8 / 56
6
VOLUME 13 NUMBER 1 • JULY 2016
RESEARCH ARTICLE
SA JOURNAL OF DIABETES & VASCULAR DISEASE
(i.e. for PAID total alpha and alpha for each subscale). Values
above 0.70 are regarded as satisfactory.
32
Concurrent validity was
assessed using Pearson’s correlation between PAID total score
and other psychosocial variables of interest. In addition, item
total correlation was computed to evaluate the degree to which
differences among patients’ responses to the items were consistent
differences in their total PAID scores. Correlations among measures
of the same attribute should be between 0.40 and 0.80.
7
A lower
correlation indicates either an unacceptably low reliability of one
of the measures, or that the measures are measuring different
phenomena.
33
Discriminant validity was examined by conducting
a multiple regression analysis to determine the extent to which
PAID scores were predicted by body mass index, depression, fear
of hypoglycaemia and diabetes self-care, after adjusting for age,
diabetes type, gender and socio-economic status.
Results
The lowest PAID item missing value accounted for 1.3% and the
highest item accounted for 3.8%. In particular, five items had the
largest missing value percentages; ‘feeling constantly burned out
by constant effort to manage diabetes’ (3.8%), ‘feeling constantly
concerned with food’ (3.8%), ‘feeling diabetes is taking up too
much of your mental and physical energy’ (2.5%), ‘not accepting
diabetes’ (2.5%) and ‘feeling overwhelmed by your diabetes
regimens’ (2.5%). These items accounted for 15.1% missing
values. The Bartlett test of sphericity was highly significant [
χ
2
(190)
= 1005.533,
p
< 0.001]. The KMO value was 0.86. These statistics
suggest that the EFA can be adequately applied.
In the second step, principal component analysis using direct
oblimin was conducted and inspection of the Eigenvalues yielded
a maximum of five factors: 6.69, 2.14, 1.35, 1.14 and 1.02 (values
before rotation). Given the ambiguity as to the number of factors
reported in the literature, we explored solutions with different
numbers of factors, working back from five factors. The five-factor
solution could not be used, as many items loaded on more than
one factor and some of the factors could not be interpreted
Consistent with Snoek and colleagues,
18
a forced four-factor
model was inspected. However, the results were not consistent
with the four-factor model described by Snoek and colleagues,
who had the following four factors: (1) emotional problems related
to diabetes, (2) treatment problems, (3) food problems, (4) lack of
social support.
18
As can be seen in Table 2, our four-factor solution
was not interpretable, with the exception of factor 1, which had
some resemblance to Snoek and colleagues’ emotional subscale;
but items from the food-related, social support and treatment-
related distress subscales loaded on the same factor.
We also inspected a three-factor and a two-factor model using
EFA; both solutions were not interpretable. The first factor of the
two-factor solution consisted of items assessing ‘diabetes stress’
and a second factor containing items covering not only ‘food-
related problems’ but also covering ‘coping with complications and
being overwhelmed with the diabetes regimens’. The second factor
had item combinations that rendered it difficult to interpret, partly
because of various secondary loadings. Therefore the two-factor
solution was discarded.
Lastly, we examined a one-factor solution. Our data provided
the strongest support for a one-factor model, although it had four
items with low loadings below 0.30 (concerned about food and
eating = 0.11, deprivation regarding food = 0.17, coping with
complications = –0.00 and feeling overwhelmed by your diabetes
= 0.23).The retained items had factor loadings ranging from 0.36
to 0.73. Internal consistency remained high even after removing
Table 2.
One-, two-, three-, four- and five-factor solution for the PAID as reported by 157 Zambian participants (aged 12–68 years) with type 1 and
2 diabetes
Factor solution
One-factor Two-factor
Three-factor Four-factor
Five-factor
Shortened item content
F1 F1 F2 F1
F2 F3 F1 F2 F3 F4 F1 F2 F3 F4 F5
Feeling depressed?
0.71 0.70
0.76
0.75
0.74
Worry about low blood sugar reactions? 0.58 0.62
0.63
0.76
0.77
Worry about complications?
0.70 0.70
0.69
0.68
0.69
Feeling angry?
0.63 0.60
0.74
–0.40 0.72
–0.40
0.71
–0.37
Feeling scared?
0.73 0.73
0.77
0.62
0.63
Feeling discouraged with treatment?
0.73 0.72
0.75
0.64
0.61
Mood related to diabetes?
0.63 0.63
0.54
0.61
0.31
0.59
Feeling ‘burned out’?
0.70 0.67
0.60
0.64
0.53
–0.40
Feelings of guilt or anxiety?
0.65 0.64
0.63
0.55
0.55
Diabetes is taking up too much energy? 0.59 0.54
0.51
0.53
0.57
Uncomfortable social situation?
0.63 0.65
0.60
0.76
0.76
Feeling that others are not supportive?
0.67 0.67
0.55
0.39
0.31 0.63
0.65 –0.31
Feeling alone with your diabetes?
0.62 0.64 0.69
0.72
0.70
Not ‘accepting’ your diabetes?
0.68 0.72
0.63
0.31 0.30
0.51
0.53
Unsatisfied with diabetes physician?
0.56 0.60
0.40
0.64
0.59 0.44
0.53 0.46
Concerned about food and eating?
0.11
0.72
0.73
0.75
0.87
Feelings of deprivation regarding food? 0.17
0.76
0.73
0.79
0.76
Coping with complications?
–0.00
0.69
0.75
0.66
–0.42
0.59 –0.42
Not having clear and concrete goal?
0.36
0.43
0.43
0.38
–0.90
Feeling overwhelmed by your diabetes? 0.23
0.59
0.41 –0.59
0.49 –0.57
–0.79 –0.32
Eigenvalue before rotation
6.69 6.69 2.14 6.69 2.14 1.35 6.69 2.14 1.35 1.14 6.69 2.14 1.35 1.14 1.02
% variance before rotation
33.45 33.45 10.70 33.45 10.70 6.73 33.45 10.70 6.73 5.71 33.45 10.70 6.73 5.71 5.12
Eigenvalue after rotation
6.65 2.39 6.58 2.41 1.78 6.03 2.39 1.55 4.08 5.91 2.11 1.41 4.17 2.22
Principle factor analysis using oblique rotation (direct oblimin).