Living Conditions and Illness among Injecting Drug Users in Montreal

Injection drug behavior constitutes a serious public health problem in the developed world, especially in North America [1-6]. At the end of the last decade in Montreal, Quebec, about 12,000 persons [7] were injecting drugs. Public health concerns related to illegal drug injection include the spread of HIV [3,8,9] and HCV infections [10-12], sexually transmitted diseases [13] and mental health problems [14,15]. In spite of their high morbidity, injection drug users (IDUs) reportedly misuse health services, particularly by overusing hospital emergency rooms [16-23].

The living conditions of IDUs result in an accumulation of health risk factors. In addition to illegal drug abuse, alcohol and cigarette consumption are common [17]. Many IDUs have a history of homelessness [4]. Substance abuse in IDUs constitutes a major determinant of unsafe sexual contacts leading to sexually transmitted diseases [13]. Drug dependency, financial strain and debt may drive them to violence or to trade sex to obtain money or drugs [20,24,25]. Such harsh living conditions may eventually lead to illegal behaviors and multiple imprisonments [26][27][28]. The social and health problems of IDUs are interwoven with poverty and social exclusion. Phelan et al. [29] have extensively studied the way ill health conditions could impact people's health status, conditions they called "fundamental causes" of social inequalities in health.
A number of health problems are associated with drug use. It has been estimated that 70 to 80% of IDUs in Montreal may be infected with HCV [10][11][12]. The estimated prevalence of HIV-infected IDUs amounts to 11% in Montreal city [8,30]. Dual diagnosis of drug addiction and psychiatric problems is frequent. At the end of the last decade, the number of illicit drug users (by injection or any other route) suffering from mental illness in Montreal city ranged from 25,000 to 40,000 [14]. A study carried out in other contexts among drug users reported a high frequency of complications stemming from drug injection, such as soft tissue infections, thrombosis, embolism and septicemia [31]. Illicit drug use is frequently associated with alcohol abuse and cigarette smoking [17], sex trade and traumas from violence [13,25,32]. One study has reported an association between overdoses from heroin and suicide attempts [33].
Despite over a decade of intensive harm reduction strategies and the many steps taken to address drug-related health issues, the health status of IDUs is still a cause for concern. In a study carried out in Vancouver, Spittal et al. [34] revealed the high health risk of female IDUs (compared to the female general population of British Columbia) by observing a fifty-fold increase in their mortality rate. Corneil et al. [35] showed an increased risk of HIV infection among Vancouver IDUs who reported living in unstable housing conditions. This study identifies both proximal and distal factors associated with recent episodes of illness among IDUs. By targeting acute manifestations of illness in IDUs, such as overdoses and soft tissue infections, intervention programs only partly address morbidity issues among this population. First, etiologic research should identify distal determinants of IDUs' morbidity, in order to implement welltailored intervention programs targeting the entire spectrum of IDUs' risk factors for morbidity. By focusing prevention on those distal determinants, or "fundamental causes" according to Phelan et al. [29], health service managers could redirect efforts to integrated care for chronic problems such as alcohol dependency, HIV infection, hepatitis C and mental health, in tandem with social services devoted to the IDU population.

Study population and data collection
The study population consisted of injecting drug users living in Montreal who were 18 years of age or older. Participants were recruited on the streets of downtown Montreal using a convenient sampling method in which selected participants could refer their IDU friend to the interviewer [36,37]. The interviewer, a former injection drug user, completed a training session before the survey and had easy access to IDUs. To avoid selection bias due to subjective selection of known IDUs, the interviewer was instructed not to contact friends or relations, but simply to inform IDUs about the study and distribute his business cards so that those interested could call to arrange an interview. Each IDU contacted was also asked to invite other known IDUs to participate in the study. The inclusion criteria were: residence in Montreal for at least one year, age 18 years or over and intravenous drug use at least once in the previous 6 months. Participants signed an informed consent form, which contained a numerical code to match with the anonymous questionnaire. Participants were interviewed. The questionnaire was filled in by the interviewer. Most of the interviews took place in our research office. But some interviews were carried out in other offices in the neighborhood of the participants, such community health centers, syringe exchange program centers, etc. Confidentiality and discretion were the conditions required to use an office for the interview. The Research Ethics Board of the University of Montreal approved the study.
On completion of the questionnaire, participants received a payment of CAN$10 to compensate them for their time. Respondents whom the interviewer judged to require particular services were also given a brief counseling session and referred to a social service. The use of a single interviewer was helpful in preventing people from answering the questionnaire more than once, since he could generally recognize those who had already participated. The study was conducted from February to September 2005.

Measurements
The dependent variable for this study was the self-reported occurrence of any illness episode in the 6 months before the interview, in response to the questions: "Have you suffered from any illness during the last 6 months? What was your health problem?" The list of potential explanatory variables was drawn from the literature on the morbidity of IDUs and included sociodemographic variables, economic conditions, marginality, risk behaviors and health status, according to the schema proposed by Estébanez et al. [38]. Socio-demographic characteristics (gender, age, education and sexual orientation) may have a direct or indirect impact (through marginality and risk behaviors) on the occurrence of episodes of illness. Sexual orientation included three categories: homosexuals, heterosexuals and bisexuals. Employment, type of housing and obvious indicators of financial strain, such as receiving regular help from a community center (i.e. clothing, food or furniture) and begging on the street, were used as a measure of economic condition. Combining the latter two factors (receiving help and begging) yielded the variable "number of indicators of financial strain". Employment status consisted of three categories: full time job, other jobs (part time job, independent job, occasional jobs), and welfare.
Marginality may act directly or indirectly through risk behaviors, on the occurrence of an episode of illness. The marginality indicators included: sex trade, fines for criminal offenses, previous imprisonment, unemployment and homelessness. Living arrangements consisted of three categories: independent living arrangements (rented apartment or house), dependent living arrangements (family house, friend's house, public shelter for homeless people), and homelessness (living on the street or in abandoned houses). The variable "number of marginality indicators" was created by combining history of imprisonment, fine for criminal offense, stealing and lack of identity cards.
Risk behaviors were measured for the six months before interview and included alcohol consumption, the type of drug injected the frequency of drug injection and sharing injection materials. Sharing injection materials was defined as giving to one another used materials (syringe, needle, filter, etc.) to inject drugs, in the prior 6 months. The injection drugs considered were mainly heroin, cocaine and their derivatives (crack cocaine, speedball). Two variableshaving participated in treatment for drug abuse and past or present participation in a needle exchange program -were used to evaluate access to social support, while prior visit to preventive health clinics was the criterion for health services utilization.
The health-related variables were chronic infectious disease (HIV infection, HCV infection) and a history of mental illness. Mental illness was defined as any psychiatric illness diagnosed by a healthcare professional (such as schizophrenia, schizophrenia spectrum disorders, bipolar disorders, mania, major depression, anxiety disorders, etc.), and not merely any self-perceived mental disorder not evaluated by a mental health professional. Depressive symptoms were assessed using a 13-item CES-D (Center for Epidemiologic Studies Depression) scale [39], scored on a scale of 0 to 39 points (i.e., 0 to 3 points per item) with a cut-off at 13 points.
All those potential risk factors for episodes of illness can be divided in two main groups: distal factors (demographic factors, socioeconomic conditions, marginality and risk behavior) and proximal factors (health status and chronic diseases).

Statistical analysis
Data quality was monitored by checking for possible duplications after listing subjects in an Excel file by name, reported age, birthday, and age calculated from reported birthday. Major discordances or incoherencies (e.g. declared age that did not match age calculated from the birthday), similarities in names or age, were then analyzed using SPSS software. Questionnaires considered to be duplicates based on the foregoing information and a comparison of participants' signatures, were excluded.
Bivariate analyses were performed for each independent variable to calculate the potential association with the occurrence of illness in the previous 6 months, and the statistical significance of the relation was assessed using Pearson's X 2 test. Multiple logistic regression models were fitted using staggered entry of variables according to the previously described schema: sociodemographic factors, economic conditions, marginality, risk behaviors, support and service use, and health status. Within each block, variables were selected by using a stepwise backward strategy in which statistical criteria for entry and retention of variables in each model were p ≤ 0.10 and p ≤ 0.05, respectively. Blockwise entry is the best strategy to highlight changes in the values of the coefficients following inclusion of explanatory variables in the model. The log-likelihood statistic and the Chi-square test were used to assess improvement in the model, while the Hosmer-Lemeshow test was used to evaluate its goodness of fit.

Results
A total of 678 subjects responded to the questionnaire. After completing the data-quality monitoring process, 12 questionnaires were excluded. Further analyses were carried out on the remaining 666 participants. Only 17% of them had completed more than secondary school (college or university); 12% had a full time job, while 70% were receiving social welfare benefits. Six percent had no identity card, 38% begged on the street, 48% had a history of imprisonment, 49% reported receiving help from a community center on a regular basis, and 20% were living strictly on the street.
Within the whole sample, 176 subjects (26%) reported an episode of illness in the previous 6 months. These episodes included drug overdoses and abscess at the site of injection; acute infections such as pneumonia, influenza, and gastroenteritis; mental illness and suicide attempts; traumas from violence; and fatigue & indigestion and herpes & sepsis. Overall, 140 participants reported 1 episode, 31 reported 2 episodes and 5 reported 3 episodes. Tables 1-3 show the distribution of the sample according to the selected risk factors for episodes of illness. Overall, 36% of females versus 25% of males had some illness in the 6-month recall period. Older IDUs and bisexuals were more likely to report illness episodes than younger IDUs and heterosexuals or homosexuals ( Table 1). Indicators of financial strain (begging on the street, receiving help in community centers) are associated with episodes of illness. Homeless IDUs were more likely to report episodes of illness than those living in the home of friends or family, or than those who lived in their own house or apartment (respectively, 34%, 28% and 24%; P=0.093).
All the marginality indicators were related to the frequency of episodes of illness: IDUs involved in sex trade (42% versus 24%; P<0.001), and those with a history of imprisonment (34% versus 21%; P<0.001) ( Table 3). The type of drug consumed and the frequency of drug injection were associated with episodes of illness. Those who injected both heroin and cocaine reported more episodes than those who injected only heroin or cocaine (37%, 31% and 23%, respectively; P=0.007). IDUs who injected drugs more than once a day were more likely to have an episode of illness than those reporting a lower frequency of injection (33% versus 21%; P<0.001). Sharing injection materials was also associated with a higher frequency of illness (45% versus 23%; P<0.001).    IDUs using community services were more likely to report illness in the previous 6 months (Table 4). In particular, those with a history of drug abuse treatment, those receiving help from community centers and those who had used preventive services in the past for STD testing, hepatitis testing, vaccination or needle exchange had more illness episodes than those who had not used these services. Chronic infections with HIV and HCV, mental illness and current high depressive symptoms were also associated with higher frequency of illness episodes in the 6 previous months.
The estimated odds ratios using multivariate analysis are shown in Table 5. Gender and age were significantly associated with the probability of episodes of illness even after adjusting for all other risk factors, and their coefficients remained stable throughout the five models. Female IDUs had a two-fold increased risk of illness compared to male IDUs. Older IDUs were more likely to have an illness than younger IDUs. Homosexuals had 60% more risk of illness than heterosexuals. This association was not significant, except after adjusting for marginality. Conversely, the likelihood of illness in bisexuals was stable up to the final model and consistently remained twice as high as in heterosexuals. Financial strain and marginality were also independently associated with illness, and had stable coefficients. Persons who injected heroin had a two-fold increase in the risk of illness compared to those who injected cocaine only. Those who injected both cocaine and heroin had the highest odds of illness compared with those who injected just one of these drugs. Sharing injection materials was associated with a two-fold increase in the risk of illness. In the final model, three chronic conditions were associated with the occurrence of illness: HIV infection, HCV infection and mental illness. As shown by the ascending values of the Chi-square test and the descending values of the log-likelihood statistic, the final model improved progressively as covariates were added to the equations.

Discussion
The purpose of this study among Montreal IDUs in 2005, 20 years after the rapid spread of HIV among IDUs in North America, was to identify distal and proximal factors associated with recent episodes of illness. These IDUs, whose mean age was 31 years (± 10), and who had been using injection drugs for an average of 9.52 years (± 7.47), constitute a cohort of people who have managed to survive amidst the HIV epidemic.
This study aimed to increase our understanding of the fundamental causes of IDUs' ill health in spite of many years of harm reduction programs. The results can be summarized as three principal findings. First, financial strain and marginality are associated with recent episodes of illness. Second, risk injecting behaviors continue to be highly prevalent and, as expected, are associated with recent episodes of illness. Third, mental illness, HIV and HCV infections are at the core of poor health in IDUs.
Multivariate analyses show the major predictors of recent illness episodes, illustrating mainly that the pathway from socioeconomic conditions to occurrence of illness in IDUs is shaped by financial strain, marginality and risk behaviors, mostly in those whose health status is already weakened by chronic viral infections.
Financial strain seems to be estimated with accuracy by the two variables "begging on the street" and "receiving help from a community center". Indeed, other variables could be considered such as "homelessness" and "having no job". But those variables are weakly associated with disease occurrence (P=0.093 and 0.150 respectively) and do not strongly illustrate the IDUs' situation of misfortune.
Indicators of marginality seem to be well represented by history of imprisonment, fine for criminal offenses, stealing and having no *P<0.05 a Begging, receiving help in community centers b Imprisonment, fine, stealing, no identity card identity card. We considered including sex trade with these factors but this was rejected in the multivariate model using backward regression, because its association with disease occurrence is mediated by other factors of marginality.
The association between heroin injection and illness was not significant but became stronger once chronic diseases were considered. Among those who injected both heroin and cocaine, the association was even stronger and increased in magnitude when chronic diseases were taken into account. This observation illustrates the synergistic effect between drug abuse and ill health, as shown in the fact that drug abuse is more detrimental in IDUs whose health status is already compromised.
Our findings are similar in some respects to those of other studies in IDU populations. The participants were predominantly male, as has been seen in many other studies [17,20,[40][41][42]. Female IDUs seemed more likely to be ill than male IDUs, as has also been shown by Chitwood et al. [43]. Bisexual IDUs seemed to be at higher risk than heterosexual IDUs. Boulton et al. reported that, while homosexual men are more likely to have protected sexual contact with their male partners, bisexual men usually engage in protected sex with men and unprotected sex with female partners [44]. In our study, the odds ratio for illness in bisexuals was significant in the first (OR: 2.11 (1.19-3.72) and all subsequent models, suggesting that unprotected sex may also be an explanatory factor.
Other studies have reported the link between financial strain and high morbidity [5,42]. In bivariate analyses, access to needle exchange programs has been associated with increased morbidity [45], a finding that may be explained by the attraction of needle exchange programs to IDUs at higher risk of HIV infection.
Selection and measurement biases may have gone undetected in our study. Selection biases could have resulted from the non-probabilistic nature of the design, yielding an unpredictable direction in the associations. In addition, self-reported illness and risk behaviors could have been influenced by social desirability, which would have reduced the magnitude of the associations. Nevertheless, previous studies have already confirmed the reliability of self-reported data in IDUs [46][47][48][49]. Subjects suffering from acute illnesses could have reported their risk factors with more precision than those who were feeling healthy at the time of the interview, a situation that could have led to a recall bias with an association towards the null. Confounding factors such as violence and lifelong victimization were not collected, although the episodes of illness related to violence were high for a 6-month recall period (10% of episodes). Like any cross-sectional survey, this study, along with the statistical inferences yielded by analyses, should be interpreted cautiously. As far as we know, this is the first population based study of IDUs in the city of Montreal. All previous studies have been based on clinic and social service attendees [3,8,12,36,44]. The findings of this study contribute to our knowledge of the relation between living conditions and morbidity among these survivors of the HIV epidemic.

Conclusions
Many harm reduction strategies have been implemented during the last decade [50][51][52]. Efforts have been made to help IDUs reduce risk-taking behaviors as regards safe injection practices and safe sex [53][54][55][56][57]. Integrated programs focused on harm reduction strategies in connection with primary care and drug abuse treatment have been proposed [42].
This study highlights the relevance of taking a broad perspective when studying determinants of morbidity in IDUs. From our analyses and other studies, there is strong evidence suggesting that the high rate of morbidity in IDUs is due to social exclusion and their extremely harsh living conditions. A better organization of primary health care would result in even greater utilization of health services unless measures are taken at the social baseline to improve the living conditions of IDUs, notably for street-entrenched, runaway and unemployed IDUs [57]. In addition, Gunn et al. [50] have proposed meaningful solutions related with harm reduction strategies, notably through improvement of access to the primary health care system. Health improvement programs should prolong downwards to the social ground where the IDU population lives, encompassing living arrangements, mental rehabilitation and occupational therapy. The harm reduction strategies proposed by Palepu et al. [42] should be considered as well. Health needs in IDUs are complex and should be addressed primarily at a more remote step, in the community, providing integrated care according to their individual conditions with the implication of outreach workers, social workers and nurses who have close ties with IDUs.