Sample and sampling
The study was based on a Women’s Health Survey conducted in Estonia between March and August 2014. The aim of the survey was to collect information about various sexual and reproductive health indicators of women aged 16–44 which are not routinely collected (e.g. the use of contraceptive methods, reproductive plans, the prevalence of infertility and violence, sexual health care services etc.). A random sample of the female population, stratified by age groups (16–17, 18–24, 25–34, and 35–44), was taken from the Population Registry. For sample size calculations, data from previous studies [23, 24] regarding the percentage of sexually active respondents in each age group and total response rates were taken into account in order to get a sufficiently large sample. The initial sample size was 5233 women, and by age group the numbers were 2112 (16–17), 1144 (18–24), 993 (25–34), and 984 (35–44). For this calculation, the OpenEpi software package  was used.
Of the initial sample, 40 women were not able to reply. Out of 5193 eligible respondents, 2708 did not return the questionnaire, and 12 refused to answer. In addition, 24 questionnaires were returned with very few answered questions, six had conflicting answers and three respondents were over 45 years old. The final response rate was 47.0%. Of the respondents, 16.6% answered electronically and 83.4% on paper. The survey methods are described in more detail in the study report, where the full questionnaire in English is also included .
For this paper, women aged 16–17 years and women who had not answered to the question about age were left out of the analysis (n=727). Respondents aged 16–17 years were not included due to the nature of the questions used to determine risky sexual behaviour, as only 50.6% in that age group had experienced sexual intercourse. Of the respondents, 37 women did not answer the questions about SV and six were left out of the analysis because their native language was neither Estonian nor Russian. The final sample size was 1670.
The questionnaire contained questions about socio-demographic background, intimate relationships and sexuality, reproductive history, satisfaction with health care services, contraception, any future plans of having children, overall health, and exposure to violence.
Sexual violence assessment instrument
To assess exposure to SV we used NorVold Abuse Questionnaire (NorAQ), which has good reliability and validity . In the analysis for this paper we used four questions from the NorAQ:
1) “SV with no genital contact”: Has anybody against your will touched parts of your body other than the genitals in a ‘sexual way’ or forced you to touch other parts of his or her body in a ‘sexual way’?
2) “Sexual humiliation”: Have you in any way been sexually humiliated; e.g., by being forced to watch a porno movie or similar against your will, forced to participate in a porno movie or similar, forced to show your body naked or forced to watch when somebody else showed his/her body naked?
3) “SV with genital contact”: Has anybody against your will touched your genitals, used your body to satisfy him/herself sexually or forced you to touch anybody else’s genitals?
4) “Rape/attempted rape”: Has anybody against your will put his penis into your vagina, mouth or rectum or tried to put an object or other part of the body into your vagina, mouth or rectum?
For each question the respondent could choose between the answers: no; yes, as a child (less than 18 years old); yes, as an adult (18 years old or older); yes, as both a child and an adult. We created an aggregated measure where all women who gave an affirmative answer to at least one of the questions about SV were considered to have been exposed to SV.
Self-reported socio-demographic factors were used as categorical variables with following categories: three groups based on age (18–24; 25–34; 35–44 years); two groups based on the language spoken (Estonian; Russian); three groups based on the level of education (basic or less; secondary/vocational secondary; vocational higher/bachelor’s degree/master’s or doctoral degree); three groups based on marital status (married/cohabiting; separated/divorced/widowed; single); four groups based on employment status (employed; pupil/student/postgraduate student; on pregnancy or parental leave; retired/disability pension); and three groups based on the reported level of difficulties in paying bills (always/often; sometimes; rarely/never).
Risky health and sexual behaviour variables
Due to the nature of binary multiple logistic regression analysis all outcome variables were dichotomised.
To determine risky health behaviour, smoking (never a smoker; past/current smoker), drinking enough alcohol to lose control of oneself (daily/weekly/monthly/less often than monthly; never) during the last 12 months, and ever using illicit drugs (no; yes) were considered in analysis.
We used four questions to evaluate risky sexual behaviour. To determine contraception use during the most recent intercourse, the respondents could choose a contraceptive method from a list or they could answer that they did not use any method. When analysing this question, we left out women who did not need contraception (were pregnant, wanted to get pregnant, or were breastfeeding a less than 6-month-old baby; n=169). Participants were asked “Have you been asked to have sex in exchange for money or other material reward?”, and respondents were assigned into two groups: the first group consisted of women who had not been asked to do so and also of those who had been asked, but refused and to the second group women who had accepted the offer. To assess whether the respondent had concurrent sexual relationships, the following question was asked “Have you had concurrent sexual relationships during your present marriage/cohabitation”, and responses were categorized into two groups (yes; no). Women who were not currently in a relationship were left out of the analysis (n=191). We asked if the respondent had ever been diagnosed with at least one of the following STIs: chlamydia; gonorrhoea; trichomoniasis; HIV/AIDS; syphilis. The respondents were categorized into two groups (yes; no). Respondents who did not know whether they had ever been diagnosed with an STI or had not been tested, were not included into analyses that involved this variable (n=116).
Differences between women exposed to SV and not exposed to SV were analysed using a chi-square test with a significance level of p<0.05. Selected socio-demographic potential confounders (age, education, marital status, occupation, difficulties with paying bills) were entered into multivariable binary logistic regression analysis models to explore associations between SV exposure and risky health and sexual behaviours. Associations are presented as odds ratios (OR) and adjusted odds ratios (AOR) with 95% confidence intervals (CI 95%).