Abstract
Keywords
Posttraumatic stress disorder was included for the first time in DSM-III [1] in 1980 and the ICD-10 [2] in the early late 1990s and this crystallized clinical observation about the impact of trauma on psychological health. There have been numerous studies of disaster-stricken populations [3], of traumatized groups [4, 5] and of clinical populations [6]. Combat has been the original and continuing generator of research in posttraumatic stress disorder [7]. All these studies produced estimates of the risk of posttraumatic stress disorder of between 9% and 24% of those exposed to trauma [8] but gave little indication of the prevalence of trauma or posttraumatic stress disorder in the wider community.
The development of criterion-based diagnosis of posttraumatic stress disorder in the standard nomenclatures opened it to study in large scale community surveys for the first time in the US (the ECA Study [9], National Comorbidity Study [10] and others [8]) and more recently in other countries. However, even studies with criterionbased diagnoses have yielded differences in prevalence rates due to drift in the criteria between versions of nosologies and also changes in questions about the traumas.
Women have been considered especially susceptible to posttraumatic stress disorder despite lower exposure to trauma [11]. This susceptibility has been found repeatedly in focused and general studies [8, 12] as well as in the ECA [9] and NCS studies [10].
Australia has contributed foundational work on posttraumatic stress disorder in the effect of trauma in natural disasters [3, 13, 14] and important contributions to the development of the concept of posttraumatic stress disorder [15]. Australian studies include the Ash Wednesday bushfires, which found posttraumatic stress disorder in 18% of victims [16], and the Newcastle earthquake studies which estimated posttraumatic stress disorder in 18.3% of those traumatized [3]. More recently, the Australian Survey of Mental Health and Wellbeing, conducted in 1997, offered unprecedented data on the Australian population [17]. This study included questions about trauma and symptoms from each of 10 641 subjects in its Australia-wide sample. This allows an unusual view of the experience of trauma in the population and the consequences of trauma and posttraumatic stress disorder.
Method
The National Survey of Mental Health and Wellbeing was conducted throughout Australia in 1997 [17, 18]. The sample survey comprised 10 641 adult residents (aged 18 or over) from 13 600 dwellings selected in a stratified multistage area sampling of the whole population. The sampling ensured that the sample represented all people aged over 18 and living in private dwellings.
Subjects were interviewed with an expanded version of the Composite International Diagnostic Interview (CIDI) adapted for the study [19]. The study instrument included questions on symptoms of posttraumatic stress disorder in the preceding 12 months and the lifetime experience of trauma. Computer algorithms were applied to all the data to produce diagnoses according to both DSM-IV and ICD-10.
Items relating to trauma (listed in Table 1) were taken from various instruments used in trauma questionnaires. They resemble, but are not identical with, the questions asked in the NCS study [10]. Rather than asked orally, the items were written on a card handed to the subject in order to minimize potential embarrassment and improve disclosure. The list included those traumas implied in the DSM-III-R. The DSMIV widened the implicit list of candidate trauma to include lifethreatening illness. This is represented here only by ‘other events’. As the questions were asked by lay interviewers to a standard format, it was not possible to amplify or discriminate the intensity or personal significance of the trauma.
Lifetime prevalence of trauma
The Australian Bureau of Statistics conducted the data collection and prepared the data in a ‘confidentialised’ format. (Certain categories of data were combined to ensure that no individual could be identified from the data released for analysis.) Age was reported in 5-year subgroups, education was recorded by eight levels of qualification and socioeconomic status of areas was reported in 10 percentiles [20]. Sample weightings were provided so that the data could be extrapolated to the whole population. The present study used sample data only and no weighting to the general population was attempted. These data were analysed using SPSS for univariate and logistic analysis.
Results
Trauma
The experience of trauma is detailed in Table 1. Fifty-seven per cent of the sample has experienced one or more of the candidate trauma. More than half (56%) of those experiencing trauma experienced more than one episode in their lifetime.
Men were more likely to experience most traumas and more episodes of trauma. Women were more likely to be sexually assaulted and to report ‘other’ trauma and trauma that ‘happened to someone close’.
The risk of trauma
The contribution of demographic factors to the odds of experiencing trauma was estimated by a logistic regression in which gender, age, socioeconomic status of the area and education qualification were predictors.
Women were significantly less likely to have experienced trauma (odds ratio (OR) = 0.56, CI = 0.52–0.61, p < 0.01). Living in areas of higher socioeconomic status decreased the risk of trauma (OR = 0.98, CI = 0.97–0.99, p < 0.01), especially of rape (non-parametric correlation = 0.044, p < 0.001) and assault (non-parametric correlation = 0.048, p < 0.001).
Age had no effect on the odds of experiencing trauma (OR = 0.996, CI = 0.98–1.01, p = 0.5), although older people were more likely to experience multiple trauma (corr = 0.021, p = 0.03).
Posttraumatic stress disorder
The prevalence of posttraumatic stress disorder in the 12 months preceding the survey was estimated according to both DSM-IV and ICD-10 criteria. Prevalence in the general population and in those experiencing trauma is shown in Table 2.
Twelve-month prevalence of posttraumatic stress disorder (PTSD)
ICD-10 yielded more than twice as many diagnoses of posttraumatic stress disorder as DSM-IV (Kappa 0.47) reflecting differences in classification and threshold between the CIDI algorithms for the ICD-10 and DSM-IV diagnoses. Women who experienced trauma were significantly more likely to develop posttraumatic stress disorder especially for the ICD diagnosis.
Risks of DSM-IV posttraumatic stress disorder
For those who had experienced trauma, the odds of diagnosis of posttraumatic stress disorder was affected by demographic and socioeconomic factors. A logistic regression of DSM-IV diagnosis of posttraumatic stress disorder showed that, among subjects who had experienced trauma, women were more likely to develop posttraumatic stress disorder after trauma (OR = 1.44, CI = 1.04–1.99, p < 0.05) while greater age (OR = 0.85, CI = 0.80–0.90, p < 0.001), higher socioeconomic status of area (OR = 0.94, CI = 0.88–0.98, p < 0.05) and higher education qualification (OR = 0.85, CI = 0.77–0.94, p < 0.01) decreased the odds.
Trauma and posttraumatic stress disorder
The nature of the trauma and the experience of that trauma will influence the development of posttraumatic stress disorder. For each trauma the conditional risk of posttraumatic stress disorder (i.e. the proportion of people with lifetime experience of that trauma who suffered posttraumatic stress disorder in the past 12 months) is shown in Table 3.
Trauma and posttraumatic stress disorder (PTSD)
Women appear to be more susceptible to posttraumatic stress disorder after most lifetime traumas, except for sexual assaults for which men are more likely to develop posttraumatic stress disorder (despite their much smaller numbers). Sexual traumas appear to be most related to subsequent posttraumatic stress disorder.
Multivariate analysis
Because all factors operate simultaneously, multivariate analysis will make better sense of the data and corrects for the simultaneous impact of a number of factors. Table 4 shows the multivariate logistic regression of the odds of posttraumatic stress disorder.
Multivariate analysis of odds of diagnosis of DSM PTSD in subjects with history of trauma
Younger age, lower socioeconomic status, less education, and multiple traumas significantly increase the odds of posttraumatic stress disorder in subjects who have experienced trauma. Particular traumas stood out in their impact. Combat was the most powerful trauma while sexual assaults are especially strong in their effect. ‘Other’ trauma was also significantly more likely to lead on to posttraumatic stress disorder.
Passage of time from the important trauma significantly reduced the impact of most trauma (although combat experience maintained its impact despite the time that has passed since most war involvements).
The experience of terror or helplessness in the trauma is a criterion of posttraumatic stress disorder and is always associated with posttraumatic stress disorder. Helplessness is the more common reaction. (Of the sample that experienced trauma, 68% felt helpless while 47% experienced terror). While posttraumatic stress disorder was not significantly more likely after helplessness than after terror, experience of both terror and helplessness together made posttraumatic stress disorder twice as likely as either emotion alone (OR = 1.86, CI = (95%)1.34–2.60).
Importantly, the effect of gender disappears when the other factors are taken into account. Sequential entry of factors into the logistic regression can demonstrate that the nature of the trauma subsumes the gender effect on the likelihood of posttraumatic stress disorder. The trauma that women are especially likely to experience appears decisive in the development from trauma to posttraumatic stress disorder.
Discussion
This paper reports the findings from the National Survey of Mental Health and Wellbeing. It concentrates on those people in the sample who have reported lifetime experience of trauma and have suffered posttraumatic stress disorder in the 12 months prior to the survey. Detailed examination has been restricted to the DSM-IV diagnosis of posttraumatic stress disorder as it is more restrictive and is used more commonly in Australia.
Creamer et al. [21] have previously published findings on the population-weighted incidence of trauma and prevalence of posttraumatic stress disorder in the National Survey of Mental Health and Wellbeing data set. The rates calculated in that study differ slightly from the present study (1.33% vs 1.5%) due to sample weighting. Creamer et al. used sample weights to extrapolate to the whole Australian population. In the present study, the data are examined on the 10 641 members of the sample without the weighted calculation to the whole population. Weights have not been applied to these data as the weighting process may bias the results in the multilevel analyses that have been conducted on these data.
The data show a surprisingly high rate of traumatic experience in the sample. The likelihood of lifetime experience of trauma was significantly higher for men and residents of poorer socioeconomic areas. Overall, people with less education are more likely to have experienced trauma. Age had little effect except that older people were a little more likely to report multiple traumas, reflecting their longer exposure to risk.
The 12-month prevalence rates of posttraumatic stress disorder according to ICD-10 and DSM-IV are quite different, indicating significant disagreement. Most of the disagreement is on the threshold or severity required for diagnosis. The ICD-10 yields prevalence of posttraumatic stress disorder more than twice the prevalence by the DSM-IV. This accords with Andrew's findings on comparisons of ICD and DSM [22]. I have concentrated on the DSM-IV for simplicity, comprehensibility and comparability with other studies. It is possible that analysis according to the ICD-10 would yield different results.
Important findings stand out from the simple analysis of the data. However, some of their implications are altered by the multivariate analysis.
Straightforward analysis shows that, following the lifetime experience of trauma, women are nearly one and a half times more likely to develop posttraumatic stress disorder. Younger and less educated people living in poorer areas are more likely to develop posttraumatic stress disorder, suggesting that gender and social conditions influence vulnerability.
However, the nature of the trauma predominated the risk. In men, as well as women, rape and sexual molestation stood out from other trauma as the precursor of posttraumatic stress disorder. This is consistent with other studies [4] which predate the recent public focus on the consequences of sexual molestation. Understandably, terror and helplessness experienced together have a greater impact than either emotion alone.
All these factors are interrelated. This is demonstrated by the multivariate analysis that yields some surprising and important findings. This analysis underlines the traumatic force of combat (albeit in smaller numbers given the long period since most had the experience) and sexual assault. All effects are moderated by time since the trauma and increased when people have experienced multiple trauma.
Gender disappears, almost completely, as the vulnerability factor. It is subsumed by the trauma – rape and molestation – that women are more likely to experience. Older age, more education and living in a higher socioeconomic area reduce the risk either by reducing exposure to particular types of trauma, or, more controversially, vulnerability.
These results resemble, but also show important differences from, the US National Comorbidity Survey (NCS) [10] which used similar sampling strategies and instruments. In that survey, 56% of the population reported lifetime experience of at least one of the candidate trauma, compared with 57% in the Australian sample. Sexual trauma in women was notably higher in the NCS. Rates of posttraumatic stress disorder cannot be compared because the NCS study estimated ‘lifetime’ rates (versus the Australian estimate of last 12 months) and used DSM-III-R criteria that are slightly different. Lifetime rates of 5% in men and 10.4% in women in the NCS are well above Australian 12-month rates and there is no reliable means of comparing these rates. In the NCS study, the susceptibility of women to posttraumatic stress disorder stood as an unchallenged finding. However, the frequency of rape in women and the strong correlation between rape and posttraumatic stress disorder suggest similar forces at work.
In the broader posttraumatic stress disorder literature, the special sensitivity of women remains undecided [11]. The Detroit Area survey of trauma [12] controlled for trauma type but still found women more liable than men to posttraumatic stress disorder. However, for this Australian sample the susceptibility of women to posttraumatic stress disorder is a consequence of the nature of the trauma that they are more likely to suffer.
This study has the strengths that come from a broad based community study but shares the limitations of such studies. Posttraumatic stress disorder was not the focus of the survey and the data were scanty about the trauma and duration of symptoms. Errors or omission of recall of events and ‘effort after meaning’ cannot be corrected. Application of diagnostic criteria was automated by computer algorithm and lacked clinical subtlety. Overall, the sensitivity of symptom detection and diagnosis is likely to be low. These data should be read as an amplification and balance to the fine-grained data that come from a more clinical focus.
Footnotes
Acknowledgements
I thank Ailsa Korten and Rick Marshall for their help.