We conducted the first population-based survey in Hong Kong involving obstetrics and gynaecology departments in seven hospitals, which collectively have about 34 000 deliveries per year. The study was conducted between 1 July 2005 and 30 April 2006. During this period, all women (≥18 years old) between 32 and 36 weeks of pregnancy were eligible for the study. The exclusion criteria were: women who were unable to be questioned without the presence of a family member or who were not competent to give informed consent.
We defined intimate partner violence as a pattern of coercive control. This included the use of physical violence, sexual assault and/or emotional abuse or the use of threats or coercive tactics against a woman by her current or former intimate partner who may be a current or former spouse, cohabiting partner, boyfriend, or dating partner.9
As previous local studies6,10
conducted in similar settings have revealed that intimate partner violence in Chinese couples is predominantly psychological in nature, ‘violence’ is therefore broadly defined in this study: it is not confined to violence in the sense of hitting, kicking or similar but also includes the more subtle form of abuse such as psychological abuse.
The Abuse Assessment Screen (AAS) was used to detect the prevalence of intimate partner violence. The AAS was developed by the Nursing Research Consortium on Violence and Abuse to determine abuse status and perpetrator within a defined period.11
In the present study, for the Chinese participants, we used a Chinese version of the AAS. The Chinese AAS addresses both emotional and physical violence separately for all time periods (lifetime, the preceding 12 months and during pregnancy). It differs from the original English AAS that treats emotional and physical violence simultaneously for the lifetime period while focusing on physical violence for the other time periods. Examples of emotional/psychological abuse commonly cited by Chinese women in previous studies6,10
were also provided to ensure that participants understood what was meant by ‘emotionally hurt’. The Chinese AAS has been validated and has demonstrated satisfactory measurement accuracy and utility for identifying intimate partner violence.8
Women who answered ‘yes’ to any of the questions on physical, emotional/psychological or sexual abuse by an intimate partner in the past year, or since becoming pregnant, were considered abused.
Mental health in this study is defined in accordance with the conceptualisation described in the World Health Report 2001.12
Mental health in this sense includes subjective wellbeing as well as perceived self-efficacy and is broader than a lack of mental disorders. Mental health and mental functioning are fundamentally interconnected with physical and social functioning. In this study, we measured mental health using the Edinburgh Postnatal Depression Scale (EPDS)13
to elicit subjective depressive symptoms and the SF-12 Health Survey (SF-12)14
to assess mental, physical and social functioning.
The Chinese version of the EPDS13
was used to detect postnatal depression among the Chinese participants. As recommended, the cutoff score of ≥10 was used for screening depressive illness in the Chinese postnatal population.13
In addition, the participants’ responses to question 10 of the EPDS were also examined (‘The thought of harming myself has occurred to me: (1) yes, quite often, (2) sometimes, (3) hardly ever, and (4) never.’)
The standard SF-12,15
an abbreviated form of the medical outcomes study (MOS) 36-item Short Form Health Survey (SF 36) was used to assess health-related quality of life. The standard SF-12 consists of 12 items grouped under the physical health summary and mental health summary (MCS) scales. The higher the scale score the better the corresponding quality of life. In this study, for the Chinese participants, we used the Chinese (HK)-specific SF-12, which has been validated and found to have satisfactory psychometric properties compared with the standard SF-12.14
We used a demographic questionnaire to collect socio-demographic information from the participants based on the known risk factors documented in the literature. The information included nationality, age, education, marital status, number of children, planned pregnancy, employment, family income, indebtness, financial assistance, social support, consumption of alcohol, chronic illness in the family and in-law conflict.
Pregnancy outcome was assessed by chart review and included the information about preterm delivery, mode of delivery, birthweight, Apgar scores and admission to neonatal intensive care unit (ICU).
Each of the seven experienced research nurses received a 2-day training workshop (conducted by the first author). The workshop dealt with ensuring privacy and safety of the participant during data collection, enhancing consistency when using the measuring instruments in face-to-face and telephone interviews, and promoting maternal and infant health by educating abused women about referrals and community resources.
The study was approved by the institutional review board of each of the participating hospitals. During the study period, women attending antenatal clinics in the seven participating hospitals and who met the inclusion criteria were invited to participate in a private area without the presence of their partners. Sampling was consecutive. It was made clear to the women that participation was entirely voluntary and informed consent was ensured. Those who agreed to participate signed a consent form. At their entry to the study, the AAS and the demographic questionnaire were administered face-to-face by one of the research nurses. Women assessed to be ‘abused’ by the AAS were counselled by the research nurse. Each was provided with a wallet-sized card detailing hotlines and community support for abused women if it was safe for her to take home such information. If not, she was provided with the telephone number of the research nurse, disguised as a nurse from the antenatal clinic, which she could call at any time. The research nurse also asked each of the abused women if she wished to call the police and/or to be referred to a medical social worker as per hospital standard policy.
Before the end of the initial interview, the research nurse obtained from the woman some means of contacting her for the postdelivery interview (home and/or mobile telephone number) and established a code for ensuring her safety when calling her in case her partner was nearby.
At 1 week postdelivery, a research nurse contacted the woman by phone. If it was safe for the woman (e.g. her partner was not nearby), the AAS would be administered again. The repeat administration was conducted because some women may not have disclosed their abusive experience when asked the first time but may do so subsequently. The SF-12 and EPDS were administered over the telephone. Before leaving the woman, our research nurse would check the safety of the woman and her baby, and whether she wished to be referred to any resources designed for abused women in the community.
Sample size calculation was based on the estimation of the prevalence of intimate partner violence. For a maximum tolerable error of 1% and an anticipated prevalence of 10%, a total of 3548 subjects were required using a 95% confidence interval.
Prevalence of abuse by an intimate partner in the past year was reported together with its exact 95% confidence interval obtained from a binomial distribution. To examine the effects of demographics, socio-economic status, history of chronic diseases and in-law conflicts on abuse in the past year, the potential factors were first examined individually by conditional logistic regression to account for the extra covariance among subjects within a hospital. Then, stepwise conditional logistic regression analysis was performed to determine independent effects of the factors. Pseudo R2
of the final model was calculated as 1 − log L
), where log L
) is the log-likelihood value of the final model and log L
) is the log-likelihood value of the initial model with the constant only.16
The impact of abuse by an intimate partner on postnatal depression (measured by EPDS) was examined using conditional logistic regression with adjustment for demographics, socio-economic status, history of chronic diseases and in-law conflict. Additionally, the impact on quality of life (measured by SF-12) was examined using a mixed effects model with a random effect for each hospital and the same adjustment for potential confounding effects.
Additional sensitivity analyses were performed to examine the effect of missing values. First, missing values of a factor were replaced by the predicted values from a regression on other observed factors. Second, missing values of a factor were simply replaced by the mean of the observed values of the factor.
The level of significance was taken as 5% and the Statistical Analysis System, version 9 was used for the statistical analysis.