Journal of Psychopathology and Behavioral Assessment, Vol. 25, No. 4, December 2003 ( C  2003) Algorithm Versus Cut-Point Derived PTSD in Ex-Prisoners of War Joan M. Cook, 1,2,6 Richard Thompson, 3 James C. Coyne, 1 and Javaid I. Sheikh 4,5 Accepted May 15, 2003 This study compared symptom-based algorithm to cut-point scoring of the PTSD Checklist (PCL) in a sample of 504 World War II ex-prisoners of war not seeking treatment. There was relatively high correspondence (κ = .80) between the scoring methods; the algorithm identified more participants as having PTSD than the standard cut-point of 50. Receiver Operator Curve analyses revealed that a cut-point of 42 distinguished optimally between the algorithm-defined PTSD and non-PTSD groups. An optimal scoring method might use both cut-point and algorithm, ensuring individuals have the symptoms necessary for a diagnosis and of sufficient severity. KEY WORDS: PTSD; PTSD Checklist; psychometrics; prisoners of war. Semistructured interviews, such as the Clinician- Administered PTSD Scale (CAPS; Blake et al., 1995), are the preferred means of assessing PTSD for clini- cal and research purposes. However, reliance on self- report instruments is quite common, both as the sole means of defining a diagnosis of PTSD and in a two- stage process of screening and interviewing of respon- dents scoring positive. Self-report measures correspond- ing to the DSM-IV (American Psychiatric Association, 1994) criteria for PTSD are plentiful (for a review see www.ncptsd.org/publications/assessment) and one of the most widely used is the PTSD Checklist (PCL; Weathers, Litz, Herman, Huska, & Keane, 1993). The PCL’s 17-items, correspond to the DSM-IV symptoms of PTSD (five re-experiencing, seven avoidance/numbing, and five hyperarousal symptoms). Participants indicate the extent of bother by each symptom in the past month from not at all (1) to extremely (5). In samples of Vietnam veterans, the PCL has high internal 1 University of Pennsylvania, Philadelphia, Pennsylvania. 2 Philadelphia VA Medical Center, University and Woodland Avenues, Philadelphia, Pennsylvania. 3 Juvenile Protective Association, Chicago, Illinois. 4 VA Palo Alto Health Care System, Palo Alto, California. 5 Stanford University, Palo Alto, California. 6 To whom correspondence should be addressed at Philadelphia VA Med- ical Center (116A), University and Woodland Avenues, Philadelphia, Pennsylvania 19104; e-mail: cook j@mail.trc.upenn.edu. consistency, test–retest reliability and concurrent validity, and correspondence with a diagnosis of PTSD based on structured interview (Weathers et al., 1993). There are two primary methods of scoring the PCL: cut-point and mapping of diagnostic criteria. The cut-point method involves adding all the items to form a total score, with scores higher than a criterion cut-point suggestive of a PTSD diagnosis (Weathers et al., 1993). The alterna- tive mapping of diagnostic criteria or algorithm method requires a score of at least “moderate” (rating of 3 or more) for at least one criteria “B” (re-experiencing), three “C” (avoidance/numbing) and two “D” (hyperarousal) symptoms. Optimal cut-point for an instrument, in terms of the balance of sensitivity and specificity, will vary with the population being assessed. Studies of male veterans have found that a cut-point of 50 best predicted PTSD diagnoses (Weathers et al., 1993; Forbes, Creamer, & Biddle, 2001). In a sample of predominately motor vehicle accident vic- tims and sexual/physical assault survivors, a cut-point score of 44 is most efficient (Blanchard, Jones-Alexander, Buckley, & Forneris, 1996). Recommendations for can- cer patients range from 35 to 50 (Andrykowski, Cordova, Studts, & Miller, 1998; Smith, Redd, DuHamel, Vickberg, & Ricketts, 1999). Two studies suggest a cut-point of 30 for patients seeking care at medical clinics (Walker, Newman, Dobie, Ciechanowski, & Katon, 2002) and 38 for female veterans (Dobie et al., 2002). 267 0882-2689/03/1200-0267/0 C  2003 Plenum Publishing Corporation