PhD/PsyD Reliable Change Index question

This forum made possible through the generous support of SDN members, donors, and sponsors. Thank you.

erg923

Regional Clinical Officer, Centene Corporation
Account on Hold
15+ Year Member
Joined
Apr 6, 2007
Messages
10,827
Reaction score
5,609
Is there a reliable change index for common measures such as PCL-M, PHQ-9, GAD-7? We use these as follow-up and outcome measures and nobody on my team can answer this for me.

Members don't see this ad.
 
Members don't see this ad :)
True, I do think you need the RCI, but you can calculate that if you have the test-retest reliability. You may have to dig and find the validation studies for said instruments, but they shouldn't be too hard to find.
 
If you just want a general guideline, the National Center for PTSD says

"Good clinical care requires that clinicians monitor patient progress. Evidence for the PCL for DSM-IV suggests that a 5-10 point change represents reliable change (i.e., change not due to chance) and a 10-20 point change represents clinically significant change. Therefore, it was recommended to use 5 points as a minimum threshold for determining whether an individual has responded to treatment and 10 points as a minimum threshold for determining whether the improvement is clinically meaningful using the PCL for DSM-IV."
 
Like WisNeuro mentioned, if you have test-retest data available (e.g., the test-retest correlation coefficient, sample size, and test-retest score means and standard deviations), you can essentially calculate this for any measure you want. You can also decide if you want to use a "standard" RCI equation (e.g., Chelune et al. developed one that built on previous work and incorporated practice effects) or regression-based reliable change formulae (I think it's Crawford who's done a lot of work in this area). You can then use this info to determine how likely it is that a change of +/- XX will occur. If Crawford is the right name, you could google him, as I believe he posts a bunch of his reliable change work on his website.

It can of course get a bit hairy if you're using test-retest data from an interval that's substantially different from the one you're examining (e.g., test-retest data is from a period of 1 month, but you're looking at 1-2 years), but hey, it's still likely better than nothing.
 
Top