Objective: The association between alliance and therapy outcome is one of the most investigated factors in psychotherapy research. However, even studies using advanced methods estimate effects over a specific time period (interval) between measurement occasions. Thus, it remains unknown how the magnitude and direction of effects depend on the considered time interval, resulting in limited comparability across studies. The current study examines the influence of time on the within-person relationship between alliance and symptom severity. Method: Alliance (WAI-SR) and symptom severity (SCL-K11) were assessed every fifth session in N = 650 patients receiving up to 100 weekly sessions (mode = 55; M = 41.03; SD = 27.23) of individual psychotherapy in a German outpatient clinic. Bivariate continuous-time (CT) structural equation modeling (SEM) was used to estimate within-person cross-and auto-effects. Results: Analysis revealed significant reciprocal within-person cross-effects with stronger relative effects of SCL-K11 on WAI-SR (a 21 ) than vice versa (a 12 ). CT analysis showed that both cross-lagged effects increased for longer time intervals with strongest effects for time intervals of about 40 sessions (a 21 = −.47; a 12 = −.19). Conclusions: Alliance and symptom severity showed a reciprocal relationship. Expanding current evidence, our analysis showed how the magnitude of these effects depends on the considered time interval. Applying CT-SEM on longitudinal data of the alliance outcome association complements current crosslagged panel analysis and allows to compare results of studies which are based on different time intervals between measurement occasions. Methodological, theoretical, and clinical implications are discussed.
Public Significance StatementThis study shows that alliance predicts subsequent therapy success and vice versa in patients with longterm therapy. Moreover, our study suggests that these associations become stronger over the course of about 40 weekly sessions in patients with long-term treatment.