood characteristics can influence the health outcomes of the resident population.23,24 For example, poor neighborhoods tend to have higher rates of premature mortality than more affluent neighborhoods.25-27 Measures of socioeconomic status are highly correlated.28 For instance, income varies by occupation, and different occupations require particular levels or types of education. Occupation has been shown in previous research to be an adequate single predictor of socioeconomic status.28 Our findings support this view, because unemployment, the lowest level on an occupation scale, emerged as the most significant predictor of late access to treatment.
Additionally, we observed that residing in a poor neighborhood was associated with an increased risk of mortality for patients on HAART. Decreased survival time was observed in neighborhoods with low levels of postsecondary education and high percentages of poverty, and these effects remained after adjusting for baseline clinical variables and patient characteristics. Residents of poor neighborhoods might be declining and/or delaying treatment because of competing problems such as drug use, lack of affordable housing, or mental illness.29 Our study did not find injection drug use to be independently associated with a higher risk of mortality. This finding suggests that the risk of mortality for people living in poor neighborhoods may be similar regardless of drug use.
The present study has several limitations. First, the use of census-based data in the absence of individual information represents a potential limitation in these analyses. Large census areas are expected to correlate poorly with individual reported measures of socioeconomic status; however, neighborhood-level data, as used in our study, have been shown to predict health outcomes adequately in British Columbia.7 This approach is not necessarily prone to ecologic bias, assuming that neighborhood characteristics are stable and homogeneous within a given geographic area.30 Individually reported socioeconomic data may also be subject to misclassification, however, because of self-reporting bias.31,32 It is also possible that using aggregate data to define neighborhoods leads to oversimplification. For example, 2 neighborhoods with the same score could differ in the parameters that contributed to that score. Census data collect explicit variables and do not provide information on social instability, for example. In addition to this, future analyses would benefit from the inclusion of other contextual factors, such as cultural beliefs, physician experience, and stigma and discrimination. Second, we must also recognize that routine HIV testing is not performed in British Columbia on individuals whose deaths are attributed to accidental causes, such as drug overdose and suicide, and that unless an individual was experiencing morbidity attributable to confirmed HIV infection, his or her death would not be coded as being associated with HIV.16 Finally, 2% of HIV-related deaths occurring during the study period and 4% of patients accessing treatment through the DTP could not be matched with the census data. Because postal code data and census data are continually updated, this is largely the result of not being able to link the 2 data sources together.
Our study used measures of neighborhood socioeconomic status to examine access to treatment, timing to treatment initiation by CD4 cell count, and survival among those who accessed treatment. Socioeconomic status was predictive of these outcomes, assuming that these neighborhoods accurately reflect the resident HIV-positive populations. Neighborhoods with low socioeconomic status were shown to be important when describing survival time and the probability of accessing treatment late. Our study draws special attention to the number of eligible persons living with HIV/AIDS who do not access treatment even though the health care system is universal and free of charge. The benefits of HAART in the management of HIV disease are well established; therefore, novel social and health policy initiatives targeting HIV-infected individuals of lower socioeconomic status, beyond the provision of free and universal health care, are required to optimize access to HAART programs.
The authors express their gratitude to Elizabeth Ferris for copyediting and to Svetlana Draskovic, Kelly Hsu, Peter Vann, and Benita Yip for research and administrative assistance.
R. Joy, E. F. Druyts, E. Wood, J. S. G. Montaner, and R. S. Hogg designed the study. R. S. Hogg participated in data gathering. R. Joy, E. F. Druyts, V. D. Lima, and W. Zhang participated in the statistical analysis. R. Joy, E. F. Druyts, E. K. Brandson, V. D. Lima, C. A. Rustad, E. Wood, J. S. G. Montaner, and R. S. Hogg contributed to the writing of the paper.
1. Smith CJ, Sabin CA, Youle MS, et al. Factors influencing increases in CD4 cell counts of HIV-positive persons receiving long-term highly active antiretroviral therapy. J Infect Dis
2. Hammer SM, Squires KE, Hughes MD, et al. A controlled trial of two nucleoside analogues plus indinavir in persons with human immunodeficiency virus infection and CD4 cell counts of 200 per cubic millimeter or less. AIDS Clinical Trials Group 320 Study Team. N Engl J Med
3. Montaner JS, Reiss P, Cooper D, et al. A randomized, double-blind trial comparing combinations of nevirapine, didanosine, and zidovudine for HIV-infected patients: the INCAS Trial. Italy, The Netherlands, Canada and Australia Study. JAMA
4. Cunningham WE, Hays RD, Duan N, et al. The effect of socioeconomic status on survival of people receiving care for HIV infection in the United States. J Health Care Poor Underserved
5. Hogg RS, Heath KV, Strathdee SA, et al. HIV/AIDS mortality in Canada: evidence of gender, regional and local area differentials. AIDS
6. Schechter MT, Hogg RS, Aylward B, et al. Higher socioeconomic status is associated with slower progression of HIV infection independent of access to health care. J Clin Epidemiol
7. Wood E, Montaner JS, Chan K, et al. Socioeconomic status, access to triple therapy, and survival from HIV-disease since 1996. AIDS
8. Wood E, Schechter MT, Tyndall MW, et al. Antiretroviral medication use among injection drug users: two potential futures. AIDS
9. Hogg RS, Yip B, Chan KJ, et al. Rates of disease progression by baseline CD4 cell count and viral load after initiating triple-drug therapy. JAMA
10. Carpenter CC, Fischl MA, Hammer SM, et al. Antiretroviral therapy for HIV infection in 1996. Recommendations of an international panel. International AIDS Society-USA. JAMA
11. Carpenter CC, Fischl MA, Hammer SM, et al. Antiretroviral therapy for HIV infection in 1997. Updated recommendations of the International AIDS Society-USA Panel. JAMA
12. Carpenter CC, Fischl MA, Hammer SM, et al. Antiretroviral therapy for HIV infection in 1998: updated recommendations of the International AIDS Society-USA Panel. JAMA
13. Carpenter CC, Cooper DA, Fischl MA, et al. Antiretroviral therapy in adults: updated recommendations of the International AIDS Society-USA Panel. JAMA
14. Yeni PG, Hammer SM, Carpenter CC, et al. Antiretroviral treatment for adult HIV infection in 2002: updated recommendations of the International AIDS Society- USA Panel. JAMA
15. Hammer SM, Saag MS, Schecter M, et al. Treatment for adult HIV infection: 2006 recommendations of the International AIDS Society-USA Panel. JAMA
16. Wood E, Montaner JS, Yip B, et al. Adherence and plasma HIV RNA responses to highly active antiretroviral therapy among HIV-1 infected injection drug users. CMAJ
17. Katz MH, Hsu L, Lingo M, et al. Impact of socioeconomic status on survival with AIDS. Am J Epidemiol
18. Hogg RS, Strathdee SA, Craib KJ, et al. Lower socioeconomic status and shorter survival following HIV infection. Lancet
19. Hosmer D, Lemeshow S. Applied Logistic Regression
. 2nd ed. New York: Wiley; 2000.
20. Allison PD. Survival Analysis Using SAS: A Practical Guide
. Cary, NC: SAS Institute; 1995.
21. Hogg RS, Yip B, Chan KJ, et al. Rates of disease progression by baseline CD4 cell count and viral load after initiating triple-drug therapy. JAMA
22. Wood E, Hogg RS, Yip B, et al. Is there a baseline CD4 cell count that precludes a survival response to modern antiretroviral therapy? AIDS
23. Lochner KA, Kawachi I, Brennan RT, et al. Social capital and neighborhood mortality rates in Chicago. Soc Sci Med
24. Macintyre S, Maciver S, Sooman A. Area, class and health: should we be focusing on places or people? J Soc Policy
25. Anderson RT, Sorlie P, Backlund E, et al. Mortality effects of community socioeconomic status. Epidemiology
26. Waitzman NJ, Smith KR. Phantom of the area: poverty-area residence and mortality in the United States. Am J Public Health
27. Cubbin C, LeClere FB, Smith GS. Socioeconomic status and injury mortality: individual and neighbourhood determinants. J Epidemiol Community Health
28. Powers M. Measures of Socioeconomic Status
. Boulder, CO: Westview Press; 1982.
29. McFarland W, Chen S, Hsu L, et al. Low socioeconomic status is associated with a higher rate of death in the era of highly active antiretroviral therapy, San Francisco. J Acquir Immune Defic Syndr
30. Demissie K, Hanley JA, Menzies D, et al. Agreement in measuring socio-economic status: area-based versus individual measures. Chronic Dis Can
31. Stewart WF, Tonascia JA, Matanoski GM. The validity of questionnaires reported work history in live respondents. J Occup Med
32. Wilkins R. Use of postal codes and addresses in the analysis of health data. Health Rep
Keywords:© 2008 Lippincott Williams & Wilkins, Inc.
access; antiretroviral therapy; HIV/AIDS; socioeconomic status; survival