Page 99 - Journal of Special Operations Medicine - Spring 2014

Prevalence of Posttraumatic Stress Disorder
in Special Operations Forces

Daniel J. Neller, PsyD, ABPP (Forensic); Jimmie J. Butcher, PhD

ing, Cabrera, Barstow, and Forsten conducted an the Hing et al. sample does not adequately represent the
Himportant study that they state “determines the in- greater SOF population.
cidence of PTSD [posttraumatic stress disorder] symp-
toms” in Special Operations Forces (SOF) Soldiers Second, the Hing et al. study results rest largely on an
assigned to the U.S. Army Special Operations Com- instrument that has questionable validity when used in
mand (USASOC) at Ft. Bragg. Based on results of an anonymous, online surveys of military personnel. At
1
anonymous, online survey, they conclude, “Our study least some findings indicate that, compared with clinical
focused on SOF Soldiers, and suggests that for this rep- interviews, self-report questionnaires such as the mili-
2
resentative sample, conservative estimates for rates of tary version of the PTSD Checklist (PCL) may overes-
PTSD range from 16% to 20%.” We commend Hing timate PTSD rates in Servicemembers. Indeed, research
2
and his colleagues for their work; however, it is our po- to date suggests that Servicemembers endorse a greater
sition that their conclusions overstate the actual results number of items on the PCL when the instrument is
of their study. used in anonymous surveys than when used as part of
4
on-the-record PTSD screenings or more in-depth inter-
5
First, the Hing et al. sample does not seem to adequately views. The PCL also lacks indicators of underreporting
represent the greater SOF population. It apparently was and overreporting, calling into question the veracity of
derived from a nonrandom procedure, and no more symptoms endorsed by survey participants. Taken to-
than 5% of all SOF Soldiers invited to participate in gether, these findings raise questions about the validity
the study actually completed the online survey instru- of PCL scores when used in anonymous, online surveys
ment. When considered alone, nonrandom selection of service members, such as the one conducted by Hing
procedures or exceedingly low response rates raise se- and his colleagues.
rious questions about sample representativeness. When
considered jointly, they strongly suggest that a sample Third, even if the points raised here are erroneous, it
differs from its greater population in important ways. nevertheless appears that Hing and his colleagues mis-
The Hing et al. sample’s poor representativeness is evi- estimated their sample’s base rate (BR) of PTSD.* This
dent by the proportions of Soldiers drawn from various is because they interpreted all positive scores on the PCL
units. The 95th Civil Affairs Brigade, USASOC and U.S. as true cases of PTSD even though the PCL, like all as-
Army Special Forces Command Headquarters, and Spe- sessment tools in the behavioral sciences, is an imperfect
cial Warfare Center and School made up approximately classification instrument. The recommended cutoff score
35%, 29%, and 5% of the Hing et al. sample, respec- from the initial validation sample was associated with a
tively. By comparison, they make up only about 5%, true-positive rate (TPR) of 0.82 and a false-positive rate
3%, and 19% of the entire Army SOF population. These (FPR) of 0.17. This combination of TPR and FPR es-
sample-population differences are both substantial and timates is not highly discrepant from estimates derived
6,7
significant (respective z-scores = –28.2, –32.1, 7.2; all from subsequent studies. If we use these estimates in
p values <.0001) and provide compelling evidence that a setting where the PTSD BR is as high as 20% and

Notes: *Hing and colleagues distinguish prevalence rates (or BRs) from incidence rates. Generally, BRs refer to the proportion of
a sample or population that has the condition of interest during a specified period of time. Incidence rates refer to the proportion
of new cases of the condition of interest in a sample or population at risk during a specified period of time. We do not assume
that identified cases are new, and therefore, we consider BRs to be the more appropriate descriptor.
†Using the TVS to estimate local BR is, of course, subject to reliability restrictions of the procedure and standard errors of pro-
portion for the TPR and FPR estimates. A TVS spreadsheet is accessible at http://www.richardfrederick.com.

91