Psychometric measurement of PEI in chronic musculoskeletal pain
Table VII. Patient Enablement Instrument (PEI) median and interquartile range (IQR) scores of participants with chronic pain based
on their answers on the Global Perceived Effect (GPE) (anchor)
Patient Enablement Instrument
All participants
Global perceived effect
WAD
CR
MixCP
n = 516 (%) Median (IQR) n = 116 (%) Median (IQR) n = 115 (%) Median (IQR)
Very much improved/ Completely recovered 14 (2.7)
Much improved
105 (20.3)
Improved
200 (38.8)
Unchanged
154 (29.8)
Worsened
31 (6.0)
Much worsened
10 (1.9)
Very much worsened
2 (0.4)
11 (9–12)
7 (6–9)
4 (2–6)
1 (0–3)
0 (0–2)
0 (0–2)
1 (0–1)
1 (0.9)
21 (18.1)
55 (47.4)
31 (26.7)
8 (6.9)
0 (0.0)
11 (11–11)
8 (6–11)
5 (3–6)
2 (1–3)
0 (0–1)
9 (7.8)
51 (44.3)
44 (38.3)
4 (3.5)
4 (3.5)
3 (2.6)
n = 285 (%) Median (IQR)
12 (9–12)
7 (6–10)
4 (2–6)
0 (0–2)
0 (0–1)
0 (0–0)
4 (1.4)
33 (11.6)
101 (35.4)
119 (41.8)
19 (6.7)
7 (2.5)
2 (0.7)
9
6
4
1
1
1
1
(9–10)
(6–7)
(2–6)
(0–3)
(0–3)
(0–3)
(0–1)
IQR: interquartile range; WAD: whiplash-associated disorders; CR: cervical radiculopathy; MixCP: mixed chronic pain.
CSQ catastrophizing after treatment was moderate
(r Sp = 0.552). All other correlations between PEI and
self-reported measures were fair (r Sp = –0.250–0.380),
except for none to little correlations between PEI score
and the change between baseline and after treatment in
the EQ-5D Index and self-reported work ability. In all
groups the correlations between PEI score and socio-
demographic variables at baseline were none to little.
Self-rated change
593
All the 516 participants included in the factor ana-
lysis had complete data for the PEI and GPE scores.
The correlation between the PEI and GPE was good
(r Sp = –0.69, p < 0.001, Table VII). Ten participants re-
ported a large deterioration and 2 participants a very
large deterioration; they were excluded from further
analyses, since this number was too small to determine
the SRC for deterioration (29). For the WAD, CR and
MixCP groups the correlations between the PEI and
GPE were r Sp = –0.72, –0.70 and –0.60, respectively.
The SRC ROC , defined as the most optimal ROC
cut-off point, was found at a PEI score of 5 points
for all participants, and had a sensitivity of 85% and
a specificity of 82% (see Figs 1a–d). These findings
correspond to a misclassification of 33% of the parti-
cipants. The positive predictive value (PPV) was 59%
and the negative predictive value (NPV) was 95%. The
overall model quality was 0.87.
For the WAD group, the SRC ROC was found at a PEI
score of 5 points and had a sensitivity of 96% and a
specificity of 75%, corresponding to a misclassification
of 30% of the participants. The PPV was 47% and the
NPV was 99%. The overall model quality was 0.86. For
the CR group, the SRC ROC was found at a PEI score of 6
points and had a sensitivity of 85% and a specificity of
69%. These findings correspond to a misclassification
of 43% of the participants. The PPV was 76% and the
NPV was 80%. The overall model quality was 0.79.
For the MixCP group, the SRC ROC was found at a PEI
score of 4 points and had a sensitivity of 89% and a
specificity of 77%, corresponding to a misclassification
of 32% of the participants. The PPV was 40% and the
NPV was 98%. The overall model quality was 0.86.
DISCUSSION
The results of this study suggest that the PEI demon-
strates fair measurement properties (content validity,
construct validity, and internal consistency) in parti-
cipants with chronic musculoskeletal pain treated in
different settings. Higher patient enablement scores
were related to better health after treatment and to
positive changes in health between baseline and after
treatment. The PEI may be a valid outcome measure for
use in the long-term management of participants with
chronic musculoskeletal pain, although the instrument
may need further development (see below).
Content validity of the Patient Enablement
Instrument
Only a few patients did not complete the questionnaire,
or found the PEI questions “not applicable”, suggesting
that most patients considered the questions relevant to
their condition. This contrasted with another Swedish
study in which many patients (39%) consulting a ge-
neral practitioner (GP) in primary care characterized at
least 1 of the questions as not applicable (19). A large
study performed in primary care in England (30) found
that a larger percentage of patients with longstanding
illnesses (53.6% vs 46.0%), or above-average consul-
tation rates (31.1% vs 24.8%) completed the PEI. It
is possible that patients who received an intervention
over time find the PEI questions more relevant compa-
red with those who completed the questionnaire after
consultation with a GP.
Construct validity of the Patient Enablement
Instrument
CFA showed the 1-factor PEI model had approximate
fit to the data for all 4 groups, with all SRMR values
< 0.08, and standardized residual correlations were
J Rehabil Med 51, 2019