Estimating post-traumatic stress disorder in the community: lifetime perspective and the impact of typical traumatic events

Abstract

Background. Community surveys have assessed post-traumatic stress disorder (PTSD) in relation to traumatic events designated by respondents as the worst they have ever experienced. An assessment of PTSD in relation to all reported traumas would impose too great a burden on respondents, a considerable proportion of whom report multiple traumas. The ‘worst event’ method is efficient for identifying persons with PTSD, but may overestimate the conditional probability of PTSD associated with the entire range of PTSD-level traumas. In this report, we evaluate this potential bias.

Method. The Detroit Area Survey of Trauma (n=2181) estimated the PTSD risk from two samples of traumas: (1) a representative sample of traumas formed by selecting a random trauma from each respondent's list of traumas; and (2) traumas designated by respondents as the worst (the standard method).

Results. Both estimation methods converged on key findings, including identifying trauma types with the highest probability of PTSD and sex differences in the risk of PTSD. Compared to the random events, the ‘worst event’ method yielded a moderately higher conditional probability for PTSD (0·136 v. 0·092). The bias was due almost entirely to the deviation of the distribution of the worst events from expected values, if all event types had equal prior selection probabilities. Direct adjustment, setting the distribution equal to expected values and applying the observed probabilities of PTSD associated with individual event types brought the estimate close to the unbiased estimate, based on the randomly selected traumas.

Conclusions. Only the ‘worst event’ method can be used as a short-cut to assessing all traumas. The bias in the estimated risk of PTSD is modest and is attenuated by direct adjustment.