Page 117 - Effects and Efficacy of (Laparoscopic) Gastrostomy Placement in Children - Josephine Franken

then reverse-scored and transformed into a scale from 0 to 100, where higher scores indicate better HRQoL.
statistical analysis
Continuous variables were expressed as mean ± standard deviations for symmetric variables or as median with interquartile ranges (IQR) for skewed variables.
A linear mixed model was used to compare pre- and postoperative HRQoL and to identify predictors of HRQoL and predictors of postoperative increase or decrease in HRQoL. Mixed models are appropriate for the analysis of repeated measurements, especially in the presence of missing data on the outcome variable. 13 Fixed effects were timing of the measurement (postoperative versus preoperative), age, neurologic impairment, cardiac disease, body mass index (BMI) and postoperative complications of GP, and a random intercept per child was included. The variables included in the mixed model were chosen based on univariate analysis. Coefficients from the mixed model represent the predictive value of the variables on the outcome variable.
To examine the effects of the examined predictors on changes in HRQoL, interactions of all variables (except for the variable ‘complications’, because no preoperative values of this variable were available) with timing of the measurement (preoperative versus postoperative) were added to the mixed model analysis. A significant interaction indicates that the variable is associated with postoperative change in HRQoL.
A small subsample of children was also asked to complete HRQoL questionnaires. For this subsample, the responses of children and caregivers were compared using a linear mixed model. Fixed effects were timing (postoperative versus preoperative) and children versus caregivers; a random intercept per child was included to account for clustering of measurements within children.
Statistical significance was defined by p-values of less than 0.05. All analyses were performed using SPSS 22.0 statistical package (IBM, USA).
ResULTs
Patient inclusion
An overview of patient inclusion is depicted in Figure 1. In 28 out of 31 patients that were excluded from the study the reason was the refusal of parents to participate in the clinical tests that this study was combined with, namely 24-hour pH-Impedance monitoring
Chapter 7
115