Abstract
Background: Measurement is critical in postnatal posttraumatic stress disorder (PTSD) because symptoms may be influenced by normal postnatal phenomena such as physiological changes and fatigue. Objective: This study examined: (1) whether hyperarousal symptoms differ between women who have traumatic or nontraumatic births; (2) whether the construct of hyperarousal is coherent in postnatal women; and (3) whether hyperarousal symptoms are useful for identifying women who have traumatic births or PTSD. Methods: A survey of PTSD symptoms in 1,078 women recruited via the community or Internet who completed an online or paper questionnaire measuring childbirth-related PTSD symptoms between 1 and 36 months after birth. Women who had a traumatic birth as defined by DSM-IV criterion A (n = 458) were compared with women who did not have a traumatic birth (n = 591). Results: A one-factor dimension of hyperarousal was identified that included all five hyperarousal items. Diagnostic criteria of two or more hyperarousal symptoms in the previous week were reported by 75.3% of women with traumatic birth and 50.5% of women with nontraumatic births. The difference in mean hyperarousal symptoms between groups was substantial at 0.76 of a standard deviation (Hedge’s g, CI = 0.64, 0.89). A larger difference was observed between women with and without diagnostic PTSD (g = 1.64, CI 1.46, 1.81). However, receiver operating characteristic analyses showed hyperarousal symptoms have poor specificity and alternative ways of calculating symptoms did not improve this. Comparison with other PTSD symptoms found re-experiencing symptoms were most accurate at identifying women with traumatic births. Conclusions: Results suggest hyperarousal symptoms are associated with traumatic birth and are a coherent construct in postnatal women. However, they have poor specificity and should only be used as part of diagnostic criteria, not as a sole indicator.
Keywords:
Notes
1. Percentages based on women recruited from the community. For more information see Ayers et al. (Citation2009).
2. One community study included in the previous analyses reported by Ayers et al. (Citation2009) is not included here because it did not include a measure of criterion A.
3. Comparison of this graded model with three other models (a rating scale model, generalised rating scale, and generalised partial credit model) showed it was a significantly better fit of the data (AIC 10270.9, BIC 10370.56, χ2 > 36.7, all p < .001).