Providing high quality care to Veterans with post-traumatic stress disorder (PTSD) is critically important to the Veterans Health. In this study we employ state-of-the-art psychometric techniques to improve the measurement of health-related quality of life (HRQOL), an important outcome measure, for Veterans with deployment-related PTSD. We build on methods established by the NIH-funded Patient-Reported Outcomes Measurement Information System (PROMIS) and Neuro-QOL (Quality Of Life in Neurological disorders) measurement systems.
1) Identify the important HRQOL domains/constructs (e.g. distress, pain, social role changes) in Veterans with deployment-related PTSD. 2) Develop new item pools and/or adapt item banks from existing NIH funded tools (e.g., PROMIS/Neuro QOL) representing uni-dimensional domains/constructs that address the unique issues facing Veterans with deployment-related PTSD. 3) Field test the newly created or adapted item pools with a sample of Veterans with deployment-related PTSD. 4) Use Item Response Theory (IRT) methods to calibrate item banks measuring the domains/constructs. 5) Create short forms from of the calibrated item banks for use with Veterans with deployment-related PTSD. 6) Conduct Classical Test Theory-based validation study of calibrated short forms to examine their validity and reliability.
The four-year prospective study employed a mixed-methods research design in three phases. In Phase I focus groups of patients and providers were conducted to identify potential HRQOL domains/ constructs related to deployment-related PTSD. Results of the qualitative data guided selection/ development of item pools designed measure the domains/constructs. Subsequently, Veterans completed cognitive interviews of the new item pools providing feedback on conceptual definitions and questionnaire items regarding the language, comprehensibility, ambiguity, and relevance of each item. In Phase II, the new item pools and selected items from existing item banks were field tested in a large sample of Veterans (n >500) with deployment-related PTSD. In Phase III, psychometric analyses employing item response theory (IRT) was completed on data collected in Phase II to create banks and short forms of the domains/constructs. Finally Classical Test Theory psychometric analyses were conducted on a sample of respondents to provide support of validity and reliability of the short forms.
In Phase I a total of 8 Veteran (n=58) and 4 professional (n=23) focus groups were conducted. Based on the results of the focus groups, potential domains/constructs identified included Anger, Violence/Aggression, Avoidance/Isolation, Control Over the Environment, Social Roles, Task Performance, Sleep Interference and Hypervigilance. The Task Performance domain, which emphasized items related to the impact of PTSD on employment, was dropped from field test because of concerns that the targeted cohort would include many retirees and might not support the psychometric analysis of this construct. A small number of items in each pool were re-worded or dropped based on results of cognitive testing with a minimum of 5 respondents per item. A total of 178 new items were field tested along with relevant PROMIS/NeuroQOL, demographic, and the PTSD symptom measure (PCL) items. Total number of new items ranged from 7 for the Anger domain to 39 for the Social Roles domain. Because of the large number of items developed for field testing, data collection for Phase II was separated into two questionnaire waves. Veterans who completed surveys in wave 1 (n= 538), had a mean (SD) age of 57.1 (13.4), were primarily male (n = 489, 91.0%), white (n= 400, 74.0%) with 216 (40.0%) having served in the OEF/OIF/OND conflicts or Gulf War. The respondents in wave 2 (n = 506) were similar with a mean (SD) age of 57.1 (13.7) and were again primarily male (n = 471, 93%) males, white (n = 354, 70%) and included (n= 203, 40.0%) who had served in the OEF/OIF/OND conflicts or Gulf War.
Statistical analyses in Phase III found evidence for creation of psychometrically-sound item banks and short forms for five of domains/constructs. The results fell into one of two broad categories. In most cases almost all of the items selected for the banks or short forms were newly developed in the study, with a small number of existing PROMIS/NeuroQOL found to belong in the bank/short form based on psychometric analyses. This suggests the domain/construct is unique to Veterans with deployment-related PTSD. Less often, a large portion of the items included in the banks and short forms were from existing PROMIS/NeuroQOL measures. This suggests an extension of the existing PROMIS/NeuroQOL domain/construct by the new items for this population, rather than development of a new one. The analyses supports that item banks and short forms for Violence/Aggression (bank n = 28, short form n = 10), Avoidance/Isolation (bank n = 35, short form n = 10), and Hypervigilance/Control Over the Environment (bank n = 33, short form n = 10) represent new domains/constructs. Conversely, the analyses support that the item banks and short forms for Social Roles/Relationships (bank n = 21, short form n = 10), Sleep Disturbance (bank n = 13, short form n = 10), Sleep-related Impairment (bank n = 9, short form n = 9) extend existing PROMIS/NeuroQOL domains/constructs. Differential item analyses supported the use of the measure in Veterans who served in Operations OEF/OIF/OND, the Gulf war and previous conflicts. Psychometric analyses of the short forms provided support for their validity and reliability.
The proposed study fills a significant void by developing valid and reliable outcome measures of HRQOL for individuals with deployment-related PTSD. The new item banks/short forms advance our knowledge of this field in clinical treatment, research and surveillance, improving effectiveness and efficiency of PTSD research designed to evaluate treatment approaches or track outcomes over time. This work complements recently-completed work on HRQOL measures for Veterans with traumatic brain injury (TBI). Together these new or expanded measures provide a framework from which expanded use of HRQOL measure can be used in research and clinical practice. Because the item banks are developed using IRT, they can be included in computer adaptive testing programs thereby minimizing respondent burden while ensuring precise measurement of the targeted domains/constructs.
- Haun JN, Patel NR, French DD, Campbell RR, Bradham DD, Lapcevic WA. Association between health literacy and medical care costs in an integrated healthcare system: a regional population based study. BMC health services research. 2015 Jun 27; 15:249.
- Haun JN. Insight into the digital divide: The relationship between computer and health literacy and patients' use of a digital health resource. Poster session presented at: Health Literacy Research Annual Conference; 2015 Nov 2; Bethesda, MD.
- Dillahunt-Aspillaga CJ, Lind JD, Haun JN, Antinori N, Luther SL. Health-related quality of life and employment concerns among Veterans with PTSD: A qualitative exploration. Poster session presented at: American Congress of Rehabilitation Medicine Annual Meeting; 2015 Oct 1; Dallas, TX.