Data Overview

see summary stats of our variables. for the table on non-missing data (2), it seems the first 16 participants (vpn_nr) literally had no ESM data, so don’t be confused when the first page of the table only shows 0 - the participants with data appear later in the table. for table 3 on Questionnaires, note that one WHODAS item (WHODAS_10) in both BL and Post has different min-max values than the others because it was an optional item > excluded this item for later analyses.

Basic Descriptives

ESM “state-to-trait”

A state-to-trait score is the within-person average of a single ESM variable over time (e.g. state psychological flexibility averaged over time as a proxy for trait psychological flexibility). n = 191, of which 38 provided no ESM data

ESM “coupling”

A coupling score is the within-person average of a bivariate relationship between two ESM variables over time. Coupling scores are derived using iARIMAX models, resulting in a score for each ESM-variable-pairing in each client > then derived pooled effects using REMA. You’ll see 12 coupling scores: one for each facet of psychological flexibility (6 facets: acceptance, defusion, present-moment-awareness, values, committed action, self-as-context) and their relationship to a) positive affect and b) negative affect. Caterpillar plots for each effect are below summary table. betas indicate moderate links on average in the expected directions but also quite a lot of heterogeneity.

ESM vs. Baseline Questionnaires

Here, Pearson correlation was used to examine the degree of convergence between ESM “state-to-trait” scores and Baseline Questionnaire scores. Hypothesis 1 is that PsyFlex captured by ESM and by questionnaire yield distinct information as reflected in a low correlation between the two. our results (table 1) indicate moderate correlation. the other pairings are exploratory analysis (e.g. ESM Trait Affect - MHC)

Incremental Utility of ESM PsyFlex Data (Predicting Raw Change Scores)

To examine whether ESM Psyflex data adds predictive utility for treatment outcome beyond baseline PsyFlex questionnaire, three baseline models were compared using elastic-net regression and 5-fold cross-validation: 1) questionnaire-only 2) ESM data (within-person means of each PF facet) 3) questionnaire + ESM data Research Questions are: Do ESM process profiles predict improvement? Do ESM process profiles outperform questionnaires? Does adding ESM to questionnaire increase out-of-sample \(R^2\)? Hypothesis 2 was that this would be the case for PsyFlex ESM data. Interestingly, PsyFlex ESM itself did not predict change but adding Psyflex ESM to PsyFlex questionnaire improved prediction. For tables below, I chose a raw change score between PsyFlex BL and PsyFlex Post as treatment outcome.

## [1] 132

(Exploratory) Incremental Utility of Process Profiles (Predicting Raw Change Scores)

Exploratory Analyses to examine whether process profiles add predictive utility for treatment outcome beyond baseline questionnaires. Again, three baseline models were compared using elastic-net regression and 5-fold cross-validation: 1) questionnaire-only 2) ESM process profile-only 3) questionnaire + ESM process profile Research Questions are: Do ESM process profiles predict improvement? Do ESM process profiles outperform questionnaires? Does adding ESM to questionnaire increase out-of-sample \(R^2\)? For tables below, I chose a raw change score between baseline and post MHC, BSCL, and WHODAS as treatment outcome.

Incremental Utility of ESM PsyFlex Data (Predicting RCI)

## [1] 0.7860972
## [1] 132
## 
##  0  1 
## 84 48

(Exploratory) Incremental Utility of Process Profiles (Predicting RCI)

##     Scale Cronbach_alpha Baseline_SD
## 1 PsyFlex      0.7860972    4.251561
## 2     MHC      0.9076661   13.658391
## 3    BSCL      0.9670588   38.721646
## 4  WHODAS      0.8935551   10.361285
##     Scale Non_responders Responders Missing
## 1 PsyFlex            110         58      33
## 2     MHC            112         58      31
## 3    BSCL             44         31     126
## 4  WHODAS            109         62      30
## Sample size: 56 
## y
##  0  1 
## 33 23
## Sample size: 56 
## y
##  0  1 
## 43 13
## Sample size: 57 
## y
##  0  1 
## 45 12
## Sample size: 56 
## y
##  0  1 
## 41 15