Design and selection of the IDU sample
A cross-sectional survey was carried out as part of UNICEF’s research on the biological and behaviours survey among injection drug users in Sarajevo, Banja Luka and Zenica in Bosnia and Herzegovina during May – July 2007 [15].
Respondent Driven Sampling (RDS) was used to obtain a sample of hidden populations and to reach populations such as IDUs [16]. It is a form of chain-referral sampling. These chains consist of waves that penetrate deeper into the population in order to reach its hidden parts. RDS begins with a set of participants (“seeds”) who recruit their peers, people whom they know and who know them. The method is coupon-based recruitment process, accounts for homophily and non-random selection of initial respondents. In all three cities, six initial respondents (“seeds”), three female and three male, were selected in cooperation with NGOs. Respondents presented two sub-sample, one of 18–24 years and sub-sample of older than 25. Those “seeds” were the first “wave” of participants, who recruited the second wave. The number of waves was 7, in order to reduce selection bias (possibly caused of some “seeds” who were contacted by NGOs’ beneficiaries) and to reach deeper into the IDUs population and provide generalizability of the results [17]. The RDS method involves financial compensation, primarily for personal participation in the survey, but also for the recruitment of new participants. Respondents were allowed to recruit a maximum of three new participants. The primary incentive was 10 EURO (E), while the secondary incentive was 5E per recruited IDU.
Before the inclusion of respondents in survey, screener who was responsible for checking the legitimacy of participation explained the details of the research, aim of the research, procedures and the content of the questions. Anonymity, confidentiality and privacy of data were explained and guaranteed. Participants gave their written informed consent. A total of 750 IDUs participated in the survey and fulfilled the criteria. Participants had to be older than 18 years, injected drugs in the previous month, and living, studying or working in the cities where the survey was conducted for at least three months. The sample of 519 IDUs answered questions related to their oral health. All participants used heroin. There were 483 (93.1%) males and 36 (6.9%) females, 213 (41.0%) were from Sarajevo, 257 (49.5%) were from Zenica and 49 (9.4%) were from Banja Luka. The survey was approved by the Ethical Committees of the Clinical Centre of the University of Sarajevo, Cantonal Hospital in Zenica and Agency for Drugs of the Republic of Srpska (Banja Luka).
Procedure
The data were obtained through face-to-face interviews using a structured questionnaire (Additional file 1: Questionnaire). There were 16 sections containing questions related to socio-demographic characteristics and HIV risk–taking behaviour among IDUs. The last section contained questions related to oral health: the number of missing teeth, problems with xerostomia, symptoms of common oral health problems, frequency of tooth brushing, frequency of visits to the dentist, the use of private or public dental services, the assessment of accessibility of dental services and self-report oral health status. A pre-test questionnaire was conducted in Banja Luka on a group of 10 IDUs, in order to determine appropriate form of the questions, their order and length of the interview. The interview took approximately 40 minutes plus time for a short introduction and counselling about blood test procedure, before moving to the next biological stage of the research.
The NGO facilities were used to interview the IDUs which were open each weekday for 6 hours and Saturdays for 4 hours. Working hours were adjusted to suit the participants and interviews were mainly conducted in the afternoon. At each research facility, trained teams were available, whose members were responsible for a particular phase of the research (e.g. practical skills and knowledge on interviewing techniques and confidentiality). NGOs which took part in this survey, mostly dealt with education, networking, exchange of information, publishing, work on improving the position of marginalized groups, specially in field of drugs, and protection of human rights and freedoms. One of NGOs provided prevention activities and psychotherapeutic services for drug users and their families.
Variables
A total of 10 variables were analyzed as predictors of oral health. These variables were: age, gender, level of education, marital status, employment status, housing, duration of drug use, frequency of drug injection in the last month, frequency of tooth brushing and regular dental checkups. Housing was classified into two categories, where the code “1” represented better housing (living in own or parents’ apartment/house), and code “0” (rented house, friends’/relatives’ house, collective residence, street/park/abandoned building, institutions for the treatment of addiction, prisons and others). Oral health of the participants was coded as: bad or good oral health. Oral health variable was composed of questions about mouth and teeth condition; missing teeth; difficulty with chewing and swallowing food due to problems with teeth and dry mouth; and frequent feeling of dryness in mouth. Score values of the answers at abovementioned first three questions were recorded on the scale from 0 (the worst value) to 4 (the best value). The last question respondent could answer with “yes” or “no” (“no” included no, I do not know and missed answer). The total score was the sum of the individual scores and ranged from 0 to 13. Applying the cut-off value (median was 7) on the total score the participants were divided into two groups, with good (values 8 and more) or bad oral health (7 and less). A variable regular checkup at the dentist was obtained by dichotomization of answers to question “In the past year, how many times have you visited a dentist?” Participants who had answered that they visited dentist one or more times during past year, were coded as participants who regularly visited a dentist. A variable teeth brushing was obtained by dichotomization of answers to question “How often do you brush your teeth?”, while regular teeth brushing implied twice a day or after every meal.
Statistical analysis
Data were expressed as frequencies (percentages) for categorical variables and as means ± SD or medians and range for continuous variables, respectively. Univariate analyses were conducted to assess the association between each independent variable and the outcome variable, bad or good oral health. All variables which were significantly associated with the outcome measure (p < 0.05) were entered into a multiple logistic regression model. The IBM SPSS Statistics 19 package was used for these analyses.