Skip to Main content Skip to Navigation
Master Thesis

Facteurs prédictifs de la réponse au MBCT pour les patients souffrant de trouble dépressif : le point de vue des statistiques et du machine learning

Abstract : Objectives: Whether there is abundant literature relating to the benefits of Mindfulness Based Interventions, data about factors associated with efficiency are scarce. Our study attempts to determine the predictors of efficacy and adherence in the Mindfulness Based Cognitive Therapy (MBCT) with a classic statistical analysis and a machine learning analysis. Methods: 76 psychiatric outpatients at “Conception Psychiatric Pole” were included. They had to answer a battery of clinical, mindfulness and psychological functioning self-report questionnaires before, after MBCT and 6 months post-MBCT. Data were compared using the Chi-square test, or Fisher's exact test when Chi-square’s conditions were not met. Machine learning analysis is based on the support vector machine method. Results: For the efficacy predictors, the classic statistical analysis finds marital status (living in couple), the presence of pharmacological treatment, a severe depressive episode assessed by the BDI (Beck> 15) and a high anxiety-state score at STAI (54 on average). The machine learning analysis finds a two-dimensional model: a severe depressive episode assessed by the BDI (Beck> 25) and a trait mindfulness score (FFMQ> 90) would be efficacy predictors (sensitivity 79.3% specificity 71.9%). For the predictors of compliance, the classic statistical analysis finds a mild to moderate depressive episode with a high Observation facet (27 on average) and an high physical limitation at SF-36 (55 on average). The machine learning analysis finds a three-dimensional model with a low or high FMI score, a low or high bodily dissociation score, and a low self-compassion score (sensitivity 48.4% specificity 71.9%). Conclusion: The predictive power of machine learning makes it possible to establish standard patient profiles to allow personalized care. It would be interesting to include in MBCT more psychoeducation in order to maximize clinical benefits and adherence to the program.
Complete list of metadatas

Cited literature [61 references]  Display  Hide  Download

https://dumas.ccsd.cnrs.fr/dumas-02921546
Contributor : Faculté de Médecine Amu <>
Submitted on : Tuesday, August 25, 2020 - 12:13:07 PM
Last modification on : Friday, October 23, 2020 - 4:58:56 PM

File

These_exemplaire_final_dethoor...
Files produced by the author(s)

Identifiers

  • HAL Id : dumas-02921546, version 1

Collections

Citation

Manon Dethoor. Facteurs prédictifs de la réponse au MBCT pour les patients souffrant de trouble dépressif : le point de vue des statistiques et du machine learning. Sciences du Vivant [q-bio]. 2020. ⟨dumas-02921546⟩

Share

Metrics

Record views

31

Files downloads

27