Methods of review
The present protocol has been written in accordance with the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) guidelines and is presented in accordance with the PRISMA-P checklist (Additional file 1). The protocol has further been registered in PROSPERO (#CRD42018116026).
The systematic review will be conducted as an individual and dual process by two researchers (SD + DBH) in regard to screening, eligibility and inclusion. Screening will be done on a title basis, followed by an abstract and full-text basis. A third researcher (RK) from the team will be consulted to resolve issues regarding disagreement of eligibility and inclusion on a full-text basis. If sufficient data can be extracted, a random-effects meta-analysis will be conducted on the extracted data.
It is expected that not all studies report the recorded data on psychological violence, but rather cluster subtypes of IPV (i.e. physical, sexual and psychological violence). In such cases, the corresponding authors will be contacted and invited to share the raw data.
Key definitions of the systematic review
Intimate partner violence
As proposed in a statistical definition by EIGE, IPV is defined as “any act of physical, sexual, psychological or economic violence that occurs between former or current spouses or partners, whether or not the perpetrator shares or has shared the same residence with the victim” [19] (p. 44). This definition has been proposed to aid the EU member states to collect and compare administrative data on violence against women in a standardized manner [19].
Psychological violence
A broad terminology for psychological violence is used in both scientific literature and in clinical practice (e.g. psychological violence, emotional abuse, coercion, psychological aggression). For clarity, this review will use the term psychological violence seeing that this definition links it directly to other types of IPV (e.g. physical or sexual violence), while emphasizing a core psychological aspect of harm in both the act of perpetration as well as the effect of victimization. The variation in terminology will shortly be addressed in the discussion.
In the current protocol, psychological violence will be defined by an overall definition combining that of both the WHO [1] (i.e. both emotional/psychological abuse and controlling behaviours) and the European Institute of Gender Equality [19]: “Any act or behaviour which causes psychological harm to the partner or former partner. Psychological violence can take the form of, among others, coercion, defamation, a verbal insult or harassment” [19] (p. 45), including belittling, constant humiliation, intimidation (e.g. destroying things), threats of harm, threats to take away children and/or isolating a person from family and friends; monitoring their movements; and restricting access to financial resources, employment, education or medical care [1] (p. 1). A broad definition will allow for studies using varying definitions to be included and will be assessed through subgroup analyses. The different definitions will be evaluated and discussed based on their effect on mental health.
Mental health in this context
As mentioned above, it has previously been argued that psychological violence alone cannot be characterized as a trauma [11]. However, recent studies have identified an association between psychological violence and PTSD [13]. Based on these findings, the authors consider psychological violence a potential traumatic event and wish to further examine the relationship between psychological violence and PTSD. Other mental health consequences identified are based on the National Institute for Health and Care Excellence (NICE) guidelines and comorbid mental health problems of PTSD [24], i.e. depression, anxiety, alcohol or drug abuse, suicidality, sexual problems, sleep problems, problems with concentration, somatization and functional problems (e.g. social, educational, or occupational) as well as feelings of shame and guilt.
Search method
A dual search will be conducted in the electronic databases PsycINFO, PubMed, EMBASE and Web of Science. Other methods used for identifying relevant research include reference checking and hand-searching of grey literature. Furthermore, the following scientific journals will be hand-searched: Journal of Interpersonal Violence and Journal of Violence and Victims.
Criteria for including studies
The review will include studies of psychological violence on mental health when controlling for other types of partner abuse. Hence, studies including an adult (≥ 18 years) population of victims of IPV (dating samples, national samples, clinical settings, etc.) that report on psychological violence specifically. Many studies are expected to include groups of comparison (e.g. non-abused or other types of abuse); however, comparisons are not required.
Furthermore, the review will only include peer-reviewed articles in English, German, Dutch or Scandinavian languages (i.e. Norwegian, Swedish or Danish).
Criteria for excluding studies
The review will exclude case studies, reviews, commentaries, editorials, letters to editorials, book chapters and other non-primary research articles.
Search string
See Table 2 for the PsycINFO search string.
Main outcome of interest
To investigate the independent effect of psychological violence on mental health. A meta-analysis will be performed to estimate the effect of psychological violence on PTSD, depression and anxiety. A narrative summary will present all related mental health problems as defined by the NICE guidelines [24].
Secondary outcomes of interest include the following:
-
How does “type” of psychological violence affect mental health?
-
How does frequency and severity affect mental health outcomes? (e.g. high frequency and/or low severity or low frequency and/or high severity).
-
Are there potential gender differences in mental health consequences with regard to psychological violence?
-
Will controlling for previous trauma affect the association between psychological violence and mental health?
-
Does sample population differ in mental health consequences? (e.g. dating samples vs. treatment samples)
Finally, the systematic review aims to evaluate included psychometric tools measuring psychological violence and how potential variations in the conceptualization of psychological violence affects results.
Data extraction
Data will be extracted with help from Endnote and Covidence. Meta-analyses will be conducted using the programming language R. The authors will design a data extraction form. The form will include authors, year, sample size, population, country, age, gender, design, IPV assessment tool(s), mental health assessment, primary outcome (effect size), secondary outcome(s), timeframe of assessment (lifetime or specified), scoring method(s), previous trauma and previous mental health problems. Data will be extracted by one reviewer (SD), and independently cross-checked by another reviewer (DBH). Inconsistencies in data extraction will be resolved between the reviewers by referring to the source study until a consensus is reached.
Quality assessment
The “Quality Assessment Tool for Quantitative Studies” developed by the Effective Public Health Practice Project [25] will be used to assess the quality of the included studies. This is in line with previous research evaluating IPV psychometric tools [26]. Assessment will be based on six components: (1) selection bias, (2) study design, (3) confounders, (4) blinding, (5) data collection methods and (6) withdrawals and dropouts [24, 25]. Two researchers (SD and DBH) will classify studies on three levels: weak, moderate and strong. If classifications are inconsistent a third researcher (RK) will be involved and classification will be discussed until consensus is reached. If possible, moderation analyses will compare studies of strong vs. weak quality.
Plan for data synthesis
As demonstrated above, studies on psychological violence are quite heterogeneous in regard to conceptualization, psychometrics, sampling, design, scoring, and so forth. Therefore, we will perform a random-effects meta-analysis, because we expect high heterogeneity in the included studies. The random-effects meta-analysis assumes variance in effect across studies due to real differences in effect as well as by chance. The meta-analysis will help estimate the common effect of psychological violence on mental health (i.e. PTSD, depression and anxiety) by synthesizing individual results. If possible, moderation analyses will compare studies according to quality assessment and varying samples. The I2 statistic will be used to test for heterogeneity, and as suggested, an I2 statistic above 75% implies considerable heterogeneity, while an I2 statistic below 40% is not considered to be a concern [27].
The included studies are expected to report effect sizes of varying types (i.e. correlation, regression, mean differences and association of categorical variables, e.g. odds ratio). For data synthesis, the reported effect sizes will be recoded into the same type of effect size using the programming language R. For this meta-analysis, we expect to perform a stepwise analysis according to the outcome of interest. For the main outcome of interest, as well as subgroup analyses, the meta-analysis will be conducted with effect sizes based on correlations. Additionally, the impact of the duration and frequency of psychological violence will be estimated by using meta-regression. The Metafor-package for the programming language R will be applied to conduct the meta-analysis [28].