Making the most out of timeseries symptom data: A machine learning study on symptom predictions of internet-based CBT

Research output: Journal contributionsJournal articlesResearchpeer-review

Standard

Making the most out of timeseries symptom data: A machine learning study on symptom predictions of internet-based CBT. / Hentati Isacsson, Nils; Zantvoort, Kirsten; Forsell, Erik et al.
In: Internet Interventions, Vol. 38, 100773, 12.2024.

Research output: Journal contributionsJournal articlesResearchpeer-review

Harvard

APA

Vancouver

Hentati Isacsson N, Zantvoort K, Forsell E, Boman M, Kaldo V. Making the most out of timeseries symptom data: A machine learning study on symptom predictions of internet-based CBT. Internet Interventions. 2024 Dec;38:100773. doi: 10.1016/j.invent.2024.100773

Bibtex

@article{0bf13c4796154bd99c60cdc7585e5412,
title = "Making the most out of timeseries symptom data: A machine learning study on symptom predictions of internet-based CBT",
abstract = "Objective: Predicting who will not benefit enough from Internet-Based Cognitive Behavioral (ICBT) Therapy early on can assist in better allocation of limited mental health care resources. Repeated measures of symptoms during treatment is the strongest predictor of outcome, and we want to investigate if methods that explicitly account for time-dependency are superior to methods that do not, with data from (a) only two pre-treatment timepoints and (b) the pre-treatment timepoints and three timepoints during initial treatment. Methods: We use 1) commonly used time-independent methods (i.e., Linear Regression and Random Forest models) and 2) time-dependent methods (i.e., multilevel model regression, mixed-effects random forest, and a Long Short-Term Memory model) to predict symptoms during treatment, including the final outcome. This is done with symptom scores from 6436 ICBT patients from regular care, using robust multiple imputation and nested cross-validation methods. Results: The models had a 14 %–12 % root mean squared error (RMSE) in predicting the post-treatment outcome, corresponding to a balanced accuracy of 67–74 %. Time-dependent models did not have higher accuracies. Using data for the initial treatment period (b) instead of only from before treatment (a) increased prediction results by 1.3 % percentage points (12 % to 10.7 %) RMSE and 6 % percentage points BACC (69 % to 75 %). Conclusion: Training prediction models on only symptom scores of the first few weeks is a promising avenue for symptom predictions in treatment, regardless of which model is used. Further research is necessary to better understand the interaction between model complexity, dataset length and width, and the prediction tasks at hand.",
keywords = "Adaptive treatment strategy, Machine learning, Precision psychiatry, Prediction, Timeseries symptom, Treatment outcome",
author = "{Hentati Isacsson}, Nils and Kirsten Zantvoort and Erik Forsell and Magnus Boman and Viktor Kaldo",
note = "Publisher Copyright: {\textcopyright} 2024 The Authors",
year = "2024",
month = dec,
doi = "10.1016/j.invent.2024.100773",
language = "English",
volume = "38",
journal = "Internet Interventions",
issn = "2214-7829",
publisher = "Elsevier B.V.",

}

RIS

TY - JOUR

T1 - Making the most out of timeseries symptom data

T2 - A machine learning study on symptom predictions of internet-based CBT

AU - Hentati Isacsson, Nils

AU - Zantvoort, Kirsten

AU - Forsell, Erik

AU - Boman, Magnus

AU - Kaldo, Viktor

N1 - Publisher Copyright: © 2024 The Authors

PY - 2024/12

Y1 - 2024/12

N2 - Objective: Predicting who will not benefit enough from Internet-Based Cognitive Behavioral (ICBT) Therapy early on can assist in better allocation of limited mental health care resources. Repeated measures of symptoms during treatment is the strongest predictor of outcome, and we want to investigate if methods that explicitly account for time-dependency are superior to methods that do not, with data from (a) only two pre-treatment timepoints and (b) the pre-treatment timepoints and three timepoints during initial treatment. Methods: We use 1) commonly used time-independent methods (i.e., Linear Regression and Random Forest models) and 2) time-dependent methods (i.e., multilevel model regression, mixed-effects random forest, and a Long Short-Term Memory model) to predict symptoms during treatment, including the final outcome. This is done with symptom scores from 6436 ICBT patients from regular care, using robust multiple imputation and nested cross-validation methods. Results: The models had a 14 %–12 % root mean squared error (RMSE) in predicting the post-treatment outcome, corresponding to a balanced accuracy of 67–74 %. Time-dependent models did not have higher accuracies. Using data for the initial treatment period (b) instead of only from before treatment (a) increased prediction results by 1.3 % percentage points (12 % to 10.7 %) RMSE and 6 % percentage points BACC (69 % to 75 %). Conclusion: Training prediction models on only symptom scores of the first few weeks is a promising avenue for symptom predictions in treatment, regardless of which model is used. Further research is necessary to better understand the interaction between model complexity, dataset length and width, and the prediction tasks at hand.

AB - Objective: Predicting who will not benefit enough from Internet-Based Cognitive Behavioral (ICBT) Therapy early on can assist in better allocation of limited mental health care resources. Repeated measures of symptoms during treatment is the strongest predictor of outcome, and we want to investigate if methods that explicitly account for time-dependency are superior to methods that do not, with data from (a) only two pre-treatment timepoints and (b) the pre-treatment timepoints and three timepoints during initial treatment. Methods: We use 1) commonly used time-independent methods (i.e., Linear Regression and Random Forest models) and 2) time-dependent methods (i.e., multilevel model regression, mixed-effects random forest, and a Long Short-Term Memory model) to predict symptoms during treatment, including the final outcome. This is done with symptom scores from 6436 ICBT patients from regular care, using robust multiple imputation and nested cross-validation methods. Results: The models had a 14 %–12 % root mean squared error (RMSE) in predicting the post-treatment outcome, corresponding to a balanced accuracy of 67–74 %. Time-dependent models did not have higher accuracies. Using data for the initial treatment period (b) instead of only from before treatment (a) increased prediction results by 1.3 % percentage points (12 % to 10.7 %) RMSE and 6 % percentage points BACC (69 % to 75 %). Conclusion: Training prediction models on only symptom scores of the first few weeks is a promising avenue for symptom predictions in treatment, regardless of which model is used. Further research is necessary to better understand the interaction between model complexity, dataset length and width, and the prediction tasks at hand.

KW - Adaptive treatment strategy

KW - Machine learning

KW - Precision psychiatry

KW - Prediction

KW - Timeseries symptom

KW - Treatment outcome

UR - http://www.scopus.com/inward/record.url?scp=85203524573&partnerID=8YFLogxK

UR - https://www.mendeley.com/catalogue/ed931dae-dfd8-3eb1-9b76-00bd47b6d332/

U2 - 10.1016/j.invent.2024.100773

DO - 10.1016/j.invent.2024.100773

M3 - Journal articles

AN - SCOPUS:85203524573

VL - 38

JO - Internet Interventions

JF - Internet Interventions

SN - 2214-7829

M1 - 100773

ER -