Table 6.
Posterior distribution [mean (SD) and 95 % equal-tailed credibility interval (CI)] of the parameters of the Bayesian multilevel probit model for the probability of changing the ordinal FEES scores of the independent rating task during the consensus panel rating task
| Piecemeal deglutition | Postswallow vallecular pooling | Postswallow pyriform sinus pooling | Laryngeal penetration/tracheal aspiration | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Mean (SD) | 95 % CI | Mean (SD) | 95 % CI | Mean (SD) | 95 % CI | Mean (SD) | 95 % CI | |||||
| Intercept | −1.09 (0.46) | −2.04 | −0.22 | −0.77 (0.24) | −1.26 | −0.30 | −0.62 (0.29) | −1.21 | −0.069 | −0.70 (0.25) | −1.19 | −0.21 |
| Observera | 0.20 (0.23) | −0.24 | 0.64 | −0.13 (0.21) | −0.54 | 0.28 | 0.73 (0.24) | 0.27 | 1.20 | 0.10 (0.21) | −0.30 | 0.52 |
| Consistencyb | 0.034 (0.23) | −0.41 | 0.48 | −0.38 (0.21) | −0.79 | 0.025 | 0.15 (0.22) | −0.29 | 0.59 | 0.30 (0.21) | −0.10 | 0.71 |
| Groupc | 0.12 (0.60) | −1.07 | 1.32 | 0.15 (0.27) | −0.38 | 0.68 | 0.21 (0.37) | −0.50 | 0.96 | 0.17 (0.29) | −0.39 | 0.76 |
| σ² | 1.38 (1.14) | 0.33 | 4.05 | 0.16 (0.15) | 0.00 | 0.53 | 0.37 (0.31) | 0.036 | 1.18 | 0.20 (0.21) | 0.002 | 0.72 |
To facilitate interpretation of the table, a more detailed description is given. Mean, SD, and 95 % CI are presented separately per FEES variable. When ‘0’ is not entailed in the 95 % CI, the difference between the predictors (observer 1 and observer 2, or thin and thick liquid consistency, or neurological and oncological group) is statistically significant. A positive mean indicates that the agreement of the predictor used as reference is lower. For instance, in the line ‘Observers,’ the comparison between the observers’ probability of changing FEES scores of the independent rating task during the consensus panel rating task is analyzed. ‘0’ is entailed in the 95 % CI of all FEES variables, except for postswallow pyriform sinus pooling (0.27, 1.20). It means that a statistically significant difference was found in the observers’ probability of changing the FEES scores of the independent rating task during the consensus panel rating task when rating this variable. The positive mean (0.73) indicates that observer 1 changed the scores more frequently than observer 2 during the panel task
SD standard deviation
The groups used as reference are:
aObserver 2 for observer effect
bThick liquid for bolus consistency
cNeurological patients for the etiological group