Target condition being diagnosed
After a traumatic event, some survivors will develop a psychiatric disorder such as post-traumatic stress disorder (PTSD). PTSD is characterised by symptoms of re-experiencing of the traumatic event, avoidance of thoughts and behaviours related to the traumatic event, emotional numbing and hyperarousal. In the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV), acute PTSD is diagnosed if symptoms are present for at least one month, and chronic PTSD is diagnosed if symptoms persist for three months or longer. In addition, the disturbance should cause clinically significant distress or functional impairment (American Psychiatric Association 1994). PTSD is associated with substantial health care and economic costs (Walker 2003; Chan 2009).
The risk of development of PTSD after trauma ranges from about 6% in accident victims to 21% in assault victims (Kessler 1995). Although incidence rates vary between populations and samples studied, longitudinal studies (e.g. O'Donnell 2003) and epidemiological surveys (e.g. Breslau 2009) generally show patterns of decreased prevalence during the first year after trauma. However, some trauma survivors do not show a decrease during the first year, and a minority experience delayed onset of PTSD, meaning that at least six months has passed between the trauma and the onset of symptoms (American Psychiatric Association 1994, p. 465).
The presence of PTSD is usually established with a comprehensive diagnostic interview such as the Clinician-Administered PTSD Scale (CAPS; Blake 1995), the Structured Clinical Interview for DSM-IV Axis I Diagnosis-Patient Edition; First 1996), the Structured Interview for PTSD (SI-PTSD; Davidson 1997) and the M.I.N.I. Neuropsychiatric Interview (Sheehan 1998), or with a full assessment of PTSD symptoms by a clinician. Interviews consist of semi-structured questions based on symptoms in DSM-IV or the tenth revision of the International Classification of Diseases (ICD-10); these assessment tools have multiple response categories and are administered face-to-face by mental health care professionals such as psychologists, psychiatrists or trained nurses. Although the administration time of these interviews varies (e.g. about 15 minutes for the PTSD module of the SCID I/P and up to 25 to 45 minutes for the CAPS), in general they are relatively time-consuming.
Efforts to prevent PTSD using brief single-session or multiple-session interventions for all victims involved in the traumatic event have been unsuccessful (see Rose 2009 and Roberts 2009 for reviews). A more beneficial prevention strategy may be to select trauma survivors with a probable clinical diagnosis of acute PTSD using a self-report instrument for further diagnostic procedures, after which they may receive treatment. A recent Cochrane review showed that trauma-focused cognitive behavioural therapy (TFCBT) is effective in the treatment of acute PTSD (Roberts 2010).
Randomised clinical trial evidence suggests that chronic PTSD can be effectively treated with psychological treatments such as TFCBT or eye movement desensitization and reprocessing (EMDR; see Bisson 2007; Bisson 2007a, Powers 2010 for reviews). Pharmacological treatments such as selective serotonin reuptake inhibitors (SSRIs) may also be effective (Stein 2009), but some clinical guidelines and consensus statements (e.g. National 2005, Institute of Medicine 2008) recommend them as a second line of treatment after TFCBT.
To facilitate more rapid identification of trauma survivors with a probable diagnosis of PTSD than is possible with semi-structured interviews, self-report instruments have been developed. Examples of such instruments are the Impact of Event Scale (IES; Horowitz 1979), the PTSD Checklist－Civilian Version (PCL-C; Weathers 1993), the Posttraumatic Stress Symptom Scale－Self-Report Version (PSS-SR; Foa 1993), the Davidson Trauma Scale (Davidson 1997a) and the Trauma Screening Questionnaire (TSQ; Brewin 2002). Although most of these instruments consist of items based on the 17 PTSD symptoms outlined in DSM-IV, shorter and longer questionnaires have been developed. Most instruments provide sum scores for each symptom cluster (re-experiencing, avoidance and hyperarousal) and a total PTSD sum score. Usually, the scores of these self-report instruments are interpreted without additional information, such as details of functional impairment or personal suffering or background characteristics.
In practice, PTSD self-report instruments are administered in specialised care as well as in public health settings. For example, they are administered by psychologists and psychiatrists in treatment settings but also by other professionals or non-professionals involved in the care of individuals exposed to a traumatic event, such as nurses, social workers or management staff in professions at high risk of experiencing traumatic events. In addition, they can be completed at home and returned by mail or administered through the Internet (e.g. Read 2009). In addition to the fact that they save time, an advantage offered by these instruments is that their administration does not require the involvement of trained clinicians. However, self-report instruments may also offer disadvantages in comparison with interviews. Items may not always be understood, or they may be understood differently by different patient groups. For instance, the PTSD DSM-IV C3 criterion "inability to recall an important aspect of the trauma" may be endorsed by many accidental injury victims as the result of unconsciousness during the event. A recent study in soldiers deployed to Afghanistan found that scores obtained with the PSS-SR overestimated true PTSD rates by a factor of about 3.5 (Engelhard 2007a), possibly because symptoms may have been endorsed that stem not from a traumatic event (according to the DSM-IV definition) but rather from another type of stressful experience (e.g. a divorce, a discharge), or symptoms may have been endorsed that were already present before the traumatic event, such as hyperarousal symptoms. In addition, several PTSD symptoms overlap with symptoms of other anxiety disorders or affective disorders, and this may result in inflated scores on self-report measures (Engelhard 2007). Finally, translation versions may not perform as well as original language instruments because of cultural differences or translation problems.
In general, PTSD self-report instruments may be used for two purposes. First, they may be useful as a triage test. Rather than undertaking a clinical interview with all individuals, the self-report instrument is administered first as a selection tool. Only those individuals who achieve a score above a threshold go on to the interview to obtain a diagnosis. If the self-report instrument is sufficiently accurate (sensitive), such a strategy saves resources and costs (Bossuyt 2006). For example, triage may be carried out in the aftermath of mass trauma. In fact, after the 2005 London bombings, survivors were sent a two-page brief questionnaire, which included the TSQ (Brewin 2010). Individuals who screened positive were invited for a more detailed assessment that included the SCID I/P. Other target groups for triage with PTSD self-report instruments may include injured trauma patients in general hospitals (e.g. O'Donnell 2008), victims applying for assistance at victim support agencies (e.g. Dekkers 2009), victims reporting a crime to the police (e.g. Wohlfarth 2003), soldiers returning from deployment in war zones (e.g. Bliese 2008), primary care patients (e.g. Ouimette 2008) and members of the general population (e.g. Terhakopian 2008).
Second, self-report instruments may replace the structured interview. Although, by definition, an index test (the self-report instrument) cannot perform better than the reference standard (the structured interview), replacement of the time-consuming interview with a much simpler self-report instrument with satisfactory accuracy may be worthwhile. For example, this approach may be useful for monitoring treatment outcomes in mental health care settings, or for research purposes.
As has been discussed, a considerable risk for development of PTSD has been noted in trauma-exposed individuals. Several strategies have been proposed for prevention or early treatment of PTSD. It is important to note that brief early psychological interventions for all, such as debriefing, have proved ineffective and in some cases even harmful (Rose 2009), whereas early treatment of PTSD patients using TFCBT has been shown to be an efficacious alternative (see Roberts 2009). Accurate self-report instruments would facilitate the identification of individuals with PTSD before they are referred for treatment.
In recent years, the number of studies evaluating the sensitivity and specificity of early screening questionnaires in identifying trauma survivors with early symptoms of PTSD has grown rapidly. 'Sensitivity' refers to the percentage of individuals with a diagnosis of PTSD who were correctly identified as such with use of the self-report instrument, whereas 'specificity' refers to the percentage of individuals without PTSD who were correctly identified as such with use of the instrument. The purpose for which the test is used determines whether sensitivity or specificity is considered more important. For instance, when a PTSD self-report instrument is to be used as a triage test, it should be very sensitive so that as many true cases as possible can be detected and referred for a more in-depth clinical interview. On the other hand, if a PTSD self-report instrument is used to replace a diagnostic interview, a more balanced trade-off between sensitivity and specificity is required.
The diagnostic accuracy of PTSD self-report instruments has been described in reviews by Brewin 2005 and Connor 2006. Brewin 2005 reviewed 13 separate instruments and found that the sensitivities of these instruments ranged between .60 and 1.00 and specificities between .60 and .99. However, until now, no meta-analysis on the diagnostic accuracy of self-report instruments has been carried out. Therefore, it is not clear whether some instruments are more accurate than others, or whether accuracy of screening instruments varies between groups of trauma-exposed individuals. It is possible that the accuracy of PTSD self-report instruments depends on the type of trauma to which individuals were exposed. Some instruments may better tap into post-traumatic stress reactions after accidental injury, whereas others may be phrased to better reflect combat stress reactions. In addition, the accuracy of self-report instruments may depend on the time between assessment and the traumatic incident. In the first months after trauma, it may be more difficult to distinguish PTSD symptoms from transient stress reactions.
We will carry out a systematic review of studies evaluating the diagnostic accuracy of PTSD self-report instruments (the index test) in relation to the reference standard, which is a clinical DSM or ICD diagnosis of PTSD made with a structured interview. We will include studies of victims of all types of traumatic events, including assault, road traffic accidents and disasters. We will focus on the accuracy of self-report instruments in diagnosing PTSD rather than predicting PTSD. DSM-IV states that symptoms should be present for at least one month before PTSD may be diagnosed; therefore we will include studies in which the self-report instrument(s) and the interview were administered at least one month after the trauma and simultaneously (i.e., within a maximum period of seven days).