Experiences of Domestic Violence and Mental DisordersA Systematic Review and Meta-Analysis

Picture a clinician sitting across from a patient who has just been diagnosed with post-traumatic stress disorder. The checklist covers symptoms, medications, and sleep disturbance. It does not include one question — whether someone at home is hurting this person. That missing question turns out to matter enormously. Women with PTSD are more than seven times as likely to be experiencing partner violence as women without a mental disorder. That single number comes from a systematic review and meta-analysis by Trevillion, Oram, Feder, and Howard — and it is not the most alarming finding in the paper. Domestic violence and mental disorder are linked in ways that clinical practice has been slow to recognize. Trevillion and colleagues frame their review against a stark backdrop: lifetime estimates of physical or sexual partner violence among women range from 15 to 71 percent globally, with past-year estimates from 4 to 54 percent. No comparable global estimates exist for men. Research on same-sex relationships is limited, though some United States evidence suggests similar prevalence across same-sex and heterosexual relationships. The economic costs alone — approximately one point seventy-three billion pounds per year in the UK and four point one billion dollars per year in the USA in direct medical and mental healthcare — point to a problem that health systems are absorbing without fully acknowledging. The question Trevillion and colleagues set out to answer is specific: how often do people presenting with particular mental disorders also turn out to be victims of domestic violence, and how do those odds compare with people who have no mental disorder? To answer that, they built the kind of evidence base that individual studies cannot provide on their own. Their systematic review followed the MOOSE and PRISMA reporting guidelines, registered a protocol with PROSPERO, and searched eighteen biomedical and social science databases from inception through March 2011 — with no language restrictions. They hand-searched three journals, tracked forward citations, and sought expert recommendations. The crucial methodological decision was this: they only included studies that used validated diagnostic instruments for mental disorder, not unvalidated screening tools. That distinction matters because a proper diagnosis and a high screening score are not the same thing, and mixing them would muddy the signal. Two independent reviewers screened all studies, extracted data, and appraised quality, reaching an inter-rater reliability of zero point ninety-five — essentially near-perfect agreement. Forty-one studies made the final cut. For the statistical pooling, they used DerSimonian-Laird random-effects models — a method suited to situations where studies are not all measuring the same underlying effect. The studies in this review varied in how they defined domestic violence, which populations they sampled, and whether they measured lifetime or past-year exposure. A random-effects model accounts for that variation rather than pretending it away. Pooled estimates were calculated separately by sex, by diagnosis, and by timeframe, but only when at least three high-quality studies were available. That constraint explains why several findings had to be reported narratively rather than as pooled numbers. Now for the findings themselves. For women with depressive disorders, the pooled odds ratio for lifetime intimate partner violence was two point seventy-seven, with a ninety-five percent confidence interval from one point ninety-six to three point ninety-two. For anxiety disorders, it was four point zero eight, with a confidence interval of two point thirty-nine to six point ninety-seven. For PTSD, it was seven point thirty-four, with a confidence interval from four point fifty to eleven point ninety-eight. Each of those confidence intervals sits well above one point zero, meaning the association is consistent across the studies that fed into each estimate. The median lifetime prevalence of any partner violence among women with PTSD was sixty-one percent. More than six in ten. PTSD carries the highest measured risk by a considerable margin — more than twice the odds ratio seen for depression. But the elevated risk extends beyond those three diagnoses. Individual studies reported raised odds across a range of other diagnostic categories where pooled estimates couldn't be calculated. One large national survey found an odds ratio of five point fifty-eight for women with dysthymic disorder and eight point fourteen for women with bipolar disorder. For men with bipolar disorder, the same survey reported an odds ratio of nine point forty-two. Men with PTSD, in one study, had an odds ratio of nine point sixty-six. The data on men are sparse — too sparse for pooled estimates — but the direction is consistent: elevated odds across every diagnostic category where studies existed. The review is clear that prevalence estimates were higher for women, but the pattern of increased risk cuts across sexes. Here is where the evidence hits a genuine wall. The association between mental disorder and domestic violence is strong and consistent. What it cannot tell us is which came first. Does having a mental disorder make someone more vulnerable to violence? Does experiencing violence cause mental disorder? Is it both, in a cycle that reinforces itself? Trevillion and colleagues found only three longitudinal studies that even tried to address this question. One three-year cohort of two hundred eighty-six women found that among those who were both depressed and in a violent relationship, only one instance of depression had predated the violence. Among the twelve women who left the violent relationship within a year, the rate of depression fell to roughly the same level as women who had never experienced violence — twenty-five percent versus twenty-three percent. That is suggestive, but a sample of fourteen women with overlapping depression and violence cannot carry the weight of a causal conclusion. A second study, using the Dunedin birth cohort, found that depression at age fifteen did not predict relationship violence at age twenty-one, while conduct disorder did. A third study, using the Christchurch cohort, reported that having any mental disorder in adolescence and early adulthood was associated with partner violence in the mid-twenties, and that association held even after adjusting for prior mental disorder and other antecedent and concurrent factors. Three studies. That's the entire longitudinal evidence base. The review concludes plainly: the direction of causality cannot be established from available data. This matters clinically, not just academically. If violence causes PTSD, then addressing the violence is part of treating the disorder. If mental disorder increases vulnerability to violence, screening and safety planning become part of the care pathway from the start. The answer may well be both — a bidirectional relationship that compounds over time — but without longitudinal studies designed to test that, treatment planning is operating without a roadmap. Trevillion and colleagues call explicitly for longitudinal research to identify pathways to victimization so that healthcare responses can be optimized. That call has been sitting in the literature since 2012. What the evidence does support — now, without waiting for that longitudinal work — is a change in clinical practice. The review found elevated odds of domestic violence across every diagnostic category studied: depression, anxiety, PTSD, psychosis, dysthymia, and bipolar disorder. That pattern holds for women and, in individual studies, for men. Any mental health assessment is therefore an opportunity to ask about domestic violence. Trevillion and colleagues are specific about what effective identification requires: professionals trained to understand the nature and long-term impact of domestic violence, the ability to ask about it safely when abuse may be ongoing, and clear referral and care pathways when abuse is identified. The World Health Organization guidelines they cite recommend that primary care and mental health services work in active partnership with domestic violence services — not as a courtesy, but as a structural requirement. The review's honest accounting of its own limits deserves to be named. Data on men are thin. Family violence — abuse perpetrated by non-partner relatives — was barely studied, and pooled estimates could not be produced for it. Same-sex relationships are nearly absent from the evidence base. The pooled estimates are unadjusted odds ratios, because primary studies rarely provided enough data to adjust for confounders. Funnel-plot asymmetry suggests the possibility of publication bias, particularly in the depression studies — meaning the true pooled estimate may be somewhat lower than what was calculated. Heterogeneity across studies was high for several of the pooled estimates, which means the numbers should be read as signals of direction and magnitude, not as precise point estimates. Even with those caveats, the signal is clear enough to act on. The clinician sitting across from a patient with PTSD, or depression, or anxiety — the one whose checklist doesn't include a question about domestic violence — is missing something the evidence now says they should be looking for. The question isn't peripheral. It belongs in the room. This lecture was created by ennepō. Go to https://ennepo.ai to Discover, Create and Follow the latest research in your field. Read when you can. Listen when you want to.