Therapeutic Alliance and the NIMH study on depression by Elkin

Krupnick et al (1994) in reviewing the NIMH study on depression (Elkin et al, 1990), found that when the treatments were pooled, the ratings of alliance were significantly higher in the most improved group compared to the least improved. However when outcome was also considered within treatments, there was only a significant difference for the interpersonal psychotherapy. Also the psychotherapy groups showed higher ratings of alliance at the .05 level but at the .01 level only interpersonal psychotherapy showed a significant difference when compared to placebo plus clinical management. Krupnick suggests that where extreme outcome groups are compared, as in their study, the impact of the alliance on outcome may also depend on the treatment that is used.

Krupnick et al (1994) in accounting for the different finding in their study and that of others in the therapeutic alliance literature, suggest that one reason for this may be that other studies used therapists and clients to rate alliance, whilst they used a rater only scale of clinical vignettes, the Vanderbilt Therapeutic Alliance Scale. Marziali et al (1981) found a divergence between client and therapist ratings where the client assessment of the alliance was the only one predictive of outcome. This finding was supported by a study into eating disorders (Gallop et al, 1994) using therapist and client ratings (using the Working Alliance Inventory) which showed little correlation between staff and patient perceptions of alliance. The patient perception of alliance was the critical factor in outcome (in this case as measured by staying in the program).

Hence these studies indicate a divergence between client and therapist measures of alliance but not, as Krupnick suggests, a divergence in outcome due to rater versus client/therapist ratings. The more important variable would seem to be the rating of the alliance by the client in relation to the outcome.

As mentioned in my first post, a meta-analysis of 24 studies by Horvath and Symonds (1991) found a correlation between alliance and outcome across all types of measure of outcome, which was not a function of type of therapy practised.The assessment of therapeutic alliance was most predictive of treatment outcome based on clients' assessments of alliance, less on ratings by therapists and least on ratings done by observers.

Replies:

Measurement of Alliance, by Chuck Elliott, 2/21/97

| Behavior OnLine Home Page | Disclaimer |

Copyright © 1996-2004 Behavior OnLine, Inc. All rights reserved.