What Predicts Psychotherapy Outcomes Across Disorders and Treatments: an Umbrella Review
Abstract
The objective of assigning patients to their most promising treatment has sparked considerable interest in contemporary research. Many studies develop models to predict treatment outcomes and stratify treatment allocation. Predictive accuracy of such models, however, has remained limited due to a lack of systematic information used for model training, given that variables included in such datasets are often selected due to their availability in routine care datasets. To increase systematic information in the data, the inclusion of promising candidate predictors, based on theoretical or empirical considerations, is crucial. Therefore, this study systematically searches for and summarizes the evidence on patient level variables that predict psychotherapy outcome. Since the literature to answer this question is vast, an umbrella review methodology (i.e., a review of reviews) was implemented. Searches in PubMed, Web of Science and the Cochrane Library with preregistered search strings yielded 5926 articles. Title and abstract screening implemented AI-based methods alongside human review. N = 51 papers were extracted and quantitative evidence was descriptively aggregated. We identified 69 potential predictors. Using post-hoc defined criteria, reliable and relevant predictors were selected. Some of these were known and frequently assessed predictors (e.g., symptom severity, diagnosis) while others are not frequently assessed in routine care. Among the latter were social support, executive functioning, therapeutic alliance and motivation for treatment. The identified variables represent promising candidate predictors, which can inform theoretical models of therapy success, guide clinicians, or, if included in future analysis pipelines, have the potential to increase predictive performance in psychotherapy studies.
Related
- Generalizable AI predicts immunotherapy outcomes across cancers and treatments
- Exercise interventions for clinical and patient important outcomes in adults with substance use disorders: An updated umbrella review
- Digital Mental Health Tools to Support Evidence-Based Treatments in Youth Psychotherapy: A Mixed Methods Study
- Motivational Incongruence Predicts Subsequent Psychological Distress in Psychotherapy: Within-Person Evidence From Three Samples
- An Individual Meta‐Analysis of Psychological Treatments for Mood and Personality Disorders Using the Severity Diagnostic System of ICD‐11
BibTeX
@misc{jacobs2026what,
title = {{What Predicts Psychotherapy Outcomes Across Disorders and Treatments: an Umbrella Review}},
author = {Niklas Jacobs and Tabea Jaleta and Kevin Hilbert},
howpublished = {PsyArXiv},
year = {2026},
month = sep,
doi = {10.31234/osf.io/s49n8_v2},
url = {https://osf.io/preprints/psyarxiv/s49n8_v2/},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - What Predicts Psychotherapy Outcomes Across Disorders and Treatments: an Umbrella Review
AU - Niklas Jacobs
AU - Tabea Jaleta
AU - Kevin Hilbert
PY - 2026
DA - 2026/09/11
PB - PsyArXiv
DO - 10.31234/osf.io/s49n8_v2
UR - https://osf.io/preprints/psyarxiv/s49n8_v2/
AB - The objective of assigning patients to their most promising treatment has sparked considerable interest in contemporary research. Many studies develop models to predict treatment outcomes and stratify treatment allocation. Predictive accuracy of such models, however, has remained limited due to a lack of systematic information used for model training, given that variables included in such datasets are often selected due to their availability in routine care datasets. To increase systematic information in the data, the inclusion of promising candidate predictors, based on theoretical or empirical considerations, is crucial. Therefore, this study systematically searches for and summarizes the evidence on patient level variables that predict psychotherapy outcome. Since the literature to answer this question is vast, an umbrella review methodology (i.e., a review of reviews) was implemented. Searches in PubMed, Web of Science and the Cochrane Library with preregistered search strings yielded 5926 articles. Title and abstract screening implemented AI-based methods alongside human review. N = 51 papers were extracted and quantitative evidence was descriptively aggregated. We identified 69 potential predictors. Using post-hoc defined criteria, reliable and relevant predictors were selected. Some of these were known and frequently assessed predictors (e.g., symptom severity, diagnosis) while others are not frequently assessed in routine care. Among the latter were social support, executive functioning, therapeutic alliance and motivation for treatment. The identified variables represent promising candidate predictors, which can inform theoretical models of therapy success, guide clinicians, or, if included in future analysis pipelines, have the potential to increase predictive performance in psychotherapy studies.
ER -