Journal of Rehabilitation Medicine 51-10 | Page 80

800
S. Thomas et al.
Statistical analyses
Descriptive and inference statistics were used depending on the
type of test and data distribution. The global alpha level was
set at 0.05. Predefined categories of walking ability (FAC ≥ 3)
were used to statistically compare all predefined physiotherapy
categories of interventions.
Based on an a priori sample size calculation, 150 patients
were needed to be recruited (14). The probability of regaining
walking function was calculated using the method of Kaplan &
Meier (30). Univariate and multivariate Cox regression analy-
ses with a selection of possible predictor variables for primary
outcome were used as described below (31, 32). The main
result of such analysis is a multivariate adjusted hazard ratio
(HR), which provides a comparison between the probability of
regaining walking function in a subgroup of the cohort and the
probability of regaining walking function in another subgroup.
The HR was used to describe whether patients receiving a
specific physiotherapy intervention progress at a different rate
from patients not receiving that type of therapy.
As an example, a HR of 1.5 means that, with a specific
treatment, patients progress 1.5 times faster than patients not
receiving that type of therapy.
Univariate analysis. Univariate Cox regression analysis of
the following variables was performed: total time (in min) of
all physiotherapy interventions per week for the first 2 weeks
of rehabilitation (total min per week) spent in walking activi-
ties, age at baseline, body mass index (BMI), sex, lower limb
strength, duration of illness, duration of mechanical ventilation,
and number (total) of secondary diagnoses.
Multivariate analysis and model building. All statistically sig-
nificant variables (alpha level 0.2 for selection) were candidate
predictor variables, and those with the highest global χ 2 score
were selected first and entered into a multivariate regression
analysis (31, 32). To remain in the multivariate model a va-
riable had to be significant at the 0.1 level (32). The multiva-
riate models were then compared with remaining variables on
global score χ 2 statistic (best subset selection) and on Akaike’s
information criterion (AIC) and Schwarz’s Bayesian criterion
(SBC) for the final multivariate model (31, 32). The effects of
the final multivariate model were expressed as HRs with 95%
confidence interval (95% CI). A Kaplan–Meier estimate of the
final multivariate adjusted Cox proportional hazards model was
provided. SAS/STAT 9.3 was used for all statistical procedures
(SAS Institute Inc., Cary, NC, USA) and proportional hazards
assumptions were tested with the implemented function.
RESULTS
A total of 150 patients with ICUAW between January
2013 and March 2015 were included in this cohort
study (Fig. 1). The demographic and clinical charac-
teristics of subjects at study onset in post-acute reha-
bilitation (T0) are shown in Table II. All therapeutic
interventions and the total time (in min) per week per
intervention category are shown in Table III, catego-
rized according to walking ability achieved.
At study onset no patients were able to walk. During
post-acute rehabilitation walking ability was achieved
after a median of 28.5 days (interquartile range (IQR)
45 days) after the start of individual rehabilitation.
www.medicaljournals.se/jrm
Fig. 1. Study flow chart. T0: baseline; T1 to T4: every 2 weeks up to
8 weeks.
The amount and content of therapy in the first 2
weeks did not differ between subgroups. However, dif-
ferent interventions were applied in patient subgroups.
The most frequent interventions in the first 2 weeks
of rehabilitation in patients who regained walking
ability were: practicing walking, sit-to-stand training,
and balance training while sitting (mean total time per
week: 48.03 (SD 41.10), 20.13 (SD 21.12) and 12.37
(SD 26.95) min, respectively).
The most frequent interventions in the first 2 weeks
of rehabilitation in those patients who did not regain
walking ability were: passive-assistive movements,
sit-to-stand training, and balance training while sitting
(mean total time per week: 15.29 (SD 22.93), 15.15
(SD 22.75) and 14.85 (SD 16.99) min, respectively).
To test the robustness of the relationship between the
amount and content of therapy in the first 2 weeks the
Table II. Baseline characteristics of study subjects (at first
admission to post-acute hospital or inpatient rehabilitation)
Variable (n  = 150) Median (IQR) Mean (SD)
Age, years
BMI, points
Duration of illness, days*
Duration of mechanical ventilation, days
APACHE II, points
Barthel Index, points
MRC total score at baseline, upper limb
MRC total score at baseline, lower limb 71 (12)
27.4 (6.7)
41 (30)
53 (42)
16 (5)
5 (25)
9.5 (3.25)
9 (3.25)
69.16 (9.02)
29.11 (8.25)
49.13 (29.13)
65.22 (45.14)
16.45 (4.08)
14.68 (19.20)
9.45 (2.6)
8.45 (2.5)
*Duration of illness was defined as the time between the first day on ICU (first
admission to the acute hospital due to the onset of primary illness) until study
onset (admission to post-acute hospital or inpatient rehabilitation).
ICU: intensive care unit; IQR: interquartile range; SD: standard deviation;
BMI: body mass index; MRC: Medical Research Council; APACHE II: Acute
Physiology and Chronic Health Evaluation II.