Study design and setting
We conducted this cross-sectional study in 2017 in the capital city and 11 provinces. A feasibility assessment was conducted before selecting the 12 sites, which consisted of 21 operational districts with a high burden of HIV and a large population of people who use drugs. People who use drugs were defined as people who have used any illicit drugs, as stated in the Cambodian Law on Control of Drugs in the past 12 months .
To be eligible for the survey, an individual must: (1) be at least 18 years old, (2) have a predetermined study coupon, (3) never participate in this survey earlier, (4) meet the definition of people who use drugs, and (5) be able and willing to provide written informed consent to participate in the survey. Details of the main survey have been published elsewhere [37–39].
Sample size calculation
In this survey, we calculated the sample size separately for people who use non-injecting drugs and people who inject drugs. Assumptions were made, including a 20% drop in the prevalence of HIV between 2012 and 2017 and a population size of 13000 people who use non-injecting drugs and 1300 people who inject drugs . Based on the most recent national survey, the prevalence of HIV was 4% among people use non-injecting drugs, and 25% among people who inject drugs . We used a 1.5% margin of error, a 95% confidence interval [CI], a 90% response rate, and a 1.5% design effect. We used 12 selected sites as strata and sampled 15% of people who use non-injecting drugs and 22% of people who inject drugs from their respective estimated population in each site acquired from the National Authority for Combatting Drugs. Figure 1 shows the flow chart of the sample selection. The survey included 1677 participants. We finally included 1598 participants in the multiple logistic regression analysis after dropping 79 observations with missing data on variables included in the model.
We adapted the Respondent-Driven Sampling (RDS) method to recruit the study participants due to the hard-to-reach nature of people who use drugs. The sampling procedure included five steps. Initially, we sought support from local NGOs in the selected 21 data collection locations to obtain four seeds with a good connection with people who use drugs in each location. A personal identification number was then assigned to each seed after receiving written informed consent from the participant. Next, we provided three coupons to each seed for referring three other people who use drugs to the study. Seeds received US$2 for a successful referral and were expected to refer three to six peers. New seeds would be selected when the recruitment tree had dried up. Finally, participants recruited for the study were invited to become seeds allowing them to recruit other people who use drugs from their networks.
Data collection training
The interviews were conducted by formally trained counselors from HIV voluntary counseling and testing centers in the respective province. All members of the data collection teams attended a three-day training on data collection covering topics such as informed consent process, study protocol, HIV and HCV testing, data collection tools, interview techniques, administration of the questionnaire, participants’ privacy and confidentiality protection, as well as data quality checks, including spot-checking and reviewing the filled questionnaires.
Variables and measurements
A structured questionnaire was developed based on standardized and validated tools adapted from previous studies on mental health among HIV key populations, including people who use drugs [21,40,42,43]. We also conducted a questionnaire validation workshop participated by representatives from communities, NGOs, development partners, and national programs working on HIV and harm reduction programs in Cambodia. The questionnaire was piloted with 20 people who use drugs residing in Phnom Penh. The questionnaire collected information on socio-demographic characteristics, drug use behaviors, sexual behaviors, HIV and other sexually transmitted infections (STIs), other substance use, adverse childhood experiences, and psychological distress.
Sociodemographic information included age (18-24, 24-35, 35+ years), gender (male, female), type of community (urban, rural), years of formal schooling attained (0-6, 7-9, 10+ years), average monthly income earned in the past six months (<100, 100-199, 200+ USD), primary occupation (entertainment workers, office workers, laborer/farmer, unemployed, other), and living arrangement (living with family/relatives, living with friends, living in own dwelling, living on the streets, other).
The information on drug use included types of illicit drugs most commonly used and frequency of use in the past three months. We also collected information regarding other substance use (i.e., alcohol drinking and binge drinking, cigarette smoking) and exposure to community-based HIV, harm reduction, and other related services in the past six months.
To measure HIV risks, we asked participants about their sexual behaviors in the past three months. The information included the number of sexual partners and condom use with commercial (defined as partners with whom the participant had sex in exchange for money or gifts) and non-commercial partners. We also collected information on HIV testing history, STI symptoms, and care-seeking behaviors for the symptoms in the past three months.
We adapted five questions on ACEs from the brief screening version of the Childhood Traumatic Questionnaire . The questions collected information on the experiences of physical abuse, emotional abuse, sexual abuse, physical neglect, and emotional neglect. All the five questions were close-ended, where respondents chose between zero (No) and one (Yes) to describe their ACEs.
The General Health Questionnaire (GHQ-12) was used to assess psychological distress . A four-point Likert-type scale, which varied from “0= less than usual” to “3= much more than usual,” was applied to each question . A dichotomous variable of “1= high psychological distress” and “0= low psychological distress” was developed based on the GHQ-12 guide. Therein, the four-point Likert-like scale was re-coded using a “0-0-1-1” method to eliminate bias . The median of the sum of the newly derived score for the sample was used as a cut-off to identify participants with high (GHQ-12> 2) and low (GHQ-12≤ 2) psychological distress . Cronbach's alpha of the scale in this study was 0.88, confirming good reliability . Besides, we added questions on participants’ suicidal thoughts and overall quality of life to the questionnaire.
Stata (StataCorp LP, version 14.2) was used for data analyses in this study. We compared sociodemographic characteristics, substance use, sexual behaviors, and ACEs of participants with low psychological distress (GHQ-12≤2) to those of participants with high psychological distress (GHQ-12>2). We used Chi-square test (or Fisher's exact test when the expected cell count was smaller than 5) for categorical variables and Student's t-test (or Mann-Whitney test when a variable was not normally distributed) for continuous variables. A multiple logistic regression model was constructed to identify factors associated with psychological distress. We included age and gender regardless their statistical significance level and other variables with a p-value <0.05 in the bivariate analyses in the model. The variables included marital status, education level, living arrangement, duration of drug use, experience at drug rehabilitation, mode of first-time drug use, HIV status, and ACEs. We obtained crude and adjusted odds ratios (AOR) of the associations and presented with 95% CIs and p-values.
This study received ethical approval from the National Ethics Committee for Health Research (NECHR) of the Ministry of Health in Cambodia (No. 420 NECHR). Participation in the study was voluntary, and all participants provided written informed consent. We protected participants’ privacy and confidentiality by collecting data in a private room and removing personal identifiers from research documents.