Measures
Based on findings from earlier studies on factors associated with IPV [
4,
19‐
24,
27,
28], the following measures on social and psychological characteristics, relationship and violence exposure were included in the study:
Demographic characteristics: age (years), place of birth (i.e. if born in Sweden, yes/no), education (Up to high school/University) and occupational status (Part- or fulltime job/Student/Retired/Sick-leave/Unemployed/Long holiday/Other).
Social life and financial situation: women’s ability to adapt to and be satisfied with her social role was measured with the Social Adjustment Scale-Self-Report [
32] (SAS-SR). SAS-SR includes 54 items addressing specific ways of behaving within a broad range of everyday social contexts, such as interaction with colleagues, neighbors and family. This instrument investigates six social role areas; work role, social and leisure, extended family, primary relationship, parental and family unit. Each item is rated and a mean score is then calculated where higher scores indicate greater impairment. A validation of the total scale has shown good internal consistency (Cronbach’s alpha = 0.74) [
33] and good psychometric quality [
32]. The social and leisure scale which measures an individual’s relationship with friends and social activities was considered to add most complementary information in relation to other measures added in this study and thus included in the present study. Also, a single item concerning economy was included regarding financial needs.
Psychological distress: the presence of psychiatric symptoms and mental health problems was measured with the Symptom Checklist-90-R [
34], a 90-item self-report questionnaire designed as a symptom inventory aimed to reflect an individual’s psychological symptom patterns. Each item is rated on a five-point scale of distress (0–4) ranging from “not at all” to “extremely”. The total mean score, Global Severity Index (GSI) which indicates an individual’s level of psychological distress was used in this study. The GSI is the recommended global measure of the Symptom Checklist [
34] and it has shown very good internal consistency (Cronbach’s alpha = 0.97) [
35]. A Swedish validation study of the instrument suggested a GSI score of 1.21 corresponding to a clinical population and a score of 0.49 to the general population [
35].
Psychosocial functioning: the Outcome Questionnaire [
36] (OQ-30.2) was used, a 30 item self-report questionnaire covering personal and social characteristics that affect an individual’s quality of life. The OQ-30.2 measures psychosocial functioning through three factors; symptom distress, interpersonal relations and social role. A total score measure has been recommended due to high correlation between its subscales [
36]. Each item is scored on a five-point likert scale (0–4) and a total score is calculated summing the ratings across all 30 items where a high score indicates more difficulties (Cronbach’s alpha = 0.93). A cut off score of 44 has been recommended to differentiate between normal and reduced psychosocial functioning [
36].
Alcohol use: alcohol consumption was measured with the Alcohol Use Disorders Identification Test (AUDIT), a 10 item self-report questionnaire developed to screen for excessive drinking [
37]. Each item response is scored 0–4 giving a total score between 0 and 40. For women, a score of 6 or above has been recommended as an indication of harmful alcohol use. This test has shown good results in differentiating between normal and problematic alcohol consumption and has been validated in a Swedish context showing very good internal consistency (Cronbach’s alpha = 0.95) [
38].
Relationship to the perpetrator: single questions about respondents’ current relationship with the perpetrator were included and concerned if the woman remained in the relationship with the perpetrator, if she had children together with him and if she maintained any contact with him (yes/no). Also a question on whether she had a new partner was answered (yes/no).
Violence: incidence of violence was measured with the Revised Conflict Tactics Scale (CTS2) [
39], a 78-item self-report questionnaire that comprises five scales; negotiation, psychological aggression, physical assault, injury and sexual coercion. Since our focus was on women’s exposure to violence the subscales concerning physical, sexual and psychological violence were used in this study. The CTS2 has been validated in different contexts and is commonly used in research to measure IPV [
39,
40]. The scales have shown good internal consistency ; with a Cronbach’s alpha of 0.79 to 0.86 for the different scales [
39]. For each item respondents are asked to indicate the number of times the violent action has occurred during the past year (never to more than 20 times, or sometime earlier in life). Items can be organized into minor and severe acts [
39]. Minor physical violence comprises throwing something, punching, slapping, grabbing or shoving. Severe physical violence includes acts like kicking, biting, hitting with a fist, threatening with a knife or fire gun. Minor psychological violence includes e.g. insults or shouting whereas severe psychological violence refers to threats of violence or destroying partner’s belongings. Minor sexual violence concerns for example insisting on having sex and severe sexual violence includes using threats or force to have sex [
39].
The survey included questions about violence exposure (psychological/physical by adults and between parents) during childhood, and violence exposure (psychological, physical and sexual) in previous relationships. The answers were coded dichotomously (yes/no). In addition, women in the non help-seeking group were asked about the last incidence of IPV exposure for each type. Due to contradictory answers between these single questions and the CTS2 on when the violence occurred, violence could not be organized into “happened the past year” or “earlier in life”. Consequently, information about the chronicity according to the CTS2, i.e. frequency of violence during the previous year, could not be calculated.
Missing data
Due to the circumstances described above, the number of missing cases was not known and thus there was no possibility to calculate an inclusion rate. The method used to recruit non help-seekers and the lack of information about help-seeking women who declined participation impeded control of missing cases.
The measures were completed by most participants, but some items in the questionnaire were not answered. The internal non-response rates for the specific variables included in the study were 0-8%. Variables social and leisure, psychological distress, alcohol use, children together with perpetrator, physical violence by adults, physical violence by adults/between parents, any violence during childhood and any violence in previous relationships, had no missing values. One response (0.2%) was missing for variables psychosocial function and physical, psychological and sexual violence in previous relationships. Variables age, economy and psychological violence by adults during childhood lacked two responses (0.4%). Three responses (0.6%) were missing concerning occupation, four (0.8%) regarding in a relationship with perpetrator and six (1.2%) about education. Two variables differentiated from the rest with 38 (8%) and 36 (7.5%) missing responses; new partner and maintain contact with perpetrator respectively. In total, the amount of missing cases in the multivariate analyses did not exceed 10%.
Statistical analyses
We included the following independent variables in the comparison between the non help-seeking and the help-seeking group: demographics,
social and leisure,
psychological distress,
psychosocial functioning,
alcohol use,
violence during childhood,
violence in previous relationships, relationship status,
children together with the perpetrator and violence conducted by the perpetrator (Tables
1,
2 and
3).
Help-seeking (yes/no) was used as the dependent variable/outcome measure. A correlation matrix with binary statistics between all variables using Spearman’s Rank Order correlation was initially conducted to get an overview of the data before further analyses. In order to control for possible independent group differences, the chi-square test of independence and the Mann–Whitney U test for variables with nonparametric score distributions were used. Prior to the multivariate analysis, linearity among the continuous variables was checked for by grouping them into categories. This procedure was followed in order to avoid false assumptions of linear correlations between continuous independent variables and the outcome (help-seeking). All continuous variables in the model showed a linear association with help-seeking and were thus included in the analyses.
Table 1
Non help-seeking and help-seeking women’s psychosocial characteristics and relationship status (n = 475)
Age (years) | 40.1 (11.71) | 34.1 (9.90) | .000a
|
Foreign born | | | .000 |
Born in Sweden | 86% | 49% | |
Born outside Sweden | 14% | 51% | |
Education | | | .000 |
Upper secondary school | 47% | 72% | |
University | 53% | 28% | |
Occupation | | | .000 |
Part- or fulltime job | 49% | 32% | |
Student | 10% | 21% | |
Retired/sick-leave | 22% | 13% | |
Unemployed/long holiday | 12% | 29% | |
Other | 7% | 5% | |
Economy | | | .009 |
Enough money for needs | 50% | 32% | |
Usually enough money | 11% | 16% | |
Not enough, did not had to borrow | 15% | 17% | |
Not enough, had to borrow | 12% | 14% | |
Great financial difficulty | 12% | 21% | |
Social and leisure (SAS) | 2.2 (0.62) | 2.6 (0.69) | .000a
|
Psychological distress (GSI) | 1.1 (0.74) | 1.5 (0.76) | .000a
|
Psychosocial function (OQ) | 45.7 (20.95) | 55.7 (18.13) | .000a
|
Alcohol use/misuse (AUDIT) | 5.1 (5.60) | 3.0 (4.91) | .000a
|
Relationship status
|
In a relationship with perpetrator | 10% | 23% | .003 |
Children together with perpetrator | 29% | 64% | .000 |
Maintain contact with perpetrator | 39% | 39% | .983 |
New partner | 36% | 10% | .000 |
Table 2
Experiences of violence among non help-seekers and help-seekers (n = 475)
Violence during childhood
|
Psychological by adults | 52% | 52% | .908 |
Physical by adults | 48% | 46% | .722 |
Psychological between parents | 47% | 47% | .987 |
Physical between parents | 28% | 33% | .325 |
Any violence during childhood | 67% | 68% | .818 |
Violence in previous intimate relationships
|
Psychological | 53% | 46% | .181 |
Physical | 42% | 34% | .104 |
Sexual | 20% | 19% | .817 |
Any | 61% | 50% | .036 |
Violence in latest violent relationship
|
Minor psychological | 100% | 100% | .543 |
Severe psychological | 88% | 95% | .004 |
Minor physical | 95% | 97% | .004 |
Severe physical | 89% | 91% | .575 |
Minor sexual | 89% | 69% | .189 |
Severe sexual | 38% | 43% | .332 |
Any severe | 95% | 98% | .080 |
Last incidence of violence | | | .000 |
During the last year | 35% | 91% | |
Earlier in life | 65% | 9% | |
Table 3
Women’s characteristics by help-seeking versus non help-seeking (n = 428)
Age (years) | 0.94 (0.91–0.96) | .000 |
Education | | |
Up to high school | 1 | |
University | 0.38 (0.22–0.67) | .001 |
Psychological distress (GSI) | 2.83 (1.87–4.29) | .000 |
Alcohol consumption (AUDIT) | 0.94 (0.89–0.99) | .029 |
In a relationship with the perpetrator |
No | 1 | |
Yes | 3.92 (1.24–12.40) | .020 |
Children together with the perpetrator |
No | 1 | |
Yes | 5.46 (2.99–9.97) | .000 |
New partner | | |
No | 1 | |
Yes | 0.19 (0.09–0.37) | .000 |
Only statistically significant variables from the binary tests, along with significant correlation variables in the initial correlation matrix were included in the multivariate analysis. Due to multicollinearity between the variables
social and leisure,
psychosocial functioning and
psychological distress, only the latter variable (
psychological distress) was kept in the multivariate analyses as it was deemed of greatest value to report. Multivariate logistic regression models with a backward stepwise approach were conducted to investigate potential differences between the groups, thus results are presented as adjusted odds ratios (Adj. OR) with 95% confidence interval (CI). The goodness-of-fit of the multivariate models were tested with the Hosmer-Lemeshow test and Nagelkerke R square [
41]. A listwise deletion approach was used to deal with missing data, meaning that only complete cases were included in the multivariate analyses. Analyses of missing data did not show any significant differences between those who responded to all items and those who did not, thus this approach was chosen since it facilitates comparability as all calculations proceed from a common base [
42]. Given the exclusion of missing cases the final sample size in the multivariate logistic regression model included 428 cases i.e. 90% of the initial sample of 475 women and the proportion of included cases were similarly distributed by group (114 non help-seekers and 314 help-seekers). All statistical analyses were performed using the statistical program package SPSS, version 20.0.
Two variables could not be included in the multivariate analysis due their interaction with other independent variables in combination with few observations in the non help-seeking group. There were too few non help-seekers born outside Sweden or with IPV exposure during the last year. Only the group of women born in Sweden was possible to analyze separately in a multivariate logistic model. However, the variables foreign born and last incidence of violence were separately analyzed together with independent variables in our final model. Separate independent sample t-tests for help-seeking with split file by foreign born were conducted on the continuous independent variables age, psychological distress and alcohol use. Analyses with discrete variables; education, if still in a relationship with the perpetrator, if children together with the perpetrator and if she has a new partner, were conducted using cross-tabulations and Chi2 with split file by foreign born. The same procedure was carried out when examining the variable last incidence of violence coded as “during the last year” or “earlier in life” i.e. cross-tabulations with Chi2 and independent sample t-tests with split file by last incidence of violence.