Page 112 - Physical activity in recipients of solid organ transplantation - Edwin J. van Adrichem

Chapter 6
level of PA or sedentary time, multiple regression analyzes were conducted. Explanatory
variables included: age, gender, type of SOT, comorbidity category (0/1-2/≥
working/not actively working/retired), PA barriers (fear of negative e
3
comorbidities), education level (low/medium/high), body mass index (BMI), time since
transplantation, length of hospital stay, exercise self-e

cacy, employment status (actively

ect/physical
limitations/low expectations and self-con

motivators (health and physical outcomes/external in
dence/lack of motivation or time), and PA

uences/group activities/psychological
outcomes). All continuous explanatory variables were centered by median subtraction to
increase clinical interpretability. Due to their substantial number, a variable selection
procedure by penalized regression according to the Smoothly Clipped Absolute Deviation
penalty was used.
regression coe

35
cients.

e latter performs well in variable selection and provides unbiased
36,37
A plot of the coe

cients’ path depending on penalty size was
given to statistically learn about the importance of each explanatory variable in predicting
the level of PA or sedentary time. A cut-o

value for the penalty parameter was
Results
Physical activity level
A total number of 656 recipients participated.
inclusion in analyzes is presented in Figure 1.
38
determined by cross validation yielding regression estimates of zero for unimportant
explanatory variables. On the basis of the selected explanatory variables, a generalized
additive model with integrated smoothness
possible non-linear e
was estimated to statistically test for a

ect of age on the level of PA and sedentary time.
e

nal model was
interpreted on the selected explanatory variables. Analyzes were performed with the
statistical programming language R (version 3.2.0) and IBM SPSS statistical software
(version 23.0). A
p-
value <.05 was considered statistically signi

cant.
e

owchart of the response rates and data

e response rates per organ groups are
corrected for non-responders due to being deceased, inability to

ll out the questionnaire,
or having an unknown address. Participant characteristics are presented in Table 1.

e PA guideline was met by 55.9% of participants.
to identify associated factors of PA (Table 3). of variance (adjusted R2). e nal model obt

e median time spent on moderate to
vigorous PA was 720 (270-1460) minutes per week. Guideline compliance and time per
transplant group is summarized in Table 2. Two multiple linear regressions were performed
ain
e model with all variables explained 21% ed after variable selection included: gend
age, time since transplantation, employment status, fear of negative e

ects, physical
limitations, and low expectations and self-con

dence.

e contribution and importance of
these associated factors in the regression model is illustrated in Supplemental Figure 1.
e
variables in the statistical learning
the far left represent the strongest predictors. As an example, the coe

gure deviating from a regression coe

cient of zero at

cient of ‘physical
limitations’ deviates the earliest from zero and is, therefore, strongly related to the level of
er,
PA. 110.