Disability and Rehabilitation
ISSN: 0963-8288 (Print) 1464-5165 (Online) Journal homepage: http://www.tandfonline.com/loi/idre20
Predictors of multidisciplinary treatment outcome in patients with chronic musculoskeletal pain Anne M. Boonstra, Michiel F. Reneman, Berend R. Waaksma, Henrica R. Schiphorst Preuper & Roy E. Stewart To cite this article: Anne M. Boonstra, Michiel F. Reneman, Berend R. Waaksma, Henrica R. Schiphorst Preuper & Roy E. Stewart (2015) Predictors of multidisciplinary treatment outcome in patients with chronic musculoskeletal pain, Disability and Rehabilitation, 37:14, 1242-1250, DOI: 10.3109/09638288.2014.961657 To link to this article: http://dx.doi.org/10.3109/09638288.2014.961657
Published online: 17 Sep 2014.
Submit your article to this journal
Article views: 137
View related articles
View Crossmark data
Full Terms & Conditions of access and use can be found at http://www.tandfonline.com/action/journalInformation?journalCode=idre20 Download by: [Australian National University]
Date: 06 November 2015, At: 19:34
http://informahealthcare.com/dre ISSN 0963-8288 print/ISSN 1464-5165 online Disabil Rehabil, 2015; 37(14): 1242–1250 ! 2014 Informa UK Ltd. DOI: 10.3109/09638288.2014.961657
RESEARCH PAPER
Predictors of multidisciplinary treatment outcome in patients with chronic musculoskeletal pain Anne M. Boonstra1, Michiel F. Reneman2, Berend R. Waaksma1, Henrica R. Schiphorst Preuper2, and Roy E. Stewart3
Downloaded by [Australian National University] at 19:34 06 November 2015
1
Department of Rehabilitation, ‘Revalidatie Friesland’ Center for Rehabilitation, Beetsterzwaag, The Netherlands, 2Department of Rehabilitation, Center for Rehabilitation, University Medical Center Groningen, University of Groningen, Groningen, The Netherlands, and 3Department of Health Sciences, Community & Occupational Medicine, University Medical Center Groningen, University of Groningen, Groningen, The Netherlands Abstract
Keywords
Purpose: The present study aimed to identify predictors of rehabilitation outcome for patients with chronic musculoskeletal pain (CMP) and psychological problems. Methods: A retrospective cohort study including 230 adult patients with CMP admitted for multidisciplinary pain rehabilitation. Potential predictors were patient characteristics, duration of complaints, baseline functioning, pain, personality, coping style, fear of movement, psychological distress and type of treatment. Outcome measures were physical functioning, mental health, pain and patient-reported effect. Multiple (logistic) regression models were used to identify predictors. Results: Patients who were more disabled and patients with more pain benefitted more from the rehabilitation treatment than less disabled patients or those with less pain. Age, work status, vitality, depression and coping style also predicted outcomes significantly. The models explained between 27 and 80% of the outcomes. There was an interaction between type of treatment, work status and the baseline pain score as regards the outcome in terms of pain. Conclusions: No strong predictors of treatment outcome were found other than the baseline scores of the respective outcome variables. More disabled patients and patients with more pain benefitted more from the rehabilitation program. Other predictors improved the prediction models slightly.
Chronic pain, clinical outcome, musculoskeletal pain History Received 15 June 2013 Revised 26 July 2014 Accepted 1 September 2014 Published online 17 September 2014
ä Implications for Rehabilitation
It remains challenging to correctly predict the outcome of treatment from patients’ baseline sociodemographic and psychological characteristics; predictors other than baseline scores of the outcome variables are only slightly associated with treatment outcome. Patients with chronic musculoskeletal pain and poor physical functioning or mental health benefit most from pain rehabilitation. Older patients benefit less from a pain rehabilitation program than younger patients in terms of physical functioning. Pain reduction during a pain rehabilitation program is greatest in patients with high pain intensity who are not at work at the start of the rehabilitation program. Coping style influences the outcome of rehabilitation of patients with chronic musculoskeletal pain.
Introduction Disability due to chronic pain is multifactorial, and a biopsychosocial perspective is needed for better treatment of the patients. The most frequently recommended approach is that of multidisciplinary rehabilitation including elements of cognitive behavioral therapy [1]. Patients are known to differ in their response to this multidisciplinary approach, but it is still unclear who will benefit the most, remain unchanged, or even worsen [2,3].
Address for correspondence: Anne M. Boonstra, Revalidatie Friesland, PO Box 2, 9244 ZN Beetsterzwaag, The Netherlands. Tel: +31 512 389329. Fax: +31 512 389244. E-mail: [email protected]
Moreover, currently available treatments provide only modest improvements in terms of pain and functioning of chronic pain patients [3]. Identifying predictors of outcomes may help clinicians to choose the right program for their patients and to develop new treatments for those for whom existing treatments are insufficiently effective. A wide range of predictors of pain rehabilitation outcome has been reported. Better outcome has been reported for patients with high scores on depression [4–6], higher fear-avoidance beliefs [6], higher scores on perceived distress [7,8] or optimistic attitudes [7] on the Multidimensional Pain Inventory (MPI), low need to socialize [7], younger age [4,7,9,10] or higher educational level [11]. Poorer outcome has been reported for patients with greater psychological dysfunction [12] or lower quality of performance
Predictors of pain rehabilitation
Downloaded by [Australian National University] at 19:34 06 November 2015
DOI: 10.3109/09638288.2014.961657
on lumbar dynamometry [8,12], or those who are unemployed [13]. Mixed findings have been reported for anxiety [4,5] and gender [14,15]. Others, however, found several of these factors, such as age, educational level, work status, depressive symptoms and fear of avoidance not to be predictors [5–7,11,16]. Recently, a review was published about predictors of multidisciplinary treatment outcome in fibromyalgia [17]. Poor outcomes were predicted by depression, by the ‘‘disturbance and pain’’ profile of the Minnesota Multiphasic Personality Inventory (MMPI), by strong beliefs in fate and by high disability. Better outcomes were predicted by a poorer baseline status, the ‘‘dysfunctional’’ and ‘‘adaptive copers’’ profiles of the MPI and high levels of pain. Inconclusive evidence was found for other demographic and clinical factors [17]. The discrepancies between these studies may be due to differences in sample characteristics, outcome measures and/or treatments [7,11,18]. Several authors [8,17,18] have recommended studies to compare different methods, settings and durations of multidisciplinary treatments and to examine their relations with patient characteristics in more detail, in order to detect differential effects. Knowing which patients may have a poor treatment outcome should help to design more effective rehabilitation programs. It may also help to refer or select patients for the most suitable treatment according to their characteristics. Moreover, knowing which variables are important in predicting the outcome of a treatment may help to inform patients more reliably about the effects that can be expected. Finally, it is important for conducting randomized controlled trials to know predictors of outcome, in order to check or control for potential confounders. Studies about predictors may therefore help in examining the effects of rehabilitation treatments in the future. Our study focused on the outcome and outcome predictors of patients treated in rehabilitation programs in usual care. Our study focused on patients with chronic musculoskeletal pain (CMP) and moderate-to-severe psychosocial problems, as this is a large subgroup treated in rehabilitation centers in Western Europe. The aim of the present study was to identify predictors of the rehabilitation outcome for those patients with CMP.
Materials and methods Patients The inclusion criteria for participation in the present study were: patients with CMP referred to the ‘‘Revalidatie Friesland’’ Rehabilitation Center in Beetsterzwaag (The Netherlands), who were given inpatient or outpatient cognitive behavioral treatment, who were aged above 18 years, and whose pain had lasted for over 3 months. Patients were referred for the inpatient treatment by rehabilitation physicians from all over the northern part of the Netherlands, mainly from the province of Friesland, while patients referred for the outpatient treatment came from the area where the center is located. Another inclusion criterion that had to be met was the involvement of a psychologist in the treatment, by way of operationalization of having moderate-to-severe psychosocial problems. Involvement of a psychologist was, e.g. indicated if the patient experienced high psychological distress, pain-related fear, mild or moderate depression due to acceptance problems, compulsive behavior, personality disorder, etc. Exclusion criteria were insufficient command of Dutch, co-morbidity with severe negative consequences for physical functioning, current major psychiatric disorder (active psychosis, severe depression, addiction, etc.) and unwillingness to provide data for research purposes. The study included patients who started treatment between September 2003 and December 2010. If patients were treated twice in the inclusion period, we used the data of the first episode for our study.
1243
Measurements Potential predictors The potential predictors we selected were those mentioned in the literature [4–15] insofar as they were relevant to our study population; they were supplemented with variables that were considered potential predictors on the basis of clinical experience. The following characteristics were assessed, using a selfconstructed questionnaire: duration of current pain, age, gender, marital status, age of children if any, educational level, being employed or self-employed, currently in work (work status, i.e. not being on sick leave, disabled for work or homemaker), outstanding litigation or unresolved workers’ compensation claims, and receiving a benefit (e.g. social benefit, unemployment benefit). Baseline functioning, general and mental health status, and pain intensity were assessed using the Dutch version of the Short Form 36 Health Survey (SF-36) [19,20]. This instrument consists of 36 questions, and measures eight dimensions: physical functioning, social functioning, physical role restriction, emotional role restriction, mental health, vitality, pain, general health and health change (scores range from 0 to 100 for each dimension; a lower score means more disability or more pain). Personality was assessed by the Dutch Personality Questionnaire (DPQ) [21]. The DPQ is based on a translated, shortened version of the California Psychological Inventory. The DPQ1 consists of 132 items, and the DPQ2 of 140 items, to which subjects respond by ‘‘correct’’, ‘‘don’t know’’ or ‘‘incorrect’’. Items are distributed over seven scales: neuroticism, social anxiety, rigidity, hostility, egoism, dominance and self-esteem. DPQ2 is a revised version [22], and this is the version which was used for the study sample after July 2008, i.e. for 22% of the study sample. The DPQ1 scores were recoded to DPQ2 scores using the formula provided by the developers [22]. Coping reactions were measured with the Coping with Pain Questionnaire (CPQ) [23], a Dutch adaptation of the Coping Strategy Questionnaire (CSQ) [24]. The CPQ assesses the use of cognitive strategies: ignoring pain sensations, coping selfstatements, re-interpreting pain sensations, catastrophizing, ‘‘praying and hoping’’, diverting attention and ability to control pain, as well as one behavioral strategy, viz. increasing one’s activity level. The ‘‘active coping’’ factor in the CSQ is based on the scores for ignoring pain sensations, coping selfstatements, reinterpreting pain sensations and increasing activity level, while the ‘‘helplessness’’ factor is based on the scores for catastrophizing, praying and hoping, and diverting attention [23]. We calculated the ‘‘active coping’’ and ‘‘helplessness’’ factors by summing the relevant scores. A higher score means that a patient is more likely to use this specific coping style predominantly. Pain-related fear was assessed by the Tampa Scale of Kinesiophobia, Dutch version (TSK) [25]. The TSK consists of 17 items, each provided with a 4-point Likert scale with scoring alternatives ranging from ‘‘strongly disagree’’ to ‘‘strongly agree’’. Higher scores indicate more fear. Psychosocial distress was assessed by the Symptom Checklist 90-Revised, Dutch version (SCL-90-R) [26]. In the SCL-90-R, patients are instructed to rate 90 distress symptoms on a 5-point Likert scale ranging from ‘‘not at all’’ to ‘‘extremely’’. The statements are assigned to eight dimensions reflecting various types of psychopathology: anxiety, agoraphobia, depression, somatization, insufficiency, sensitivity, hostility and insomnia. The SCL-90-R also includes a total score for ‘‘psychoneuroticism’’, which is a global measure of distress. Higher scores indicate higher levels of distress.
1244
A. M. Boonstra et al.
Outcome variables Outcome variables were physical functioning, mental health and pain measured by the SF-36, and patient-reported effect (PRE). PRE was assessed after treatment by asking the patients to report the global perceived effect of the treatment on their symptoms, using a Likert scale: completely recovered, very much improved, much improved, unchanged, worse, much worse, very much worse. These four outcome variables reflect important rehabilitation goals [27] and have been recommended by the Initiative on Methods, Measurement, and Pain Assessment in Clinical Trials (IMMPACT) [28].
Downloaded by [Australian National University] at 19:34 06 November 2015
Procedure and rehabilitation program Patients were referred to multidisciplinary rehabilitation for their pain-related disabilities. Referral was done by rehabilitation physicians working in the northern part of the Netherlands. Eligibility for rehabilitation treatment was judged by a rehabilitation physician, and the aims and contents of the treatment were explained to the patients in general terms. Patients also consulted the psychologist at the pain department of the rehabilitation center before the treatment started. The final choice to enter a program was a shared decision by patient, physician and psychologist. Treatments were based on cognitive behavioral concepts [1]. The team consisted of a rehabilitation physician, a physiotherapist, an occupational therapist and a psychologist, and often also a social worker. Program goals depended on the characteristics of the complaints and the aims of the patient, and included reduction of activity limitations, participation problems and psychological distress. The program was delivered individually as well as in groups. The most commonly used treatment modalities were teaching ergonomic principles, graded activity and behavioral therapy. Treatment generally focused on optimization of functioning. Outpatient programs were mostly provided on 3 days per week for one to four hours a day, over a period of approximately 10 to 16 weeks. The inpatient program was provided on Monday through Friday, with patients going home for the weekend. This program lasted approximately 6 to 10 weeks and was sometimes followed by outpatient treatment. The length and intensity of the program were determined by the characteristics of the complaints and the progress made by the patient during the treatment, as is usual in clinical practice. If a patient started an outpatient program, and it turned out during the observation period, or in the first period of treatment, that the treatment targets were not going to be met, patients could still be referred to an inpatient program. The SF-36 was administered at the beginning (baseline) and at the end of the program (discharge). The first assessment (baseline) took place just before the start or during the first 2 weeks in the case of an outpatient program, or during the first week in the case of an inpatient program. The discharge assessment of patients in outpatient programs took place in the final week or during the first 4 weeks after discharge, while the assessment of patients in inpatient programs took place 2 to 4 weeks after discharge or, in the case of continuation as outpatient treatment, in the final week or during the first weeks after the end of the program. The other assessments were administered in the first or second week of the program as part of regular clinical procedures. Statistical analysis Demographic characteristics are presented as mean and standard deviation (age) or percentages (other variables). Since the duration of complaints and the scores on the SF-36, DPQ, CPQ, TSK and SCL-90-R were ordinal and/or had a skewed distribution, medians and quartiles were calculated. To enable
Disabil Rehabil, 2015; 37(14): 1242–1250
comparison with other studies, we also calculated means and standard deviations for the duration of complaints and SF-36 scores. Differences between outpatients and inpatients were tested with the t-test for normally distributed data (checked with the Kolmogorov–Smirnov test), the Mann–Whitney U-test for nonnormally distributed data, and the Chi-square test for dichotomous data. The effect sizes (ES) of the SF-36 scores were calculated using the following formula [29]:
1 X 2 pffiffiffiffiffiffiffiffiffiffiffiffiffiffi X = ð1 rÞ SDpooled The following formula was used for SDpooled:
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ððSDbaseline Þ2 þ ðSDdischarge Þ2 Þ 2 where X1 is the baseline value, X2 is the value at discharge, r is the correlation coefficient between X1 and X2, and SD the standard deviation. The ES interpretation was as follows: 50.20 trivial effect, 0.20 to 0.50 small effect, 0.50 to 0.80 medium effect and 0.80 large effect [29]. Variables assessed as potential predictors were: age (in years), gender (male ¼ 0), marital status (single (0) versus married or living with a partner (1)), educational level (low: primary school to intermediate vocational education (0) versus high: higher vocational education to university (1)), employment (employed or self–employed (0) versus unemployed, including students and homemakers (1)), work status (not currently in work (0) versus in regular or adapted work (1)), benefit (no benefit (0) versus receiving a benefit, e.g. social benefit or unemployment benefit (1)), age of youngest child (child 12 years (0) versus child 412 years or no children living at home (1)), being involved in an ongoing litigation or unresolved workers’ compensation claim procedure (no procedure ¼ 0), duration of current complaints (years), baseline scores on the SF-36, DPQ, CPQ, TSK and SCL-90-R, or domains thereof, and type of treatment (inpatient (0) or outpatient (1)). Outcome parameters were PRE and change scores of SF-36 between admission and discharge (discharge score minus baseline score) for the domains of physical functioning, mental health and pain. In view of the small number of patients reporting worsening of their complaints, ‘‘PRE’’ was dichotomized into no improvement (no effect or worsening (0)) and improvement (1). Regression analyses were conducted to identify predictors of outcomes in terms of physical functioning, mental health, and pain (linear regression) and to analyze predictors of PRE (binary logistic regression). The analysis of each outcome measure was done in seven steps. Step 1 involved univariate analyses. Variables with p values 50.20 were identified as potential predictors. Step 2 involved analyses of blocks of related variables. In view of the large number of variables relative to the number of patients we performed a connecting analysis. Related potential predictors were clustered into five groups of variables: (a) work-related variables (employment, work status, benefit), (b) SF-36 scores excluding the pain score, (c) DPQ scores, (d) CPQ scores and (e) SCL-90 scores. Multiple regression or logistic analyses per block were performed with the above-mentioned variables as independent variables and the outcome variables as dependent variables. In the block analysis of the SF-36 scores, the model did not include the score for the same domain of the outcome measure (for example, the SF-36 score for ‘‘physical functioning’’ was not entered as an independent variable in the model with the outcome measure of physical functioning as the dependent variable).
Predictors of pain rehabilitation
Downloaded by [Australian National University] at 19:34 06 November 2015
DOI: 10.3109/09638288.2014.961657
In step 3, variables whose p values remained 50.20 in the block analyses, and potential predictors that were not entered in the block analyses (characteristics, SF-36 pain score, SF-36 score for the same domain of the outcome measure, TSK score and type of treatment) were entered in the next model. In step 4, variables with a p value50.05 in step 3 were entered in the final model. The next two steps were done to examine interactions between type of treatment (inpatient/outpatient: inpatient program ¼ 0) and (potential) predictors. In step 5, the variables entered in step 3, as well as the potential interaction variables with the type of treatment as independent variables, were entered in the regression models with the outcome variables as dependent variables. In step 6, variables with a p value50.05 in step 5 were entered in the final interaction models. Since a relatively large percentage of the eligible patients did not complete the outcome assessment, we compared eligible and included patients regarding the variables that we found to be outcome predictors. We used the t-test for normally distributed data (checked with the Kolmogorov–Smirnov test), the Mann– Whitney U-test for non-normally distributed data, and the Chi-square test for dichotomous data. The study covered a period of seven years. Since it was theoretically possible that the outcome changed over this period, we checked the distributions of the outcome visually using boxplots (continuous outcome variables) or histograms (patient-reported effect, i.e. dichotomous variable).
Results Study population Three hundred and sixty-four patients were eligible for this study (n ¼ 364). Psychological data of eight of the patients were missing, as were data on potential predictors for two of the patients. Five patients started treatment twice. Two hundred and forty-one patients had an outcome assessment (response rate 66% for outpatients and 67% for inpatients). Nine patients did not give permission to use their data for research purposes, and so were excluded. Finally, 230 patients were included in the analysis, 107 outpatients and 123 inpatients. Eight of these inpatients had a (short) outpatient treatment before being admitted to an inpatient program, and 10 of the inpatients continued as outpatients after completing the inpatient program. They were included in the ‘‘inpatient group’’. The total duration of treatment (excluding the hours spent by the nursing staff and physician or trainee) was 150 h (SD 35 h) for the inpatient group, including 17 h (SD 5 h) of treatment by the psychologist and social worker. The corresponding values for the outpatient group were 101 h (SD 37 h) and 24 h (SD 13 h), respectively. Most patients suffered from low back pain (31%), fibromyalgia (23%), widespread pain (13%) or neck pain (16%). Characteristics of patients are presented in Table 1. Outcomes Most patients in both types of program improved, with ESs for the total group of 0.99 for physical functioning; 0.24 for mental functioning; and 1.65 for pain. A subgroup comprising 4% of the sample reported worsening (Table 2). Predictors of outcome Better outcomes were associated with several patient characteristics as well as baseline pain intensity and various baseline physical and psychosocial variables (Table 3). Table 4 presents the results of the multiple (logistic) regression analyses. Due to missing data and list-wise deletion, the multiple
1245
regression analyses included fewer than the total number of respondents. Greater improvement in physical functioning was associated with an active coping style, and smaller improvement with older age and better physical functioning at the start of the program. The adjusted R2 in the unadjusted model with baseline physical functioning as predictor and physical outcome as dependent variable was 0.19, compared to 0.27 in the final model with physical functioning, age and active coping as predictors and physical outcome as dependent variable. Smaller improvement in mental health was associated with better mental health and a ‘‘praying and hoping’’ coping style, vitality and depression at baseline. The adjusted R2 in the unadjusted model with baseline mental functioning as predictor and mental health outcome as dependent variable was 0.78, compared to 0.80 in the model with baseline mental health functioning, ‘‘praying and hoping’’, vitality and depression as predictors and mental health outcome as dependent variable. Less pain reduction was associated with being currently in work, having less pain at baseline. The adjusted R2 in the unadjusted model with baseline pain as predictor and pain outcome as dependent variable was 0.66, compared to 0.68 in the model with baseline pain and work status as predictors and pain outcome as dependent variable. Higher PRE was associated with being employed and less ‘‘active coping’’. Interaction We explored a two-way interaction model, which indicated significant effects of pain, work status and treatment program, and interactions between the type of treatment and pain, and between pain and work status. As regards the association between baseline pain score and outcome in terms of pain, an interaction was observed with type of treatment and with work status (Table 5). Included and not included patients The predicting variables did not differ significantly between the patients included (n ¼ 230) and those not included in the study (n ¼ 132): baseline physical functioning (mean (SD): 41 (19) and 39 (24), respectively); baseline mental health (60 (19) and 59 (18), respectively); baseline pain (32 (16) and 31 (18), respectively); age (43 (12) and 41 (12), respectively); active coping (99 (37) and 91 (32), respectively), helplessness (61 (27) and 57 (26), respectively); ‘‘praying and hoping’’ (18 (12) and 17 (13), respectively), fear (37(7) and 36 (7), respectively), employment (53 and 51%, respectively); child 12 yrs (22 and 21%, respectively). Outcome variables during the years of study According to the boxplots or histogram, the outcome variables did not show a trend towards better or worse outcome over the years of the study.
Discussion The aim of the study was to identify predictors of rehabilitation treatment outcomes for patients with CMP and moderateto-severe psychosocial problems. The outcomes in terms of physical functioning, mental health and pain were significantly predicted by their respective baseline scores: lower scores predicted greater effects. This may not be surprising, as lower scores indicate a poorer status and thus more room for improvement. The variances in the final models were mainly explained by the baseline scores, as shown by the adjusted R2 of the various models. However, other predictors might, although of less significance, be clinically relevant as well. Coping style was a predictor of the outcomes in terms of physical functioning and
1246
A. M. Boonstra et al.
Disabil Rehabil, 2015; 37(14): 1242–1250
Table 1. Demographic and clinical characteristics of patients with chronic musculoskeletal pain (baseline data).
Characteristics Age (yrs; mean (SD), range) Gender (% male) Marital status (% single) Educational level (% low) Age of youngest child (% 12 yr) Ongoing procedure (% no) Employed (% yes) Work status (% not in work) Benefit (% no) Duration of current complaints in years Mean (standard deviation) Median (quartiles)
Outpatient program (n ¼ 107)
Inpatient program (n ¼ 123)
p Value
43 (10,19–69) 36 21 76 36 62 66 61 40
43 (13,18–76) 21 27 83 11 76 42 80 34
0.97a 0.02 0.48 0.15 50.001 0.03 0.001 0.002 0.33
4.9 (5.3) 3.0 (1.5–6.0)
5.9 (5.8) 3.8 (2.0–9.5)
0.09b
Downloaded by [Australian National University] at 19:34 06 November 2015
Median (quartiles) Pain (SF-36 scores, range 0–100) –Pain 33 (22–45) Functioning and health (SF-36 scores, range of each item 0-100) – Physical functioning 50 (30–65) – Social functioning 50 (38–75) – Physical role 0 (0–13) – Emotional role 33 (0–100) – Mental health 68 (52–76) – Vitality 40 (25–45) – General health 50 (35–65) – Health change 25 (25–50) Personality (DPQ scores, range of each item 0–40) – Neuroticism 15 (7–23) – Social anxiety 9 (4–19) – Rigidity 27 (21–31) – Hostility 11 (6–19) – Egoism 6 (3–10) – Dominance 19 (14–24) – Self-esteem 28 (23–33) Coping (CPQ) (scoring range) – Ignoring pain sensations (0–60) 31 (19–39) – Coping self-statements (0–60) 35 (30–45) – Reinterpreting pain sensations (0–60) 8 (4–19) – Catastrophizing (0–60) 22 (14–34) – Praying and hoping (0–60) 15 (7–23) – Diverting attention (0–60) 20 (9–28) – Ability to control pain (0–20) 8 (3–11) – Increasing activity level (0–60) 22 (17–30) Composite scores – Active coping (0–180) 103 (77–119) – Helplessness (0–240) 58 (41–78) Fear of movement (TSK) (scoring range 17–61) – Fear 37 (33–40) Psychological distress (SCL-90-R) (scoring range) – Anxiety (10–50) 16 (12–20) – Phobic anxiety (agoraphobia) (7–35) 7 (7–9) – Depression (16–80) 30 (26–39) – Somatization (12–60) 28 (22–33) – Insufficiency (9–45) 21 (17–26) – Sensitivity (18–90) 27 (22–33) – Hostility (6–30) 8 (7–11) – Insomnia (3–15) 7 (5–12) – Psychoneuroticism (GSI) (90–450) 161 (140–194)
Mean (SD)
Median (quartiles)
Mean (SD)
35 (15)
33 (12–45)
28 (17)
47 54 10 45 64 38 51 35
35 38 0 33 56 30 40 25
(25–45) (25–50) (0–0) (0–100) (44–72) (18–43) (30–55) (0–50)
35 39 7 47 58 31 43 27
16 14 27 14 6 17 27
(9–26) (6–22) (22–33) (7–20) (4–9) (12–23) (20–32)
0.11b 0.01b 0.28b 0.17b 0.91b 0.58a 0.10b
28 38 8 24 16 19 5 24
(17–38) (26–45) (1–18) (15–34) (10–25) (10–27) (0–10) (17–30)
0.59a 0.77a 0.30b 0.74a 0.30b 0.64b 0.03b 0.37a
(20) (24) (22) (43) (17) (16) (19) (24)
(17) (25) (18) (44) (20) (18) (17) (25)
38 (30–43) 18 9 35 29 24 27 8 9 181
(14–23) (7–12) (26–44) (24–36) (18–27) (23–39) (7–11) (6–12) (146–212)
50.001b 50.001b 0.27b 0.86b 0.01b 0.001b 0.003b 0.02b
0.82a 0.64a
99 (73–127) 59 (40–82) 36 (6.4)
0.001b
37 (8.2)
0.46a 0.01b 0.001b 0.04b 0.07b 0.06b 0.16b 0.34b 0.11b 0.02b
Differences between outpatients and inpatients were tested with the t-test for normally distributed data, the Mann–Whitney U-test for non-normally distributed data, and the Chi-square test for dichotomous data. a t-test; bMann–Whitney U-test.
mental health. Active coping was positively correlated with a better physical functioning outcome, but patients with a more active coping style were less likely to perceive improvement according to the PRE. An active coping style may help patients to
adopt the exercises and improve their physical condition, and hence increase their physical activities. On the other hand, patients with an active coping style may be disappointed in the degree of overall improvement in view of their efforts,
Predictors of pain rehabilitation
DOI: 10.3109/09638288.2014.961657
1247
Table 2. Outcomes in terms of physical functioning, mental health and pain at admission and discharge, and difference between discharge and admission, with effect sizes and patient-reported effect at discharge. Outpatient treatment
Downloaded by [Australian National University] at 19:34 06 November 2015
Pain (SF-36 scores) – Pain Functioning (SF-36 scores) – Physical functioning – Mental health Patient reported effect (%) Completely recovered Very much improved Much improved Unchanged Much worse Very much worse Far worse
Admission (mean, SD)
Discharge (mean, SD)
Difference (mean, SD)
35 (15)
54 (17)
18.5 (28.5)
47 (20) 64 (17)
57 (20) 64 (9)
9.7 (17.9) 0.12 (16.6)
0 41 34 19 5 1 0
and may therefore describe the complaints as unchanged. The ‘‘praying and hoping’’ coping style was negatively associated with an improvement in mental health. ‘‘Praying and hoping’’ is a passive coping style, and patients with this coping style may have benefitted less from the treatment because their attitude prevented them from adopting the recommendations in their daily lives. Personality was not found to be a predictor of treatment outcome. This is in agreement with Michaelson et al. [7]. However, patients with personality disorders were excluded from the rehabilitation programs and were therefore not included in our study. Our finding cannot be generalized to patients beyond the scope of the study. As regards the association between baseline pain score and outcome in terms of pain, an interaction was observed with type of treatment (inpatient versus outpatient). The improvement in pain score between baseline and discharge (pain outcome) was larger for patients with more severe baseline pain (low SF-36 score), and the difference in improvement between inpatients and outpatients, as well as between patients currently in work and those not currently in work changed with increasing baseline pain levels. We did not find a clear explanation for this phenomenon. Perhaps inpatients who had been in work before admission had more trouble adjusting to a new daily structure without work than outpatients. Inpatients may have allowed themselves to feel the pain and stress after admission, more so than outpatients, as many patients are used to ignoring their feelings, including feelings of pain. Further research needs to be done to explain this phenomenon. The data were collected over many years. Neither the type of treatment, i.e. cognitive behavioral therapy, nor the outcome of the treatment programs, changed over time. However, it is very likely that the content of the treatment programs changed in these years as the expertise of the therapists developed further during the study. This may have influenced the predictors of outcome. Patients with less favorable outcomes may have been excluded over time or treated in a better way. There is a possible bias in the fact that no predictors were found, because of increasing homogeneity during the years of this study. However, we found no reason to assume that the predictors found in our study were biased, i.e. were not really predictors. The effect sizes we found for physical functioning and pain are within the range of those found in other studies [14,30–35], while those for mental health were smaller than those reported by others [14,30]. This small effect size for the mental functioning outcome may be caused by
Inpatient treatment Admission (mean, SD)
Discharge (mean, SD)
Difference (mean, SD)
Effect size
1.64
28 (17)
54 (18)
25.4 (27.4)
2.14
0.78 0.02
35 (17) 58 (20)
53 (22) 63 (9)
17.8 (21.6) 5.4 (20.5)
1.44 0.44
Effect size
1 49 31 17 1 1 1
the fact that the mental functioning score at baseline was relatively high in comparison with those in other studies [14,30]. It is unclear whether this small effect size has influenced the identification of predictors. The effect sizes resulting from inpatient treatment were larger than those for outpatient treatment, which may be caused by the fact that patients benefit more from inpatient than outpatient treatment, or by the fact that the levels at the beginning of the program were lower among the inpatients. Since this was not one of our research questions, we did not explore this finding any further. Bremander et al. [4] found anxiety and depression to be predictors of outcome. This was not confirmed by our study. Bremander et al. [4] used the Hospital Anxiety and Depression Scale (HADS) for their assessment, whereas we used the SCL-90-R, which may not be specific enough to assess this variable. Another possible explanation is that Bremander et al. included fewer, and different, potential predictors in their study (i.e. age, gender, antidepressant medication, pain intensity and pain distribution) than we did. This demonstrates once more how difficult it is to compare studies. No overview of variables associated with outcome in patients with CMP is available in the literature, except for fibromyalgia [17], making comparisons with other studies difficult. Moreover, the considerable differences in study design also makes it difficult to compare some of the other findings of our study with those of others. For example, our study found age to be only a predictor of the outcome in terms of physical functioning, while gender was not found to be an outcome predictor at all, in contrast to what was found in other studies [7,10,11,14–16]. We found a proportion of explained variance between 0.23 and 0.79, which was much larger than that reported by, e.g. Van der Hulst et al. [6]. This may have been caused by differences in study sample and/or differences in outcome measures. To unravel the complex association between the large number of possible predictors and the large variety of outcome variables, there is an urgent need for more studies on this issue, finally followed by systematic reviews to combine the findings. Our findings may add to such reviews. Just like Morley et al. [32], who evaluated usual care, we found that a very small percentage of the patients perceived worsening of their complaints. It would be very interesting to find predictors that could identify patients who are likely to get worse, as this might help physicians decide when to advise against cognitive behavioral treatment. However, the small percentage of worsening patients did not enable us to identify these predictors.
1248
A. M. Boonstra et al.
Disabil Rehabil, 2015; 37(14): 1242–1250
Table 3. Results of univariable analyses in total group (inpatients and outpatients). Physical functioning outcome
Mental health outcome
Downloaded by [Australian National University] at 19:34 06 November 2015
n Characteristics Age Gender Marital status Educational level Age of youngest child Ongoing procedure Duration of current complaints Work Employed Work status Benefit Pain (SF-36 scores) – Pain Functioning and health (SF-36 scores) – Physical functioning – Social functioning – Physical role – Emotional role – Mental health – Vitality – General health – Health change Personality (DPQ) – Neuroticism – Social anxiety – Rigidity – Hostility – Egoism – Dominance – Self-esteem Coping (CPQ) – Ignoring pain sensations – Coping self-statements – Reinterpreting pain sensations – Catastrophizing – Praying and hoping – Diverting attention – Ability to control – Increasing activity level Composite scores – Active coping – Helplessness Fear (TSK) – Fear Psychological distress (SCL-90-R) – Anxiety – Phobic anxiety (agoraphobia) – Depression – Somatization – Insufficiency – Sensitivity – Hostility – Insomnia – Psychoneuroticism Type of treatment
0.004 – 0.15 – – – –
220 220 220 220 217 218 218 221
– – –
Pain outcome
n – – – 0.10 – – –
216 216 216 216 214 216 213 216
– 0.13 –
Patient-reported effect
n 0.15 0.18 0.10 0.006 0.08 –
222 222 222 222 221 222 213 222
0.08 50.001 0.13
n – – – – 0.14 0.03 –
229 229 229 229 227 229 226 229
0.01 0.03 –
0.001
221
0.09
216
50.001
222
–
222
50.001 0.04 – – – – 0.19 0.05
221 218 221 219 219 219 218 220 215
– 50.001 0.004 50.001 50.001 50.001 50.001 –
216 215 216 214 216 216 215 215 210
50.001 50.001 50.001 0.006 0.006 0.002 0.014 50.001
221 219 222 220 220 220 219 221 216
0.18 – – – 0.08 – 0.09 –
221 221 222 220 220 220 221 221 223
50.001 0.01 0.05 50.001 – 0.003 0.003
– – – – – – – 219 0.04 0.03 0.01
– – – 0.007 0.08 – – 214
– – – 0.15 0.12 – – 220
227
0.03 – – –
– – – 5 0.001 0.14 0.03 0.001 0.01
0.10 – – 0.14 – – 5 0.001 –
0.08 0.09 – 0.16 – – – 0.20
0.01 0.11
– –
– –
0.07 –
172 –
168 0.06
222 – – – 0.08 – – – – – 0.003
223
172 0.05
215 5 0.001 5 0.001 50.001 0.002 50.001 50.001 0.001 0.01 50.001 0.04
217
179 0.09
221 0.04 0.01 0.001 50.001 0.001 0.15 0.04 0.001 0.001 0.07
222
227 – – – – – 0.20 – – – 0.29
229
–:p Value 40.20. Independent variables are those in the first column, outcome measures are the changes in scores between discharge and admission for physical functioning, mental health and pain, measured with the SF–36, and the patient-reported effect of the treatment (n¼number included in the analysis).
Study limitations The first limitation of our study was that we did not include improved social functioning as an outcome measure, although this has been found to be important [27]. Since any increase in social
activities (e.g. reintegration in work activities) had not yet been implemented at the time of our assessments immediately after the treatment, it was not useful to include this in our study. Secondly, the retrospective design of the study gave inherent limitations.
Predictors of pain rehabilitation
DOI: 10.3109/09638288.2014.961657
1249
Table 4. Results of the multiple linear regression analyses (for the outcomes in terms of physical and mental functioning and pain) and logistic regression analysis (for patient-reported effect) of the final models in the total group (inpatients and outpatients); independent variables are the scores at admission (baseline scores) and type of treatment in- or out-patient, dependent variables are the outcome scores (scores at discharge minus those at admission) and patient-reported effect. Unstandardized coefficient B
Downloaded by [Australian National University] at 19:34 06 November 2015
Physical functioning outcome Physical functioning Age Active coping Mental health outcome Mental health Vitality Coping style ‘‘praying and hoping’’ depression (SCL-90-R score) Pain outcome Pain Work status
Standard error
p Value
0.50 0.42 0.08
0.06 0.10 0.03
50.001 50.001 0.01
0.90 0.08 0.16 0.14
0.04 0.04 0.05 0.06
50.001 0.046 0.003 0.022
1.36 9.08
0.07 2.36
50.001 50.001
p Value
Nagelkerke R2 0.14
OR (95% CI) Patient-reported effect Employment Active coping Fear
0.31 (0.14–0.69) 0.98 (0.97–0.99) 0.93 (0.88–0.99)
Table 5. Result of the interaction analysis for the outcome in terms of pain: Univariate analysis of variance with baseline pain score (SF-36), work status, treatment program and interaction variables as independent variables, and the outcome in terms of pain (scores at discharge minus those at admission) as the dependent variable (n¼223). Unstandardized coefficient
Outcome pain Pain (SF-36) Work status Treatment program Treatment program *pain Treatment program *work status Pain*work status
p Value Adjusted R2
B
Standard error
1.29 23.14 15.22 0.33
0.14 6.46 6.17 0.14
50.001 50.001 0.01 0.02
0.48
4.83
0.92
0.39
0.15
0.01
0.70
Adjusted R2
n
0.27
220
0.80
214
0.68
222
n 179
0.008 0.010 0.113
samples. However, a predictor can be found only when there is a certain distribution in the scores.
Conclusions No strong predictors of treatment outcome were found other than the baseline scores of the respective outcome variables. More disabled patients and patients with more pain benefitted more from the rehabilitation program. Other variables, and the inclusion of interaction variables in the prediction model, did improve the outcome prediction, but only slightly. It thus remains difficult to predict the outcome of a treatment before its start.
Declaration of interest The authors report no declarations of interest.
References One of these is that the inclusion criterion of having psychosocial problems was not well defined. However, in usual care, a psychologist will only be involved when psychosocial problems are moderate or severe, so there is no doubt that our patients met this inclusion criterion. The low scores of the patients on mental health and vitality of the SF-36 and the high scores on the SCL-90 confirm this. Because, in a retrospective study, the inclusion of patients is less controlled, it is not known whether patients were erroneously not enrolled and bias was present. In addition, the exact content of the treatment program was not assessed either, due to the retrospective nature of the study. However, an important advantage of the current study was that it was conducted in usual care, as an experimental design may also introduce bias. The response rate in our study was low, possibly because the outcome assessment was made in the context of usual care, and over a period of several years. This meant there was no researcher who did his utmost to collect the data, as is inherent to the retrospective design. The low response rate may have caused some bias, but the included and not included eligible patients did not appear to differ in terms of predictive variables. Thirdly, the patients in our study differed largely in characteristics. This may complicate the external validity to specific homogeneous patient
1. Turk DC, Wilson HD, Cahana A. Treatment of chronic non-cancer pain, part 2. Lancet 2011;25:2226–34. 2. Vlaeyen FW, Morley S. Cognitive-behavioral treatments for chronic pain: what works for whom? Clin J Pain 2005;21:1–8. 3. Turk DC, Wilson HD, Cahana A. Treatment of chronic non-cancer pain. Lancet 2011;337:2226–35. 4. Bremander AB, Holmstro¨m G, Bergman S. Depression and age as predictors of patient-reported outcome in a multidisciplinary rehabilitation programme for chronic musculoskeletal pain. Musculoskelet Care 2011;9:41–8. 5. Orenius TI, Koskela T, Koho P, et al. Anxiety and depression are independent predictors of quality of life in patients with chronic musculoskeletal pain. J Health Psychol 2013;18:167–75. 6. Van der H, Vollenbroek-Hutten MM, Groothuis-Oudshoorn KGM, Hermens HJ. Multidisciplinary rehabilitation treatments with chronic low back pain: a prognostic model for its outcome. Clin J Pain 2008;24:421–30. 7. Michaelson P, Sjo¨lander P, Johansson H. Factors predicting pain reduction in chronic back and neck pain after multimodal treatment. Clin J Pain 2004;20:447–54. 8. Vollenbroek-Hutten MMR, Hermens HJ, Wever D, et al. Differences in outcome of a multidisciplinary treatment between subgroups of chronic low back pain patients defined using two multiaxial assessment instruments: the multidimensional pair inventory and lumbar dynamometry. Clin Rehabil 2004;18:566–79. 9. Buchner M, Neubauer E, Zahlten-Hinguranage A, Schiltenwolf M. Age as a predicting factor in the therapy outcome of
1250
10.
11.
12.
13.
Downloaded by [Australian National University] at 19:34 06 November 2015
14.
15.
16.
17.
18.
19.
20.
21.
A. M. Boonstra et al.
multidisciplinary treatment of patients with chronic low back pain – a prospective longitudinal clinical study in 405 patients. Clin Rheumatol 2007;26:385–92. Persson E, Lexell J, Eklund M, Rivano-Fischer M. Positive effects of a musculoskeletal pain rehabilitation program regardless of pain duration or diagnosis. PM&R 2012;4:355–66. Keel PJ, Wittig R, Deutschmann R, et al. Effectiveness of in-patient rehabilitation for sub-chronic and chronic low back pain by an integrative group treatment program (Swiss multicentre study). Scand J Rehab Med 1998;30:211–19. Hutten MMR, Hermens H, Zilvold G. Differences in treatment outcome between subgroups of patients with chronic low back pain using lumbar dynamometry and psychological aspects. Clin Rehabil 2001;15:479–88. Stigmar KGE, Petersson IF, Jo¨ud A, Grahn BEM. Promoting work ability in a structured national rehabilitation program in patients with musculoskeletal disorders: outcomes and predictors in a prospective cohort study. BMC Musculoskel Disord 2013;14:57. Keogh E, McCracken LM, Eccleston C. Do men and women differ in their response to interdisciplinary chronic pain management? Pain 2005;114:37–46. Hampel P, Graef T, Krohn-Grimberghe B, Tlach L. Effects of gender and cognitive-behavioral management of depressive symptoms on rehabilitation outcome among orthopaedic patients with chronic low back pain: a 1 year longitudinal study. Eur Spine J 2009;181: 1867–80. de Williams ACC, Richardson PH, Nicholas MK, et al. Inpatient vs. outpatient pain management: results of a randomized controlled trial. Pain 1996;66:13–22. de Rooij A, Roorda LD, Otten RHJ, et al. Predictors of multidisciplinary treatment outcome in fibromyalgia: a systematic review. Disabil Rehabil 2013;35:437–49. Scascighini L, Toma V, Dober-Spielmann S, Sprott H. Multidisciplinary treatment for chronic pain: a systematic review of interventions and outcomes. Rheumatology (Oxford) 2008;47: 670–8. Aaronson NK, Muller M, Cohen PDA, et al. Translation, validation, and norming of the Dutch language version of the SF-36 Health Survey in community and chronic disease populations. J Clin Epidemiol 1998;51:1055–68. Bergman S, Jacobsson LTH, Herrstro¨m P, Petorsson IF. Health status as measured by SF-36 reflects changes and predicts outcome in chronic musculoskeletal pain: a 3-year follow up study in the general population. Pain 2004;108:115–23. Barelds DPH, Luteijn F. Measuring personality: a comparison of three personality questionnaires in the Netherlands. Personal Individ Differ 2002;33:499–510.
Disabil Rehabil, 2015; 37(14): 1242–1250
22. Barelds DPH, Luteijn F, Van Dijk H, Starren J. Nederlandse Persoonlijkheids Vragenlijst-2, handleiding [manual for Dutch Personality Questionnaire, revised]. Lisse: Swets & Zeitlinger;2007 [in Dutch]. 23. Spinhoven P, ter Kuile MM, Linssen ACG. Coping met pijn vragenlijst; handleiding [The Coping with Pain Questionnaire: manual]. Lisse: Swets & Zeitlinger; 1994 [in Dutch]. 24. Rosenstiel AK, Keefe FJ. The use of coping strategies in chronic low back pain patients: relationship to patient characteristics and current adjustment. Pain 1983;17:33–44. 25. Vlaeyen JW, Kole-Snijders AM, Boeren RG, Van Eek H. Fear of movement/(re)injury in chronic low back pain and its relation to behavioral performance. Pain 1995;62:363–72. 26. Arrindell W, Ettema H, Groenman N, et al. De groeiende inbedding van de Nederlandse SCL-90-R; psychodiagnostisch gereedschap. [Growing use of Dutch version of SCL-90-R psychodiagnostic instrument] De Psycholoog 2003;38: 576–82. 27. Chapman JR, Norvell DC, Hermsmeyer JT, et al. Evaluating common outcomes for measuring treatment success for chronic low back pain. Spine 2011;36:s54–68. 28. Turk DC, Dworkin RH, Burke LB, et al. Developing patientreported outcome measures for pain clinical trials: IMMPACT recommendations. Pain 2006;215:208–15. 29. Cohen J. Statistical power analysis for the behavioral sciences. New York: Academic Press, Inc; 1977. 30. Angst F, Brioschi R, Main CJ, et al. Interdisciplinary rehabilitation in fibromyalgia and chronic back pain: a prospective outcome study. J Pain 2006;7:807–15. 31. Keller A, Hayden J, Bombardier C, Van Tulder M. Effect sizes of non-surgical treatments of non-specific low-back pain. Eur Spine J 2007;16:1776–88. 32. Morley S, Williams A, Hussain S. Estimating the clinical effectiveness of cognitive behavioural therapy in the clinic: evaluation of a CBT informed pain management programme. Pain 2008;137: 670–80. 33. Lunde L-H, Nordhus IH, Pallesen S. The effectiveness of cognitive and behavioural treatment of chronic pain in the elderly: a quantitative review. J Clin Psychol Med Settings 2009;16: 254–62. 34. Mangels M, Schwartz S, Worringen U, et al. Evaluation of a behavioral-medical inpatient rehabilitation treatment including booster sessions: a randomized controlled trial. Clin J Pain 2009; 25:356–64. 35. Boonstra AM, Reneman MF, Schiphorst Preuper HR, et al. Differences between patients with chronic musculoskeletal pain treated in an inpatient or outpatient multidisciplinary rehabilitation program. IJRR 2014;37:187–91.
ISSN: 0963-8288 (Print) 1464-5165 (Online) Journal homepage: http://www.tandfonline.com/loi/idre20
Predictors of multidisciplinary treatment outcome in patients with chronic musculoskeletal pain Anne M. Boonstra, Michiel F. Reneman, Berend R. Waaksma, Henrica R. Schiphorst Preuper & Roy E. Stewart To cite this article: Anne M. Boonstra, Michiel F. Reneman, Berend R. Waaksma, Henrica R. Schiphorst Preuper & Roy E. Stewart (2015) Predictors of multidisciplinary treatment outcome in patients with chronic musculoskeletal pain, Disability and Rehabilitation, 37:14, 1242-1250, DOI: 10.3109/09638288.2014.961657 To link to this article: http://dx.doi.org/10.3109/09638288.2014.961657
Published online: 17 Sep 2014.
Submit your article to this journal
Article views: 137
View related articles
View Crossmark data
Full Terms & Conditions of access and use can be found at http://www.tandfonline.com/action/journalInformation?journalCode=idre20 Download by: [Australian National University]
Date: 06 November 2015, At: 19:34
http://informahealthcare.com/dre ISSN 0963-8288 print/ISSN 1464-5165 online Disabil Rehabil, 2015; 37(14): 1242–1250 ! 2014 Informa UK Ltd. DOI: 10.3109/09638288.2014.961657
RESEARCH PAPER
Predictors of multidisciplinary treatment outcome in patients with chronic musculoskeletal pain Anne M. Boonstra1, Michiel F. Reneman2, Berend R. Waaksma1, Henrica R. Schiphorst Preuper2, and Roy E. Stewart3
Downloaded by [Australian National University] at 19:34 06 November 2015
1
Department of Rehabilitation, ‘Revalidatie Friesland’ Center for Rehabilitation, Beetsterzwaag, The Netherlands, 2Department of Rehabilitation, Center for Rehabilitation, University Medical Center Groningen, University of Groningen, Groningen, The Netherlands, and 3Department of Health Sciences, Community & Occupational Medicine, University Medical Center Groningen, University of Groningen, Groningen, The Netherlands Abstract
Keywords
Purpose: The present study aimed to identify predictors of rehabilitation outcome for patients with chronic musculoskeletal pain (CMP) and psychological problems. Methods: A retrospective cohort study including 230 adult patients with CMP admitted for multidisciplinary pain rehabilitation. Potential predictors were patient characteristics, duration of complaints, baseline functioning, pain, personality, coping style, fear of movement, psychological distress and type of treatment. Outcome measures were physical functioning, mental health, pain and patient-reported effect. Multiple (logistic) regression models were used to identify predictors. Results: Patients who were more disabled and patients with more pain benefitted more from the rehabilitation treatment than less disabled patients or those with less pain. Age, work status, vitality, depression and coping style also predicted outcomes significantly. The models explained between 27 and 80% of the outcomes. There was an interaction between type of treatment, work status and the baseline pain score as regards the outcome in terms of pain. Conclusions: No strong predictors of treatment outcome were found other than the baseline scores of the respective outcome variables. More disabled patients and patients with more pain benefitted more from the rehabilitation program. Other predictors improved the prediction models slightly.
Chronic pain, clinical outcome, musculoskeletal pain History Received 15 June 2013 Revised 26 July 2014 Accepted 1 September 2014 Published online 17 September 2014
ä Implications for Rehabilitation
It remains challenging to correctly predict the outcome of treatment from patients’ baseline sociodemographic and psychological characteristics; predictors other than baseline scores of the outcome variables are only slightly associated with treatment outcome. Patients with chronic musculoskeletal pain and poor physical functioning or mental health benefit most from pain rehabilitation. Older patients benefit less from a pain rehabilitation program than younger patients in terms of physical functioning. Pain reduction during a pain rehabilitation program is greatest in patients with high pain intensity who are not at work at the start of the rehabilitation program. Coping style influences the outcome of rehabilitation of patients with chronic musculoskeletal pain.
Introduction Disability due to chronic pain is multifactorial, and a biopsychosocial perspective is needed for better treatment of the patients. The most frequently recommended approach is that of multidisciplinary rehabilitation including elements of cognitive behavioral therapy [1]. Patients are known to differ in their response to this multidisciplinary approach, but it is still unclear who will benefit the most, remain unchanged, or even worsen [2,3].
Address for correspondence: Anne M. Boonstra, Revalidatie Friesland, PO Box 2, 9244 ZN Beetsterzwaag, The Netherlands. Tel: +31 512 389329. Fax: +31 512 389244. E-mail: [email protected]
Moreover, currently available treatments provide only modest improvements in terms of pain and functioning of chronic pain patients [3]. Identifying predictors of outcomes may help clinicians to choose the right program for their patients and to develop new treatments for those for whom existing treatments are insufficiently effective. A wide range of predictors of pain rehabilitation outcome has been reported. Better outcome has been reported for patients with high scores on depression [4–6], higher fear-avoidance beliefs [6], higher scores on perceived distress [7,8] or optimistic attitudes [7] on the Multidimensional Pain Inventory (MPI), low need to socialize [7], younger age [4,7,9,10] or higher educational level [11]. Poorer outcome has been reported for patients with greater psychological dysfunction [12] or lower quality of performance
Predictors of pain rehabilitation
Downloaded by [Australian National University] at 19:34 06 November 2015
DOI: 10.3109/09638288.2014.961657
on lumbar dynamometry [8,12], or those who are unemployed [13]. Mixed findings have been reported for anxiety [4,5] and gender [14,15]. Others, however, found several of these factors, such as age, educational level, work status, depressive symptoms and fear of avoidance not to be predictors [5–7,11,16]. Recently, a review was published about predictors of multidisciplinary treatment outcome in fibromyalgia [17]. Poor outcomes were predicted by depression, by the ‘‘disturbance and pain’’ profile of the Minnesota Multiphasic Personality Inventory (MMPI), by strong beliefs in fate and by high disability. Better outcomes were predicted by a poorer baseline status, the ‘‘dysfunctional’’ and ‘‘adaptive copers’’ profiles of the MPI and high levels of pain. Inconclusive evidence was found for other demographic and clinical factors [17]. The discrepancies between these studies may be due to differences in sample characteristics, outcome measures and/or treatments [7,11,18]. Several authors [8,17,18] have recommended studies to compare different methods, settings and durations of multidisciplinary treatments and to examine their relations with patient characteristics in more detail, in order to detect differential effects. Knowing which patients may have a poor treatment outcome should help to design more effective rehabilitation programs. It may also help to refer or select patients for the most suitable treatment according to their characteristics. Moreover, knowing which variables are important in predicting the outcome of a treatment may help to inform patients more reliably about the effects that can be expected. Finally, it is important for conducting randomized controlled trials to know predictors of outcome, in order to check or control for potential confounders. Studies about predictors may therefore help in examining the effects of rehabilitation treatments in the future. Our study focused on the outcome and outcome predictors of patients treated in rehabilitation programs in usual care. Our study focused on patients with chronic musculoskeletal pain (CMP) and moderate-to-severe psychosocial problems, as this is a large subgroup treated in rehabilitation centers in Western Europe. The aim of the present study was to identify predictors of the rehabilitation outcome for those patients with CMP.
Materials and methods Patients The inclusion criteria for participation in the present study were: patients with CMP referred to the ‘‘Revalidatie Friesland’’ Rehabilitation Center in Beetsterzwaag (The Netherlands), who were given inpatient or outpatient cognitive behavioral treatment, who were aged above 18 years, and whose pain had lasted for over 3 months. Patients were referred for the inpatient treatment by rehabilitation physicians from all over the northern part of the Netherlands, mainly from the province of Friesland, while patients referred for the outpatient treatment came from the area where the center is located. Another inclusion criterion that had to be met was the involvement of a psychologist in the treatment, by way of operationalization of having moderate-to-severe psychosocial problems. Involvement of a psychologist was, e.g. indicated if the patient experienced high psychological distress, pain-related fear, mild or moderate depression due to acceptance problems, compulsive behavior, personality disorder, etc. Exclusion criteria were insufficient command of Dutch, co-morbidity with severe negative consequences for physical functioning, current major psychiatric disorder (active psychosis, severe depression, addiction, etc.) and unwillingness to provide data for research purposes. The study included patients who started treatment between September 2003 and December 2010. If patients were treated twice in the inclusion period, we used the data of the first episode for our study.
1243
Measurements Potential predictors The potential predictors we selected were those mentioned in the literature [4–15] insofar as they were relevant to our study population; they were supplemented with variables that were considered potential predictors on the basis of clinical experience. The following characteristics were assessed, using a selfconstructed questionnaire: duration of current pain, age, gender, marital status, age of children if any, educational level, being employed or self-employed, currently in work (work status, i.e. not being on sick leave, disabled for work or homemaker), outstanding litigation or unresolved workers’ compensation claims, and receiving a benefit (e.g. social benefit, unemployment benefit). Baseline functioning, general and mental health status, and pain intensity were assessed using the Dutch version of the Short Form 36 Health Survey (SF-36) [19,20]. This instrument consists of 36 questions, and measures eight dimensions: physical functioning, social functioning, physical role restriction, emotional role restriction, mental health, vitality, pain, general health and health change (scores range from 0 to 100 for each dimension; a lower score means more disability or more pain). Personality was assessed by the Dutch Personality Questionnaire (DPQ) [21]. The DPQ is based on a translated, shortened version of the California Psychological Inventory. The DPQ1 consists of 132 items, and the DPQ2 of 140 items, to which subjects respond by ‘‘correct’’, ‘‘don’t know’’ or ‘‘incorrect’’. Items are distributed over seven scales: neuroticism, social anxiety, rigidity, hostility, egoism, dominance and self-esteem. DPQ2 is a revised version [22], and this is the version which was used for the study sample after July 2008, i.e. for 22% of the study sample. The DPQ1 scores were recoded to DPQ2 scores using the formula provided by the developers [22]. Coping reactions were measured with the Coping with Pain Questionnaire (CPQ) [23], a Dutch adaptation of the Coping Strategy Questionnaire (CSQ) [24]. The CPQ assesses the use of cognitive strategies: ignoring pain sensations, coping selfstatements, re-interpreting pain sensations, catastrophizing, ‘‘praying and hoping’’, diverting attention and ability to control pain, as well as one behavioral strategy, viz. increasing one’s activity level. The ‘‘active coping’’ factor in the CSQ is based on the scores for ignoring pain sensations, coping selfstatements, reinterpreting pain sensations and increasing activity level, while the ‘‘helplessness’’ factor is based on the scores for catastrophizing, praying and hoping, and diverting attention [23]. We calculated the ‘‘active coping’’ and ‘‘helplessness’’ factors by summing the relevant scores. A higher score means that a patient is more likely to use this specific coping style predominantly. Pain-related fear was assessed by the Tampa Scale of Kinesiophobia, Dutch version (TSK) [25]. The TSK consists of 17 items, each provided with a 4-point Likert scale with scoring alternatives ranging from ‘‘strongly disagree’’ to ‘‘strongly agree’’. Higher scores indicate more fear. Psychosocial distress was assessed by the Symptom Checklist 90-Revised, Dutch version (SCL-90-R) [26]. In the SCL-90-R, patients are instructed to rate 90 distress symptoms on a 5-point Likert scale ranging from ‘‘not at all’’ to ‘‘extremely’’. The statements are assigned to eight dimensions reflecting various types of psychopathology: anxiety, agoraphobia, depression, somatization, insufficiency, sensitivity, hostility and insomnia. The SCL-90-R also includes a total score for ‘‘psychoneuroticism’’, which is a global measure of distress. Higher scores indicate higher levels of distress.
1244
A. M. Boonstra et al.
Outcome variables Outcome variables were physical functioning, mental health and pain measured by the SF-36, and patient-reported effect (PRE). PRE was assessed after treatment by asking the patients to report the global perceived effect of the treatment on their symptoms, using a Likert scale: completely recovered, very much improved, much improved, unchanged, worse, much worse, very much worse. These four outcome variables reflect important rehabilitation goals [27] and have been recommended by the Initiative on Methods, Measurement, and Pain Assessment in Clinical Trials (IMMPACT) [28].
Downloaded by [Australian National University] at 19:34 06 November 2015
Procedure and rehabilitation program Patients were referred to multidisciplinary rehabilitation for their pain-related disabilities. Referral was done by rehabilitation physicians working in the northern part of the Netherlands. Eligibility for rehabilitation treatment was judged by a rehabilitation physician, and the aims and contents of the treatment were explained to the patients in general terms. Patients also consulted the psychologist at the pain department of the rehabilitation center before the treatment started. The final choice to enter a program was a shared decision by patient, physician and psychologist. Treatments were based on cognitive behavioral concepts [1]. The team consisted of a rehabilitation physician, a physiotherapist, an occupational therapist and a psychologist, and often also a social worker. Program goals depended on the characteristics of the complaints and the aims of the patient, and included reduction of activity limitations, participation problems and psychological distress. The program was delivered individually as well as in groups. The most commonly used treatment modalities were teaching ergonomic principles, graded activity and behavioral therapy. Treatment generally focused on optimization of functioning. Outpatient programs were mostly provided on 3 days per week for one to four hours a day, over a period of approximately 10 to 16 weeks. The inpatient program was provided on Monday through Friday, with patients going home for the weekend. This program lasted approximately 6 to 10 weeks and was sometimes followed by outpatient treatment. The length and intensity of the program were determined by the characteristics of the complaints and the progress made by the patient during the treatment, as is usual in clinical practice. If a patient started an outpatient program, and it turned out during the observation period, or in the first period of treatment, that the treatment targets were not going to be met, patients could still be referred to an inpatient program. The SF-36 was administered at the beginning (baseline) and at the end of the program (discharge). The first assessment (baseline) took place just before the start or during the first 2 weeks in the case of an outpatient program, or during the first week in the case of an inpatient program. The discharge assessment of patients in outpatient programs took place in the final week or during the first 4 weeks after discharge, while the assessment of patients in inpatient programs took place 2 to 4 weeks after discharge or, in the case of continuation as outpatient treatment, in the final week or during the first weeks after the end of the program. The other assessments were administered in the first or second week of the program as part of regular clinical procedures. Statistical analysis Demographic characteristics are presented as mean and standard deviation (age) or percentages (other variables). Since the duration of complaints and the scores on the SF-36, DPQ, CPQ, TSK and SCL-90-R were ordinal and/or had a skewed distribution, medians and quartiles were calculated. To enable
Disabil Rehabil, 2015; 37(14): 1242–1250
comparison with other studies, we also calculated means and standard deviations for the duration of complaints and SF-36 scores. Differences between outpatients and inpatients were tested with the t-test for normally distributed data (checked with the Kolmogorov–Smirnov test), the Mann–Whitney U-test for nonnormally distributed data, and the Chi-square test for dichotomous data. The effect sizes (ES) of the SF-36 scores were calculated using the following formula [29]:
1 X 2 pffiffiffiffiffiffiffiffiffiffiffiffiffiffi X = ð1 rÞ SDpooled The following formula was used for SDpooled:
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ððSDbaseline Þ2 þ ðSDdischarge Þ2 Þ 2 where X1 is the baseline value, X2 is the value at discharge, r is the correlation coefficient between X1 and X2, and SD the standard deviation. The ES interpretation was as follows: 50.20 trivial effect, 0.20 to 0.50 small effect, 0.50 to 0.80 medium effect and 0.80 large effect [29]. Variables assessed as potential predictors were: age (in years), gender (male ¼ 0), marital status (single (0) versus married or living with a partner (1)), educational level (low: primary school to intermediate vocational education (0) versus high: higher vocational education to university (1)), employment (employed or self–employed (0) versus unemployed, including students and homemakers (1)), work status (not currently in work (0) versus in regular or adapted work (1)), benefit (no benefit (0) versus receiving a benefit, e.g. social benefit or unemployment benefit (1)), age of youngest child (child 12 years (0) versus child 412 years or no children living at home (1)), being involved in an ongoing litigation or unresolved workers’ compensation claim procedure (no procedure ¼ 0), duration of current complaints (years), baseline scores on the SF-36, DPQ, CPQ, TSK and SCL-90-R, or domains thereof, and type of treatment (inpatient (0) or outpatient (1)). Outcome parameters were PRE and change scores of SF-36 between admission and discharge (discharge score minus baseline score) for the domains of physical functioning, mental health and pain. In view of the small number of patients reporting worsening of their complaints, ‘‘PRE’’ was dichotomized into no improvement (no effect or worsening (0)) and improvement (1). Regression analyses were conducted to identify predictors of outcomes in terms of physical functioning, mental health, and pain (linear regression) and to analyze predictors of PRE (binary logistic regression). The analysis of each outcome measure was done in seven steps. Step 1 involved univariate analyses. Variables with p values 50.20 were identified as potential predictors. Step 2 involved analyses of blocks of related variables. In view of the large number of variables relative to the number of patients we performed a connecting analysis. Related potential predictors were clustered into five groups of variables: (a) work-related variables (employment, work status, benefit), (b) SF-36 scores excluding the pain score, (c) DPQ scores, (d) CPQ scores and (e) SCL-90 scores. Multiple regression or logistic analyses per block were performed with the above-mentioned variables as independent variables and the outcome variables as dependent variables. In the block analysis of the SF-36 scores, the model did not include the score for the same domain of the outcome measure (for example, the SF-36 score for ‘‘physical functioning’’ was not entered as an independent variable in the model with the outcome measure of physical functioning as the dependent variable).
Predictors of pain rehabilitation
Downloaded by [Australian National University] at 19:34 06 November 2015
DOI: 10.3109/09638288.2014.961657
In step 3, variables whose p values remained 50.20 in the block analyses, and potential predictors that were not entered in the block analyses (characteristics, SF-36 pain score, SF-36 score for the same domain of the outcome measure, TSK score and type of treatment) were entered in the next model. In step 4, variables with a p value50.05 in step 3 were entered in the final model. The next two steps were done to examine interactions between type of treatment (inpatient/outpatient: inpatient program ¼ 0) and (potential) predictors. In step 5, the variables entered in step 3, as well as the potential interaction variables with the type of treatment as independent variables, were entered in the regression models with the outcome variables as dependent variables. In step 6, variables with a p value50.05 in step 5 were entered in the final interaction models. Since a relatively large percentage of the eligible patients did not complete the outcome assessment, we compared eligible and included patients regarding the variables that we found to be outcome predictors. We used the t-test for normally distributed data (checked with the Kolmogorov–Smirnov test), the Mann– Whitney U-test for non-normally distributed data, and the Chi-square test for dichotomous data. The study covered a period of seven years. Since it was theoretically possible that the outcome changed over this period, we checked the distributions of the outcome visually using boxplots (continuous outcome variables) or histograms (patient-reported effect, i.e. dichotomous variable).
Results Study population Three hundred and sixty-four patients were eligible for this study (n ¼ 364). Psychological data of eight of the patients were missing, as were data on potential predictors for two of the patients. Five patients started treatment twice. Two hundred and forty-one patients had an outcome assessment (response rate 66% for outpatients and 67% for inpatients). Nine patients did not give permission to use their data for research purposes, and so were excluded. Finally, 230 patients were included in the analysis, 107 outpatients and 123 inpatients. Eight of these inpatients had a (short) outpatient treatment before being admitted to an inpatient program, and 10 of the inpatients continued as outpatients after completing the inpatient program. They were included in the ‘‘inpatient group’’. The total duration of treatment (excluding the hours spent by the nursing staff and physician or trainee) was 150 h (SD 35 h) for the inpatient group, including 17 h (SD 5 h) of treatment by the psychologist and social worker. The corresponding values for the outpatient group were 101 h (SD 37 h) and 24 h (SD 13 h), respectively. Most patients suffered from low back pain (31%), fibromyalgia (23%), widespread pain (13%) or neck pain (16%). Characteristics of patients are presented in Table 1. Outcomes Most patients in both types of program improved, with ESs for the total group of 0.99 for physical functioning; 0.24 for mental functioning; and 1.65 for pain. A subgroup comprising 4% of the sample reported worsening (Table 2). Predictors of outcome Better outcomes were associated with several patient characteristics as well as baseline pain intensity and various baseline physical and psychosocial variables (Table 3). Table 4 presents the results of the multiple (logistic) regression analyses. Due to missing data and list-wise deletion, the multiple
1245
regression analyses included fewer than the total number of respondents. Greater improvement in physical functioning was associated with an active coping style, and smaller improvement with older age and better physical functioning at the start of the program. The adjusted R2 in the unadjusted model with baseline physical functioning as predictor and physical outcome as dependent variable was 0.19, compared to 0.27 in the final model with physical functioning, age and active coping as predictors and physical outcome as dependent variable. Smaller improvement in mental health was associated with better mental health and a ‘‘praying and hoping’’ coping style, vitality and depression at baseline. The adjusted R2 in the unadjusted model with baseline mental functioning as predictor and mental health outcome as dependent variable was 0.78, compared to 0.80 in the model with baseline mental health functioning, ‘‘praying and hoping’’, vitality and depression as predictors and mental health outcome as dependent variable. Less pain reduction was associated with being currently in work, having less pain at baseline. The adjusted R2 in the unadjusted model with baseline pain as predictor and pain outcome as dependent variable was 0.66, compared to 0.68 in the model with baseline pain and work status as predictors and pain outcome as dependent variable. Higher PRE was associated with being employed and less ‘‘active coping’’. Interaction We explored a two-way interaction model, which indicated significant effects of pain, work status and treatment program, and interactions between the type of treatment and pain, and between pain and work status. As regards the association between baseline pain score and outcome in terms of pain, an interaction was observed with type of treatment and with work status (Table 5). Included and not included patients The predicting variables did not differ significantly between the patients included (n ¼ 230) and those not included in the study (n ¼ 132): baseline physical functioning (mean (SD): 41 (19) and 39 (24), respectively); baseline mental health (60 (19) and 59 (18), respectively); baseline pain (32 (16) and 31 (18), respectively); age (43 (12) and 41 (12), respectively); active coping (99 (37) and 91 (32), respectively), helplessness (61 (27) and 57 (26), respectively); ‘‘praying and hoping’’ (18 (12) and 17 (13), respectively), fear (37(7) and 36 (7), respectively), employment (53 and 51%, respectively); child 12 yrs (22 and 21%, respectively). Outcome variables during the years of study According to the boxplots or histogram, the outcome variables did not show a trend towards better or worse outcome over the years of the study.
Discussion The aim of the study was to identify predictors of rehabilitation treatment outcomes for patients with CMP and moderateto-severe psychosocial problems. The outcomes in terms of physical functioning, mental health and pain were significantly predicted by their respective baseline scores: lower scores predicted greater effects. This may not be surprising, as lower scores indicate a poorer status and thus more room for improvement. The variances in the final models were mainly explained by the baseline scores, as shown by the adjusted R2 of the various models. However, other predictors might, although of less significance, be clinically relevant as well. Coping style was a predictor of the outcomes in terms of physical functioning and
1246
A. M. Boonstra et al.
Disabil Rehabil, 2015; 37(14): 1242–1250
Table 1. Demographic and clinical characteristics of patients with chronic musculoskeletal pain (baseline data).
Characteristics Age (yrs; mean (SD), range) Gender (% male) Marital status (% single) Educational level (% low) Age of youngest child (% 12 yr) Ongoing procedure (% no) Employed (% yes) Work status (% not in work) Benefit (% no) Duration of current complaints in years Mean (standard deviation) Median (quartiles)
Outpatient program (n ¼ 107)
Inpatient program (n ¼ 123)
p Value
43 (10,19–69) 36 21 76 36 62 66 61 40
43 (13,18–76) 21 27 83 11 76 42 80 34
0.97a 0.02 0.48 0.15 50.001 0.03 0.001 0.002 0.33
4.9 (5.3) 3.0 (1.5–6.0)
5.9 (5.8) 3.8 (2.0–9.5)
0.09b
Downloaded by [Australian National University] at 19:34 06 November 2015
Median (quartiles) Pain (SF-36 scores, range 0–100) –Pain 33 (22–45) Functioning and health (SF-36 scores, range of each item 0-100) – Physical functioning 50 (30–65) – Social functioning 50 (38–75) – Physical role 0 (0–13) – Emotional role 33 (0–100) – Mental health 68 (52–76) – Vitality 40 (25–45) – General health 50 (35–65) – Health change 25 (25–50) Personality (DPQ scores, range of each item 0–40) – Neuroticism 15 (7–23) – Social anxiety 9 (4–19) – Rigidity 27 (21–31) – Hostility 11 (6–19) – Egoism 6 (3–10) – Dominance 19 (14–24) – Self-esteem 28 (23–33) Coping (CPQ) (scoring range) – Ignoring pain sensations (0–60) 31 (19–39) – Coping self-statements (0–60) 35 (30–45) – Reinterpreting pain sensations (0–60) 8 (4–19) – Catastrophizing (0–60) 22 (14–34) – Praying and hoping (0–60) 15 (7–23) – Diverting attention (0–60) 20 (9–28) – Ability to control pain (0–20) 8 (3–11) – Increasing activity level (0–60) 22 (17–30) Composite scores – Active coping (0–180) 103 (77–119) – Helplessness (0–240) 58 (41–78) Fear of movement (TSK) (scoring range 17–61) – Fear 37 (33–40) Psychological distress (SCL-90-R) (scoring range) – Anxiety (10–50) 16 (12–20) – Phobic anxiety (agoraphobia) (7–35) 7 (7–9) – Depression (16–80) 30 (26–39) – Somatization (12–60) 28 (22–33) – Insufficiency (9–45) 21 (17–26) – Sensitivity (18–90) 27 (22–33) – Hostility (6–30) 8 (7–11) – Insomnia (3–15) 7 (5–12) – Psychoneuroticism (GSI) (90–450) 161 (140–194)
Mean (SD)
Median (quartiles)
Mean (SD)
35 (15)
33 (12–45)
28 (17)
47 54 10 45 64 38 51 35
35 38 0 33 56 30 40 25
(25–45) (25–50) (0–0) (0–100) (44–72) (18–43) (30–55) (0–50)
35 39 7 47 58 31 43 27
16 14 27 14 6 17 27
(9–26) (6–22) (22–33) (7–20) (4–9) (12–23) (20–32)
0.11b 0.01b 0.28b 0.17b 0.91b 0.58a 0.10b
28 38 8 24 16 19 5 24
(17–38) (26–45) (1–18) (15–34) (10–25) (10–27) (0–10) (17–30)
0.59a 0.77a 0.30b 0.74a 0.30b 0.64b 0.03b 0.37a
(20) (24) (22) (43) (17) (16) (19) (24)
(17) (25) (18) (44) (20) (18) (17) (25)
38 (30–43) 18 9 35 29 24 27 8 9 181
(14–23) (7–12) (26–44) (24–36) (18–27) (23–39) (7–11) (6–12) (146–212)
50.001b 50.001b 0.27b 0.86b 0.01b 0.001b 0.003b 0.02b
0.82a 0.64a
99 (73–127) 59 (40–82) 36 (6.4)
0.001b
37 (8.2)
0.46a 0.01b 0.001b 0.04b 0.07b 0.06b 0.16b 0.34b 0.11b 0.02b
Differences between outpatients and inpatients were tested with the t-test for normally distributed data, the Mann–Whitney U-test for non-normally distributed data, and the Chi-square test for dichotomous data. a t-test; bMann–Whitney U-test.
mental health. Active coping was positively correlated with a better physical functioning outcome, but patients with a more active coping style were less likely to perceive improvement according to the PRE. An active coping style may help patients to
adopt the exercises and improve their physical condition, and hence increase their physical activities. On the other hand, patients with an active coping style may be disappointed in the degree of overall improvement in view of their efforts,
Predictors of pain rehabilitation
DOI: 10.3109/09638288.2014.961657
1247
Table 2. Outcomes in terms of physical functioning, mental health and pain at admission and discharge, and difference between discharge and admission, with effect sizes and patient-reported effect at discharge. Outpatient treatment
Downloaded by [Australian National University] at 19:34 06 November 2015
Pain (SF-36 scores) – Pain Functioning (SF-36 scores) – Physical functioning – Mental health Patient reported effect (%) Completely recovered Very much improved Much improved Unchanged Much worse Very much worse Far worse
Admission (mean, SD)
Discharge (mean, SD)
Difference (mean, SD)
35 (15)
54 (17)
18.5 (28.5)
47 (20) 64 (17)
57 (20) 64 (9)
9.7 (17.9) 0.12 (16.6)
0 41 34 19 5 1 0
and may therefore describe the complaints as unchanged. The ‘‘praying and hoping’’ coping style was negatively associated with an improvement in mental health. ‘‘Praying and hoping’’ is a passive coping style, and patients with this coping style may have benefitted less from the treatment because their attitude prevented them from adopting the recommendations in their daily lives. Personality was not found to be a predictor of treatment outcome. This is in agreement with Michaelson et al. [7]. However, patients with personality disorders were excluded from the rehabilitation programs and were therefore not included in our study. Our finding cannot be generalized to patients beyond the scope of the study. As regards the association between baseline pain score and outcome in terms of pain, an interaction was observed with type of treatment (inpatient versus outpatient). The improvement in pain score between baseline and discharge (pain outcome) was larger for patients with more severe baseline pain (low SF-36 score), and the difference in improvement between inpatients and outpatients, as well as between patients currently in work and those not currently in work changed with increasing baseline pain levels. We did not find a clear explanation for this phenomenon. Perhaps inpatients who had been in work before admission had more trouble adjusting to a new daily structure without work than outpatients. Inpatients may have allowed themselves to feel the pain and stress after admission, more so than outpatients, as many patients are used to ignoring their feelings, including feelings of pain. Further research needs to be done to explain this phenomenon. The data were collected over many years. Neither the type of treatment, i.e. cognitive behavioral therapy, nor the outcome of the treatment programs, changed over time. However, it is very likely that the content of the treatment programs changed in these years as the expertise of the therapists developed further during the study. This may have influenced the predictors of outcome. Patients with less favorable outcomes may have been excluded over time or treated in a better way. There is a possible bias in the fact that no predictors were found, because of increasing homogeneity during the years of this study. However, we found no reason to assume that the predictors found in our study were biased, i.e. were not really predictors. The effect sizes we found for physical functioning and pain are within the range of those found in other studies [14,30–35], while those for mental health were smaller than those reported by others [14,30]. This small effect size for the mental functioning outcome may be caused by
Inpatient treatment Admission (mean, SD)
Discharge (mean, SD)
Difference (mean, SD)
Effect size
1.64
28 (17)
54 (18)
25.4 (27.4)
2.14
0.78 0.02
35 (17) 58 (20)
53 (22) 63 (9)
17.8 (21.6) 5.4 (20.5)
1.44 0.44
Effect size
1 49 31 17 1 1 1
the fact that the mental functioning score at baseline was relatively high in comparison with those in other studies [14,30]. It is unclear whether this small effect size has influenced the identification of predictors. The effect sizes resulting from inpatient treatment were larger than those for outpatient treatment, which may be caused by the fact that patients benefit more from inpatient than outpatient treatment, or by the fact that the levels at the beginning of the program were lower among the inpatients. Since this was not one of our research questions, we did not explore this finding any further. Bremander et al. [4] found anxiety and depression to be predictors of outcome. This was not confirmed by our study. Bremander et al. [4] used the Hospital Anxiety and Depression Scale (HADS) for their assessment, whereas we used the SCL-90-R, which may not be specific enough to assess this variable. Another possible explanation is that Bremander et al. included fewer, and different, potential predictors in their study (i.e. age, gender, antidepressant medication, pain intensity and pain distribution) than we did. This demonstrates once more how difficult it is to compare studies. No overview of variables associated with outcome in patients with CMP is available in the literature, except for fibromyalgia [17], making comparisons with other studies difficult. Moreover, the considerable differences in study design also makes it difficult to compare some of the other findings of our study with those of others. For example, our study found age to be only a predictor of the outcome in terms of physical functioning, while gender was not found to be an outcome predictor at all, in contrast to what was found in other studies [7,10,11,14–16]. We found a proportion of explained variance between 0.23 and 0.79, which was much larger than that reported by, e.g. Van der Hulst et al. [6]. This may have been caused by differences in study sample and/or differences in outcome measures. To unravel the complex association between the large number of possible predictors and the large variety of outcome variables, there is an urgent need for more studies on this issue, finally followed by systematic reviews to combine the findings. Our findings may add to such reviews. Just like Morley et al. [32], who evaluated usual care, we found that a very small percentage of the patients perceived worsening of their complaints. It would be very interesting to find predictors that could identify patients who are likely to get worse, as this might help physicians decide when to advise against cognitive behavioral treatment. However, the small percentage of worsening patients did not enable us to identify these predictors.
1248
A. M. Boonstra et al.
Disabil Rehabil, 2015; 37(14): 1242–1250
Table 3. Results of univariable analyses in total group (inpatients and outpatients). Physical functioning outcome
Mental health outcome
Downloaded by [Australian National University] at 19:34 06 November 2015
n Characteristics Age Gender Marital status Educational level Age of youngest child Ongoing procedure Duration of current complaints Work Employed Work status Benefit Pain (SF-36 scores) – Pain Functioning and health (SF-36 scores) – Physical functioning – Social functioning – Physical role – Emotional role – Mental health – Vitality – General health – Health change Personality (DPQ) – Neuroticism – Social anxiety – Rigidity – Hostility – Egoism – Dominance – Self-esteem Coping (CPQ) – Ignoring pain sensations – Coping self-statements – Reinterpreting pain sensations – Catastrophizing – Praying and hoping – Diverting attention – Ability to control – Increasing activity level Composite scores – Active coping – Helplessness Fear (TSK) – Fear Psychological distress (SCL-90-R) – Anxiety – Phobic anxiety (agoraphobia) – Depression – Somatization – Insufficiency – Sensitivity – Hostility – Insomnia – Psychoneuroticism Type of treatment
0.004 – 0.15 – – – –
220 220 220 220 217 218 218 221
– – –
Pain outcome
n – – – 0.10 – – –
216 216 216 216 214 216 213 216
– 0.13 –
Patient-reported effect
n 0.15 0.18 0.10 0.006 0.08 –
222 222 222 222 221 222 213 222
0.08 50.001 0.13
n – – – – 0.14 0.03 –
229 229 229 229 227 229 226 229
0.01 0.03 –
0.001
221
0.09
216
50.001
222
–
222
50.001 0.04 – – – – 0.19 0.05
221 218 221 219 219 219 218 220 215
– 50.001 0.004 50.001 50.001 50.001 50.001 –
216 215 216 214 216 216 215 215 210
50.001 50.001 50.001 0.006 0.006 0.002 0.014 50.001
221 219 222 220 220 220 219 221 216
0.18 – – – 0.08 – 0.09 –
221 221 222 220 220 220 221 221 223
50.001 0.01 0.05 50.001 – 0.003 0.003
– – – – – – – 219 0.04 0.03 0.01
– – – 0.007 0.08 – – 214
– – – 0.15 0.12 – – 220
227
0.03 – – –
– – – 5 0.001 0.14 0.03 0.001 0.01
0.10 – – 0.14 – – 5 0.001 –
0.08 0.09 – 0.16 – – – 0.20
0.01 0.11
– –
– –
0.07 –
172 –
168 0.06
222 – – – 0.08 – – – – – 0.003
223
172 0.05
215 5 0.001 5 0.001 50.001 0.002 50.001 50.001 0.001 0.01 50.001 0.04
217
179 0.09
221 0.04 0.01 0.001 50.001 0.001 0.15 0.04 0.001 0.001 0.07
222
227 – – – – – 0.20 – – – 0.29
229
–:p Value 40.20. Independent variables are those in the first column, outcome measures are the changes in scores between discharge and admission for physical functioning, mental health and pain, measured with the SF–36, and the patient-reported effect of the treatment (n¼number included in the analysis).
Study limitations The first limitation of our study was that we did not include improved social functioning as an outcome measure, although this has been found to be important [27]. Since any increase in social
activities (e.g. reintegration in work activities) had not yet been implemented at the time of our assessments immediately after the treatment, it was not useful to include this in our study. Secondly, the retrospective design of the study gave inherent limitations.
Predictors of pain rehabilitation
DOI: 10.3109/09638288.2014.961657
1249
Table 4. Results of the multiple linear regression analyses (for the outcomes in terms of physical and mental functioning and pain) and logistic regression analysis (for patient-reported effect) of the final models in the total group (inpatients and outpatients); independent variables are the scores at admission (baseline scores) and type of treatment in- or out-patient, dependent variables are the outcome scores (scores at discharge minus those at admission) and patient-reported effect. Unstandardized coefficient B
Downloaded by [Australian National University] at 19:34 06 November 2015
Physical functioning outcome Physical functioning Age Active coping Mental health outcome Mental health Vitality Coping style ‘‘praying and hoping’’ depression (SCL-90-R score) Pain outcome Pain Work status
Standard error
p Value
0.50 0.42 0.08
0.06 0.10 0.03
50.001 50.001 0.01
0.90 0.08 0.16 0.14
0.04 0.04 0.05 0.06
50.001 0.046 0.003 0.022
1.36 9.08
0.07 2.36
50.001 50.001
p Value
Nagelkerke R2 0.14
OR (95% CI) Patient-reported effect Employment Active coping Fear
0.31 (0.14–0.69) 0.98 (0.97–0.99) 0.93 (0.88–0.99)
Table 5. Result of the interaction analysis for the outcome in terms of pain: Univariate analysis of variance with baseline pain score (SF-36), work status, treatment program and interaction variables as independent variables, and the outcome in terms of pain (scores at discharge minus those at admission) as the dependent variable (n¼223). Unstandardized coefficient
Outcome pain Pain (SF-36) Work status Treatment program Treatment program *pain Treatment program *work status Pain*work status
p Value Adjusted R2
B
Standard error
1.29 23.14 15.22 0.33
0.14 6.46 6.17 0.14
50.001 50.001 0.01 0.02
0.48
4.83
0.92
0.39
0.15
0.01
0.70
Adjusted R2
n
0.27
220
0.80
214
0.68
222
n 179
0.008 0.010 0.113
samples. However, a predictor can be found only when there is a certain distribution in the scores.
Conclusions No strong predictors of treatment outcome were found other than the baseline scores of the respective outcome variables. More disabled patients and patients with more pain benefitted more from the rehabilitation program. Other variables, and the inclusion of interaction variables in the prediction model, did improve the outcome prediction, but only slightly. It thus remains difficult to predict the outcome of a treatment before its start.
Declaration of interest The authors report no declarations of interest.
References One of these is that the inclusion criterion of having psychosocial problems was not well defined. However, in usual care, a psychologist will only be involved when psychosocial problems are moderate or severe, so there is no doubt that our patients met this inclusion criterion. The low scores of the patients on mental health and vitality of the SF-36 and the high scores on the SCL-90 confirm this. Because, in a retrospective study, the inclusion of patients is less controlled, it is not known whether patients were erroneously not enrolled and bias was present. In addition, the exact content of the treatment program was not assessed either, due to the retrospective nature of the study. However, an important advantage of the current study was that it was conducted in usual care, as an experimental design may also introduce bias. The response rate in our study was low, possibly because the outcome assessment was made in the context of usual care, and over a period of several years. This meant there was no researcher who did his utmost to collect the data, as is inherent to the retrospective design. The low response rate may have caused some bias, but the included and not included eligible patients did not appear to differ in terms of predictive variables. Thirdly, the patients in our study differed largely in characteristics. This may complicate the external validity to specific homogeneous patient
1. Turk DC, Wilson HD, Cahana A. Treatment of chronic non-cancer pain, part 2. Lancet 2011;25:2226–34. 2. Vlaeyen FW, Morley S. Cognitive-behavioral treatments for chronic pain: what works for whom? Clin J Pain 2005;21:1–8. 3. Turk DC, Wilson HD, Cahana A. Treatment of chronic non-cancer pain. Lancet 2011;337:2226–35. 4. Bremander AB, Holmstro¨m G, Bergman S. Depression and age as predictors of patient-reported outcome in a multidisciplinary rehabilitation programme for chronic musculoskeletal pain. Musculoskelet Care 2011;9:41–8. 5. Orenius TI, Koskela T, Koho P, et al. Anxiety and depression are independent predictors of quality of life in patients with chronic musculoskeletal pain. J Health Psychol 2013;18:167–75. 6. Van der H, Vollenbroek-Hutten MM, Groothuis-Oudshoorn KGM, Hermens HJ. Multidisciplinary rehabilitation treatments with chronic low back pain: a prognostic model for its outcome. Clin J Pain 2008;24:421–30. 7. Michaelson P, Sjo¨lander P, Johansson H. Factors predicting pain reduction in chronic back and neck pain after multimodal treatment. Clin J Pain 2004;20:447–54. 8. Vollenbroek-Hutten MMR, Hermens HJ, Wever D, et al. Differences in outcome of a multidisciplinary treatment between subgroups of chronic low back pain patients defined using two multiaxial assessment instruments: the multidimensional pair inventory and lumbar dynamometry. Clin Rehabil 2004;18:566–79. 9. Buchner M, Neubauer E, Zahlten-Hinguranage A, Schiltenwolf M. Age as a predicting factor in the therapy outcome of
1250
10.
11.
12.
13.
Downloaded by [Australian National University] at 19:34 06 November 2015
14.
15.
16.
17.
18.
19.
20.
21.
A. M. Boonstra et al.
multidisciplinary treatment of patients with chronic low back pain – a prospective longitudinal clinical study in 405 patients. Clin Rheumatol 2007;26:385–92. Persson E, Lexell J, Eklund M, Rivano-Fischer M. Positive effects of a musculoskeletal pain rehabilitation program regardless of pain duration or diagnosis. PM&R 2012;4:355–66. Keel PJ, Wittig R, Deutschmann R, et al. Effectiveness of in-patient rehabilitation for sub-chronic and chronic low back pain by an integrative group treatment program (Swiss multicentre study). Scand J Rehab Med 1998;30:211–19. Hutten MMR, Hermens H, Zilvold G. Differences in treatment outcome between subgroups of patients with chronic low back pain using lumbar dynamometry and psychological aspects. Clin Rehabil 2001;15:479–88. Stigmar KGE, Petersson IF, Jo¨ud A, Grahn BEM. Promoting work ability in a structured national rehabilitation program in patients with musculoskeletal disorders: outcomes and predictors in a prospective cohort study. BMC Musculoskel Disord 2013;14:57. Keogh E, McCracken LM, Eccleston C. Do men and women differ in their response to interdisciplinary chronic pain management? Pain 2005;114:37–46. Hampel P, Graef T, Krohn-Grimberghe B, Tlach L. Effects of gender and cognitive-behavioral management of depressive symptoms on rehabilitation outcome among orthopaedic patients with chronic low back pain: a 1 year longitudinal study. Eur Spine J 2009;181: 1867–80. de Williams ACC, Richardson PH, Nicholas MK, et al. Inpatient vs. outpatient pain management: results of a randomized controlled trial. Pain 1996;66:13–22. de Rooij A, Roorda LD, Otten RHJ, et al. Predictors of multidisciplinary treatment outcome in fibromyalgia: a systematic review. Disabil Rehabil 2013;35:437–49. Scascighini L, Toma V, Dober-Spielmann S, Sprott H. Multidisciplinary treatment for chronic pain: a systematic review of interventions and outcomes. Rheumatology (Oxford) 2008;47: 670–8. Aaronson NK, Muller M, Cohen PDA, et al. Translation, validation, and norming of the Dutch language version of the SF-36 Health Survey in community and chronic disease populations. J Clin Epidemiol 1998;51:1055–68. Bergman S, Jacobsson LTH, Herrstro¨m P, Petorsson IF. Health status as measured by SF-36 reflects changes and predicts outcome in chronic musculoskeletal pain: a 3-year follow up study in the general population. Pain 2004;108:115–23. Barelds DPH, Luteijn F. Measuring personality: a comparison of three personality questionnaires in the Netherlands. Personal Individ Differ 2002;33:499–510.
Disabil Rehabil, 2015; 37(14): 1242–1250
22. Barelds DPH, Luteijn F, Van Dijk H, Starren J. Nederlandse Persoonlijkheids Vragenlijst-2, handleiding [manual for Dutch Personality Questionnaire, revised]. Lisse: Swets & Zeitlinger;2007 [in Dutch]. 23. Spinhoven P, ter Kuile MM, Linssen ACG. Coping met pijn vragenlijst; handleiding [The Coping with Pain Questionnaire: manual]. Lisse: Swets & Zeitlinger; 1994 [in Dutch]. 24. Rosenstiel AK, Keefe FJ. The use of coping strategies in chronic low back pain patients: relationship to patient characteristics and current adjustment. Pain 1983;17:33–44. 25. Vlaeyen JW, Kole-Snijders AM, Boeren RG, Van Eek H. Fear of movement/(re)injury in chronic low back pain and its relation to behavioral performance. Pain 1995;62:363–72. 26. Arrindell W, Ettema H, Groenman N, et al. De groeiende inbedding van de Nederlandse SCL-90-R; psychodiagnostisch gereedschap. [Growing use of Dutch version of SCL-90-R psychodiagnostic instrument] De Psycholoog 2003;38: 576–82. 27. Chapman JR, Norvell DC, Hermsmeyer JT, et al. Evaluating common outcomes for measuring treatment success for chronic low back pain. Spine 2011;36:s54–68. 28. Turk DC, Dworkin RH, Burke LB, et al. Developing patientreported outcome measures for pain clinical trials: IMMPACT recommendations. Pain 2006;215:208–15. 29. Cohen J. Statistical power analysis for the behavioral sciences. New York: Academic Press, Inc; 1977. 30. Angst F, Brioschi R, Main CJ, et al. Interdisciplinary rehabilitation in fibromyalgia and chronic back pain: a prospective outcome study. J Pain 2006;7:807–15. 31. Keller A, Hayden J, Bombardier C, Van Tulder M. Effect sizes of non-surgical treatments of non-specific low-back pain. Eur Spine J 2007;16:1776–88. 32. Morley S, Williams A, Hussain S. Estimating the clinical effectiveness of cognitive behavioural therapy in the clinic: evaluation of a CBT informed pain management programme. Pain 2008;137: 670–80. 33. Lunde L-H, Nordhus IH, Pallesen S. The effectiveness of cognitive and behavioural treatment of chronic pain in the elderly: a quantitative review. J Clin Psychol Med Settings 2009;16: 254–62. 34. Mangels M, Schwartz S, Worringen U, et al. Evaluation of a behavioral-medical inpatient rehabilitation treatment including booster sessions: a randomized controlled trial. Clin J Pain 2009; 25:356–64. 35. Boonstra AM, Reneman MF, Schiphorst Preuper HR, et al. Differences between patients with chronic musculoskeletal pain treated in an inpatient or outpatient multidisciplinary rehabilitation program. IJRR 2014;37:187–91.
