A machine learning ensemble to predict treatment outcomes following an Internet intervention for depression

Rahel Pearson, Derek Pisner, Björn Meyer, Jason Shumake, Christopher Beevers

Research output: Contribution to journalArticle

1 Citation (Scopus)

Abstract

Background Some Internet interventions are regarded as effective treatments for adult depression, but less is known about who responds to this form of treatment.Method An elastic net and random forest were trained to predict depression symptoms and related disability after an 8-week course of an Internet intervention, Deprexis, involving adults (N = 283) from across the USA. Candidate predictors included psychopathology, demographics, treatment expectancies, treatment usage, and environmental context obtained from population databases. Model performance was evaluated using predictive R2 (R2pred) the expected variance explained in a new sample, estimated by 10 repetitions of 10-fold cross-validation.Results An ensemble model was created by averaging the predictions of the elastic net and random forest. Model performance was compared with a benchmark linear autoregressive model that predicted each outcome using only its baseline. The ensemble predicted more variance in post-treatment depression (8.0% gain, 95% CI 0.8-15; total R2pred = 0.25), disability (5.0% gain, 95% CI -0.3 to 10; total R2pred = 0.25), and well-being (11.6% gain, 95% CI 4.9-19; total R2pred = 0.29) than the benchmark model. Important predictors included comorbid psychopathology, particularly total psychopathology and dysthymia, low symptom-related disability, treatment credibility, lower access to therapists, and time spent using certain Deprexis modules.Conclusion A number of variables predict symptom improvement following an Internet intervention, but each of these variables makes relatively small contributions. Machine learning ensembles may be a promising statistical approach for identifying the cumulative contribution of many weak predictors to psychosocial depression treatment response.

Original languageEnglish (US)
Pages (from-to)2330-2341
Number of pages12
JournalPsychological Medicine
Volume49
Issue number14
DOIs
StatePublished - Oct 1 2019

Fingerprint

Internet
Psychopathology
Depression
Benchmarking
Linear Models
Demography
Databases
Machine Learning
Population
Forests

Keywords

  • Depression treatment
  • Internet interventions
  • machine learning

ASJC Scopus subject areas

  • Applied Psychology
  • Psychiatry and Mental health

Cite this

A machine learning ensemble to predict treatment outcomes following an Internet intervention for depression. / Pearson, Rahel; Pisner, Derek; Meyer, Björn; Shumake, Jason; Beevers, Christopher.

In: Psychological Medicine, Vol. 49, No. 14, 01.10.2019, p. 2330-2341.

Research output: Contribution to journalArticle

Pearson, Rahel ; Pisner, Derek ; Meyer, Björn ; Shumake, Jason ; Beevers, Christopher. / A machine learning ensemble to predict treatment outcomes following an Internet intervention for depression. In: Psychological Medicine. 2019 ; Vol. 49, No. 14. pp. 2330-2341.
@article{628d7e2cfe814deb8d283d0b12343bc1,
title = "A machine learning ensemble to predict treatment outcomes following an Internet intervention for depression",
abstract = "Background Some Internet interventions are regarded as effective treatments for adult depression, but less is known about who responds to this form of treatment.Method An elastic net and random forest were trained to predict depression symptoms and related disability after an 8-week course of an Internet intervention, Deprexis, involving adults (N = 283) from across the USA. Candidate predictors included psychopathology, demographics, treatment expectancies, treatment usage, and environmental context obtained from population databases. Model performance was evaluated using predictive R2 (R2pred) the expected variance explained in a new sample, estimated by 10 repetitions of 10-fold cross-validation.Results An ensemble model was created by averaging the predictions of the elastic net and random forest. Model performance was compared with a benchmark linear autoregressive model that predicted each outcome using only its baseline. The ensemble predicted more variance in post-treatment depression (8.0{\%} gain, 95{\%} CI 0.8-15; total R2pred = 0.25), disability (5.0{\%} gain, 95{\%} CI -0.3 to 10; total R2pred = 0.25), and well-being (11.6{\%} gain, 95{\%} CI 4.9-19; total R2pred = 0.29) than the benchmark model. Important predictors included comorbid psychopathology, particularly total psychopathology and dysthymia, low symptom-related disability, treatment credibility, lower access to therapists, and time spent using certain Deprexis modules.Conclusion A number of variables predict symptom improvement following an Internet intervention, but each of these variables makes relatively small contributions. Machine learning ensembles may be a promising statistical approach for identifying the cumulative contribution of many weak predictors to psychosocial depression treatment response.",
keywords = "Depression treatment, Internet interventions, machine learning",
author = "Rahel Pearson and Derek Pisner and Bj{\"o}rn Meyer and Jason Shumake and Christopher Beevers",
year = "2019",
month = "10",
day = "1",
doi = "10.1017/S003329171800315X",
language = "English (US)",
volume = "49",
pages = "2330--2341",
journal = "Psychological Medicine",
issn = "0033-2917",
publisher = "Cambridge University Press",
number = "14",

}

TY - JOUR

T1 - A machine learning ensemble to predict treatment outcomes following an Internet intervention for depression

AU - Pearson, Rahel

AU - Pisner, Derek

AU - Meyer, Björn

AU - Shumake, Jason

AU - Beevers, Christopher

PY - 2019/10/1

Y1 - 2019/10/1

N2 - Background Some Internet interventions are regarded as effective treatments for adult depression, but less is known about who responds to this form of treatment.Method An elastic net and random forest were trained to predict depression symptoms and related disability after an 8-week course of an Internet intervention, Deprexis, involving adults (N = 283) from across the USA. Candidate predictors included psychopathology, demographics, treatment expectancies, treatment usage, and environmental context obtained from population databases. Model performance was evaluated using predictive R2 (R2pred) the expected variance explained in a new sample, estimated by 10 repetitions of 10-fold cross-validation.Results An ensemble model was created by averaging the predictions of the elastic net and random forest. Model performance was compared with a benchmark linear autoregressive model that predicted each outcome using only its baseline. The ensemble predicted more variance in post-treatment depression (8.0% gain, 95% CI 0.8-15; total R2pred = 0.25), disability (5.0% gain, 95% CI -0.3 to 10; total R2pred = 0.25), and well-being (11.6% gain, 95% CI 4.9-19; total R2pred = 0.29) than the benchmark model. Important predictors included comorbid psychopathology, particularly total psychopathology and dysthymia, low symptom-related disability, treatment credibility, lower access to therapists, and time spent using certain Deprexis modules.Conclusion A number of variables predict symptom improvement following an Internet intervention, but each of these variables makes relatively small contributions. Machine learning ensembles may be a promising statistical approach for identifying the cumulative contribution of many weak predictors to psychosocial depression treatment response.

AB - Background Some Internet interventions are regarded as effective treatments for adult depression, but less is known about who responds to this form of treatment.Method An elastic net and random forest were trained to predict depression symptoms and related disability after an 8-week course of an Internet intervention, Deprexis, involving adults (N = 283) from across the USA. Candidate predictors included psychopathology, demographics, treatment expectancies, treatment usage, and environmental context obtained from population databases. Model performance was evaluated using predictive R2 (R2pred) the expected variance explained in a new sample, estimated by 10 repetitions of 10-fold cross-validation.Results An ensemble model was created by averaging the predictions of the elastic net and random forest. Model performance was compared with a benchmark linear autoregressive model that predicted each outcome using only its baseline. The ensemble predicted more variance in post-treatment depression (8.0% gain, 95% CI 0.8-15; total R2pred = 0.25), disability (5.0% gain, 95% CI -0.3 to 10; total R2pred = 0.25), and well-being (11.6% gain, 95% CI 4.9-19; total R2pred = 0.29) than the benchmark model. Important predictors included comorbid psychopathology, particularly total psychopathology and dysthymia, low symptom-related disability, treatment credibility, lower access to therapists, and time spent using certain Deprexis modules.Conclusion A number of variables predict symptom improvement following an Internet intervention, but each of these variables makes relatively small contributions. Machine learning ensembles may be a promising statistical approach for identifying the cumulative contribution of many weak predictors to psychosocial depression treatment response.

KW - Depression treatment

KW - Internet interventions

KW - machine learning

UR - http://www.scopus.com/inward/record.url?scp=85056151201&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85056151201&partnerID=8YFLogxK

U2 - 10.1017/S003329171800315X

DO - 10.1017/S003329171800315X

M3 - Article

C2 - 30392475

AN - SCOPUS:85056151201

VL - 49

SP - 2330

EP - 2341

JO - Psychological Medicine

JF - Psychological Medicine

SN - 0033-2917

IS - 14

ER -