I am considering simplifying the Impact of Events Scale, a scale used widely in making assessments of Post Traumatic Stress Disorder. It is my expectation that PTSD symptoms occur in clusters and are more or less hierarchically related. I do not have enough data to begin adequate modeling. I think that it is possible to do sample-free item or symptom calibration and small and or mixed data sets would be adequate for the initial scale reconstruction.
Collaborators are welcome, particularly if you can contribute deidentified item level data from those who have PTSD (Civilian or combat) or those with other acute stress or anxiety conditions.
Some people criticize using the IES to detect PTSD. Is it your best choice to do this?
Thank you Professor fielding and Dr. Kratzer for your comments.
It appears that this investigation would be considered exempt research not requiring or even eligible for review by review boards for the protection of human subjects in the United States. The principle issue that makes this investigation exempt is that the data from living subjects would not be identifiable by the researcher and therefore unlikely to cause harm to any unknowing participant. Notwithstanding that regulatory provision, data from multiple sources would further obfiscuate its origin as to a particular research subject. If the data were used contrary to any informed consent, then I think that the consent should take precedence and the data should not be used. Informed consent would appear not to be required under the United States code of federal regulations governing human subjects, however. (see 45 CFR 46 102(f)(2). See the charts at the following:
http://www.hhs.gov/ohrp/regulations-and-policy/decision-charts/#c1
If you have other ethical concerns, please let me know so that appropriate resolution can be made.
Either or both the IES or IES-R would be useful in this very preliminary investigation. The issue is whether there is a hierarchical dependence on the order in which symptoms are either obtained or given up. I really want to know whether the probability of having a particular symptom is defied by the level or severity of PTSD. If the data can be described by a Rasch model, then many other possibilities exist in terms of finely tracking progress of therapy. Additionally, one can see what symptoms are unexpectedly present when they would not be likely or which symptoms are absent when they would be expected to be present. The underlying estimate of Rasch parameters (person and item) are statistically independent of one another so that sampling at this level is not really an issue. Missing data is also easily accommodated unlike models that rely on co-variance structures such as factor analysis. This is a "shot in the dark" based on the assumption that all symptoms ar not equivalent. This investigation will not establish and test a developmental sequence, but will, if successful, identify whether it is reasonable to look further an test developmental hypotheses. I don't think that another calculation of reliability estimates or a factor analysis will yiield particularly useful new knowledge. Can we engage in a fruitful exploration with available resources?
I am not a strong proponent of the IES or IES-R, though I have used the latter for many years as a tracking instrument rather than a screening instrument. I only chose it for this project because it is in widespread use and may contain necessary data to begin the exploration I wish to pursue. In clinical practice, I much prefer the CAPS-5 (or older CAPS-IV.) .
- Badges
- Science topic
- Similar topics
- Psychology