Journal of Rehabilitation Medicine 51-1CompleteIssue | Page 37
34
J. Ponsford et al.
of 2, 3, or 4 were given, reflecting novel mild, moderate or severe
post-injury symptoms respectively (range 2–64).
Secondary outcomes. The Anxiety scale of the Hospital Anxiety
and Depression Scale (HADS) (31) was used to assess anxiety
symptoms post-injury, since it has been shown to be sensitive in
TBI populations (32, 33). HADS-Anxiety requires participants
to rate on a 4-point scale (0–3) their experience of 7 symptoms
over the past week, with total possible Anxiety score ranging
from 0 to 21. A score >7 indicates clinically significant anxiety,
with scores 8–10 indicating mild, 11–14 moderate and ≥15
severe anxiety.
The Quality of Life – Short Form (QoL SF-12) (34) ques-
tionnaire was used to assess quality of life post-injury. This
includes 12 questions, 10 on a 5-point scale and 2 on a 3-point
scale. Higher scores indicate better quality of life than lower
scores. Physical functioning composite scores (PFCS) and
Mental functioning composite scores (MCS) can be derived,
with higher scores indicating greater functioning (mean 50
(standard deviation 10)).
Return to work/study outcomes were also documented. This
included questions on whether participants were working/stud-
ying pre-injury (and for what hours) and at the time of the inter-
view, and whether their duties or enrolment status had changed.
Data analysis
All analyses were conducted with SPSS v22 (SPSS Inc., Chi-
cago, IL, USA) and Stata v12. For the RPQ, a total PCS score
was computed for each participant, whereby scores of 0 and
1 were excluded based on the procedure adopted by King et
al. (30) as 0 = symptom not present and 1 = symptom no more
of a problem than pre-injury. An overall RPQ score was also
computed including all scores. Overall RPQ scores were then
dichotomized into no new post-injury symptoms vs. mild, mo-
derate and severe symptoms for the prediction analysis. HADS
anxiety scale scores were summed from response scores on
HADS anxiety items. For the QoL questionnaire, separate phy-
sical functioning and mental functioning composite scores were
generated and standardized according to Australian norms (35).
Descriptive statistics, including frequencies, were compu-
ted for all measures. χ 2 analysis was conducted to determine
whether there were differences in PCS reporting, based on
recall of receipt of information at discharge from ED. Pearson’s
correlation coefficient was calculated to estimate the linear
correlation between PCS, HADS Anxiety and SF-12 QoL. A
series of logistic regression models, accounting for the clustered
nature of the sample arising from potential correlated responses
of individuals within EDs, were conducted to identify factors
associated with PCS reporting on the RPQ (i.e. mild, moderate
or severe) vs. no new post-injury symptom reporting. Stata
logistic regression with clustered/robust variance estimation
is equivalent to performing a generalized estimated equations
(GEE) analysis assuming an independence correlation structure
(36). Predictive variables were selected based on previous re-
search and available data, and included demographic variables
(age and sex) in Model 1, premorbid psychological history/
substance abuse variables in Model 2, injury-related variables
(GCS, presence of LOC, time post-injury and other injury) in
Model 3, and whether participants recalled having received
information on discharge from ED in Model 4. Statistically
significant predictors in each of these models, defined using
a cut-point of p < 0.05, were then entered together into a final
model (Model 5). Each of these models also controlled for which
intervention the ED was allocated in the NET trial.
www.medicaljournals.se/jrm
RESULTS
Sample characteristics and descriptive on outcome
measures
Of the 1,943 participants completing the NET trial,
536 consented to be contacted for the NET-Plus arm.
Of these, 35 declined to proceed with interview, 16
were deemed not competent, and 119 were lost to
follow-up (no response after 4 attempts). Interviews
were completed by 366 participants, but data from 23
participants were excluded due to co-morbid neuro-
logical conditions. The subsequent NET-Plus sample
included 343 participants. Details of this sample are
shown in Table I. The demographic profile did not
differ from the broader NET trial, where the mean
age of participants was 51 years (for control group)
and 54 years (for intervention group), and 45% were
male. The predominant cause of injury for these 343
participants was incidental falls. In comparison with
the rest of the NET trial sample, the NET-Plus sample
were more likely to have a GCS of 15/15 (95.6% vs
87.6%; p = 0.000), but also more likely to have LOC
(19.8% vs 17.1%; p = 0.001). Other injuries were do-
Table I. Participant characteristics of the NET Plus study. n = 343
unless otherwise specified
Characteristic
Age, years, mean (SD),
[range; IQR]
Sex (male), %
Time post-injury, days, mean (SD),
[range; IQR]
Cause of injury, %
54 (21)
[18–99; 36–71]
54.5
210 (39)
[130–321; 181–239]
Fall 51.9
Violence/assault 14.9
Sport-related injury
Road traffic accident
Miscellaneous*
GCS, %
Score of 14
Score of 15
Loss of consciousness, %
Yes
Unclear or not recorded on file
No
Skull fracture, %
Present
Absent
Not scanned
Reporting other injuries, %
Employed prior to injury (n = 207), %
Employed at follow-up (for those employed prior to
injury; n = 192)
8.2
6.1
19
4.4
95.6
19.8
7.3
72.9
4.7
9.6
85.7
49
60.3
92.8
Reported change in duties due to head injury (n = 21) 10.1
9
Studying prior to injury (n = 31), %
Studying at follow-up (for those studying prior to
injury; n = 18) 58.1
Reported change in enrolment (n = 3)
Pre-injury psychological issues (n = 331), %
Pre-injury illicit substance use issues (n = 331), %
Pre-injury alcohol issues (n = 331), % 9.7
26.3
13
6.3
*Head strike on shelf, kicked by horse.
SD: standard deviation; IQR: interquartile range.