Page 191 - PhD GT

change are repeated using criteria for statistically reliable and clinically significant change.
9.1.2 Statistically reliable and clinically significant change in dependence
The use of change scores has been criticised for failing to take account of the difference between individuals in the scores with which they start out and finish up and the reliability of the instruments used to measure the condition which is being investigated. (This debate is further discussed in Chapter 10). Addressing this criticism, Jacobson et al. (1999) proposed two criteria for asserting that clinically significant change, based upon the use of change scores, could be said to have occurred. These were: i) that the magnitude of change had to be statistically reliable and ii) that individuals end up in a range that renders them indistinguishable from well-functioning people. The calculation of statistically reliable change takes account of the reliability of the measurement instrument and thus avoids the pitfall of interpreting measurement error as change.
Jacobson and Truax (1991) proposed the following formula for the calculation of statistically reliable change (RC):
RC=x2 –x1 /Sdiff
and state that a reliable change score greater than 1.96 is unlikely to occur (p<.05) without actual change.
Using the above formula, x1 represents the t1 total dependence (in this case) score, x2 represents the t2 total dependence score and Sdiff is the standard error of the difference between two test scores. Sdiff is computed from the standard error of measurement according to the formula:
and SE is calculated as:
Sdiff = √2(SE)2
SE = S1√1-rxx
where rxx is the test re-test reliability of the measure. This was calculated at 0.95 in the validation study of the LDQ (see Raistrick et al. 1994). S1 is the standard deviation for the pretreatment experimental group or in the present case, for the mean total dependence score in Sample 6c at t1 and is calculated at 7.6. Thus the standard error of measurement SE is calculated to be 1.7 and the standard error of the difference between the two test scores (Sdiff) is 2.4. A reliable change score was then calculated for each participant using this formula.
Jacobson et al. (1999) propose three methods for fulfilling the second of their criteria for clinically significant change, which have varying degrees of stringency. These methods are briefly described below and further discussed in Chapter 10. In the three methods, cut-off points are proposed for asserting that change from the dysfunctional to the functional has occurred. In the first
186