Abstract
Objective: The therapeutic alliance is a well-established predictor of psychotherapy outcome, yet much research has shown that therapists' and patients' views of the alliance can diverge substantially. Therapists systematically underestimate their patients' perceived level of alliance, and the correlation between therapist and patient estimates of patient alliance is only moderate. The present study explored the divergence between therapists' and patients' perspectives on patients' alliance experience, and its relations to therapists' concurrent work involvement and session process experiences. Method: Data from 98 patients treated by 26 therapists with psychodynamic psychotherapy were analyzed. Results: Therapist-patient divergence was significantly related to therapists' case-wise work involvement, but not to therapist's views of session process. The best predictor of therapist-patient divergence was therapists experiencing a “distressed practice” work involvement pattern. Conclusion: Although therapists' work involvement experiences are not commonly investigated, they can be a relevant predictor of therapy processes.
Funding
This work was partially supported by the Heigl Stiftung, Düsseldorf.
Notes
1 The meta-analysis of Tryon et al. (Citation2007) did not take this difference into account and integrated studies using all kinds of alliance measures and perspectives.
2 Fitting a spline instead of a linear regression line is very common (two-dimensional data). The line of the spline can be set to be stiff or flexible. Response surface models are an extension of this method: they don't fit a “rope” on the floor but rather a “flying carpet” into the air. An inflexibly stiff fit of three-dimensional data generates a flat “shelf” floating in space and, at the other extreme, a fully flexible fit generates a “fluttering flag” in the wind of noisy data. The flexibility of the spline cannot be fitted. Its grade is chosen by the data analyst, and all significance tests thus depend on the analyst's intuition and ability to select a reasonable spline.
3 Given that Wald's Z is a two-sided test but variances can never be lower than zero, it is justified to use it as a one-sided test and to divide the significance level probability by two.