Lifetime Prevalence of Gender-Based Violence in Women and the Relationship With Mental Disorders and Psychosocial Function

Context Intimate partner physical violence, rape, sexual assault, and stalking are pervasive and co-occurring forms of gender-based violence (GBV). An association between these forms of abuse and lifetime mental disorder and psychosocial disability among women needs to be examined.

Objectives To assess the association of GBV and mental disorder, its severity and comorbidity, and psychosocial functioning among women.

Design, Setting, and Participants A cross-sectional study based on the Australian National Mental Health and Well-being Survey in 2007, of 4451 women (65% response rate) aged 16 to 85 years.

Main Outcome Measures The Composite International Diagnostic Interview version 3.0 of the World Health Organization's World Mental Health Survey Initiative was used to assess lifetime prevalence of any mental disorder, anxiety, mood disorder, substance use disorder, and posttraumatic stress disorder (PTSD). Also included were indices of lifetime trauma exposure, including GBV, sociodemographic characteristics, economic status, family history of mental disorder, social supports, general mental and physical functioning, quality of life, and overall disability.

Results A total of 1218 women (27.4%) reported experiencing at least 1 type of GBV. For women exposed to 3 or 4 types of GBV (n = 139), the rates of mental disorders were 77.3% (odds ratio [OR], 10.06; 95% confidence interval [CI], 5.85-17.30) for anxiety disorders, 52.5% (OR, 3.59; 95% CI, 2.31-5.60) for mood disorder, 47.1% (OR, 5.61; 95% CI, 3.46-9.10) for substance use disorder, 56.2% (OR, 15.90; 95% CI, 8.32-30.20) for PTSD, 89.4% (OR, 11.00; 95% CI, 5.46-22.17) for any mental disorder, and 34.7% (OR, 14.80; 95% CI, 6.89-31.60) for suicide attempts. Gender-based violence was associated with more severe current mental disorder (OR, 4.60; 95% CI, 2.93-7.22), higher rates of 3 or more lifetime disorders (OR, 7.79; 95% CI, 6.10-9.95), physical disability (OR, 4.00; 95% CI, 1.82-8.82), mental disability (OR, 7.14; 95% CI, 2.87-17.75), impaired quality of life (OR, 2.96; 95% CI, 1.60-5.47), an increase in disability days (OR, 3.14; 95% CI, 2.43-4.05), and overall disability (OR, 2.73; 95% CI, 1.99-3.75).

Conclusion Among a nationally representative sample of Australian women, GBV was significantly associated with mental health disorder, dysfunction, and disability.

Violence against women is a major public health concern,¹ contributing to high levels of morbidity and mortality worldwide.² Our study examines the mental health associations of 4 types of violence that are commonly perpetrated against women, namely physical forms of intimate partner violence (IPV), rape, other forms of sexual assault, and stalking, and were specifically included in the Australian National Mental Health and Well-being Survey. These 4 types of violence are referred to collectively herein as gender-based violence (GBV).³

In the United States, 17% of women report rape or attempted rape¹ and more than one-fifth of women report IPV, stalking, or both.^4-6 There is mounting evidence that each of these forms of GBV is associated with mental disorder among women,^3,7 although methodological shortcomings of existing studies constrain the inferences that can be drawn.³ Limitations of the majority of studies include a failure to use random and nationally representative samples⁸; the tendency to focus on 1 abuse rather than a range of interrelated abuses, an important consideration because women who experience 1 form of GBV are at greater risk of incurring other types^9,10; and the use of symptom checklists rather than structured diagnostic interviews that identify the prevalence of common mental disorders.⁸ In general, studies have not reported associations of GBV with measures of severity and comorbidity of mental disorders.³ In addition, past studies have given insufficient attention to the socioeconomic status and psychosocial functioning of women affected by GBV. This study is novel in its scope in that it attempts to assess in a comprehensive manner whether it is possible to identify a population of women with a concentration of mental disorders and wider psychosocial problems associated with GBV.

The specific goals of our study were to assess cross-sectionally in a nationally representative sample the association of a composite index of GBV (rape, sexual abuse, IPV, and stalking) with a range of lifetime mental disorders, including indices of severity and comorbidity. We applied standardized diagnostic measures for mental disorder. In addition, we examined whether GBV was associated with reported poor health, disability, and reduced quality of life.

The Australian Bureau of Statistics (ABS), the national government statistics agency, conducted the second National Mental Health and Well-being Survey in 2007. The ABS provided ethical review and approval for the survey, including voluntary recruitment, rigorous confidentiality provisions, and written informed consent. The ABS operates under Australian National Legislation that mandates strict provisions for the ethical conduct of the agency's research.

Random-stratified, multistage area probability sampling was used to select a nationally representative sample. In-scope persons aged 16 to 85 years were the usual residents of the private dwellings.^11,12 A total of 14 805 dwellings were included from the initial pool of 17 352 households, after excluding those that were ineligible or out of scope. An algorithm was applied to randomly select 1 person from each household without replacement for refusals. A total of 8841 respondents completed the full interview, representing an overall response rate of 60%. A response rate of 65% was imputed for women participants based on the unweighted data and the gender distribution of the Australian population. Of the total number of nonresponders (n = 5964), 61% (n = 3638) were full refusals, household information was obtained but not an interview in 27% (n = 1610), and the remaining 12% (n = 716) provided partial or incomplete information. A follow-up study of nonresponders suggested that there were slight biases in the proportional representation of men, young adults, and the population of 1 city (Perth), but the effects on overall prevalence rates of mental disorder were assessed to be insignificant.¹² The ABS-trained interviewers conducted face-to-face computer-assisted interviews with respondents.¹²

The Composite International Diagnostic Interview version 3.0 of the World Health Organization's World Mental Health Survey Initiative (WMH-CIDI 3.0)¹³ was used to assign lifetime diagnoses according to the Diagnostic and Statistical Manual of Mental Disorders (Fourth Edition) (DSM-IV). We aggregated respondents into 6 broad classes: (1) no disorder, (2) any mental disorder, (3) anxiety disorders (panic disorder, agoraphobia, social phobia, generalized anxiety disorder, obsessive-compulsive disorder, and posttraumatic stress disorder [PTSD]), (4) mood disorders (major depressive episode, dysthymia, and bipolar affective disorder), (5) substance use disorders (harmful alcohol use, alcohol dependence, and drug use disorders), and (6) PTSD.¹³ Posttraumatic stress disorder was the only individual disorder analyzed given its specific association with exposure to traumatic events such as GBV. There was a risk of the results being confounded by including the PTSD category in 2 sets of analysis, that is within the broader domain of anxiety disorders of which it forms part, and then on its own. Preliminary testing indicated that removal or retention of PTSD within the anxiety domain had only a negligible effect on the results. We therefore retained the PTSD subcategory in all analyses involving the anxiety disorder domain. All DSM-IV diagnoses were included in the analyses without applying hierarchical rules giving priority to one diagnosis over another.

An index of the clinical severity of mental disorder (mild, moderate, and severe)¹¹ was assigned to each person with any mental disorder. The index is defined by a composite set of characteristics that refer to the duration of symptoms and associated psychosocial impairment based on an algorithm developed by the World Mental Health Survey Initiative team and adapted to Australian conditions.¹² The WMH-CIDI 3.0 recorded a history of past suicide attempts. The instrument also documented 29 types of traumatic events (referred to as potentially traumatic events [PTEs]),^13,14 including the 4 GBV categories of interest (physical violence IPV, rape, other forms of sexual assault, and stalking). For physical violence IPV, the inquiry was whether the respondent was ever badly beaten up by a spouse or romantic partner (respondent answered positively to the following question: “Were you ever badly beaten up by a spouse or romantic partner?”). Rape was defined as sexual intercourse or penetration with a finger or object against the person's will, or by use of threat or force, or when the person was too young to understand what was happening. Sexual assault referred to experiences of sexual assault and molestation that were not within the definition provided for rape (respondent answered positively to the following question: “Other than rape, were you ever sexually assaulted, where someone touched you inappropriately, or when you did not want them to?”). Stalking was defined as being followed or kept track of in a manner that led to feelings of serious danger.

Information was recorded for the age of first exposure and frequency of occurrence of each PTE. Most women who have experienced GBV recorded a lifetime frequency of between 1 and 3 exposures. Because the frequency range was wide for the minority participants who reported higher exposure, and certain GBVs such as stalking tend to be repeated over time, we adopted a conservative approach by analyzing each type of GBV categorically as present or absent. Two indices of GBV were used in the analyses: (1) a categorical index (any GBV), reflecting those women who had experienced at least 1 of the 4 types (scored 0 or 1); and (2) a measure of incremental exposure to different types of GBV (scored 1, 2, and 3-4), where the highest category (3-4) was collapsed because of the relatively low rate of endorsement of all 4 GBV categories (3 GBV categories [n = 113]; 4 GBV categories [n = 26]).

Nonsexual childhood abuse was recorded by 2 trauma items that referred to being beaten by a caregiver or having witnessed serious physical violence at home as a child. The remainder of the list of other PTEs covered civilian violence, war, natural disasters, unintentional injuries, and experiences associated with settings of mass conflict.¹¹

Sociodemographic characteristics included age, sex, household composition, marital status, labor force participation, social support, number of reported physical health conditions, and an index of relative socioeconomic disadvantage (based on geographic area) developed by the ABS.^11,15 This index provides a composite ranking based on the location of residence, taking into account employment levels, income, home ownership, and education for the geographic area as a whole; the lower the numerical ranking of an area, the greater the level of socioeconomic disadvantage.

The Australian Assessment of Quality of Life (AQoL) Instrument¹⁶ was used to assess quality of life. The AQoL has undergone extensive development and psychometric testing.¹⁶ The instrument has a coherent principal component structure with 5 orthogonal factors (illness, independent living, physical ability, psychological well-being, and social relationships).¹⁶ Internal reliability of the items for the whole measure was found to be high with Cronbach α = .80. The AQoL showed sound convergence with an independent index of quality of life (comparative fit index = 0.90), with high correlations being evident for the psychological state (0.87) and social relationships (0.77) dimensions, respectively.¹⁶ The 12-item version World Health Organization Disability Assessment Schedule (WHODAS 2.0)¹⁷ and the Short-Form Disability Module¹⁸ were used to assess a multidimensional construct of disability based on the International Classification of Functioning, Disability, and Health. Two standard questions assessing disability days have been used widely across epidemiological studies worldwide.^11,12

Tetrachoric correlations were calculated to examine for associations among the 4 types of GBV. Logistic regression models were calculated to examine associations between GBV and the categories of lifetime mental disorders, and the strength of the association between mental disorder and incremental exposure to GBV, childhood nonsexual trauma, and other traumas. A cumulative logistic regression model examined associations between any GBV and indices of (1) severity and comorbidity of mental disorder, (2) general health status, and (3) quality of life and disability. All tests involved unadjusted univariate models and multivariate analyses controlling for the effects of age, marital status, country of birth, education, labor force participation, index of socioeconomic disadvantage, number of reported physical health conditions, number of relatives with a mental disorder, and social support (number of family and friends the participant could rely on).

SAS version 9.13 (SAS Institute Inc, Cary, North Carolina) was used to analyze data from the female sample only (weighted n = 4451). Full poststratification weights were computed by the ABS to reflect the demographic structure of the population, the probability of being sampled, and the differential response pattern across the population.¹¹ Sixty replicate weights were computed to standardize the sample against national census data. The characteristics that were standardized included sex, geographical location, age, household composition, educational attainment, and labor force participation. Normalized poststratification weighting was used for all statistical modeling. The delete-a-group jackknife variance technique was used to calculate the standard errors for estimates and the associated 95% confidence intervals (CIs).¹¹ Prevalence data are shown as unweighted and weighted estimates. Tests of associations are based on the weighted data and are reported as odds ratios (ORs) with 95% CIs. P < .05 was used to indicate statistical significance in all analyses.

The weighted sample included 4451 women. The lifetime prevalence for any mental disorder was 37.8%; for anxiety disorder, 24.6%; for mood disorder, 18.3%; for substance use disorder, 13.9%; and for PTSD, 9.8%.

A total of 1218 women (27.4%) reported experiencing at least 1 of the types of GBV assessed in this study. The lifetime prevalence rates were highest for sexual assault (14.7%), followed by stalking (10.0%), rape (8.1%), and physical violence IPV (7.8%). The median (interquartile range) age of the first occurrence of rape was 13 (8-18) years; of sexual assault, 12 (7-17) years; of physical violence IPV, 22 (12-27) years; and of stalking, 22 (17-30) years.

Table 1 shows the tetrachoric correlations comparing the 4 types of GBV with each other. All correlations were positive and statistically significant at .05 level, ranging from a correlation of 0.59 (95% CI, 0.53-0.64) for rape and sexual assault to a correlation of 0.33 (95% CI, 0.26-0.40) for physical violence IPV and sexual assault.

Table 2 shows the lifetime prevalence of GBV was highest (35.8%) for women aged 30 to 49 years and lowest (14.5%) for women aged 65 years or older. Married women had the lowest prevalence of GBV (22.9%) compared with those who never married (29.1%) and those not classified according to conventional categories (widowed, divorced, or separated; 37.8%). Australian-born women and immigrants from English-speaking countries reported higher exposure to GBV (28.6% and 30.2%, respectively) than immigrants from non–English-speaking countries (20.2%). Gender-based violence was more prevalent among women recording the greatest level of socioeconomic disadvantage.

The analysis assessed incremental exposure to GBV (0, 1, 2, or 3-4) with the 4 domains of mental disorder. No exposure to GBV was associated with a prevalence rate for any mental disorder of 28.0% compared with 57.3% for exposure to a single form of GBV and 89.4% for exposure to 3 to 4 types of GBV. Table 3 shows that both the unadjusted univariate and multivariate analyses yielded positive and statistically significant associations between the number of types of GBV reported for women and the prevalence of mental disorder. Specifically, multivariate analysis found that women who have been exposed to 1 form of GBV reported a high rate of lifetime mood disorder (weighted, 30.7%; OR, 2.26; 95% CI, 1.59-3.20), lifetime anxiety disorder (38.5%; OR, 2.41; 95% CI, 1.84-3.15), lifetime substance use disorder (23.0%; OR, 2.65; 95% CI, 1.88-3.74), lifetime PTSD (15.2%; OR, 2.82; 95% CI, 2.01-3.95), and any lifetime mental disorder (57.3%; OR, 2.60; 95% CI, 1.99-3.40). The association was particularly strong for exposure to 3 to 4 types of GBV (lifetime anxiety disorders: weighted, 77.3%; OR, 10.06; 95% CI, 5.85-17.30; lifetime mood disorder: 52.5%; OR, 3.59; 95% CI, 2.31-5.60; lifetime substance use disorder: 47.1%; OR, 5.61; 95% CI, 3.46-9.10; lifetime PTSD: 56.2%; OR, 15.90; 95% CI, 8.32-30.20; and any lifetime mental disorder: 89.4%; OR, 11.00; 95% CI, 5.46-22.17). Women exposed to a single GBV reported increased rates of suicide attempts (weighted, 6.6%; OR, 3.39; 95% CI, 1.89-6.05), with the highest rate of suicide attempts for those who reported exposure to 3 to 4 types of GBV (34.7%; OR, 14.80; 95% CI, 6.89-31.60).

Table 4 shows the associations between mental health disorder and (1) any GBV, (2) childhood nonsexual trauma without GBV, and (3) any other forms of PTE alone. The adjusted multivariate analysis for any GBV yielded ORs that ranged from 3.60 (95% CI, 2.68-4.84) for a lifetime mood disorder to 4.44 (95% CI, 3.37-5.85) for any lifetime mental health disorder. For the any other PTE alone category, the associations were an OR of 1.76 (95% CI, 1.40-2.20) for any lifetime mental health disorder, 1.87 (95% CI, 1.42-2.46) for any lifetime mood disorder, and 1.78 (95% CI, 1.33-2.37) for lifetime anxiety disorder. After adjusting for confounding variables, the association between childhood nonsexual trauma showed only nonsignificant trends with mental health disorders (any lifetime mental disorder: OR, 1.43; 95% CI, 0.93-2.18; lifetime mood disorder: OR, 1.55; 95% CI, 0.99-2.40; lifetime anxiety disorder: OR, 1.41; 95% CI, 0.89-2.24; and lifetime substance use disorder: OR, 1.63; 95% CI, 0.97-2.74).

Table 5 shows the results of the cumulative logistic regression analyses examining the associations between GBV and indices of severity and comorbidity of mental disorder based on the DSM-IV, as well as general health and psychosocial functioning. Gender-based violence was associated with a higher prevalence of severe current mental disorders based on the Australian DSM-IV severity measure (OR, 4.60; 95% CI, 2.93-7.22), 3 or more lifetime DSM-IV disorders (OR, 7.79; 95% CI, 6.10-9.95), 2 or more disorders from the same diagnostic class (OR, 3.07; 95% CI, 2.18-4.33), and 2 or more disorders from different classes (OR, 5.19; 95% CI, 4.12-6.54). Gender-based violence was also associated with poor self-reported physical health (OR, 4.00; 95% CI, 1.82-8.82) and poor self-reported mental health (OR, 7.14; 95% CI, 2.87-17.75). In relation to psychosocial functioning, women who reported experiencing GBV reported lower quality of life (OR, 2.96; 95% CI, 1.60-5.47); more than 7 disability days in the past 30 days (OR, 3.14; 95% CI, 2.43-4.05), and higher levels of overall disability based on the WHODAS 2.0¹⁷ (OR, 2.73; 95% CI, 1.99-3.75).

Our study is novel in that it incorporates a range of the key types of GBV and offers a comprehensive account of the mental health, general health, and psychosocial correlates of that form of abuse in a nationally representative sample of women. It focuses not only on the prevalence of common mental disorders, but also on their severity and comorbidity. In addition, it reveals a pattern of social disadvantage, disability, and impaired quality of life among women who have experienced GBV. In general, the study highlights the intersection of a key area of human rights concern—violence against women¹⁹—and a complex public health problem that may contribute to mental health disability worldwide.²⁰

Our findings support the view that women who experience 1 form of GBV may be more likely to experience other forms over the course of their lives.²¹ Gender-based violence was associated with all 3 broad classes (mood, anxiety, and substance use disorders) of lifetime mental health disorders with higher rates of disorder in those women experiencing the greatest exposure. In particular, the prevalence of PTSD increased by a factor of 2, according to the level of exposure to GBV. Women who had experienced GBV reported a higher level of severity and comorbidity of mental disorder, increased rates of physical disorders, greater mental health–related dysfunction, general disability, and impaired quality of life. In addition, women who had experienced GBV reported higher rates of past suicide attempts. Therefore, our study offers a comprehensive picture of the extent of impairment in key domains of health and social functioning associated with GBV among women.

This study should be interpreted in light of its limitations. The cross-sectional design of the study precludes any inferences being drawn about possible causal relationships involving GBV and mental health. Further research will be needed to map the chronological sequencing of GBV exposure and the onset and course of mental disorder. Because women can be exposed sequentially to different types of GBV over their lifespan and women who are affected are likely to accrue multiple forms of mental disorder, the relationship between the 2 domains is likely to be complex. As a severe form of trauma, GBV may predispose to mental disorder; conversely, mental disorder may increase the vulnerability of women to incurring further GBV. Social and interpersonal factors are likely to make a contribution. For example, there is evidence that male perpetrators commonly target vulnerable women.²² In addition, genetic and neurohormonal factors may be relevant to women's psychological vulnerability to trauma, a disposition that may be compounded by epigenetic changes arising from maltreatment, especially in early development.²³ Only longitudinal studies will be able to trace the chronological sequencing of these relationships, although even these designs may not be able to overcome all the methodological challenges, particularly in relation to the accurate recording of early forms of abuse against children at a population level.

Other strengths and limitations of our study need to be considered. Our study was based on a large, nationally representative sample of women in a high-income country. Our study excluded the homeless, those individuals residing in institutions, and the severely mentally ill, all of whom may experience higher rates of GBV.¹² In addition, populations living in remote and rural areas were underrepresented. The response rate (60% overall and 65% imputed for women) is consistent with recent trends for lower participation rate in studies of this kind, and falls within the mid-range of comparable mental health surveys worldwide.¹² The WMH-CIDI 3.0 has been used in several national surveys worldwide and diagnoses have been validated against other measures.²⁴ In addition, we controlled for a range of variables such as social support and socioeconomic status that could influence risk of both mental disorders and GBV.

There are substantial differences across studies in the types of abuse encompassed under the rubric of GBV and its subcategories.^25-27 In our study, the measurement of IPV was restricted to severe physical abuse only and did not include psychological abuse. That form of trauma, although considered to be a significant contributor to psychological distress,²⁵ is difficult to measure accurately using a brief set of items that can be included in a general mental health survey of the type reported herein.^3,12 In addition, extant research does not indicate a clear pattern of gender difference in the prevalence of psychological abuse as it does for other forms of GBV.²⁵

A general concern in the field is that GBV may be underreported²⁸; if so, this would have had the effect of underestimating the associations reported in our study. Concerns have been raised that individuals with disorders such as PTSD may overreport past trauma, although data are mixed.²⁹ In relation to the motivation to report adversity, our study involved a non–help-seeking sample interviewed anonymously by lay interviewers on 1 occasion only, with respondents gaining no overt benefit from reporting mental disorders or GBV exposure. The lower rate of GBV reported by older women in our sample might reflect the attenuation of memories over time or reticence to report these events in that age group.

An important question raised by our research is whether existing services address the multiple domains of disability and dysfunction associated with GBV identified by the data.^30-33 Gender-based violence services (whether for IPV, rape, or sexual assault) have tended to be established separately from mental health services and vice versa, a trend that may help to reduce stigma but that also may restrict ready access to a full array of mental health interventions. Our data underline the observation that mental health disorder in women who have experienced GBV tends to be more severe and associated with comorbidity, characteristics that require expert and comprehensive approaches to treatment.³⁴ Therefore, there is a need to ensure that expert mental health care is a central component of GBV programs.

Similarly, psychiatric services need to be better equipped to assist women with mental health disorders who have experienced GBV. For example, it may be problematic for women who have experienced GBV to attend some mixed-gender services, such as group therapies and in-patient care. Personnel need training in strategies to engage and interview women in a gender-sensitive manner, to build trust, and to ensure safety, privacy, and confidentiality in all interactions. In addition, existing treatment programs may need to be modified to incorporate strategies that explicitly focus on GBV. For example, cognitive behavioral treatments that traditionally deal with memories of past traumas in the treatment of PTSD³⁴ may need to focus more on problem solving strategies aimed at preventing future exposure. Ultimately, GBV survivors with severe and comorbid mental health disorders will need treatment plans that are multidimensional and that involve professionals from an array of disciplines to address all their needs. For the same reasons, intervention research requires the formation of multidisciplinary teams of experts, drawing for example from the fields of social science, human rights, public health, and mental health.

The need for a public health focus on changing societal attitudes toward women is a vital component of any preventive strategy.²⁸ Whereas public health campaigns have tended to target specific forms of GBV, such as the prevention of IPV, our study underscores the need to alter attitudes and mores that sanction violence against women at a more general level. Education campaigns to improve attitudes toward women, to promote sex equality, and to convey the health and social concomitants of these abuses are required to achieve progress at a society-wide level in this area.²⁸

Future studies need to assess whether the association between GBV and mental health disorder involves women in low- and middle-income countries to the same extent as in the economically developed countries in which this field of research has largely been pursued. Epidemiological studies should investigate these associations in a more diverse range of societies. In general, renewed impetus is needed to develop an evidence base for prevention and clinical interventions that reflect a comprehensive view of GBV and its association with wide-ranging mental and psychosocial disabilities among women.

Corresponding Author: Susan Rees, PhD, School of Psychiatry, University of New South Wales, and Center for Population Mental Health Research, Level 1 Mental Health Center, Liverpool Hospital, Sydney South West Area Health Network, Sydney, New South Wales, Australia 2170 (s.j.rees@unsw.edu.au).

Author Contributions: Dr Rees had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Rees, Silove, Steel, Creamer, Bryant, Mills, O’Donnell.

Acquisition of data: Creamer, Teesson, Bryant, Slade.

Analysis and interpretation of data: Rees, Silove, Chey, Ivancic, Steel, McFarlane, Carragher, Forbes.

Drafting of the manuscript: Rees, Silove, Ivancic, Steel.

Critical revision of the manuscript for important intellectual content: Rees, Silove, Chey, Steel, Creamer, Teesson, Bryant, McFarlane, Mills, Slade, Carragher, O’Donnell, Forbes.

Statistical analysis: Chey, Ivancic, Steel.

Obtained funding: Silove, Creamer, Teesson, Bryant.

Administrative, technical, or material support: Silove, Ivancic, Creamer, Bryant, McFarlane.

Study supervision: Rees, Silove, Bryant.

Funding/Support: This work was supported in part by grant funding from the National Health and Medical Research Council of Australia.

Role of the Sponsor: The funding organization had no role in the design and conduct of the study, in the collection, analysis, and interpretation of the data, or in the preparation, review, or approval of the manuscript.