Background
It has been estimated that there are 170,000 people who inject drugs (PWID) in Malaysia [
1]. HIV prevalence among PWID in Malaysia ranges from 25 % to 45 % [
1]. Further, people who inject drugs (PWID) account for 39 % of all new reported cases of people living with HIV in Malaysia, deeming drug risk behavior a serious public health concern in the country [
2]. Malaysian fishermen have not only grown by 22 % in the past decade, they have been identified as a key-affected HIV population with rates 10 times higher than national rates [
3‐
5]. Previous HIV research among fishermen has mainly focused on sexual risk behaviors (see review [
6]). Recent research examining injection drug use behavior among fishermen is scarce, although growing (e.g.,[
7,
8]), and indicates that injection drug use and risky injection drug use (i.e., receptive and non-receptive needle/syringe sharing, frontloading and back-loading, sharing equipment, sharing drugs from a common container, or adding blood to the drug solution before injecting increasing one’s risk of HIV infection) may play a central role in the high rate of HIV among fishermen.
Psychosocial and structural-level stressors may contribute to HIV injection drug risk behavior among this key-affected population. Our study is guided by the ecological framework perspective, which suggests that multi-level factors contribute to an individual’s health and well-being at micro, meso and macro levels [
9]. Based on this theoretical approach and empirical research conducted in Western and non-Western settings, depressive symptoms (individual/micro) [
10‐
15], adverse interactions with police (social/meso) [
16], and poverty-related stressors (structural/macro) [
17‐
19] have been noted to be associated with HIV injection drug use and injection risk behavior (e.g. rushed injections, needle sharing). Further, while originally developed to understand the pathway between sexual abuse and HIV sexual and drug risk behavior, Miller’s [
20] conceptual model may be adapted to understand stressors/potentially traumatic events and HIV drug risk behavior among Malaysian fishermen. The model posits that stressors contribute to the use of maladaptive coping mechanisms leading to HIV risk behaviors [
20]. Drug use behaviors may also be understood as a way to self-medicate in response to individual, social, and structural stressors, and may also be more easily available and effective as an immediate coping method [
14,
20]. The model also suggests that feelings of hopelessness and despair, which may be associated with stressors at the individual, social or structural levels, are related to unsafe injection practices [
14,
20,
21].
Research has identified a number of multi-level stressors, which are associated with injection drug use and risky injection drug use. On an individual level, numerous studies conducted in Western (e.g.,[
15,
22,
23]) and non-Western settings (e.g.,[
10,
21,
24]) suggest moderate to severe levels of depressive symptoms among injection drug users. Data also suggests that PWIDs are at risk for suicidal ideation and suicide attempts [
10,
25,
26]. Research has also indicated a consistent relationship between depressive symptoms and risky injection drug practices [
15,
27‐
29]. For example, in a meta-analysis, Conner and colleagues [
27] found a significant relationship between depression and needle sharing among injection drug users, and suggests that moderate to severe levels of depression may increase needle sharing due to feelings of hopelessness or as an attempt to cope with negative emotions by engaging in social interactions with others.
At the social level, the association between police interactions and risky injection practices, such as rushed injections, needle sharing and use of shooting galleries, has been noted among a number of studies [
17,
30,
31]. Specifically, risky injection practices and HIV prevalence are associated with arrests by police and police removal of syringes [
32‐
34]. This may be due to rushed and hurried injections without clean needles for fear of coming in contact with the police or an increased likelihood in borrowing needles from others when police remove syringes [
35]. A perceived increase in police presence was noted to be a risk factor for midazolam injection among injection drug users in Thailand [
36]. Unlawful harassment and abuse by police has also been noted to increase risky injection practice among injection drug users [
37,
38]. Similar to depression, abuse by police may increase the likelihood of risky drug injection practices due to feelings of hopelessness and despair as a result of trauma and increase maladaptive coping mechanisms [
20]. As such, it is critical to examine arrests and abuse by police and the relationship with risky injection drug use practices among the key-affected population of Malaysian fishermen.
At the structural level, a number of studies have indicated that poverty-related stressors, (i.e., substandard housing conditions, homelessness, and food insecurity) are related to injection drug use [
17‐
19]. Similar to the relationship with symptoms of depression, injection drug use may be a way to cope with daily and ongoing financial stressors. Substandard housing and homelessness has been found to be associated with both injection drug use [
39‐
42] and engaging in risky drug use behaviors such as needle sharing [
10,
18,
19]. Further, homeless PWIDs are more at risk for HIV infection compared to those who are housed [
19]. In a U.S based longitudinal study among PWIDs, Aidala and colleagues [
39] found that a change from homelessness to sufficient housing significantly reduced risk of injection drug use, needle use and needle sharing. Additionally, Weiser and colleagues [
43] found a relationship between food insecurity and HIV risk behavior. Poverty-related stressors, namely homelessness, poverty, financial stress and food insecurity have not been examined specifically among Malaysian fishermen. As poverty-related stressors have been noted to be associated with injection drug use and risky drug use practices among other populations, it is critical to understand this relationship among this key HIV-affected population for prevention efforts.
The aim of this paper is to address the gap in research on depression, policing and poverty-related stressors as predictors for injection drug use and risky injection drug use among Malaysian fishermen. Examining stressors at the individual, social, and structural level among Malaysian fishermen will contribute to the development of more effective prevention and intervention efforts, which will potentially address multiple risk factors for HIV. As such, in the current study, we examine: 1) the prevalence of depression, policing, poverty-related stressors (i.e., homelessness, poverty and food insecurity), injection drug use, risky injection drug use, and HIV infection among a sample of 406 Malaysian fishermen; 2) the relationship between depression, policing, poverty and injection drug use and risky injection drug use among this sample of Malaysian fishermen; and 3) the relationship between depression, policing (specifically, adverse interactions with police), poverty-related stressors and risky injection drug use among a sub-sample of 154 Malaysian fishermen with a history of injection drug use. Based on previous research among other populations, we hypothesize that after controlling for demographic variables, depression, policing and poverty will predict injection drug use and risky injection drug use among the sample. Further, among fishermen who endorse a history of injection drug use, we hypothesize that after controlling for demographic variables, depression, experiencing abuse by police and poverty will be associated with HIV risky injection drug use behaviors.
Methods
This paper used data from Project WAVES, a project designed to assess HIV prevalence, risk behavior, and characteristics of HIV risk behaviors among fishermen in Malaysia. The study utilized a mixed methods approach with in-depth interviews and a quantitative cross-sectional survey collected between 2009 and 2011. The current study uses data solely from the quantitative survey collected from July to December 2011.
Study population
In 2009, 3,720 men were registered fishermen in Kuantan [
3]. However, these numbers do not include those who are unregistered and unlicensed in the region, typically small-scale artisanal fishermen. As there is no universal definition and agreement regarding what is deemed a small-scale fishery, there is consensus in the underestimation of the number of small-scale fishermen [
44]. Globally, the prevalence of small-scale artisanal fishermen is understood to be significantly larger than large-scale commercial fishermen and contribute to more than half of fish catch worldwide [
45].
Study site
The study was conducted in Pahang state, which is around the Kuantan jetty. This is one of the busiest fishing jetties in the country and central to commercial and small-scale fishing in Malaysia. In 2006, Kuantan was one of the first towns in Malaysia to implement the Needle and Syringe Exchange Program deeming this an appropriate and relevant place for the study. Sampling locations comprised two government-owned commercial fishermen’s wharfs in Kuantan and a fishing village within a 100 KM radius of Kuantan.
Sampling and data collection procedures
Recruitment for the survey was based on Respondent Driven Sampling (RDS), a network, chain referral sampling methodology utilized for the recruitment of hard to reach, vulnerable or mobile populations [
46,
47]. While fishermen, in general, are not a hidden population per se, they are highly mobile. There is also no adequate sampling frame that exists for fishermen in Malaysia that includes both large and small-scale fisheries. Current official lists are restricted to Malaysian vessel owners (both commercial and non-commercial with no database specifically capturing non-vessel owning fishermen (i.e. deckhands aboard vessels) as well as non-Malaysian nationals who tend to operate illegally. Furthermore, for the purpose of this estimation, it is important to adequately recruit fishermen who are involved in illicit (i.e., substance use) behaviours, which may not be easily captured by standard surveillance methods [
48].
High mobility and the stigma of illicit behaviors makes sampling a challenge among Malaysian fishermen. Respondent-driven sampling (RDS), a coupon-based chain-referral method, was used in the current study to address some of the aforementioned obstacles [
46,
48‐
50]. RDS was chosen as an economical and time-efficient recruitment method which has been successful in recruiting PWID [
51‐
53]. In addition, RDS uses a statistical technique that adjusts for potential over-reliance on characteristics if the initial sample and bias towards participants who may be typically cooperative [
49]. Informed by data from the social network of the participants, prevalence estimates are determined of specific traits in the target population [
50].
In RDS, a small number of participants are recruited as ‘seeds,’ representing the characteristics of the population of interest and socially well-connected to the target population. After the seeds complete the survey, they are then provided a fixed number of coupons (3 in our study) to distribute to social network members who meet inclusion criteria. The coupon is required for screening into the study after the initial seed. Each participant is then subsequently provided the same number of coupons. Data are collected in the social networks of each participant with anonymous identification numbers to link the recruitment chains throughout the study. With each wave, new recruits become more independent of the index participants of each wave, decreasing bias and eventually reaching equilibrium [
46,
49].
In the current study, eight seeds were recruited based on their motivation to participate and how socially connected they were to the fishing community. Three initial seeds reported they were fishermen who used drugs and three reported non-drug use. Two additional drug-using fishermen seeds were added because of lack of recruitment by two of the initial seeds. Recruited participants who completed the survey received RM50 as compensation for their time. Upon completion of the survey each participant received three coupons to recruit other fishermen. As secondary incentive, the recruiting participant received an additional RM25 for a successful recruit.
Eligibility criteria included being: a) male, b) 18 years or older, c) working as a fisherman full-time (six months or more in the past year), d) speaks Malay, and e) provided written informed consent. At each sampling location, a person who injected drugs (PWID) and a non-injector were recruited (with a total of six initial participants). Every participant who completed the study received three coupons to recruit peers from their social network. Recruitment was complete when the sample reached equilibrium, which was determined when key variables (HIV, injection drug use status, and time spent at sea) changed by less than 2 % between recruitment waves [
46]. This was determined from previous research which indicates that the leading wave approximates equilibrium within 2 % after 6 waves when respondents recruit three peers [
46,
49].
As previously noted, quantitative data for Project Waves were collected between July and December 2011. The questionnaire was self-administered on a laptop computer using Questionnaire Development System (QDS) software (Nova Research Company, Maryland, USA). All questions were in Malay. All data for the current study can be found in the Additional File
1.
Measures
All sections of the survey were translated to Malay and pilot tested among fishermen from Kuantan, Pahang to ensure both face and content validity of the measures.
Demographic variables
Demographic variables examined in this study included, age, ethnicity, marital status, religion, and education. Participants were also characterized by the type of vessel in which they worked (commercial vs. traditional) and their occupational role (captain vs. deckhand). Working on a commercial vessel indicated large-scale fisheries. All demographic variables were dichotomized.
Depression
Depression symptoms were assessed using the Brief Symptoms Inventory (BSI; [
54]) 6-item depression subscale (including symptoms related to dysphoric mood and affect, lack of motivation, and loss of interest in life). Participants were asked how much each symptom bothered them in the past 7 days on a Likert scale from 0 = not at all to 4 = extremely. Scores from the 6 items were summed into a total score and then converted into a T-score with a mean of 50 and standard deviation of 10. A cutoff of 63 or above indicated severe depressive symptoms, based on previous research with a normative non-patient population [
55].
The BSI depression subscale has been used widely among adults in both western [
56,
57] and non-western settings [
56,
58]. Previous research has demonstrated adequate internal reliability of the depression subscale ranging from Cronbach’s alpha = 0.85 [
54] to 0.96 [
59]. The BSI has also demonstrated adequate criterion and construct validity [
54]. In the current study, reliability was good with Cronbach’s alpha = .85 (one-sided 95 % confidence interval = .83).
HIV
During the informed consent process, individuals were informed that the study would involve a rapid test for HIV. Willingness to undergo HIV testing was an eligibility criterion for enrolling in the study. All participants underwent pre-test and post-test counselling by research assistants trained in HIV counselling, testing, and referral. Participants were also asked if they had been previously tested for HIV and, if so, asked the result of their most recent test. HIV serology was determined using a rapid test (ACON HIV Rapid Test kit, ACON Laboratories, California, USA); reactive results were confirmed with a second rapid test (Intec Products Inc., Xiamen, China). Two reactive results were required for subjects to be classified HIV positive. Two participants had discordant results between the two tests and were classified as HIV negative.
Policing
Participants were asked if they had “ever been arrested” (with 0 = never been arrested and 1 = been arrested 1 or more times). Among those who reported a prior arrest, the “number of times arrested” (continuous variable), as well as whether they had “ever been brought to police lock-up” were asked (with 0 = never been brought to police lock-up and 1 = been brought to police lock-up 1 or more times). Among those who had been brought to police lock up, “number of days detained” (continuous variable) was obtained. Participants were also asked if they had “ever gone to prison” (with response categories of 0 = never gone to prison and 1 = been to prison 1 or more times) and, if so, the “number of times been to prison.”
Participants who endorsed “any previous drug use” were also asked an additional 7 questions related to negative police interactions. These additional questions were formulated for the Project Waves study and based on formative qualitative research we conducted among fishermen in the Kuantan region. Specifically, participants were asked if “fear of the police ever caused a hurried or rushed injection” (0 = No and 1 = Yes), if “the police ever demanded money to buy back drugs that had been confiscated” (0 = No and 1 = Yes), if they had “ever been beaten or tortured by the police” (0 = No and 1 = Yes), if the “police ever planted drugs on them” (0 = No and 1 = Yes), if the “police ever took their syringes” (0 = No and 1 = Yes), if the participant ever “avoided carrying syringes for fear of the police” (0 = No and 1 = Yes), and “in the past 6 months, the number of times the boat or place of work was raided by the police” (0 = No and 1 = Yes).
Poverty-related stressors included: “being at or below the poverty line” (i.e., a monthly income at or below RM820, approximately 273 USD), or “borrowing money in the past 3 months.” In addition, food insecurity was defined as “having enough money to buy food in the past 3 months.” Homelessness was defined as “having a consistent place to stay in the past 3 months.” All poverty-related variables were dichotomized (No = 0, Yes = 1).
Injection drug use
Injection drug use in our study was defined as ever intravenously injecting a substance. Injection drug use was determined using a series of questions covering 8 distinct substances, with an additional “other” category. Specifically, participants were asked if they had ever injected the following: 1) Subutex, Suboxone, Buprenorphine; 2) Ketamine; 3) Pil kuda; 4) heroine; 5) Ice, Syabu Crystal Meth; 6) Methadon; 7) Ecstasy; and 8) Dormicum, Benzodiazapene. Each respondent was asked whether they had ever injected the drug, and was coded as ‘1’ if they reported injecting ANY drug (PWID) and ‘0’ if they had never injected any of the substances (non-injector).
Risky injection drug use
Risky injection drug use behaviors were assessed among participants who endorsed injection drug use. Risky injection drug use behavior was defined as unsafe practices when intravenously injecting a substance. Unsafe injection practices in the past month was a dichotomous measure based on the Risk Behaviour Assessment [
60], which asks a series of eight questions on injection-related risk, covering receptive and non-receptive needle/syringe sharing, frontloading and back-loading, sharing equipment, sharing drugs from a common container, or adding blood to the drug solution before injecting. The respondent was coded as ‘1’ if they had engaged in any of these behaviors one or more times in the past month, indicating unsafe injection practices, and ‘0’ if they reported zero times for all behaviors.
Data analysis
All analyses were conducted in STATA SE13 [
61]. Fisher’s exact tests or χ
2 tests were used to examine group differences among those who reported injection drug use and those who reported no injection drug use for categorical variables, and t-tests were used to examine differences between groups for continuous variables. Missing data was minimal across all measures; the vast majority had 0 % missingness, with a few variables that had 1 % missingness due to nonresponse. Where missing, figures are reported using available data. All variables of interest were examined as potential confounders (e.g., HIV status and depression) and all predictors were assessed for multi-colinearity before being entered in multivariate logistic regression models. Any variables that were collinear were removed and one variable was chosen as a marker for the construct. Multivariate logistic regressions were conducted controlling for demographic variables with psychosocial and structural-level HIV risk factors as predictors of injection drug use behavior and risky injection drug use (compared to both non-risky injection drug users and non-injection drug users). Finally, a sub-sample multivariate logistic regression was conducted among injection drug users examining psychosocial and structural-level HIV risk factors as predictors of risky injection drug use behavior (compared to non-risky injection drug use behavior only). Individual sampling weights based on self-reported network data were not used for the analyses of the current study because regression is fairly immune to weights [
62].