Journal of Rehabilitation Medicine 51-1CompleteIssue | Page 8

Predictors of dropout: a systematic review
Table II. Results for (potential) predictors for dropout in simple analysis
Studies
(Potential) Predictor n Significant
Sociodemographic (19)
Age
Sex a
Education
Social status
Ethnicity
Job code (blue collar vs white collar)
Pre-treatment work status
Original job available
Pre-treatment case settlement
Job satisfaction
Job demand 7
6
2
3
2
1
2
1
1
1
2 27, 28, 29, 30, 32, 33 25
30, 32
24, 27 b , 29, 33 b
36, 37
27 b , 29, 33 b
32
24
24
24
33 b
24
24
24
c
27
24
1
1
1
1
1
1
1
1 27 c
Vibrations in job
Supervisor support
Management control
Work pressure
Height
Weight
No shows
Sick leave days
Patient (21)
Maintenance
Fear of injury
Dysthymia
Pain distress
Depression a
Catastrophizing
MMPI Disability Profile
Axis 2 disorder
Substance use disorder
Anxiety disorder
Opioid dependency
Any cluster A Dx
Any cluster B Dx
Any cluster C Dx
Any cluster D Dx
Self-efficacy
Return to work expectation
Walk distance
Pre-contemplation
Action
Somatization
Disease (21)
Pain intensity a
Age first low back pain
Smoking
ADL scores
Sport activities
Aerobic capacity
Mobility
Isometric abdominal endurance
Isometric back endurance
Compensable body parts
Area of injury
Pain site
Chronicity
Disability
Ability to work
Variability in pain
Meds too long
Length of disability
Duration of work disability
Pre-treatment surgery
Pain behavior
Treatment (2)
Type of institution
Phase of treatment
1
1
1
1
3
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
Non-significant
29
29
29
27
27
30
27 c
28, 24 d
28
29
29
30
24 e , 30
30
24
24
24
24
24
24
24
24
24 b
30
29
30
28
28
28
6
1
1
1
1
1
1
1
1
1
1
1
1
4
1
1
1
1
1
1
1 28 39, 30
1
1 32
32
5
Population
Four studies included patients with chronic
low back pain only (27, 29, 32, 33), the other
4 studies included patients with chronic pain
in various parts of the body (24, 28, 30, 31).
One study conducted a thorough investigation
of predictors of dropout with multiple logistic
regression analyses in a group of 300 (16%)
dropouts from a total sample of 1,845 patients
(16%) (24). In 6 studies the dropout group va-
ried from 14 (23%) to 685 patients (16%) (24,
33) and in 2 studies it was unclear how many
dropouts were investigated (29, 32).
There were no important differences in sex
and in age between 5 of the studies. In 2 studies
70% of the population was female (27, 28) and
one study contained a veteran population with
26% females (33). Most studies described pa-
tients between 40 and 50 years old, apart from
2 studies, which included patients with a mean
age of 35 (29) and 57 years (33).
Reported pain duration varied between a mean
of 16 months and 19 years (24, 33) and was not
reported in 3 studies (24, 29, 32). The mean pain
intensity before treatment ranged from 5 to 8
on a 0–10 NRS scale. One study included only
patients with chronic pain who reported worst
pain levels of 5 and higher (31).
Terms and definitions and incidence of dropout
27 b , 24, 31 b
27
27
27
27
27
27
27
27
24
24
30
30
24, 28, 29, 33 c
27 c
31 c
33 c
24
29
24
33
a
Conflicting results. b Simple analysis not reported, considered to be non-significant based on
description in methods section in original article. c Simple analysis not reported, considered to be
significant based on description in methods section in original article. d Beck Depression Index (BDI)
self-report measure of depression. e Structured interview for DSMIV major depressive disorder.
ADL: activities of daily living; Any cluster A Dx: paranoid; schizoid; schizotypal; Any cluster B Dx;
antisocial; borderline; histrionic; narcissistic; Any cluster C Dx: avoidant; dependent; obsessive-
compulsive; Any Cluster D Dx: otherwise; MMPI: Minnesota Multiphasic Personality Inventory.
Different terms were used to describe dropout:
(i) withdrawal from treatment; (ii) early dischar-
ge; (iii) non-completion and treatment dropout.
Only 4 studies described a definition of dropout
(24, 29, 32, 33). The incidence of dropout ranged
from 10% in a study in the UK to 51% in a study
in the USA (29, 31). Two studies did not report
the percentage of dropout (25, 30).
Methodological quality
Analysis of the overall agreement with re-
gard to the methodological quality between
the reviewers revealed a Cohen’s weighted
kappa of 0.70 (confidence interval (CI) 0.52;
0.88) indicating substantial agreement (34).
All studies were considered to be of low qua-
lity according to Hayden’s recommendations
(23) (Table I). This was due to a serious risk
of bias in most studies for study attrition,
prognostic factor measurement and statistical
analyses and reporting. For most studies the
reporting of the first step in the statistical analy-
ses, the univariate analyses were poor in contrast
J Rehabil Med 51, 2019