Early Mental Health Treatment and Domestic Violence Outcomes in Psychosis: A Population-Based Cohort Study

Abstract

Background:

Evidence for effective interventions among individuals with psychosis who perpetrate domestic violence (DV) remains limited. In this population, justice system involvement frequently occurs during periods of inadequately managed mental health symptoms.

Aims:

We examined whether early mental health treatment following a DV charge reduces subsequent DV reoffending among individuals with psychosis.

Methods:

We conducted a population-based retrospective cohort study of individuals in New South Wales, Australia, with a recorded diagnosis of psychosis who were subsequently charged with a DV offense. The primary exposure was receipt of mental health treatment within 1 month following the index DV charge. The primary outcome was time to the first subsequent DV charge. Secondary outcomes included time to first DV conviction, violent DV charge, and violent DV conviction. Time-to-event analyses were performed using competing risks regression.

Results:

During a median follow-up of 4 years, 3,513 of 8,458 individuals (41.5%) were charged with a subsequent DV offense. Among men, early mental health treatment was associated with reduced risk across all DV outcomes, including any DV charge (adjusted sub distribution hazard ratio [sHR] 0.86, 95% CI [0.79, 0.93]), conviction (sHR 0.81, 95% CI [0.73, 0.89]), violent charge (sHR 0.83, 95% CI [0.75, 0.93]), and violent conviction (sHR 0.78, 95% CI [0.68, 0.89]). There was no evidence of an association between early treatment and DV reoffending among women.

Conclusions:

These findings support integrating timely psychiatric care into DV offender management for men with psychosis, while indicating that alternative, sex-responsive strategies may be required for women with psychosis.

Keywords

domestic violence psychosis mental health treatment

Introduction

Domestic violence (DV) affects approximately one in three women globally and remains a leading cause of morbidity and mortality among women of reproductive age (Sardinha et al., 2022). In Australia, 1 in 6 women and 1 in 18 men aged 15 years and above have experienced physical and/or sexual violence by a current or previous cohabiting partner (Australian Bureau of Statistics, 2025). Many countries report similar statistics (European Union Agency for Fundamental Rights, 2025; Sardinha et al., 2022; Smith et al., 2018; World Health Organization, 2021). The annual financial cost of DV against women and their children in Australia is estimated to be $22 billion (KPMG, 2016); however, this is a conservative estimate as it captures costs from reported incidents and does not account for unreported cases or the long-term, cascading impacts of DV perpetration. Justice responses to domestic violence vary. In Australia, among police-recorded DV-related assaults, 28% resulted in no legal action; among the cases that did proceed, only 39% led to a proven court outcome (Gilbert, 2025).

Several studies have examined associations between mental health treatment and general reoffending among individuals with psychosis. A previous study demonstrated a 17% reduction in reoffending risk among people diagnosed with psychosis (Adily et al., 2020). Similarly, court diversion into mental health treatment has been associated with reductions in reoffending among people with psychotic illness (Albalawi et al., 2019; Weatherburn et al., 2021; Akpanekpo et al., 2026). However, research examining DV reoffending among individuals with psychotic disorders remains limited. Theoretically, early psychiatric intervention could interrupt escalating symptoms and address mental health needs predisposing to repeated DV offending; however, without empirical evidence this remains speculative. Furthermore, individuals with psychosis may have distinct clinical needs that mainstream DV interventions are ill-equipped to address. Standard perpetrator programs typically emphasize accountability and behavioral change (Pence & Paymar, 1993), yet acute psychotic symptoms (Arciniegas, 2015), including threat/control-override symptoms, persecutory delusions, and hallucinations, may manifest clinically as aggression or violence that is more likely to occur in domestic settings where individuals spend most of their time.

Population-based research examining the relationship between early mental health intervention and DV reoffending among individuals with psychosis could inform more integrated and effective responses. The present study was conducted among individuals with psychotic disorders who were charged with DV offenses in New South Wales (NSW), Australia. The primary objective was to examine the association between early mental health treatment and DV reoffending. We hypothesized that early mental health treatment following a DV charge would be associated with reduced DV reoffending. Given established sex differences in psychosis presentation and treatment response (Barajas et al., 2015; Fazel et al., 2014; Irving et al., 2021; Thorup et al., 2014), we stratified analyses by sex.

Methods

Study Design

This study employed a population-based retrospective cohort design using de-identified administrative data from the health and justice sectors in NSW, Australia, linked via probabilistic matching methods by the Center for Health Record Linkage (CHeReL) using established protocols.

Study Cohort

The study cohort comprised 8,646 adults (aged 18 and above) who received their first recorded diagnosis of psychosis between July 1, 2001, and June 9, 2018, and subsequently had their first recorded DV charge (see Supplemental Table 1) in the New South Wales (NSW) Re-offending Database (ROD) between May 16, 2008, and June 30, 2020. Psychosis diagnoses were identified through records from the NSW Ministry of Health’s Admitted Patient Data Collection (APDC) and the Emergency Department Data Collection (EDDC).

Psychosis diagnoses were identified using ICD-9 and ICD-10 codes from APDC and EDDC for schizophrenia-related psychoses, affective psychoses, and substance-induced psychoses (see Supplemental Table 2). SNOMED-CT codes were mapped to corresponding ICD codes. In cases of multiple recorded psychosis diagnoses for an individual, a diagnostic hierarchy was applied (Adily et al., 2020; Albalawi et al., 2019). In cases where individuals had both schizophrenia-related and other psychosis diagnoses, they were classified as having schizophrenia-related psychoses. When individuals had both affective and substance-induced psychoses (but no schizophrenia-related diagnosis), they were classified as having affective psychoses. Classification as substance-induced psychoses was applied only when individuals had substance-induced psychoses without any schizophrenia-related or affective psychoses diagnoses.

Data Sources

This study utilized data from several NSW data collections. APDC provided data on hospital admissions. EDDC provided data on emergency department presentations. The Mental Health Ambulatory Data Collection (MH-AMB) provided data on outpatient-based contacts with mental health services. ROD provided data on court appearances, charges, convictions, and sentencing, as well as demographic variables including date of birth, sex, Indigenous status, and area-level socioeconomic disadvantage. Mortality data used for censoring were obtained from the NSW Registry of Births, Deaths, and Marriages (RBDM).

Exposure

The primary exposure was the first documented contact with mental health services within 1 month following the index (first) DV charge. This 1 month window represented the median time from the index charge to the first mental health treatment episode within this study cohort and aligns with methodological precedents in related research (Adily et al., 2020; Akpanekpo et al., 2025). Mental health treatment was defined as either a recorded hospital admission in the APDC with a mental health-related diagnostic code or a documented outpatient-based clinical contact with mental health services recorded in the MH-AMB. Administrative contacts recorded in the MH-AMB were excluded.

Outcomes

Offending outcome data were ascertained from the ROD. The primary outcome was the time from the index (first recorded) DV charge to the next DV charge. Secondary outcomes were the time from the index DV charge to: (i) the first subsequent DV conviction; (ii) the first subsequent violent DV charge; and (iii) the first subsequent violent DV conviction. DV offenses were identified using official ROD flags indicating the offense occurred within a domestic context (see Supplemental Table 1 for corresponding DV-related law part codes). Such offenses include violent acts (homicide/murder, assault, kidnapping), property damage, or breaches of apprehended DV orders. Violent DV charges and convictions included violent acts, excluding both property-related offenses and breaches of apprehended DV orders.

Statistical Analysis

We compared baseline characteristics of individuals with and without mental health treatment within 1 month of the index DV charge. Continuous variables were summarized using means (standard deviations, SD) or medians (interquartile ranges, IQR). Categorical variables were summarized using count and percentages.

To mitigate immortal time bias with respect to the exposure definition, follow-up for all time-to-event analyses was left-truncated at 1 month post-index DV charge. Follow-up continued until the earliest of the outcome event, death, or the study end date (June 30, 2020). Incarceration periods were removed from follow-up time because individuals were not at risk of community-based reoffending while in custody. Each individual’s at-risk time therefore reflects cumulative time spent in the community. This was implemented using a start-stop data structure in which custodial periods were excluded and the remaining community-based segments were concatenated to form each individual’s total at-risk interval.

Fine and Gray sub distribution hazard models were used to estimate unadjusted and adjusted sub distribution hazard ratios (sHRs) with 95% confidence intervals (CIs) for the association between mental health treatment and each outcome. For the primary outcome (time to first subsequent DV charge), death was the competing event. Cumulative Incidence Functions (CIFs) were estimated to describe the probability of subsequent DV charge for each exposure group, stratified by sex. For each secondary outcome, relevant competing events were specifically defined. For time to first subsequent DV conviction, competing events included death and charges that did not result in conviction. For time to first subsequent violent DV charge, competing events included death and occurrence of a non-violent DV charge. For time to first subsequent violent DV conviction, competing events included death, charges that did not result in conviction, and occurrence of a non-violent DV conviction. Covariates for all adjusted models were selected based on literature review and clinical relevance (Adily et al., 2020; Fazel et al., 2014; Weatherburn et al., 2021). A detailed list of covariates included in the adjusted models is presented in Supplemental Table 2.

We conducted several sensitivity analyses. First, we examined alternative exposure windows of 2 weeks and 2 months post-index DV charge to evaluate the sensitivity of results to the timing of mental health treatment. Second, we excluded individuals whose mental health treatment consisted solely of involuntary hospitalization to address potential confounding by illness severity. Third, we repeated Fine–Gray models with additional adjustment for recency of mental health contact prior to the first DV charge to examine whether pre-existing engagement patterns could account for the observed associations. Finally, we conducted subgroup analyses stratified by psychosis type (schizophrenia and related psychoses, affective psychoses, and substance-induced psychoses), with treatment-by-psychosis-type interaction tests to examine whether treatment effects varied by diagnostic category.

Missing data patterns were assessed using Little’s test for missing completely at random (MCAR), which was statistically significant (p < .05), indicating data were not MCAR. Assuming data were missing at random (MAR), we used multiple imputation by chained equations (MICE) to address missing values in approximately 10% of cases. Ten imputed datasets were generated, with imputation models including multivariable analysis variables, auxiliary variables, censoring indicators, and Nelson–Aalen cumulative hazard estimates. All regression analyses used pooled estimates from these datasets. The proportional sub distribution hazards assumption for all Fine and Gray models was assessed based on time-dependent covariates. No violations were detected. All statistical tests were two-tailed, with p < .05 considered statistically significant. Data management was performed using SAS version 9.4 (SAS Institute, Cary, NC) and statistical analyses using Stata version 19.5 (Stata Corp, College Station, TX).

Results

Characteristics of the Study Population

A total of 8,646 individuals were included. Baseline characteristics, stratified by sex and early mental health treatment status, are detailed in Table 1. The median interval from psychosis diagnosis to the index DV charge was 4.0 years (IQR 1.0–8.0), similar in men and women. Treated individuals had a lower median age at psychosis diagnosis and at the index DV charge than untreated individuals in both men (26.8 vs. 29.1 years; 32.5 vs. 34.0 years) and women (27.7 vs. 28.9 years; 33.0 vs. 34.0 years). A higher proportion of women received early mental health treatment than men (790 [40.6%] of 1,947 vs. 2,312 [34.5%] of 6,699).

Table 1.

Characteristics of the Study Population.

Characteristic	Total population (N = 8,646)	Men			Women
Characteristic	Total population (N = 8,646)	Overall (N = 6,699)	Mental health treatment (n = 2,312)	No mental health treatment (n = 4,387)	Overall (N = 1,947)	Mental health treatment (n = 790)	No mental health treatment (n = 1,157)
Age at psychosis diagnosis, years
Median (IQR)	28.2 (22.0, 35.8)	28.2 (22.2, 35.5)	26.8 (21.2, 34.0)	29.1 (22.7, 36.2)	28.4 (21.6, 36.4)	27.7 (20.9, 36.4)	28.9 (22.1, 36.5)
Age at index DV offense, years
Median (IQR)	34 (27, 41)	34 (27, 41)	32.5 (26, 40)	34 (27, 41)	34 (26, 42)	33 (25, 42)	34 (27, 42)
Psychosis type, n (%)
Schizophrenia & related	5,587 (64.6)	4,291 (64.1)	1,643 (71.1)	2,648 (60.4)	1,296 (66.6)	551 (69.8)	745 (64.4)
Affective psychosis	703 (8.1)	486 (7.3)	164 (7.1)	322 (7.3)	217 (11.1)	105 (13.3)	112 (9.7)
Substance-induced psychosis	2,356 (27.3)	1,922 (28.7)	505 (21.8)	1,417 (32.3)	434 (22.3)	134 (17.0)	300 (25.9)
Comorbid diagnoses, n (%)
Mood disorders	2,573 (29.8)	1,839 (27.5)	652 (28.2)	1,187 (27.1)	734 (37.7)	314 (39.8)	420 (36.3)
Substance use disorders	4,714 (54.5)	3,638 (54.3)	1,405 (60.8)	2,233 (50.9)	1,076 (55.3)	463 (58.6)	613 (53.0)
Anxiety disorders	2,224 (25.7)	1,548 (23.1)	608 (26.3)	940 (21.4)	676 (34.7)	296 (37.5)	380 (32.8)
Trauma/stress-related	2,866 (33.2)	2,005 (29.9)	726 (31.4)	1,279 (29.2)	861 (44.2)	383 (48.5)	478 (41.3)
Prior self-harm	2,262 (26.2)	1,596 (23.8)	574 (24.8)	1,022 (23.3)	666 (34.2)	298 (37.7)	368 (31.8)
Indigenous status, n (%)
Non-indigenous	5,314 (61.5)	4,242 (63.3)	1,444 (62.5)	2,798 (63.8)	1,072 (55.1)	404 (51.1)	668 (57.7)
Indigenous	2,548 (29.5)	1,962 (29.3)	574 (24.8)	1,388 (31.6)	586 (30.1)	215 (27.2)	371 (32.1)
Missing	784 (9.1)	495 (7.4)	294 (12.7)	201 (4.6)	289 (14.8)	171 (21.7)	118 (10.2)
Socio-economic disadvantage, n (%)
Q1 (most disadvantaged)	2,494 (31.9)	2,006 (32.5)	645 (32.2)	1,361 (32.7)	488 (30.7)	176 (28.4)	312 (30.2)
Q2	2,305 (29.5)	1,788 (29.0)	552 (27.6)	1,236 (29.7)	517 (32.5)	197 (31.8)	320 (31.0)
Q3	2,014 (25.7)	1,595 (25.9)	503 (25.1)	1,092 (26.2)	419 (26.4)	160 (25.9)	259 (25.0)
Q4 (least disadvantaged)	1,009 (12.9)	780 (12.6)	303 (15.1)	477 (11.4)	229 (14.4)	86 (13.9)	143 (13.8)
Missing	824 (9.5)	530 (7.9)	309 (13.4)	221 (5.0)	294 (15.1)	171 (21.6)	123 (10.6)

IQR = inter-quartile range; Q = quartiles.

The median time since the most recent mental health contact before the first DV charge was 100 days overall (IQR 9–516; female 80 [IQR 7–435]; male 107 [IQR 9–542]; Supplemental Table 3), and differed by treatment status: median 6 days (IQR 0–54) among treated individuals vs. 269 days (IQR 64–848) among untreated individuals. Recent contact within 30 days was present in 69.3% of treated individuals compared with 15.9% of untreated men and 17.6% of untreated women (Supplemental Table 4).

Study Follow-Up

The cohort for all time-to-event analyses comprised 8,458 individuals (Figure 1). This analytical cohort was established after excluding 188 individuals who experienced a subsequent DV charge or were censored within 1 month of their index DV charge, prior to the commencement of the at-risk follow-up period. Reasons for exclusion are summarized in Supplemental Table 5. Baseline characteristics of excluded versus included individuals were similar across measured variables (all SMDs <0.10; Supplemental Table 6). The median follow-up time was 4 years (IQR, 2–7). Follow-up time was similar for individuals who had mental health treatment (n = 3,036; median 4 years, IQR 2–7) and those who did not (n = 5,422; median 4 years, IQR 2–6). A total of 4,945 individuals were censored: 345 due to death (median follow-up 3 years, IQR 1–5) and 4,600 at the end of study (median follow-up 5 years, IQR 4–8).

Figure 1.

STROBE flow diagram.

Among those who received early mental health treatment (n = 3,036), 2,997 (98.7%) had outpatient contact, 1,675 (55.2%) had inpatient admissions, and 1,107 (36.5%) had emergency presentations. In mutually exclusive setting combinations, outpatient contact only accounted for 1,110 (36.6%), outpatient plus admissions for 797 (26.3%), outpatient plus emergency for 240 (7.9%), and all three settings for 850 (28.0%). Contacts without any outpatient component were uncommon (admissions only: 22 [0.7%]; emergency only: 11 [0.4%]; admissions plus emergency: 6 [0.2%]). The distribution was similar for men and women (Supplemental Table 7).

Primary Outcome

During follow-up, 3,513 individuals (41.5%) had a subsequent DV charge. This occurred in 1,173 (38.6%) who had mental health treatment versus 2,340 (43.2%) without. At end of follow-up, the adjusted cumulative incidence of a subsequent DV charge was 5.22% points lower in treated versus untreated men (49.6% vs. 54.8%; 95% CI [–7.77, –2.54]); there was no evidence of a difference in women (see Supplemental Table 24). Adjusted cumulative incidence functions are shown in Figure 2 and sex-stratified results for all outcomes are presented in Table 2 (men) and Table 3 (women).

Figure 2.

Adjusted cumulative incidence of subsequent DV outcomes by early mental health treatment status, stratified by sex: (a) male and (b) female.

Table 2.

Association Between Mental Health Treatment and DV Outcomes among Individuals with Psychosis (Men).

Outcome	Group	N at risk	No. of events, n (%)	Unadjusted sHR (95% CI)^a	p-value	Adjusted sHR (95% CI)^a	p-value
Primary outcome
Subsequent DV charge	MH treatment	2,263	900 (39.8%)	0.84 (0.78–0.91)	<.001	0.86 (0.79–0.93)	<.001
Subsequent DV charge	No MH treatment	4,294	1,958 (45.6%)	1.0 (Reference)		1.0 (Reference)
Secondary outcomes
Subsequent DV conviction	MH treatment	2,263	645 (28.5%)	0.79 (0.72–0.87)	<.001	0.81 (0.73–0.89)	<.001
Subsequent DV conviction	No MH treatment	4,294	1,491 (34.7%)	1.0 (Reference)		1.0 (Reference)
Subsequent violent DV charge	MH treatment	2,263	473 (20.9)	0.83 (0.74–0.93)	.001	0.83 (0.75–0.93)	.001
Subsequent violent DV charge	No MH treatment	4,294	1,062 (24.7%)	1.0 (Reference)		1.0 (Reference)
Subsequent violent DV conviction	MH treatment	2,263	315 (13.9%)	0.77 (0.68–0.88)	<.001	0.78 (0.68–0.89)	<.001
Subsequent violent DV conviction	No MH treatment	4,294	761 (17.7%)	1.0 (Reference)		1.0 (Reference)

CI = confidence interval; DV = domestic violence; MH = mental health; sHR = sub distribution hazard ratio.

Sub distribution hazard ratios (unadjusted and adjusted) and corresponding p-values were derived from Fine and Gray competing risks regression models. Competing risks for each outcome were defined as described in the Methods section.

Table 3.

Association Between Mental Health Treatment and DV Outcomes among Individuals With Psychosis (Women).

Outcome	Group	N at risk	No. of events, n (%)	Unadjusted sHR (95% CI)^a	p-value	Adjusted sHR (95% CI)^a	p-value
Primary outcome
Subsequent DV charge	MH treatment	773	273 (35.3%)	1.05 (0.90 –1.23)	.542	1.03 (0.88 –1.21)	.689
Subsequent DV charge	No MH treatment	1,128	382 (33.9%)	1.0 (Reference)		1.0 (Reference)
Secondary outcomes
Subsequent DV conviction	MH treatment	773	179 (23.2%)	0.95 (0.79–1.15)	.588	0.94 (0.77–1.14)	.523
Subsequent DV conviction	No MH treatment	1,128	273 (24.2%)	1.0 (Reference)		1.0 (Reference)
Subsequent violent DV charge	MH treatment	773	146 (18.9%)	1.03 (0.83–1.27)	.775	1.03 (0.83–1.28)	.783
Subsequent violent DV charge	No MH treatment	1,128	207 (18.4%)	1.0 (Reference)		1.0 (Reference)
Subsequent violent DV conviction	MH treatment	773	90 (11.6%)	0.98 (0.75–1.29)	.911	1.00 (0.76–1.31)	.973
Subsequent violent DV conviction	No MH treatment	1,128	133 (11.8%)	1.0 (Reference)		1.0 (Reference)

CI = confidence interval; DV = domestic violence; MH = mental health; sHR = sub distribution hazard ratio.

Secondary Outcomes

Several secondary DV-related outcomes were assessed. A subsequent DV conviction occurred in 2,588 individuals (30.6%); 824 (27.1%) with mental health treatment and 1,764 (32.5%) without. At end of follow-up, the adjusted cumulative incidence of a subsequent DV conviction was 6.27% points lower in treated versus untreated men (36.0% vs. 42.3%; 95% CI [–9.07, –3.42]); there was no evidence of a difference in women. A subsequent violent DV charge was recorded in 1,888 individuals (22.3%); 619 (20.4%) with treatment and 1,269 (23.4%) without. Subsequent violent DV conviction was observed in 1,299 individuals (15.4%); 405 (13.3%) with treatment and 894 (16.5%) without.

Association Between Early Mental Health Treatment and Study Outcomes

Unadjusted and adjusted sHRs are presented in Table 2 (men) and Table 3 (women). Full model results are presented in Supplemental Tables 8 and 9. Among men, after adjustment for covariates, mental health treatment was associated with a lower hazard of experiencing a subsequent DV charge (adjusted sHR, 0.86; 95% CI [0.79, 0.93]). Mental health treatment was similarly associated with lower hazards for all secondary outcomes: subsequent DV conviction (adjusted sHR, 0.81; 95% CI [0.73, 0.89]), subsequent violent DV charge (adjusted sHR, 0.83; 95% CI [0.75, 0.93]), and subsequent violent DV conviction (adjusted sHR, 0.78; 95% CI [0.68, 0.89]).

In contrast, among women, after adjustment for covariates, mental health treatment was not statistically significantly associated with the sub distribution hazard of a subsequent DV charge (adjusted sHR, 1.03; 95% CI [0.88, 1.21]) or with any of the secondary outcomes (Table 3).

Sensitivity and Subgroup Analyses

The study’s findings were consistent across sensitivity analyses. Using a 2-week exposure window, early mental health treatment was associated with a lower hazard of a subsequent DV charge for men (adjusted sHR 0.84, 95% CI [0.77, 0.91]; Supplemental Table 10), whereas for women there was no evidence of a difference in hazard (adjusted sHR 1.00, 95% CI [0.85, 1.17]; Supplemental Table 11). Similar estimates were observed with a 2 month exposure window for men (adjusted sHR 0.90, 95% CI [0.83, 0.97]; Supplemental Table 12) and women (adjusted sHR 1.09, 95% CI [0.92, 1.29]; Supplemental Table 13). After excluding individuals whose treatment was an involuntary hospitalization, the estimates for men (adjusted sHR 0.86, 95% CI [0.79, 0.93]; Supplemental Table 14) and women (adjusted sHR 1.03, 95% CI [0.88, 1.21]; Supplemental Table 15) were similar to the primary analysis.

In subgroup analyses stratified by psychosis type, the association between early mental health treatment and reoffending differed by diagnosis. Among men with schizophrenia and related psychoses, early treatment was associated with a lower hazard of a subsequent DV charge (adjusted sHR 0.83, 95% CI [0.75, 0.91]; Supplemental Table 16). However, there was no evidence of a difference in hazard for men with affective psychoses (adjusted sHR 1.05, 95% CI [0.78, 1.42]; Supplemental Table 17) or substance-induced psychoses (adjusted sHR 0.90, 95% CI [0.77, 1.05]; Supplemental Table 18). Similarly, there was no evidence of a difference in hazard for women across any of the psychosis subgroups (Supplemental Tables 19 and 20, and 21). Treatment-by-psychosis-type interaction tests did not support effect modification by diagnostic subgroup (Supplemental Table 22).

In a sensitivity analysis additionally adjusting Fine-Gray models for recency of mental health contact prior to the first DV charge, treatment effect estimates were not materially altered. Among men, sub distribution hazard ratios for early mental health treatment remained below the null across all outcomes (subsequent DV charge: adjusted sHR 0.85, 95% CI [0.78, 0.94]; Supplemental Table 23). Recent contact within 30 days was independently associated with a higher hazard of subsequent DV charge relative to disengagement exceeding 1 year (adjusted sHR 1.14, 95% CI [1.03, 1.27]). Among women, treatment estimates remained compatible with no association.

Discussion

In this large, population-based cohort study, early mental health treatment following a DV charge was associated with a lower hazard of subsequent DV reoffending among men with psychosis, but there was no evidence of a difference in hazard for women. Men who received early treatment had a 14% lower hazard of a subsequent DV charge and a 22% lower hazard of a subsequent violent DV conviction. In absolute terms, treated men had a 5.22% point lower adjusted cumulative incidence of subsequent DV charge and 6.27% points lower for subsequent DV conviction at end of follow-up. These findings were consistent across sensitivity analyses using different exposure windows and when excluding involuntary hospitalizations. In subgroup analyses, the association was observed among men with schizophrenia and related psychoses but not among men with affective or substance-induced psychoses; formal interaction tests did not support effect modification by psychosis type. In women, cumulative incidence differences were small, and across outcomes and sensitivity analyses the estimates were not statistically significant; confidence intervals spanned the null, indicating no clear evidence of an association.

The findings from the present study align with previous research reporting associations between mental health treatment and reduced reoffending in justice-involved populations (Adily et al., 2020; Fazel et al., 2014; Weatherburn et al., 2021). In a recent meta-analysis, mental health court participation reduced recidivism by 42% (Jalain et al., 2024), while court diversion into mental health treatment has been associated with reduced reoffending both generally (Soon et al., 2024) and specifically among individuals with psychosis (Albalawi et al., 2019; Weatherburn et al., 2021; Akpanekpo et al., 2026). Consistent with our focus on early intervention, mental health treatment within 1 month of offense reduced reoffending among men with psychosis (Adily et al., 2020). Pharmacological interventions have demonstrated measurable effects on violence reduction in this population (Leucht et al., 2013; Swanson et al., 2008). For example, in a within-individual comparison of people prescribed antipsychotics or mood stabilizers, periods of medication use were associated with a 45% reduction in violent crime compared to periods without medication (Fazel et al., 2014).

The null findings for women in the present study may reflect several factors. First, studies have shown that men with psychosis more commonly present with negative symptoms and substance use, while women more commonly present with affective symptoms (Barajas et al., 2015; Irving et al., 2021; Køster et al., 2008; Thorup et al., 2014). Evidence from pharmaco-epidemiological studies also suggests differential treatment responses for violent outcomes based on sex and clinical phenotype. For example, one study found periods of adherence to pharmacological treatment was associated with reduced violent crime in women with bipolar disorder but not in women with schizophrenia-spectrum disorders or men with bipolar disorder (Fazel et al., 2014). Our findings are consistent with this evidence. Second, women who perpetrate violence, including those with mental illness, commonly report prior victimization from childhood experiences and intimate partners (Goodman et al., 2001; Hester, 2013; Stuart et al., 2006). Such violence may occur within the context of threat, self-defense, and trauma (Iratzoqui, 2018). The absence of evidence of a difference in hazard for women in this study may suggest that routine mental health treatment may be insufficient without wrap-around support and trauma-informed care.

Strengths

The strengths of this study include alignment of the estimate, exposure definition, time origin, and at-risk time scale to the question of community reoffending after a DV charge among people with psychosis. Left truncation at 1 month post-charge reduced immortal time bias. Custodial periods were removed so custody time was not misclassified as time at risk. We treated death and, for conviction-based endpoints, non-convicted charges as competing events rather than non-informative censoring, as death precludes subsequent DV reoffending and a non-convicted charge reflects informative justice-system contact that alters the subsequent probability of a convicted outcome, making non-informative censoring implausible in a setting where attrition from charge to conviction is high.

Linked administrative records reduced recall bias and differential reporting by treatment status. Reporting both charge-based and conviction-based outcomes improved construct validity, as charges are more sensitive but more surveillance-prone, whereas convictions are more specific but subject to evidentiary thresholds and case attrition. Population coverage, long follow-up, sex-stratified analyses, and multiple sensitivity analyses strengthen inference.

Limitations

Several limitations warrant consideration. The observational design precludes causal inference and residual confounding is possible, particularly through factors that drive service engagement near the index charge, such as acute symptom exacerbation, crisis presentations, diversion or referral pathways, and local service accessibility. More recent mental health contact before the first DV charge was associated with higher subsequent incidence of DV charges. Because treated individuals had more recent contact than untreated individuals, this would be expected to bias treatment associations toward the null. However, in sensitivity analyses adjusting for recency of mental health contact: (1) the observed association in men is in the opposite direction to this potential bias; (2) treatment effect estimates were not materially altered.

Differential detection is possible if post-index service engagement increases opportunities for DV incidents to be identified. We attempted to reduce this risk in three ways. First, the index DV charge was defined as the first observed DV charge after the first recorded psychosis diagnosis, which reduces confounding by post-diagnosis DV history. Second, the median 4 year interval from diagnosis to the index charge makes a simple newly-visible-to-services explanation less plausible. Third, we reported both charge-based and conviction-based endpoints. Charges are more sensitive to surveillance and reporting effects, whereas convictions are more specific but depend on downstream justice processes and case attrition. The broadly similar pattern across both endpoint types reduces the likelihood that surveillance alone explains the observed association in men, although ascertainment-related bias cannot be eliminated.

The exposure captured a first mental health service contact rather than a standardized intervention. The data did not include the content, modality, intensity, pharmacological versus psychosocial composition, or adherence to treatment, limiting clinical translation to any specific component. Official records likely underestimate true reoffending and may misclassify DV context. Administrative diagnostic codes may not capture clinical complexity or severity. Generalizability beyond NSW may be limited because legal definitions, policing practices, service pathways, and availability of mental health care vary across jurisdictions. The cohort was defined by an initial DV charge, so results do not inform primary prevention. Findings apply only to the subset of individuals with psychosis who perpetrate DV and should not be interpreted as evidence that mental illness is inherently linked to violence. We also lacked information on engagement with perpetrator programs, psychosocial supports, medication dispensing, and community supervision conditions after the index charge, which may act as co-interventions or effect modifiers. We could not distinguish crisis-driven contacts from planned care beyond available coding, and some outpatient encounters may reflect brief assessments rather than sustained treatment.

Conclusion

Early mental health service contact following a DV charge was associated with lower hazard of subsequent DV outcomes among men with psychosis. There was no evidence of a corresponding association in women. These findings suggest that integrating timely psychiatric care into DV offender management could be a useful component of strategies for men with psychosis. Given the heterogeneity of the treatment exposure across settings and modalities, these findings should be interpreted as evidence for the potential value of early service engagement broadly, rather than as support for any specific intervention, which would be the subject of future research. For women, the lack of an observed association suggests that different, sex-responsive approaches may be required.

Supplemental Material

sj-docx-1-isp-10.1177_00207640261446101 – Supplemental material for Early Mental Health Treatment and Domestic Violence Outcomes in Psychosis: A Population-Based Cohort Study

Supplemental material, sj-docx-1-isp-10.1177_00207640261446101 for Early Mental Health Treatment and Domestic Violence Outcomes in Psychosis: A Population-Based Cohort Study by Emaediong I. Akpanekpo, Nabila Z. Chowdhury, Armita Adily, George Karystianis and Tony Butler in International Journal of Social Psychiatry

Footnotes

Acknowledgements

The authors acknowledge the Center for Health Record Linkage (CHeReL) for conducting the data linkage that made this study possible. We thank the data custodians for providing access to the administrative data used in this study.

ORCID iDs

Emaediong I. Akpanekpo

Nabila Z. Chowdhury

Armita Adily

George Karystianis

Tony Butler

Ethical Considerations

This study was conducted according to the guidelines of the Declaration of Helsinki. All procedures involving human participants were approved by the NSW Population & Health Services Research Ethics Committee (PHSREC; 2019/ETH01721), the NSW Aboriginal Health and Medical Research Council (1089/15), the Justice Health and Forensic Mental Health Network (G324/14), and the Corrective Services NSW Ethics Committee (D2023/1555831).

Consent to Participate

As the study involved the use of existing de-identified data, a waiver of the requirement to seek consent was granted by the NSW Population & Health Services Research Ethics Committee.

Author Contributions

Emaediong I. Akpanekpo: Conceptualization, Methodology, Formal analysis, Investigation, Data curation, Writing – original draft, Writing – review & editing, Visualization, Project administration.

Nabila Z. Chowdhury: Conceptualization, Investigation, Data curation, Writing – review & editing.

Armita Adily: Investigation, Writing – review & editing.

George Karystianis: Investigation, Writing – review & editing.

Tony Butler: Conceptualization, Investigation, Resources, Writing – review & editing, Supervision.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data Availability Statement

The data that support the findings of this study are available through the Center for Health Record Linkage (CHeReL) and the respective data custodians. Researchers interested in accessing these data should contact CHeReL (). The authors are unable to share the data directly due to privacy and confidentiality restrictions governing the use of linked administrative health and justice data.

Supplemental Material

Supplemental material for this article is available online.

References

Adily

Albalawi

Kariminia

Wand

Chowdhury

N. Z.

Allnutt

Schofield

Sara

Ogloff

J. R. P.

O’Driscoll

Greenberg

D. M.

Grant

Butler

(2020). Association between early contact with mental health services after an offense and reoffending in individuals diagnosed with psychosis. JAMA Psychiatry, 77(11), 1137–1146. https://doi.org/10.1001/jamapsychiatry.2020.1255

Akpanekpo

E. I.

Chowdhury

N. Z.

Butler

(2025). Early mental health treatment and reoffending among justice-involved youth diagnosed with psychosis. Schizophrenia Research, 284, 16–22. https://doi.org/10.1016/j.schres.2025.07.025

Akpanekpo

E. I.

Kasinathan

Gaskin

Greenberg

Schofield

P. W.

Butler

. (2026). Mental health diversion and criminal justice outcomes: A population-based cohort study of 53,605 justice-involved adolescents. Health & Justice, 14(1), 22. https://doi.org/10.1186/s40352-026-00413-4

Albalawi

Chowdhury

N. Z.

Wand

Allnutt

Greenberg

Adily

Kariminia

Schofield

Sara

Hanson

O’Driscoll

Butler

(2019). Court diversion for those with psychosis and its impact on re-offending rates: Results from a longitudinal data-linkage study. BJPsych Open, 5(1), e9. https://doi.org/10.1192/bjo.2018.71

Arciniegas

D. B.

(2015). Psychosis. Continuum: Lifelong Learning in Neurology, 21(3), 715–736. https://doi.org/10.1212/01.CON.0000466662.89908.e7

Australian Bureau of Statistics. (2025). Personal safety, Australia. Australian Bureau of Statistics. https://www.abs.gov.au/statistics/people/crime-and-justice/personal-safety-australia/latest-release

Barajas

Ochoa

Obiols

J. E.

Lalucat-Jo

(2015). Gender differences in individuals at high-risk of psychosis: A comprehensive literature review. The Scientific World Journal, 2015, 430735. https://doi.org/10.1155/2015/430735

European Union Agency for Fundamental Rights. (2025). Violence against women: An EU-wide survey. Main results. Publications Office of the European Union. https://fra.europa.eu/en/publication/2014/violence-against-women-eu-wide-survey-main-results-report

Fazel

Zetterqvist

Larsson

Långström

Lichtenstein

(2014). Antipsychotics, mood stabilisers, and risk of violent crime. Lancet, 384(9949), 1206–1214. https://doi.org/10.1016/s0140-6736(14)60379-2

10.

Gilbert

(2025). Attrition of domestic assaults from the New South Wales criminal justice system (no. 176). NSW Bureau of Crime Statistics and Research.

11.

Goodman

L. A.

Salyers

M. P.

Mueser

K. T.

Rosenberg

S. D.

Swartz

Essock

S. M.

Osher

F. C.

Butterfield

M. I.

Swanson

& Five site health and risk study research committee. (2001). Recent victimization in women and men with severe mental illness: Prevalence and correlates. Journal of Traumatic Stress, 14(4), 615–632. https://doi.org/10.1023/A:1013026318450

12.

Hester

(2013). Who does what to whom? Gender and domestic violence perpetrators in English police records. European Journal of Criminology, 10(5), 623–637. https://doi.org/10.1177/1477370813479078

13.

Iratzoqui

(2018). Domestic violence and the victim/offender overlap across the life course. International Journal of Offender Therapy and Comparative Criminology, 62(9), 2801–2816. https://doi.org/10.1177/0306624X17734993

14.

Irving

Colling

Shetty

Pritchard

Stewart

Fusar-Poli

McGuire

Patel

(2021). Gender differences in clinical presentation and illicit substance use during first episode psychosis: A natural language processing, electronic case register study. BMJ Open, 11(4), e042949. https://doi.org/10.1136/bmjopen-2020-042949

15.

Jalain

C. I.

Lucas

P. A.

Higgins

G. E.

(2024). Assessing the effectiveness of mental health courts in reducing recidivism: A systematic review with meta-analysis. Justice Evaluation Journal, 7(2), 212–230. https://doi.org/10.1080/24751979.2024.2373245

16.

Køster

Lajer

Lindhardt

Rosenbaum

(2008). Gender differences in first episode psychosis. Social Psychiatry and Psychiatric Epidemiology, 43(12), 940–946. https://doi.org/10.1007/s00127-008-0384-3

17.

KPMG. (2016). The cost of violence against women and their children in Australia (final report). Department of Social Services.

18.

Leucht

Cipriani

Spineli

Mavridis

Orey

Richter

Samara

Barbui

Engel

R. R.

Geddes

J. R.

Kissling

Stapf

M. P.

Lässig

Salanti

Davis

J. M.

(2013). Comparative efficacy and tolerability of 15 antipsychotic drugs in schizophrenia: A multiple-treatments meta-analysis. Lancet, 382(9896), 951–962. https://doi.org/10.1016/S0140-6736(13)60733-3

19.

Pence

Paymar

(1993). Education groups for men who batter (pp. 7991–7993). Springer Publishing Company. https://doi.org/10.1891/9780826179913

20.

Sardinha

Maheu-Giroux

Stöckl

Meyer

S. R.

García-Moreno

(2022). Global, regional, and national prevalence estimates of physical or sexual, or both, intimate partner violence against women in 2018. Lancet, 399(10327), 803–813. https://doi.org/10.1016/s0140-6736(21)02664-7

21.

Smith

S. G.

Zhang

Basile

K. C.

Merrick

M. T.

Wang

Kresnow

Chen

(2018). The national intimate partner and sexual violence survey: 2015 data brief–updated release (CDC:60893). National Center for Injury Prevention and Control, Centers for Disease Control and Prevention. https://stacks.cdc.gov/view/cdc/60893

22.

Soon

Y. L.

Singh

Greenberg

Rae

Korobanova

Dixon

Dean

(2024). Impact of mental health court diversion on reoffending: A direct comparison of diverted and un-diverted groups. International Journal of Forensic Mental Health, 23, 484–498. https://doi.org/10.1080/14999013.2024.2390107

23.

Stuart

G. L.

Moore

T. M.

Gordon

K. C.

Ramsey

S. E.

Kahler

C. W.

(2006). Psychopathology in women arrested for domestic violence. Journal of Interpersonal Violence, 21(3), 376–389. https://doi.org/10.1177/0886260505282888

24.

Swanson

J. W.

Swartz

M. S.

Van Dorn

R. A.

Volavka

Monahan

Stroup

T. S.

McEvoy

J. P.

Wagner

H. R.

Elbogen

E. B.

Lieberman

J. A.

(2008). Comparison of antipsychotic medication effects on reducing violence in people with schizophrenia. British Journal of Psychiatry, 193(1), 37–43. https://doi.org/10.1192/bjp.bp.107.042630

25.

Thorup

Albert

Bertelsen

Petersen

Jeppesen

Le Quack

Krarup

Jørgensen

Nordentoft

(2014). Gender differences in first-episode psychosis at 5-year follow-up–two different courses of disease? Results from the OPUS study at 5-year follow-up. European Psychiatry, 29(1), 44–51. https://doi.org/10.1016/j.eurpsy.2012.11.005

26.

Weatherburn

Albalawi

Chowdhury

Wand

Adily

Allnutt

Butler

(2021). Does mental health treatment reduce recidivism among offenders with a psychotic illness? Journal of Criminology, 54(2), 239–258. https://doi.org/10.1177/0004865821996426

27.

World Health Organization. (2021). Violence against women prevalence estimates, 2018. World Health Organization. https://www.who.int/publications/i/item/9789240022256

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.15 MB

0.00 MB