Abstract

Background We assessed the impact of education on long-term overall and cause-specific mortality among 6575 injecting drug users (IDUs) according to HIV status and introduction of highly active antiretroviral therapy (HAART).

Methods Community-based cohort study of IDUs recruited in three AIDS prevention centres (1987–1996). Causes of death were ascertained in clinical centres and Mortality Registry and classified as AIDS, drug use related, injuries, or liver diseases. Poisson regression models including education and calendar period interaction and adjusted by sex, age, and HIV were used.

Results In 73 901 person-years of follow-up, there were 1493 deaths (20.2/1000 person-years): 761 related to AIDS, 234 to drug use, 179 to injuries, and 93 to liver diseases. IDUs with university studies had a lower risk of death (RR 0.52; 95% CI 0.36–0.77) than those without studies: this difference was higher after (RR 0.45; 95% CI 0.25–0.80) than before 1997 (RR 0.68; 95% CI 0.41–1.13). Compared to before 1997, while decreases in the risk of AIDS mortality were seen during 1997–2004 for both lower (RR 0.49; 95% CI 0.41–0.58) and higher (RR 0.33; 95% CI 0.23–0.48) educated, only those higher educated experienced a reduction in drug-use mortality (RR 0.54; 95% CI 0.28–1.05) and death from injuries (RR 0.52; 95% CI 0.23–1.21).

Conclusions Independently of HIV status, lower education predicts a higher risk of death in IDUs and its impact is stronger after 1997. Education has a protective effect on most causes of death and it cannot be entirely attributable to the access or use of HAART.

Introduction

Lower education has been associated with higher death rates of most specific causes, in western countries.1–2 Although educational level has been used as a measure of socio-economic position in public health research, it has several limitations and cannot be considered the best measure of social class.3,4 However, in the study of mortality inequalities in injecting drug users (IDUs), a population characterized by poverty and precariousness, educational level might be more heterogeneous than other indices of socio-economic position such as income or occupation.5 It is also more strongly related to a behavioural pathway, which plays an important role in the study of mortality among young people addicted to drugs.6 Moreover, educational level is a relatively stable variable in adulthood, whereas many other indices of socio-economic position are more often subject to change. Thus, differences in educational attainment reflect differences in socio-economic status and are a measure of social inequality within this population.7

AIDS affects mainly the younger people and relative inequalities in mortality by educational level are generally larger for the younger population than for the middle-aged and the elderly.8 In Spain, as in many developed societies, drug abuse can be considered indirectly the first cause of AIDS cases.9–11 It has been reported that drug abuse affects mainly those with low-educational level.12,13

However, it remains unclear whether among IDUs, who are characterized by their disadvantaged social situation, economic difficulties, and lack of family and social support, lower education still predicts a higher mortality.14–16

With the introduction of the highly active antiretroviral therapy (HAART), some studies have found that markers for low socio-economic status (SES) are associated with higher rates of disease progression and death17–19 among HIV-infected IDUs. This suggests that inequalities in health-care access or in medical management, or poor adherence to treatment—factors associated with SES17—could explain the observed heterogeneity. However, it does not fully explain the difference in progression found in groups with different socio-economic status since some studies have reported that the effect was also present before HAART.20

To clarify whether in the context of free and universal access to HIV/AIDS care (in Spain since 199721) higher progression rates of HIV in people with a lower educational level are only attributable to the availability of HAART, it is necessary to analyse temporal trends in overall and cause-specific mortality for both HIV-positive and negative IDUs.

We have analysed the data from a well-characterized population-based cohort study of IDUs, set up in the late 1980s, with the following objectives: (i) to investigate the impact of education on long-term overall and cause-specific mortality, (ii) to test whether the effect of education on mortality has changed over the last decades, and (iii) whether any increases in the inequalities in mortality by educational level might be attributed to the introduction of HAART.

Methods

Study population

We prospectively recruited IDUs who attended any of the three AIDS Prevention and Information Centres (CIPS) of Valencia (Spain) from 1987 to 1996. These centres provide information and advice about HIV/AIDS infection to those who request it and carry out serological tests anonymously, confidentially, and free of charge. The coverage of the CIPS is estimated to be between 60 and 70%. All subjects were asked for informed consent and were given an anonymous identification code.22,23

Of a total of 7269 IDUs recruited during the study period, information on educational level was available for 6575 (90.5%) who thus became our study population. Socio-demographic characteristics, HIV status, and drug use habits of those with data on education did not differ from those excluded.

Data collection

Information was collected through a structured questionnaire administered in a face-to-face interview by trained personnel working at the centres. For each participant, socio-demographic characteristics (educational level, sex, age at first visit), HIV status, and drug-use habits (age and year at starting IDU, duration of IDU prior to study entry, and sharing drug-use equipment) were recorded at time of enrolment.

Educational level refers to the highest academic degree achieved, classified into four educational categories: no education (people with incomplete elementary education or no education), primary studies (8 years of schooling), secondary studies (9–12 years) and university degree (≥13 years).

Follow-up and censoring strategy

Follow-up information was conducted at the recruiting centres and referral hospitals. In order to guarantee the completeness of data, cross-checks with the Mortality Registry were periodically performed using the initial of both surnames (included in the identification code), the date of birth, and sex. The last cross-check was carried out in October 2005 but analyses were censored by 31 December 2004 to allow for reporting delay. Patients were assumed to be alive on 31 December 2004 if there was no evidence of death in the Mortality Registry.

Individuals were followed up from study entry to death or 31 December 2004, whichever arose first. Entry date for the analyses was the first visit to the centres, except for 289 (4.4%) subjects who seroconverted to HIV during the study, in whom entry date was their first HIV-positive antibody test.

We analysed mortality from the four leading groups of causes of death, as defined by the International Classification of Diseases: AIDS (ICD-9 011.9 if death happened before December 1993, 117.5, 130, 136.3, 180.9 if death happened after December 1993, 200.2, 279 and 795.8, ICD-10 B20–B24); death related to drug use that included drug overdose (ICD-9 E850–858, ICD-10 X40–X44), septicaemia (ICD-9 038, ICD-10 A40–A41), endocarditis (ICD-9 421, 424.9, ICD-10 I33), lung oedema (ICD-9 518.4, ICD-10 J81) and drug dependence (ICD-9 304, ICD-10 F11–F16, F18-F19); death from injuries that included suicide (ICD-9 E950–E959, X60–X84 and Y87.0), accidental injuries (ICD-9 E800–E849, E860–E949, ICD-10 V01–X59 and Y40–Y86) and homicide (ICD-9 E960–E999, ICD-10 X85–Y34 and Y87.1); and death from liver diseases (ICD-9 070, 155–156, 570–573, ICD-10 K70, K73-K74). These four groups of causes are the most frequent in IDUs and represent almost 86% of the deaths within this cohort.

Statistical analysis

Descriptive analysis of the socio-demographic characteristics and drug-use habits for individuals in the four educational levels was carried out using frequency tables or median and interquartile range (IQR) when appropriate. Differences by educational level were analysed using the non-parametric Kruskal–Wallis test for continuous variables and the χ2 test for categorical data.

To asses the impact of HAART on the differences on global and cause-specific mortality by educational level, follow-up was divided into two calendar periods (1987–96 and 1997–2004), before and after the introduction of HAART in our setting. Each person contributed to the analyses with as many registers of ‘time periods’ and ‘age periods’ he/she had been at risk.

Poisson regression models were used to examine the differences on the risk of death by educational level. Models were adjusted for the following potential confounders: sex, age, and HIV status. In order to determine the modifying effect of calendar period and these variables on the effect of education on mortality, interaction terms were included in the multivariate models. We used likelihood ratio tests to derive P-values.

Due to the small number of deaths among the group with the highest educational level, in the cause-specific analysis we grouped persons with secondary and university studies (labelled as higher education) and persons with primary or no education (labelled as lower education).

Analyses were performed in Stata 8.0 (StataCorp, College Station, TX, USA) using robust methods to estimate confidence intervals.

Results

Socio-demographic and drug-use characteristics (Table 1)

Most individuals had primary studies (55.8%) or no education (21.9%), had used drugs for less than 120 months (79.7%) and started using drugs in their twenties (median age: 19, IQR: 16–20 years). Of 6575 IDUs, 77.2% were men, 80.3% aged 30 years or less at study entry, and 47.2% HIV positive. Individuals with higher education were more likely to be women, older at baseline, and HIV negative than those with primary studies or no education. A higher proportion of more educated individuals had not shared injection drug-use equipment and had used drugs for longer. The median follow-up time was 10 years (IQR 7.6–12.2).

Table 1

Characteristics of 6575 IDUs according to educational level at study entry: 1987–2004

n (%) Total 6575 (100.0) No education 1438 (21.9) Primary 3670 (55.8) Secondary 1260 (19.2) University 207 (3.2) 
Sex**      
    Men 5077 (77.2) 1198 (83.3) 2874 (78.3) 865 (68.7) 140 (67.6) 
    Women 1498 (22.8) 240 (16.7) 796 (21.7) 395 (31.4) 67 (32.4) 
Age**      
    ≤25 3149 (47.9) 829 (57.7) 1821 (49.6) 467 (37.1) 32 (15.5) 
    26–30 2130 (32.4) 418 (29.1) 1182 (32.2) 480 (38.1) 50 (24.2) 
    ≥31 1296 (19.7) 191 (13.3) 667 (18.2) 313 (24.8) 125 (60.4) 
Duration of addiction (months)**      
    ≤24 1542 (23.5) 386 (26.8) 809 (22.0) 302 (24.0) 45 (21.7) 
    25–72 2035 (31.0) 516 (35.9) 1126 (30.7) 346 (27.5) 47 (22.7) 
    73–120 1658 (25.2) 326 (22.7) 950 (25.9) 321 (25.5) 61 (29.5) 
    >120 1094 (16.6) 183 (12.7) 639 (17.4) 228 (18.1) 44 (21.3) 
    Unknown 246 (3.7) 27 (1.9) 146 (4.0) 63 (5.0) 10 (4.8) 
HIV status**      
    Negative 3471 (52.8) 734 (51.0) 1868 (50.9) 735 (58.3) 134 (64.7) 
    Positive 3104 (47.2) 704 (49.0) 1802 (49.1) 525 (41.7) 73 (35.3) 
Sharing drug-use equipment**      
    No 4317 (65.7) 964 (67.0) 2310 (62.9) 885 (70.2) 158 (76.3) 
    Yes 2054 (31.2) 435 (30.3) 1251 (34.1) 325 (25.8) 43 (20.8) 
    Unknown 204 (3.10) 39 (2.7) 109 (3.0) 50 (4.0) 6 (2.9) 
Age at starting IDU* (years) (Median, IQR) 19 (16–22) 18 (16–22) 19 (16–22) 20 (17–24) 23 (19–29) 
Year at starting IDU (years)* (Median, IQR) Jul85 Feb82–May89 Feb86 Jan83–May89 Jun85 Dec81–Jun89 Feb85 Mar81–Mar89 Jul84 May80–Feb89 
Follow-up time (years)** (Median, IQR) 10.0 (7.6–12.2) 10.3 (8.5–12.1) 9.6 (7.3–12.2) 10.2 (8.0–12.6) 9.9 (8.0–12.8) 
n (%) Total 6575 (100.0) No education 1438 (21.9) Primary 3670 (55.8) Secondary 1260 (19.2) University 207 (3.2) 
Sex**      
    Men 5077 (77.2) 1198 (83.3) 2874 (78.3) 865 (68.7) 140 (67.6) 
    Women 1498 (22.8) 240 (16.7) 796 (21.7) 395 (31.4) 67 (32.4) 
Age**      
    ≤25 3149 (47.9) 829 (57.7) 1821 (49.6) 467 (37.1) 32 (15.5) 
    26–30 2130 (32.4) 418 (29.1) 1182 (32.2) 480 (38.1) 50 (24.2) 
    ≥31 1296 (19.7) 191 (13.3) 667 (18.2) 313 (24.8) 125 (60.4) 
Duration of addiction (months)**      
    ≤24 1542 (23.5) 386 (26.8) 809 (22.0) 302 (24.0) 45 (21.7) 
    25–72 2035 (31.0) 516 (35.9) 1126 (30.7) 346 (27.5) 47 (22.7) 
    73–120 1658 (25.2) 326 (22.7) 950 (25.9) 321 (25.5) 61 (29.5) 
    >120 1094 (16.6) 183 (12.7) 639 (17.4) 228 (18.1) 44 (21.3) 
    Unknown 246 (3.7) 27 (1.9) 146 (4.0) 63 (5.0) 10 (4.8) 
HIV status**      
    Negative 3471 (52.8) 734 (51.0) 1868 (50.9) 735 (58.3) 134 (64.7) 
    Positive 3104 (47.2) 704 (49.0) 1802 (49.1) 525 (41.7) 73 (35.3) 
Sharing drug-use equipment**      
    No 4317 (65.7) 964 (67.0) 2310 (62.9) 885 (70.2) 158 (76.3) 
    Yes 2054 (31.2) 435 (30.3) 1251 (34.1) 325 (25.8) 43 (20.8) 
    Unknown 204 (3.10) 39 (2.7) 109 (3.0) 50 (4.0) 6 (2.9) 
Age at starting IDU* (years) (Median, IQR) 19 (16–22) 18 (16–22) 19 (16–22) 20 (17–24) 23 (19–29) 
Year at starting IDU (years)* (Median, IQR) Jul85 Feb82–May89 Feb86 Jan83–May89 Jun85 Dec81–Jun89 Feb85 Mar81–Mar89 Jul84 May80–Feb89 
Follow-up time (years)** (Median, IQR) 10.0 (7.6–12.2) 10.3 (8.5–12.1) 9.6 (7.3–12.2) 10.2 (8.0–12.6) 9.9 (8.0–12.8) 

*P-value < 0.05; ** P-value < 0.01.

All-cause mortality (Tables 2 and 3)

In 73 901 person-years of follow-up, there were 1493 deaths, resulting in an all-cause mortality rate of 20.2 per 1000 person-years (95% CI: 19.2–21.3). Educational level was found to be inversely related to mortality rates; compared to individuals with no education, rate ratios for those with secondary and university studies were, respectively, 0.71 (95% CI 0.60–0.84) and 0.52 (95% CI 0.36–0.77). Women experienced a 31% lower level of mortality than men (RR 0.69; 95% CI 0.60–0.79). Mortality rates increased significantly as age increased; compared with individuals aged less than 25 years, rate ratios for those aged 26–30 and more than 31 were, respectively, 1.46 (95% CI 1.19–1.78) and 1.77 (95% CI 1.48–2.13). As for duration of injecting drug use, a 2-fold increase in mortality (RR 2.07; 95% CI 1.76–2.44) was observed in those subjects who had injected drug use for longer than 120 months compared with those with short duration of injection (≤24 months, baseline). Compared with HIV negative, those HIV-positive IDUs had 3-fold increase in the risk of death (RR 3.33; 95% CI 2.97–3.74). Sharing injection drug-use equipment was also associated to death (RR 1.32; 95% CI 1.19–1.47). All-cause mortality rates dropped by 22% during 1997–2004 compared with the years before 1997 (RR 0.78; 95% CI 0.71–0.86).

Table 2

Number of deaths and mortality rates per 1000 person-years according to educational level and selected variables in a cohort of 6575 IDUs, 1987–2004

 Total 73 901 No education 16 289 Primary 40 518 Secondary 14 656 University 2437 
Person-years (p-y) at risk Deaths Rate Deaths Rate Deaths Rate Deaths Rate Deaths Rate 
Educational level 1493 20.2 357 21.9 880 21.7 228 15.6 28 11.5 
Sex           
    Men 1227 21.8 311 23.1 722 23.1 170 17.2 24 14.8 
    Women 266 15.0 46 16.2 158 17.1 58 12.1 4.9 
Age           
    ≤25 131 12.8 35 10.9 85 14.8 11 8.9 
    26–30 361 18.6 104 22.4 205 18.3 51 15.3 3.7 
    ≥31 1001 22.6 218 25.8 590 25.0 166 16.5 27 12.9 
Duration of addiction (months)           
    ≤24 255 14.6 75 16.7 138 15.4 37 10.5 9.8 
    25–72 434 18.3 127 21.1 238 18.4 64 15.3 9.1 
    73–120 417 22.3 98 27.2 254 24.0 57 15.1 10.7 
    >120 340 30.3 54 29.1 218 33.8 59 24.3 18.0 
    Unknown 47 16.7 9.1 32 20.1 11 14.6 7.8 
HIV status           
    Negative 410 9.9 104 11.7 229 10.5 69 7.7 4.9 
    Positive 1083 33.2 253 34.1 651 34.8 159 27.7 20 24.6 
Sharing drug-use equipment           
    No 883 18.1 224 20.2 487 19.0 150 14.5 22 11.9 
    Yes 549 23.8 121 25.4 357 25.7 67 17.6 7.4 
    Unknown 61 30.3 12 27.0 36 35.9 11 21.4 39.2 
Calendar period           
    <1997 708 23.2 164 23.6 404 24.3 124 20.6 16 16.8 
    1997–2004 785 18.1 193 20.6 476 19.9 104 12.0 12 8.1 
 Total 73 901 No education 16 289 Primary 40 518 Secondary 14 656 University 2437 
Person-years (p-y) at risk Deaths Rate Deaths Rate Deaths Rate Deaths Rate Deaths Rate 
Educational level 1493 20.2 357 21.9 880 21.7 228 15.6 28 11.5 
Sex           
    Men 1227 21.8 311 23.1 722 23.1 170 17.2 24 14.8 
    Women 266 15.0 46 16.2 158 17.1 58 12.1 4.9 
Age           
    ≤25 131 12.8 35 10.9 85 14.8 11 8.9 
    26–30 361 18.6 104 22.4 205 18.3 51 15.3 3.7 
    ≥31 1001 22.6 218 25.8 590 25.0 166 16.5 27 12.9 
Duration of addiction (months)           
    ≤24 255 14.6 75 16.7 138 15.4 37 10.5 9.8 
    25–72 434 18.3 127 21.1 238 18.4 64 15.3 9.1 
    73–120 417 22.3 98 27.2 254 24.0 57 15.1 10.7 
    >120 340 30.3 54 29.1 218 33.8 59 24.3 18.0 
    Unknown 47 16.7 9.1 32 20.1 11 14.6 7.8 
HIV status           
    Negative 410 9.9 104 11.7 229 10.5 69 7.7 4.9 
    Positive 1083 33.2 253 34.1 651 34.8 159 27.7 20 24.6 
Sharing drug-use equipment           
    No 883 18.1 224 20.2 487 19.0 150 14.5 22 11.9 
    Yes 549 23.8 121 25.4 357 25.7 67 17.6 7.4 
    Unknown 61 30.3 12 27.0 36 35.9 11 21.4 39.2 
Calendar period           
    <1997 708 23.2 164 23.6 404 24.3 124 20.6 16 16.8 
    1997–2004 785 18.1 193 20.6 476 19.9 104 12.0 12 8.1 
Table 3

Effect of education on overall mortality before and after 1997 and effect of calendar period on overall mortality for each educational level

 Effect of educational levela Effect of calendar periodb 1997–2004/Before 1997 
 Before 1997 1997–2004  
No education 1.00 1.00 0.74 (0.60–0.92) 
Primary studies 0.98 (0.82–1.18) 0.97 (0.82–1.15) 0.73 (0.63–0.85) 
Secondary studies 0.88 (0.69–1.11) 0.63 (0.50–0.81) 0.54 (0.41–0.70) 
University degree 0.68 (0.41–1.13) 0.45 (0.25–0.80) 0.49 (0.23–1.03) 
 Effect of educational levela Effect of calendar periodb 1997–2004/Before 1997 
 Before 1997 1997–2004  
No education 1.00 1.00 0.74 (0.60–0.92) 
Primary studies 0.98 (0.82–1.18) 0.97 (0.82–1.15) 0.73 (0.63–0.85) 
Secondary studies 0.88 (0.69–1.11) 0.63 (0.50–0.81) 0.54 (0.41–0.70) 
University degree 0.68 (0.41–1.13) 0.45 (0.25–0.80) 0.49 (0.23–1.03) 

Data are rate ratios (RR; 95% CI) of a multivariate Poisson model that includes educational level, calendar period, interaction between educational level and calendar period, sex, HIV status, and age.

aIn each calendar period, rate ratios have been calculated considering the individuals with no education as the reference category.

bFor each educational level, the rates ratios of calendar period have been calculated, considering the period before 1997 as the reference category.

As shown in the last two rows of Table 2, while before 1997 there was little difference in mortality by educational level, this effect was much stronger during 1997–2004 (interaction between education and calendar period, P-value = 0.046).

In multivariate Poisson analysis, shown in Table 3, the association between lower education and higher mortality remained significant after controlling for sex, HIV status, and age. Before 1997, a 32% decrease in mortality rates was seen in individuals with university studies compared with those with no education (RR 0.68; 95% CI 0.41–1.13), although it did not reach the statistical significance. After 1997 a 37% decrease in mortality was observed in those with secondary studies (RR 0.63; 95% CI 0.50–0.81) and a 55% decrease in those with a university degree (RR 0.45; 95% CI 0.25–0.80) compared with individuals with no education. A decrease in the risk of death was seen during 1997–2004 compared with years before 1997 for all four educational levels, but the reduction was greater for individuals with secondary (RR 0.54; 95% CI 0.41–0.70) and university studies (RR 0.49; 95% CI 0.23–1.03) than for those with lower education.

Cause-specific mortality (Tables 4 and 5)

The commonest cause of death was AIDS with 761 cases (10.3 per 1000 p-y) followed by drug use related with 234 (3.2 per 1000 p-y). One hundred and ninety seven (2.4 per 1000 p-y) deaths were attributable to injuries and 93 (1.3 per 1000 p-y) to liver diseases.

Table 4

Number of deaths and mortality rates per 1000 person-years of each specific group of causes of death, according to calendar period and educational level

 Total Before 1997 1997–2004 
 Low56 807 High17 093 Low23 572 High6977 Low33 236 High10 116 
p-y at risk Deaths Rate Deaths Rate Deaths Rate Deaths Rate Deaths Rate Deaths Rate 
AIDS 632 11.1 129 7.5 346 14.7 85 12.2 286 8.6 44 4.3 
Drug use 195 3.4 39 2.3 83 3.5 24 3.4 112 3.4 15 1.5 
Injuries 155 2.7 24 1.4 57 2.4 14 2.0 98 2.9 10 1.0 
Liver diseases 77 1.4 16 0.9 30 1.3 0.1 47 1.4 15 1.5 
 Total Before 1997 1997–2004 
 Low56 807 High17 093 Low23 572 High6977 Low33 236 High10 116 
p-y at risk Deaths Rate Deaths Rate Deaths Rate Deaths Rate Deaths Rate Deaths Rate 
AIDS 632 11.1 129 7.5 346 14.7 85 12.2 286 8.6 44 4.3 
Drug use 195 3.4 39 2.3 83 3.5 24 3.4 112 3.4 15 1.5 
Injuries 155 2.7 24 1.4 57 2.4 14 2.0 98 2.9 10 1.0 
Liver diseases 77 1.4 16 0.9 30 1.3 0.1 47 1.4 15 1.5 
Table 5

Effect of education on cause-specific mortality before and after 1997 and effect of calendar period in each cause-specific relative risk of death for both lower and higher educated

 Effect of educational level: high/lowa Effect of calendar period: 1997–2004/before 1997b 
 Before 1997 1997–2004 Low High 
AIDS 0.84 (0.66–1.06) 0.57 (0.41–0.78) 0.49 (0.41–0.58) 0.33 (0.23–0.48) 
Drug use 1.21 (0.75–1.93) 0.52 (0.30–0.89) 1.27 (0.90–1.79) 0.54 (0.28–1.05) 
Injuries 0.93 (0.52–1.68) 0.37 (0.19–0.71) 1.32 (0.90–1.92) 0.52 (0.23–1.21) 
Liver diseases 0.10 (0.01–0.75) 1.13 (0.63–2.01) 0.70 (0.44–1.13) 7.82 (1.06–57.87) 
 Effect of educational level: high/lowa Effect of calendar period: 1997–2004/before 1997b 
 Before 1997 1997–2004 Low High 
AIDS 0.84 (0.66–1.06) 0.57 (0.41–0.78) 0.49 (0.41–0.58) 0.33 (0.23–0.48) 
Drug use 1.21 (0.75–1.93) 0.52 (0.30–0.89) 1.27 (0.90–1.79) 0.54 (0.28–1.05) 
Injuries 0.93 (0.52–1.68) 0.37 (0.19–0.71) 1.32 (0.90–1.92) 0.52 (0.23–1.21) 
Liver diseases 0.10 (0.01–0.75) 1.13 (0.63–2.01) 0.70 (0.44–1.13) 7.82 (1.06–57.87) 

Data are rate ratios (RR; 95% CI) of different multivariate Poisson models for each group of causes of death, that includes educational level, calendar period, interaction between educational level and calendar period, sex, HIV status, and age.

aIn each calendar period, rates ratios have been calculated considering the individuals with lower education as the reference category.

bFor each educational level, the rates ratios of calendar period have been calculated, considering the period before 1997 as the reference category.

Table 4 shows cause-specific death rates for both lower and higher educated in the total population and in the time before and after HAART. The inverse relation between education and mortality was especially high and statistically significant for injuries (RR 0.51; 95% CI 0.33–0.79), drug use (RR 0.66; 95% CI 0.47–0.94) and AIDS (RR 0.68; 95% CI 0.56–0.82). While before 1997 it was difficult to see substantial differentials by education, the inverse gradient with education was greater during 1997–2004.

Multivariate analyses adjusted by sex, age, and HIV status (Table 5) showed that before 1997 higher education did not translate into statistically significant decreases in the risk of death for AIDS (RR 0.84; 95% CI 0.66–1.06), drug use (RR 1.21; 95% CI 0.75–1.93) and injuries (RR 0.93; 95% CI 0.52–1.68), but did for liver diseases (RR 0.10; 95% CI 0.01–0.75). During 1997–2004, IDUs higher educated had a lower risk of death for AIDS (RR 0.57; 95% CI 0.41–0.78), drug use (RR 0.52; 95% CI 0.30–0.89) and injuries (RR 0.37; 95% CI 0.19–0.71), compared with those with lower education. For liver disease mortality, those with higher studies had a higher risk of death, although the difference did not reach statistical significance (RR 1.13; 95% CI 0.63–2.01). Focusing on the impact of calendar period on cause-specific mortality, decreases in the risk of AIDS mortality were seen during 1997–2004 for both lower (RR 0.49; 95% CI 0.41–0.58) and higher (RR 0.33; 95% CI 0.23–0.48) educated; however, only those higher educated experienced a reduction in the risk of death for drug use (RR 0.54; 95% CI 0.28–1.05) and injuries (RR 0.52; 95% CI 0.23–1.21), compared with 1987–96.

Stratified analyses by HIV status showed that the observed effect of education on overall and cause-specific mortality, before and after the wider use of HAART, was not different for either HIV negative or positive. Similarly, we failed to find any evidence suggesting that the effect of educational level on mortality differed by sex (data not shown).

Discussion

We have shown the existence of inequalities in mortality according to educational level among IDUs, independently of HIV status. This inverse (lower education, higher mortality) gradient is evident across both calendar periods studied (before and after 1997) but is stronger after the wider use of HAART in our setting. The cause-specific analyses showed that the availability of the highly active antiretroviral therapy resulted in reductions in AIDS-mortality for both lower and higher educated, with a slightly greater improvement for IDUs with higher education. Interestingly, this lower mortality rate after 1997 among the higher educated was also observed and was particularly strong for deaths related to drug use and injuries.

Several methodological limitations of this study have to be addressed. One is the extensively reported loss of follow-up in cohorts of IDUs.23 Consequently we have combined the record linkage with the Registry of Mortality and the follow-up in Centres for AIDS Information and Prevention (CIPS) and area hospitals. Although some losses to follow-up are inevitable, we do not think that they are related to educational level. However, we have to assume a certain degree of underestimation of the overall mortality rates.

Regarding misclassification of exposure, certain bias may have taken place, as individuals with lower education at entry might have become higher educated during follow-up. This misclassification would tend to ascribe a decreased risk of death in those with lower education and would decrease the difference between lower and higher educated subjects, biasing the results towards the null. In spite of the bias, a clearly differentiated mortality by educational level was detected, emphasizing the strong impact that having lower education has on mortality in IDUs. Another type of bias that may result as a consequence of the mixed design (combination of and open and a close cohort) is the survivorship bias, as people lower educated could have worse progression than those higher educated and thus could not contribute into the analyses after 1997. However, this survivorship bias would tend to ascribe a lower mortality in those lower educated, decreasing the difference between lower and higher educated individuals. As described in the methods section, information on HIV status refers to the first visit to the centres, except for 289 (4.4%) subjects who seroconverted to HIV during the study period, in whom entry date was their first HIV-positive antibody test. We performed analyses with and without this group and the results were unchanged (data not shown). It is worth noting that the incidence of seroconversion to HIV in our cohort of seronegative IDUs was not associated with educational level.

There are few cohort studies of young adults that have examined social inequalities in mortality. A reason for this is the low death rates in this population group.6 Mortality rates observed in this study (21/1000 person-years) are extremely high, about 20 times higher than those of the general population of the same age and sex.14,24–26 Among IDUs, research on the association between educational level and mortality has produced inconsistent results: while some have reported no statistically significant association possibly because of the limited sample size,16,27 others have found lower death rates among the higher educated.14,28,29 However, the lack of information on HIV status prevented an assessment of whether the observed differences were attributable to a higher proportion of HIV infection among lower educated IDUs. With the wider use of HAART since 1996, some studies have reported lower socio-economic status or lower incomes to be associated with a higher risk of AIDS and death. This was explained in San Francisco by a lower use of or access to HAART17 and in France—in HIV subjects—by a lower adherence and compliance to treatment.18 In our setting, where there is free and universal access to HIV/AIDS care better educational level has been associated with better adherence to antiretroviral drugs in HIV-positive people,30 which appears to suggest only limited financial barriers to health-care access. Thus, it is unlikely that socio-economic status fully explains the differences in mortality for two main reasons: firstly, decreases in the risk of AIDS mortality were seen during 1997–2004 for both lower and higher educated, compared to the years before 1997. And secondly, the reduction in the risk of death during 1997–2004 among the higher educated IDUs was also observed in injuries and drug use-related causes. Duration of HIV infection, if linked to educational level, could partially explain the lower AIDS mortality among the better educated. However, adjusting for duration of injecting drug use (used as a marker of duration of HIV infection) did not change the effect of education on mortality.

Several explanations have been proposed to explain the potential mediating factors that might account for the inverse education-mortality gradient. Education achieved by a person might be a marker of living conditions at the parental home, thus reflecting early life conditions.31 It has been postulated that besides socio-economic circumstances, other characteristics in the parental home may exert an influence on trajectories in youth which lead to later afflictions causing higher risk of premature death from injuries.32,33 However, a study conducted in Finland6 showed that social class differences in total mortality among men in their middle adulthood were modestly determined by the parental home—with the exception of cardiovascular diseases and alcohol-related causes—but they were mainly attributable to educational, marital, and employment paths in youth. Nevertheless, the causes of death that mostly accounted for this association,34 are not relevant among the IDUs included in our cohort. Failures or inadequate support in early life or at vulnerable periods may give rise to an increased tendency to substance abuse or unnecessary risk taking.35 However, once drug abuse is deeply established, particularly intravenous drug abuse, as a result of lifestyle of IDUs, it is unlikely that living conditions at the parental house could contribute to explain differences in mortality.

In the general population, the pathway from level of education to mortality has been mostly attributed to material factors, partly via psychosocial and behavioural factors.36,37 In IDUs, differences in mortality could be attributed to a better access to or use of the HAART30 and harm reduction programmes among the higher educated. In this sense, the educational level of IDUs could be a marker of residence in neighbourhoods with better accessibility to the available resources. Ecological studies have shown that receiving social support36 or living in poor neighbourhoods is associated with an increasing risk of death.16,38,39 Moreover, the knowledge achieved through education may affect a person's cognitive functioning, making them more receptive to health education messages. However, behavioural factors related to health would not explain the important reduction observed during 1997–2004 for injuries.

In summary, differences in the risk of death between IDUs with lower and higher education indicate that even among IDUs, there is an excess of risk attributable to social inequalities. Independently of the relative contribution that each of the factors above mentioned has on the higher risk of death among the lower educated, recommendations on improving the material situation of IDUs with lower education will have, at most, a limited effect in reducing the strong educational inequalities in mortality. Our results highlight the importance of the education per se as a tool that provides people not only with more ability to cope with stressful events but also with a great capacity to deal with conflictive situations that increase the risk of death.

Reducing socio-economic inequalities in health is an important challenge for health policy worldwide. Some authors recommend the development and implementation of interventions aimed at improving the access, use, and adherence of HIV treatments and harm reduction programmes in low-educational groups. Although these policies could have a positive effect, our results show that promoting high quality education for everyone is probably a suitable strategy for reducing socio-economic inequalities in mortality among IDUs. Nevertheless, it is remarkable that for implementing education to be efficient, it has to be complemented with general availability of those services that are effective in reducing mortality by the diverse specific cause (i.e. specialist drug treatment, harm reduction programs, HIV clinics). At present, when inequalities in childhood health are increasing in Europe (UNESCO), we would like to underline that equal opportunities should start with a wider investment in education.

This work was financed through grants from Ministerio de Sanidad y Consumo, Instituto de Salud Carlos III (Ayudas para contratos post Formación Sanitaria Especializada), Consellería de Sanidad (orden, 21 febrero 2005) and ISCIII, Red de Centros RCESP C03/09. We acknowledge Johnattan Whitehead for revising the English version of the manuscript. Authorship criteria: IJ, BL, and IH-A designed the study. IH-A, SP-H, and IJ applied for funding. IH-A was responsible for the study concept and was project coordinator. IH and IF were responsible for data acquisition and data management. IJ was responsible for the analysis and interpretation. SP-H did the supervision of quality and accuracy of data and with I Ferreros made the cross-linkage of data with mortality registers. IJ, BL, and IH-A drafted the manuscript. All authors contributed to the interpretation of the results and improvements of the manuscript.

Conflict of interest None declared.

KEY MESSAGES

  • Inequalities in mortality according to educational level exist among injecting drug users, independently of HIV status, indicating an excess of risk attributable to social inequalities.

  • Among HIV-positive injecting drug users, the availability of highly active antiretroviral therapy results in reductions in AIDS mortality for both lower and higher educated, with a slightly greater improvement for injecting drug users with higher education.

  • This lower mortality rate after 1997 among the higher educated is particularly strong for deaths related to drug use and injuries.

  • Promoting high quality education for everyone would contribute to reduce mortality inequalities among injecting drug users.

References

1
Huisman
M
Kunst
AE
Bopp
M
, et al.  . 
Educational inequalities in cause-specific mortality in middle-aged and older men and women in eight western European populations
Lancet
 , 
2005
, vol. 
365
 (pg. 
493
-
500
)
2
Steenland
K
Henley
J
Thun
M
All-cause and cause-specific death rates by educational status for two million people in two American Cancer Society cohorts, 1959–1996
Am J Epidemiol
 , 
2002
, vol. 
156
 (pg. 
11
-
21
)
3
Krieger
N
Williams
DR
Moss
NE
Measuring social class in US public health research: concepts, methodologies, and guidelines
Annu Rev Public Health
 , 
1997
, vol. 
18
 (pg. 
341
-
78
)
4
Galobardes
B
Davey
Smith G
Jeffreys
M
McCarron
P
Childhood socioeconomic circumstances predict specific causes of death in adulthood: the Glasgow student cohort study
J Epidemiol Community Health
 , 
2006
, vol. 
60
 (pg. 
527
-
29
)
5
Dray-Spira
R
Lert
F
Marimoutou
C
Bouhnik
AD
Obadia
Y
Socio-economic conditions, health status and employment among persons living with HIV/AIDS in France in 2001
AIDS Care
 , 
2003
, vol. 
15
 (pg. 
739
-
48
)
6
Pensola
T
Martikainen
P
Life-course and mortality by adult social class among young men
Soc Sci Med
 , 
2004
, vol. 
58
 (pg. 
2149
-
70
)
7
Lynch
J
Kaplan
G
Berkman
LF
Kawachi
I
Socioeconomic position
Social epidemiology
 , 
2000
New York
Oxford University Presss
(pg. 
13
-
35
)
8
Borrell
C
Regidor
E
Arias
LC
, et al.  . 
Inequalities in mortality according to educational level in two large Southern European cities
Int J Epidemiol
 , 
1999
, vol. 
28
 (pg. 
58
-
63
)
9
Noguer
I
Castilla
J
Gutiérrez
MA
, et al.  . 
Epidemiología del SIDA en España
 , 
1995
Madrid
Ministerios de Sanidad y Consumo
10
García-Olalla
P
Caylà
JA
Jansà
JM
, et al.  . 
SIDA a Barcelona. Vigilancia Epidemiológica
1997
Barcelona
Institut Municipal de la Salut
11
European Centre for the Epidemiological Monitoring of AIDS: AIDS surveillance in Europe
Quarterly Rep
 , 
1996
, vol. 
49
 pg. 
23
 
12
Simon
PA
Hu
DJ
Kerndt
PR
Income and death rates in Los Angeles County
AIDS
 , 
1995
, vol. 
9
 (pg. 
281
-
84
)
13
Wilkinson
J
Lawes
G
Unell
I
Bradbury
J
Maclean
P
Problematic drug use and social deprivation
Public Health
 , 
1987
, vol. 
101
 (pg. 
165
-
68
)
14
European Monitoring Centre for Drugs and Drug Addiction, (EMCDDA)
2002
Lisbon
EMCCDA
 
Mortality of drug users in the EU: co-ordination of implementation of new cohort studies, follow-up and analysis of existing cohorts and development of new methods and outputs
15
Van Haastrecht
HJ
van Ameijden
EJ
van den Hoek
JA
Mientjes
GH
Bax
JS
Coutinho
RA
Predictors of mortality in the Amsterdam cohort of human immunodeficiency virus (HIV)-positive and HIV-negative drug users
Am J Epidemiol
 , 
1996
, vol. 
143
 (pg. 
380
-
91
)
16
Vlahov
D
Wang
CL
Galai
N
, et al.  . 
Mortality risk among new onset injection drug users
Addiction
 , 
2004
, vol. 
99
 (pg. 
946
-
54
)
17
McFarland
W
Chen
S
Hsu
L
Schwarcz
S
Katz
M
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
 , 
2003
, vol. 
33
 (pg. 
96
-
103
)
18
Piketty
C
Castiel
P
Giral
P
, et al.  . 
Lack of legal income is strongly associated with an increased risk of AIDS and death in HIV-infected injecting drug users
AIDS Care
 , 
1999
, vol. 
11
 (pg. 
429
-
36
)
19
Rapiti
E
Porta
D
Forastiere
F
Fusco
D
Perucci
CA
Socioeconomic status and survival of persons with AIDS before and after the introduction of highly active antiretroviral therapy. Lazio AIDS Surveillance Collaborative Group
Epidemiology
 , 
2000
, vol. 
11
 (pg. 
496
-
501
)
20
Schechter
MT
Hogg
RS
Aylward
B
Higher socio-economic status is associated with slower progression of HIV infection independent of access to health care
J Clin Epidemiol
 , 
1994
, vol. 
47
 (pg. 
59
-
67
)
21
Ministerio de Sanidad y consumo Recomendaciones del Consejo Asesor Clínico del Plan Nacional sobre el SIDA
2001
, vol. 
10
 
Madrid
Ministerio de Sanidad y consumo
(pg. 
1
-
16
)
22
Hernández-Aguado
I
Avino
MJ
Pérez-Hoyos
S
, et al.  . 
Human immunodeficiency virus (HIV) infection in parenteral drug users: evolution of the epidemic over 10 years
Int J Epidemiol
 , 
1999
, vol. 
28
 (pg. 
335
-
40
Valencian Epidemiology and Prevention of HIV Disease Study Group
23
Aviño Rico
MJ
Hernández Aguado
I
Pérez Hoyos
S
, et al.  . 
La incidencia de la infección por el virus de inmunodeficiencia humana tipo 1 (VIH-1) en usuarios de drogas por vía parenteral
Med Clin
 , 
1994
, vol. 
102
 (pg. 
369
-
73
(Barc)
24
Goedert
JJ
Pizza
G
Gritti
FM
, et al.  . 
Mortality among drug users in the AIDS era
Int J Epidemiol
 , 
1995
, vol. 
24
 (pg. 
1204
-
10
)
25
Hulse
GK
English
DR
Milne
E
Holman
CD
The quantification of mortality resulting from the regular use of illicit opiates
Addiction
 , 
1999
, vol. 
94
 (pg. 
221
-
29
)
26
Orti
RM
Domingo-Salvany
A
Muñoz
A
Macfarlane
D
Suelves
JM
Anto
JM
Mortality trends in a cohort of opiate addicts, Catalonia, Spain
Int J Epidemiol
 , 
1996
, vol. 
25
 (pg. 
545
-
53
)
27
Davoli
M
Perucci
CA
Forastiere
F
, et al.  . 
Risk factors for overdose mortality: a case-control study within a cohort of intravenous drug users
Int J Epidemiol
 , 
1993
, vol. 
22
 (pg. 
273
-
77
)
28
Schechter
MT
Hogg
RS
Aylward
B
Craib
KJ
Le
TN
Montaner
JS
Higher socioeconomic status is associated with slower progression of HIV infection independent of access to health care
J Clin Epidemiol
 , 
1994
, vol. 
47
 (pg. 
59
-
67
)
29
Borrell
C
Rodríguez-Sanz
M
Pasarin
MI
, et al.  . 
AIDS mortality before and after the introduction of highly active antiretroviral therapy: does it vary with socioeconomic group in a country with a National Health System?
Eur J Public Health
 , 
2006
 
[Epub ahead of print]
30
Gordillo
V
Del Amo
J
Soriano
V
Gonzalez-Lahoz
J
Sociodemographic and psychological variables influencing adherent to antiretroviral therapy
AIDS
 , 
1999
, vol. 
13
 (pg. 
1763
-
69
)
31
Davey Smith
G
Hart
C
Hole
D
, et al.  . 
Education and occupational social class: which is the more important indicator of mortality risk?
J Epidemiol Community Health
 , 
1998
, vol. 
52
 (pg. 
153
-
60
)
32
Pasarin
MI
Borrell
C
Brugal
MT
Diaz-Quijano
E
Weighing social and economic determinants related to inequalities in mortality
J Urban Health
 , 
2004
, vol. 
81
 (pg. 
349
-
62
)
33
Galobardes
B
Lynch
JW
Davey Smith
G
Childhood socioeconomic circumstances and cause-specific mortality in adulthood: systematic review and interpretation
Epidemiol Rev
 , 
2004
, vol. 
26
 (pg. 
7
-
21
)
34
Tyndall
MW
Craib
KJ
Currie
S
Li
K
O'Shaughnessy
MV
Schechter
MT
Impact of HIV infection on mortality in a cohort of injection drug users
J Acquir Immune Defic Syndr
 , 
2001
, vol. 
28
 (pg. 
351
-
57
)
35
Rutter
M
Pathways from childhood to adult life
J Child Psychiat
 , 
1989
, vol. 
30
 (pg. 
23
-
51
)
36
Lundberg
O
The impact of childhood living conditions on illness and mortality in adulthood
Soc Sci Med
 , 
1993
, vol. 
36
 (pg. 
1047
-
52
)
37
Van Oort
FV
van Lenthe
FJ
Mackenbach
JP
Material, psychosocial, and behavioural factors in the explanation of educational inequalities in mortality in the Netherlands
J Epidemiol Community Health
 , 
2005
, vol. 
59
 (pg. 
214
-
20
)
38
Smith
GD
Gunell
D
Ben-Shlomo
Y
Leon
DA
Walt
G
Life-course approaches to socioeconomic differentials in cause-specific adult mortality
Poverty, Inequality and Health. An International Perspective
 , 
2001
Oxford
Oxford University Press
(pg. 
88
-
124
)
39
Hembree
C
Galea
S
Ahern
J
, et al.  . 
The urban built environment and overdose mortality in New York City neighborhoods
Health Place
 , 
2005
, vol. 
11
 (pg. 
147
-
56
)