Working alliance predicts psychotherapy outcome even while controlling for prior symptom improvement

Abstract

Objective: Although the working alliance as been found to be a robust predictor of psychotherapy outcome, critics have questioned the causal status of this effect. Specifically, the effect of the alliance may be confounded with the effect of prior symptom improvement. The objective of the present study was to test this possibility. Method: A large dataset from primary care psychotherapy was used to study relationships between alliance and outcome using piecewise multilevel path analysis. Results: Initial symptom level and symptom change up to session three predicted the alliance at session three. Working alliance significantly predicted symptom change rate from session three to termination, even while controlling for several possible confounds. Conclusions: The alliance predicts outcome over and above the effect of prior symptom improvement, supporting a reciprocal influence model of the relationship between alliance and symptom change.

Obiettivo: sebbene l'alleanza di lavoro sia considerata un valido predittore di esito in psicoterapia, i critici si sono interrogati sul reale valore causale di questa variabile. Nello specifico, l'effetto dell'alleanza potrebbe essere confuso con il miglioramento iniziale dei sintomi principali. L'oggetto di questo studio è quello di valutare questa possibilità.

Metodo: è stato usato un ampio database di psicoterapia di base allo scopo di studiare la relazione tra l'alleanza e l'esito attraverso una multilevel path analysis.

Risultato: il livello dei sintomi iniziali e il cambiamento sintomatologico alla terza seduta sono predittori dell'alleanza in terza seduta. L'alleanza di lavoro predice significativamente il cambiamento sintomatologico dalla terza seduta alla fine del trattamento, se vengono controllate alcune variabili confusive.

Conclusioni: l'alleanza è un predittore dell'esito in aggiunta all'effetto del miglioramento sintomatologico iniziale, a supporto del modello che prevede un'influenza reciproca nella relazione tra l'alleanza e il cambiamento sintomatologico.

Objetivo: Embora a aliança de trabalho se tenha revelado um preditor robusto do resultado da psicoterapia, o estatuto causal deste efeito tem sido questionado. Especificamente, o efeito da aliança pode ser confundido com o efeito da melhoria prévia dos sintomas. O objetivo do presente estudo foi testar esta possibilidade. Método: Foi usada uma extensa base de dados de cuidados primários de psicoterapia para estudar as relações entre a aliança e os resultados, utilizando uma piecewise multilevel path analysis Resultados: O nível inicial de sintomas e a mudança sintomática até à sessão três mostraram-se preditores da aliança na sessão três. A aliança de trabalho emergiu como preditor significativo da taxa de mudança sintomática da sessão três ao final, mesmo depois de controlar vários factores passíveis de influenciar os dados. Conclusões: A aliança prediz o resultado para além do efeito da melhoria sintomática prévia, apoiando um modelo de influência recíproca da relação entre aliança e mudança sintomática.

Ziel: Obwohl die therapeutische Arbeitsallianz als ein robuster Prädiktor für Psychotherapie Ergebnisse identifiziert wurde, haben Kritiker den kausalen Zusammenhang dieses Effekts in Frage gestellt. Im Speziellen könnte der Effekt der Allianz konfundiert sein mit dem Einfluss vorhergegangener Symptomverbesserung. Das Ziel der vorliegenden Studie war, diese Möglichkeit zu überprüfen. Methode: Anhand eines großen Datensatzes aus psychotherapeutischer Primärversorgung wurde der Zusammenhang zwischen Allianz und Outcome mittels abschnittsweiser mehrebenen Pfadanalysen untersucht. Ergebnis: Initiales Symptomlevel und Symptomveränderung bis zur dritten Sitzung sagten die Allianz in der dritten Sitzung vorher. Die Arbeitsbeziehung sagte die Symptomveränderungsrate von der dritten Sitzung bis zur Beendigung vorher, auch wenn für diverse potenziell konfundierende Variablen kontrolliert wurde. Ergebnis: Therapeutische Allianz sagt das Therapieergebnis vorher, auch über die Effekte vorrausgehender Symptomverbesserung hinaus. Dies unterstützt das reziproke Einflussmodell der Beziehung zwischen Allianz und Symptomveränderung.

雖然工作同盟被視為治療成效的重要預測因子，但是仍有許多人質疑二者之間是否有因果關係；具體而言，工作同盟所帶來的效果是否和先前症狀改善的效果混在一起。研究方法：因此，本研究從主要的心理治療龐大資料庫採用多階層徑路分析探究工作同盟及治療成效間的關聯。研究發現：初始的症狀嚴重程度與案主到第三次會談時的症狀改變程度可用來預測第三次會談中的工作同盟。在第三次會談至會談結束的這段期間，在許多干擾因素都被控制的情況下，工作同盟仍可顯著用來預測症狀改變的速度。結論：不管先前的症狀改善程度如何，工作同盟皆可用來預測治療效果，此發現證實了工作同盟與症狀改變間的交互影響模式。

Keywords:

• working alliance
• psychotherapy outcome
• therapy process
• multilevel modeling
• path analysis

Acknowledgments

The authors are indebted to Bruce Wampold for helpful suggestions concerning this study.

Notes

¹ This sample has been used previously in a separate paper in which within-patient effects of the alliance on symptom change to the next session were analyzed (see Falkenström et al., 2013, March 18). Although using the same sample, between- and within-patient analyses are orthogonal to each other, so results from these analyses are statistically independent.

² Conventional multilevel modeling estimates the relationships between random effects as covariances (or correlations). In multilevel path analysis it is possible to specify these relationships as regression paths instead of correlations. Still, it is our impression that most researchers prefer to model them as covariances, so we chose to do so as well. However, we tested our models both ways, i.e., by also specifying regression paths among random effects (Intercept → Early Change, Intercept → Late Change, Early Change → Late Change). The models turned out to be equivalent in terms of model fit, and none of the central parameter estimates of interest differed.

³ It may seem odd that we pursued analyses of WAI subscales, when the principal component analysis yielded only one principal component with eigenvalue larger than one. The reason that we still performed these analyses was that reviewers asked for subscale analyses, and we believe other readers would also wonder if we did not present them.

⁴ The idea of ceiling effects being particularly marked for the Bond subscale was suggested by an anonymous reviewer.