Development and Validation of a PTSD-Related Functional Impairment Scale

In the first phase of this project, we will convene a series of focus groups, whose members will be a convenience sample (n = 53) selected from veterans at the Boston VAMC. These groups will be used to set forth formal definitions of the PTSD-related functional impairment variables, create an item pool, and refine that pool with emphasis on standards of content validity, both content relevance and content breadth.

Using what in psychometrics is called the rational method to instrument construction (Edwards, 1970; Hulin, Drasgow, \& Parsons, 1983; Nunnally, 1978), the first step will be to carefully delineate complete and clearly expressed written definitions of the PTSD-related dimensions of functioning and quality of life. Another important source of information will be veterans of various conflicts with military-related PTSD in four focus group sessions. The intent will be to pose a series of structured questions aimed at eliciting group members' functional impairments and quality of life issues related to their PTSD. Next, a table of specifications (Aiken, 1994) will be employed to aid in the orderly construction of items across content areas. In this regard, for each of the functional impairment dimensions, separate aspects of the definition will be identified, and items will be written so as to systematically represent these aspects.

The end result of these steps should be an initial pool of 30 to 35 items for each construct. Consistent with classic psychometric guidelines (e.g., Nunnally, 1978), this number of items should be adequate to produce individual scales with 15 to 20 items each, which, in turn (if the rational approach described above has been faithfully followed), should be a sufficient number to exhibit acceptable psychometric characteristics.

Phase 2 has the objective of selecting the best final sets of items to measure the identified dimensions of PTSD-related functional impairment as well as to assess criterion-related validity. Using empirically derived item and scale characteristics, we will trim the item pool to arrive at smaller, higher-quality, and more parsimonious sets. For all scales, content relevance and content breadth, in line with the formal definition of each factor, will be maintained.

Data collection and additional instrumentation. The test development sample will be selected in accordance with the sampling plan. Two hundred and eighty six consenting veterans at the Boston VAMC will participate in interviews, each lasting approximately two hours. Eighty-five percent of participants will be male and 15% will be female. We aim to collect data from veterans who meet criteria for current PTSD (one-third of sample) as well as those who meet criteria for lifetime but not current PTSD (one-third of sample) and those who have been traumatized but do not meet criteria for current or lifetime PTSD (no PTSD; one-third of sample). Further, we will collect data on a diverse sample of veterans, with representation from Vietnam, Persian Gulf, and OEF/OIF conflicts.

Historical/demographic information will be obtained during the interview as well as through chart review concerning the following variables: age, race, socioeconomic status, current annual income, employment, education, and marital status. In addition, historical data regarding illegal behaviors, suicide attempts, psychiatric hospitalizations, prescription medications, interpersonal conflicts, and service connection disability status will be collected. We will also obtain pertinent information from veterans' DD-214, including declaration of combat theatre duty, military occupational specialty, and combat medals awarded.

PTSD diagnosis will be assessed using the Clinician-Administered PTSD Scale for DSM-IV (CAPS). The CAPS was designed to improve the reliability and validity of PTSD assessment. Since its development, the CAPS has become the gold standard for arriving at a PTSD diagnosis, either current or lifetime, and has been used in more than 200 published studies.

Functional impairment will be assessed using several established measures. The Medical Outcomes Study 36-item Short Form Health Survey-veterans version (SF-36V; Ware \& Sherbourne, 1992) provides eight domain scores indexing physical functioning, physical role, bodily pain, general health, vitality, social functioning, emotional role, and mental health; in addition, summary physical and mental health scores may be computed. The reliability and validity of the SF-36 are well documented (e.g., McHorney, Ware, \& Raczek, 1993).

The World Health Organization Disability Assessment Schedule II (WHODAS II) is a measure of functional disability which assesses a wide-range of impairment and disability dimensions using multi-item scales including pain, concentration, understanding and communicating, mobility, self care, family burden, getting along with others, household and work activities and work loss, and participation in society (Appendix A:WHODAS). The WHODAS-II is becoming widely used in investigations of functional disability across wide-ranging populations, including the physically ill (i.e. rheumatology, pulmonary, primary care cohorts) and severely mentally ill (schizophrenic cohorts). Although patients with PTSD and related anxiety disorders were included in the original multi-center development of the instrument, the WHODAS-II has not been employed specifically with trauma survivors. For this study, we will use the 36-item self-report version of this scale. Scores for each of the 6 individual domains, as well as an overall score of global functional disability can be calculated.

The Global Assessment of Functioning (GAF) is the standard method for representing a clinician's judgment of a patient's overall level of psychosocial functioning. As such, it is probably the single most widely used method for assessing impairment among patients with psychiatric or substance use disorders or both. The GAF requires a clinician to make an overall judgment about a patient's current psychological, social, and occupational functioning. In DSM-IV-TR, this rating is made on a scale from 1 to 100, with ratings of 1 to 10 indicating severe impairment and ratings of 91 to 100 indicating superior functioning. GAF ratings of impairment are modestly associated with some indexes of social functioning, such as the extent of social networks and the need for support, and with residential instability, lack of employment, and poor work adjustment. In general, however, these relationships are relatively weak; GAF ratings tend to be more closely associated with diagnoses and psychiatric symptoms than with social and occupational functioning.

General life satisfaction, defined as "an individual's evaluation of the degree to which his or her most important needs, goals, and wishes have been fulfilled" (Frisch, Cornell, Villanueva, \& Retzlaff, 1992, p. 93) will be measured using the Quality of Life Inventory (Frisch, 1992), a 16-item instrument employing a Likert-type response format and producing a weighted summative score across multiple facets of life satisfaction (e.g., standard of living, work, home, and relationships with relatives). The psychometric qualities for this measure are strong, with high test-retest and internal consistency reliability and strong validity coefficients for clinical and community samples, including veterans.

To assess malingering, participants will be completing the M-FAST (Miller, 1995). The M-FAST is a brief 25-item screening interview for individuals ages 18 years and older that provides preliminary information regarding the probability that he/she is feigning psychiatric illness. Most malingering and symptom validity instruments assess malingered cognitive and/or neuropsychological deficits. The M-FAST focuses exclusively on malingered psychiatric illness including Depression, Schizophrenia, Hypervigilance, Personality Disorder, Nightmares, and a Suggestibility Interview.

Analytic procedures and interpretation. Classical test theory-oriented item characteristics will be computed (Aiken, 1994; Anastasi, 1982; Nunnally, 1978). For the items that are accompanied by multipoint Likert-type response formats ("strongly disagree" to "strongly agree"), frequency distributions and descriptive statistics first will be calculated. Items having a symmetric response distribution will be preferred over items having a skewed distribution. Finally, corrected item-total correlations, the correlations of each item's score with the sum of scores on all other items measuring that construct, will be computed, where appropriate to the construct. Item-total correlations are indices of an item's ability to distinguish among those high and low on the attribute and thus are indicative of the item's precision of measurement. Items with higher item-total correlations typically take precedence over those with lower item-total correlations. Using judgments of content validity, distributional and endorsement patterns and statistics, and item-total correlations, a final set of items for each construct will be chosen. Estimates of internal consistency reliability (Cronbach's alpha) will then be calculated. At the close of Phase 2, the product will be a refined inventory of PTSD-related functional impairment.

Phase 3 involves the collection of data from a second-stage test sample to support cross validation and criterion-related validity. Refined measures of the previously identified dimensions of PTSD-related functional impairment, along with measures of PTSD and other health outcomes will be administered to an active duty sample and another veteran sample.

Data collection and additional instrumentation. Seventy-five percent of participants will be male and 25% will be female. Participants will be administered the refined and shortened version of the inventory as well as scales intended to assess physical and mental health, health-related quality of life, functional impairment, and social desirability. Data collection is expected to last 60-75 minutes for each participant.

With the exception of the CAPS and MFAST, all of the measures used in Phase 2 will be employed in Phase 3. In place of the CAPS, we will use the PTSD Checklist (PCL; Weathers et al., 1993) to assess PTSD symptoms. The PCL has good sensitivity and specificity and is positively correlated with standard measures of PTSD. In place of the MFAST, we will the SIMS.

Analytic procedures and interpretation. Data from this test validation phase will offer a second opportunity to compute internal consistency reliability (Cronbach's alpha) for the measures of functional impairment. It is important to re-estimate and reexamine internal consistency on a fresh sample since the estimates derived on the same sample from which the items were selected (Phase 2) take advantage of sample-specific covariation and may be somewhat inflated. It is also critical to cross-validate the initial findings with an independent sample in order to ensure that the items which comprise the developed functional impairment scale are optimal.