Heterogeneity in patient-reported outcomes following low-intensity mental health interventions: a multilevel analysis

Shehzad Ali (Lead / Corresponding author), Elizabeth Littlewood, Dean McMillan, Jaime Delgadillo, Alfonso Miranda, Tim Croudace, Simon Gilbody

Research output: Contribution to journalArticle

8 Citations (Scopus)
113 Downloads (Pure)

Abstract

Background

Variability in patient-reported outcomes of psychological treatments has been partly attributed to therapists ? a phenomenon commonly known as therapist effects. Meta-analytic reviews reveal wide variation in therapist-attributable variability in psychotherapy outcomes, with most studies reporting therapist effects in the region of 5% to 10% and some finding minimal to no therapist effects. However, all except one study to date have been conducted in high-intensity or mixed intervention groups; therefore, there is scarcity of evidence on therapist effects in brief low-intensity psychological interventions.

Objective

To examine therapist effects in low-intensity interventions for depression and anxiety in a naturalistic setting.

Data and Analysis

Session-by-session data on patient-reported outcome measures were available for a cohort of 1,376 primary care psychotherapy patients treated by 38 therapists. Outcome measures included PHQ-9 (sensitive to depression) and GAD-7 (sensitive to general anxiety disorder) measures. Three-level hierarchical linear modelling was employed to estimate therapist-attributable proportion of variance in clinical outcomes. Therapist effects were evaluated using the intra-cluster correlation coefficient (ICC) and Bayesian empirical predictions of therapist random effects. Three sensitivity analyses were conducted: 1) using both treatment completers and non-completers; 2) a sub-sample of cases with baseline scores above the conventional clinical thresholds for PHQ-9 and GAD-7; and 3) a two-level model (using patient-level pre- and post-treatment scores nested within therapists).

Results

The ICC estimates for all outcome measures were very small, ranging between 0% and 1.3%, although most were statistically significant. The Bayesian empirical predictions showed that therapist random effects were not statistically significantly different from each other. Between patient variability explained most of the variance in outcomes.

Conclusion

Consistent with the only other study to date in low intensity interventions, evidence was found to suggest minimal to no therapist effects in patient-reported outcomes. This draws attention to the more prominent source of variability which is found at the between-patient level.
Original languageEnglish
Article numbere99658
Number of pages13
JournalPLoS ONE
Volume9
Issue number9
DOIs
Publication statusPublished - 10 Sep 2014

Fingerprint

Multilevel Analysis
mental health
Mental Health
Health
Bayes Theorem
Psychotherapy
psychotherapy
Outcome Assessment (Health Care)
Depression
Psychology
anxiety
Therapeutics
Anxiety Disorders
Primary Health Care
Anxiety
prediction
Patient Reported Outcome Measures
pretreatment

Cite this

Ali, Shehzad ; Littlewood, Elizabeth ; McMillan, Dean ; Delgadillo, Jaime ; Miranda, Alfonso ; Croudace, Tim ; Gilbody, Simon. / Heterogeneity in patient-reported outcomes following low-intensity mental health interventions : a multilevel analysis. In: PLoS ONE. 2014 ; Vol. 9, No. 9.
@article{b4b87240b4bf49949f8a06e780d9f0ea,
title = "Heterogeneity in patient-reported outcomes following low-intensity mental health interventions: a multilevel analysis",
abstract = "BackgroundVariability in patient-reported outcomes of psychological treatments has been partly attributed to therapists ? a phenomenon commonly known as therapist effects. Meta-analytic reviews reveal wide variation in therapist-attributable variability in psychotherapy outcomes, with most studies reporting therapist effects in the region of 5{\%} to 10{\%} and some finding minimal to no therapist effects. However, all except one study to date have been conducted in high-intensity or mixed intervention groups; therefore, there is scarcity of evidence on therapist effects in brief low-intensity psychological interventions.ObjectiveTo examine therapist effects in low-intensity interventions for depression and anxiety in a naturalistic setting.Data and AnalysisSession-by-session data on patient-reported outcome measures were available for a cohort of 1,376 primary care psychotherapy patients treated by 38 therapists. Outcome measures included PHQ-9 (sensitive to depression) and GAD-7 (sensitive to general anxiety disorder) measures. Three-level hierarchical linear modelling was employed to estimate therapist-attributable proportion of variance in clinical outcomes. Therapist effects were evaluated using the intra-cluster correlation coefficient (ICC) and Bayesian empirical predictions of therapist random effects. Three sensitivity analyses were conducted: 1) using both treatment completers and non-completers; 2) a sub-sample of cases with baseline scores above the conventional clinical thresholds for PHQ-9 and GAD-7; and 3) a two-level model (using patient-level pre- and post-treatment scores nested within therapists).ResultsThe ICC estimates for all outcome measures were very small, ranging between 0{\%} and 1.3{\%}, although most were statistically significant. The Bayesian empirical predictions showed that therapist random effects were not statistically significantly different from each other. Between patient variability explained most of the variance in outcomes.ConclusionConsistent with the only other study to date in low intensity interventions, evidence was found to suggest minimal to no therapist effects in patient-reported outcomes. This draws attention to the more prominent source of variability which is found at the between-patient level.",
author = "Shehzad Ali and Elizabeth Littlewood and Dean McMillan and Jaime Delgadillo and Alfonso Miranda and Tim Croudace and Simon Gilbody",
note = "M1 - Article",
year = "2014",
month = "9",
day = "10",
doi = "10.1371/journal.pone.0099658",
language = "English",
volume = "9",
journal = "PLoS ONE",
issn = "1932-6203",
publisher = "Public Library of Science",
number = "9",

}

Heterogeneity in patient-reported outcomes following low-intensity mental health interventions : a multilevel analysis. / Ali, Shehzad (Lead / Corresponding author); Littlewood, Elizabeth; McMillan, Dean; Delgadillo, Jaime; Miranda, Alfonso; Croudace, Tim; Gilbody, Simon.

In: PLoS ONE, Vol. 9, No. 9, e99658, 10.09.2014.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Heterogeneity in patient-reported outcomes following low-intensity mental health interventions

T2 - a multilevel analysis

AU - Ali, Shehzad

AU - Littlewood, Elizabeth

AU - McMillan, Dean

AU - Delgadillo, Jaime

AU - Miranda, Alfonso

AU - Croudace, Tim

AU - Gilbody, Simon

N1 - M1 - Article

PY - 2014/9/10

Y1 - 2014/9/10

N2 - BackgroundVariability in patient-reported outcomes of psychological treatments has been partly attributed to therapists ? a phenomenon commonly known as therapist effects. Meta-analytic reviews reveal wide variation in therapist-attributable variability in psychotherapy outcomes, with most studies reporting therapist effects in the region of 5% to 10% and some finding minimal to no therapist effects. However, all except one study to date have been conducted in high-intensity or mixed intervention groups; therefore, there is scarcity of evidence on therapist effects in brief low-intensity psychological interventions.ObjectiveTo examine therapist effects in low-intensity interventions for depression and anxiety in a naturalistic setting.Data and AnalysisSession-by-session data on patient-reported outcome measures were available for a cohort of 1,376 primary care psychotherapy patients treated by 38 therapists. Outcome measures included PHQ-9 (sensitive to depression) and GAD-7 (sensitive to general anxiety disorder) measures. Three-level hierarchical linear modelling was employed to estimate therapist-attributable proportion of variance in clinical outcomes. Therapist effects were evaluated using the intra-cluster correlation coefficient (ICC) and Bayesian empirical predictions of therapist random effects. Three sensitivity analyses were conducted: 1) using both treatment completers and non-completers; 2) a sub-sample of cases with baseline scores above the conventional clinical thresholds for PHQ-9 and GAD-7; and 3) a two-level model (using patient-level pre- and post-treatment scores nested within therapists).ResultsThe ICC estimates for all outcome measures were very small, ranging between 0% and 1.3%, although most were statistically significant. The Bayesian empirical predictions showed that therapist random effects were not statistically significantly different from each other. Between patient variability explained most of the variance in outcomes.ConclusionConsistent with the only other study to date in low intensity interventions, evidence was found to suggest minimal to no therapist effects in patient-reported outcomes. This draws attention to the more prominent source of variability which is found at the between-patient level.

AB - BackgroundVariability in patient-reported outcomes of psychological treatments has been partly attributed to therapists ? a phenomenon commonly known as therapist effects. Meta-analytic reviews reveal wide variation in therapist-attributable variability in psychotherapy outcomes, with most studies reporting therapist effects in the region of 5% to 10% and some finding minimal to no therapist effects. However, all except one study to date have been conducted in high-intensity or mixed intervention groups; therefore, there is scarcity of evidence on therapist effects in brief low-intensity psychological interventions.ObjectiveTo examine therapist effects in low-intensity interventions for depression and anxiety in a naturalistic setting.Data and AnalysisSession-by-session data on patient-reported outcome measures were available for a cohort of 1,376 primary care psychotherapy patients treated by 38 therapists. Outcome measures included PHQ-9 (sensitive to depression) and GAD-7 (sensitive to general anxiety disorder) measures. Three-level hierarchical linear modelling was employed to estimate therapist-attributable proportion of variance in clinical outcomes. Therapist effects were evaluated using the intra-cluster correlation coefficient (ICC) and Bayesian empirical predictions of therapist random effects. Three sensitivity analyses were conducted: 1) using both treatment completers and non-completers; 2) a sub-sample of cases with baseline scores above the conventional clinical thresholds for PHQ-9 and GAD-7; and 3) a two-level model (using patient-level pre- and post-treatment scores nested within therapists).ResultsThe ICC estimates for all outcome measures were very small, ranging between 0% and 1.3%, although most were statistically significant. The Bayesian empirical predictions showed that therapist random effects were not statistically significantly different from each other. Between patient variability explained most of the variance in outcomes.ConclusionConsistent with the only other study to date in low intensity interventions, evidence was found to suggest minimal to no therapist effects in patient-reported outcomes. This draws attention to the more prominent source of variability which is found at the between-patient level.

U2 - 10.1371/journal.pone.0099658

DO - 10.1371/journal.pone.0099658

M3 - Article

C2 - 25207881

VL - 9

JO - PLoS ONE

JF - PLoS ONE

SN - 1932-6203

IS - 9

M1 - e99658

ER -