Abstract
Objective
To establish the predictive validity of a French translation of the Örebro Musculoskeletal Pain Screening Questionnaire (OMPSQ), a screening tool assessing the risk of chronicity in patients with back pain.
Methods
Prospective follow-up study. Assessment was performed at inclusion and 6 months later with the OMPSQ and the Oswestry Disability Index (ODI). Four outcome variables (pain index, two functional variables and work absence) were defined.
Results
Ninety-one patients were included, of whom 80% completed the study. Depending on the outcome variable considered, 42 to 82% of the patients recovered within 6 months. ROC AUC, a global measure of the performance of the questionnaire integrating sensitivity and specificity data, ranged from 0.73 to 0.83. When considering the functional outcome variable derived from the ODI, a low cut-off score of 71 (corresponding to 80% sensitivity) and a high cut-off score of 106 (corresponding to 80% specificity) can be used to distinguish three groups of patients: low, intermediate and high risk of chronicity.
Conclusion
The predictive value of the French version of the OMPSQ is reasonably good, in line with the studies in other languages. This questionnaire may be particularly valuable in secondary care settings.
Résumé
Objectifs
Évaluer la valeur prédictive d’une traduction française du questionnaire Örebro Musculoskeletal Pain Screening Questionnaire (OMPSQ), un outil de dépistage du risque de chronicité chez des patients souffrant de rachialgie aiguë.
Méthodes
Étude prospective. Les patients ont été évalués à l’inclusion et six mois plus tard, au moyen des questionnaires OMPSQ et Oswestry Disability Index (ODI). Quatre variables de suivi ont été définies (index de douleur, deux variables fonctionnelles et absence au travail).
Résultats
Quatre-vingt-onze patients ont été inclus, 80 % ont complété l’étude. Selon la variable de suivi considérée, 42 à 82 % des patients peuvent être considérés guéris à six mois. L’aire sous la courbe ROC, une mesure globale de la performance du questionnaire intégrant sensibilité et spécificité, se situe entre 0,73 et 0,83. Considérant l’évolution fonctionnelle calculée selon le questionnaire ODI, des scores seuils de 71 (correspondant à une sensibilité de 80 %) et 106 (correspondant à une spécificité de 80 %) peuvent être utilisés pour délimiter trois groupes de patients : risque de chronicité bas, intermédiaire et élevé.
Conclusion
La valeur prédictive de la version française du questionnaire OMPSQ est raisonnablement bonne, en accord avec les études réalisées dans les autres langues. Ce questionnaire est particulièrement utile dans les structures de soins de santé de seconde ligne.
1
English version
1.1
Introduction
Back pain is the most prevalent musculoskeletal condition . Its human and economic burden is high and the small subpopulation of chronic low back pain patients accounts for most of the expenses in this field . As a consequence, identifying and treating acute back pain patients that are at risk of developing chronic problems could be very valuable.
Numerous psychosocial variables are involved in the transition from acute to chronic back pain and evidence-based guidelines stress the importance of assessing psychosocial factors . These “yellow flags” include psychological distress, illness beliefs and coping, occupational, social and treatment provider factors . Taken in isolation, their prognostic value is low, emphasising the need for a multidimensional assessment .
Several screening tools exist , among which the Örebro Musculoskeletal Pain Screening Questionnaire (OMPSQ) . This questionnaire has the advantage of covering most of the identified psychosocial risk factors (pain, psychological distress, coping, beliefs, work perception, work absence, functional limitations) while being relatively short. It has been validated in a number of different clinical settings and in several countries in Northern Europe . The New Zealand guidelines for acute low back pain recommend the use of the OMPSQ. A recent systematic review confirmed a moderate predictive ability and recommended its use, but suggested to continue validation efforts in different clinical and cultural settings.
The aim of this study was to assess the predictive validity of a French translation of the OMPSQ. The protocol closely followed that of the initial studies , in order to facilitate comparison. To provide an external measure of the functional capacities, we added the Oswestry Disability Index (ODI) .
1.2
Patients and methods
1.2.1
Questionnaire translation
As recommended , an initial English-to-French translation was performed by three independent individuals (two medical doctors specialised in Physical Medicine and Rehabilitation and one psychologist with expertise in cognitive-behavioural treatment of chronic pain) who are native French speakers but speak English fluently. Collation of the three versions was carried out to compose a single text. Some minor issues were solved via e-mail correspondence with Dr Linton and Dr Boersma. A native English speaker then back-translated the final text from French to English. No significant difference in meaning was observed when comparing the result with the original questionnaire. The resulting French questionnaire and response categories are presented in the first three columns of Appendix 1 . To facilitate comparison, this table has the same format as in Linton and Boersma , but questions 3 and 4 are inverted (because they are not always presented in the same order in the original papers).
1.2.2
Inclusion and exclusion criteria
Patients presenting during office hours at the emergency facility or at the outpatient clinic of the Department of Physical Medicine and Rehabilitation at the Cliniques universitaires Saint-Luc (Brussels, Belgium), with nonspecific acute or subacute low back or neck pain, were considered for inclusion. Doctors from the emergency facility did not always inform our team of the presence of eligible patients, therefore, our sample is not composed of strictly consecutive patients. Inclusion criteria were the presence of pain lasting for less than three months and a cumulated sick leave because of the pain of less than 6 months in the past year. Exclusion criteria were the inability to read and understand French and the presence of “red flags”, i.e. symptoms and signs suggestive of a potentially serious cause for the pain .
1.2.3
Procedure
During their first consultation (t 0 ), patients were informed about the study and asked to sign an informed consent. To avoid bias, the objectives of the study were described only very broadly (“we want to better understand back pain”), with no reference to evolution, chronicity or risk factors.
Patients agreeing to participate were then handed the questionnaires, which they usually filled out in the waiting room. In some instances (e.g., when the patient did not have enough time), the questionnaires and a prepaid envelope were given to the patient for return within the next few days. Health caregivers following the patients did not have access to the results of the questionnaires. Patients were treated as usual. No attempt was made to record the treatments provided to the patients.
A second evaluation was performed 6 months later (t 6 ). After a telephone reminder, questionnaires were sent by post with a prepaid return envelope, two weeks before the assessment time. If no answer was received within 2 weeks, a reminder was sent by post; if no answer was received within another 2 weeks, the patient was again contacted by phone by the first author.
The study was approved by the local Ethics Committee.
1.2.4
Questionnaires
A first sheet contained questions about the native language of the subject, his or her professional status (paid job as blue collar, white collar or self-employed worker), the existence of a work-related injury, the number of years of education after the age of six, and pain localisation (back only, back and leg above the knee or back and leg below the knee).
Two questionnaires were used, both at the first contact (t 0 ) and 6 months later (t 6 ): the French translation of the OMPSQ and a French version of the Oswestry Disability Questionnaire (ODI) . In the OMPSQ sent to the patients at 6 months, item #6 (cumulated sick leave) was adapted to cover the period of interest, i.e. “during the past 6 months” (instead of “during the past 12 months”). At the time this study began, no validated French translation of the ODI was available; we therefore used a non-validated French translation available in our department.
1.2.5
Data analysis
Data were encoded in a spreadsheet file (Microsoft Excel).
The OMPSQ was scored according to previous studies , summing scores for individual items 5 to 25. Items 16, 17 and 21 to 25 were inverted before scoring. For item #5, the score was the number of painful sites (0 to 5) multiplied by 2.
For the ODI, the answers to each item were summed and a relative score was calculated to take missing data into account: relative score = total score × 100/(5 × number of questions answered) .
Four outcome variables were defined ( Table 1 ), three of which were derived from the OMPSQ, as in the initial studies , the fourth being an external validation with the ODI. Cut-off scores for dichotomization of “recovered” versus “non-recovered” patients were the same as in the initial studies for the three outcome variables derived from the OMPSQ. For the ODI, a relative score less than or equal to 20% was considered normal . The four outcome variables covered the three domains of the International Classification of Functioning, Disability and Health (ICF) .
| Outcome variable | ICF domain | Method of computation | Cut-off scores for dichotomization | % “recovered” patients at t 6 | |
|---|---|---|---|---|---|
| “Recovered” | “Non-recovered” | ||||
| OMPSQ Pain index | Body functions and structures | Product of scores for items 10 (pain intensity) and 11 (frequency of pain episodes) | ≤ 16 | > 16 | 66 |
| OMPSQ Functional index | Activity | Sum of scores for items 21 to 25 (before inversion) | ≥ 45 | < 45 | 42 |
| OMPSQ Work absence | Participation | Item 6 (number of days off work during the past 6 months) | ≤ 5 (scores corresponding to ≤ 30 days) | > 6 (scores corresponding to > 30 days) | 82 |
| ODI | Activity | Relative score, see Methods | ≤ 20% | > 20% | 73 |
1.2.6
Statistics
OMPSQ and ODI scores were considered ordinal, because these scores are ordered but interval differences between scores cannot be assumed to be equal .
Correlation between OMPSQ and ODI scores was measured using Kendall rank-order correlation coefficient τ. Comparison between scores at t 0 and t 6 were performed with Wilcoxon signed-rank tests.
Comparison between initial OMPSQ or ODI scores split by outcome were performed with Mann-Whitney U tests.
In the univariate analysis, correlation of individual items of the questionnaires with outcome was carried out using unpaired t -tests for most items or Fisher’s exact test for binomial variables.
Receiver operating characteristic (ROC) curves and the area under the curve (AUC) were computed for each outcome variable. Sensitivity and specificity were calculated for different cut-off values of the total OMPSQ score, for each outcome variable.
Demographic characteristics comparing drop-out patients with patients having completed the study were assessed using unpaired t -tests for continuous normally distributed variables (age, years of education), Mann-Whitney U tests for not normally distributed or ordinal variables (number of painful sites, OMPSQ and ODI scores) and Fisher’s exact test for binomial variables (gender, language, work status, work-related injury).
Statistical analysis was performed with R .
1.3
Results
1.3.1
Demographic data
Ninety-one patients entered the study at t 0 , and 73 (80%) sent completed questionnaires at t 6 . As seen in Table 2 , demographic variables were not significantly different between drop-out patients and patients having completed the study, suggesting no significant inclusion bias. However, a smaller proportion of drop-out patients were native French speakers than patients who completed the study. In the group who completed the study, 47% of the patients had lumbar pain only, 8% had cervical pain and 45% had pain on more than one site.
| Patients having completed the study | Drop-out patients | P | |
|---|---|---|---|
| Number of subjects | 73 | 18 | |
| Age (years) mean (SD) | 42.2 (10.7) | 38 (8.6) | 0.128 |
| Sex ratio (% female) | 56 | 45 | 0.445 |
| Percentage of native French speaking | 86 | 56 | * 0.037 |
| Percentage with paid job | 88 | 83 | 0.382 |
| Percentage with work-related injury | 16 | 28 | 0.314 |
| Percentage with pain of less than 3 weeks | 58 | 58 | – |
| Number of years of education after age 6 mean (SD) | 15.3 (3.5) | 14.5 (2.8) | 0.409 |
| Number of painful sites median [range] | 1 [1–5] | 1 [1–3] | 0.326 |
1.3.2
Patients evolution
Scores measured at t 0 and t 6 are given in Table 3 . OMPSQ and ODI total scores were significantly lower at 6 months than at inclusion, indicating overall improvement. The only OMPSQ outcome variable which was significantly improved was the OMPSQ Functional index. OMPSQ Pain index and Work absence were were low at inclusion (we included only patients with (sub)acute pain) and not significantly different between t 0 and t 6 .
| Theoretical maximum | t 0 | t 6 | Wilcoxon signed-rank test P | |
|---|---|---|---|---|
| OMPSQ total score | 210 | 85 [25–171] | 66 [12–151] | < 0.0001 |
| OMPSQ Pain index | 100 | 12 [0–80] | 12 [0–80] | 0.201 |
| OMPSQ Functional index | 100 | 31 [0–50] | 43 [11–50] | < 0.0001 |
| OMPSQ Work absence | 10 | 2 [1–7] | 2 [1–9] | 0.223 |
| ODI relative score | 100 | 30 [0–72] | 12 [0–54] | < 0.0001 |
The proportion of patients considered as recovered (last column of Table 1 ) was 32% when all four outcome variables were considered. Twenty-five percent of patients showed poor evolution on one outcome variable only, 21% on two outcome variables, 13% on three outcome variables and 8% of the patients show persistent problems with each of the four outcome variables.
1.3.3
Correlation of Örebro Musculoskeletal Pain Screening Questionnaire with Oswestry Disability Index
Kendall rank-order correlation showed that the ODI relative score and OMPSQ total score were weakly correlated (τ = 0.46 and 0.49 at t 0 and t 6 respectively, P < 0.0001). As expected, the correlation was slightly higher when comparing ODI with the OMPSQ functional subscore (τ = 0.55 at t 0 and at t 6 , P < 0.0001).
1.3.4
Correlation of Örebro Musculoskeletal Pain Screening Questionnaire individual items with outcome
In univariate analysis, most items were significantly related to at least one of the outcome variables ( Appendix 1 , right panel). Three items emerged as being correlated with all outcome variables: depression (#14), perceived likelihood of working in 6 months (#16) and ability to perform light work (#21). Several items were correlated with three outcome variables: frequency of pain episodes (#11), stress (#13), perceived risk of chronicity (#15) and sleep (#25). Employment status (#4), coping (#12), job satisfaction (#17) and shopping (#24) showed no correlation with outcome.
1.3.5
Predictive value of Örebro Musculoskeletal Pain Screening Questionnaire total score for chronicity
Mann-Whitney U tests comparing OMPSQ scores between the group of patients who could be considered as recovered and the group of patients showing significant pain and/or impairment at 6 months were highly significant ( P ≤ 0.003) for all outcome variables. However, the distribution frequency of OMPSQ scores showed a large overlap between these two groups, whatever the outcome variable considered (data not shown). ROC AUC and the sensitivity and specificity of different cut-off scores are shown in Table 4 . The best ROC AUC result was obtained for prediction of work absence. As expected from the large group overlap observed in the frequency distribution analysis, it is difficult to select a single cut-off score showing simultaneously good sensitivity and specificity. Therefore, it seems reasonable to suggest determining three risk categories: low, intermediate and high. As proposed by Jellema et al. , the cut-off for “low risk” was set to the point corresponding to 80% sensitivity while the cut-off for “high risk” was set at the point corresponding to 80% specificity ( Table 5 ).
| Outcome variable | ROC AUC | Cut-off score | 70 | 80 | 90 | 100 | 110 | 120 |
|---|---|---|---|---|---|---|---|---|
| OMPSQ | 0.73 | Sensitivity | 88 | 72 | 68 | 56 | 36 | 32 |
| Pain index | Specificity | 35 | 54 | 75 | 81 | 90 | 98 | |
| Predictive positive value | 42 | 45 | 59 | 61 | 64 | 89 | ||
| OMPSQ | 0.787 | Sensitivity | 88 | 73 | 51 | 46 | 32 | 22 |
| Functional index | Specificity | 50 | 73 | 77 | 87 | 97 | 100 | |
| Predictive positive value | 71 | 79 | 75 | 83 | 93 | 100 | ||
| OMPSQ | 0.834 | Sensitivity | 100 | 85 | 69 | 62 | 62 | 46 |
| Work absence | Specificity | 32 | 51 | 66 | 75 | 90 | 95 | |
| Predictive positive value | 25 | 28 | 31 | 35 | 57 | 67 | ||
| ODI | 0.748 | Sensitivity | 86 | 73 | 64 | 50 | 41 | 36 |
| Specificity | 30 | 51 | 70 | 74 | 89 | 98 | ||
| Predictive positive value | 37 | 41 | 50 | 48 | 64 | 89 | ||
| Cut-off scores | ||
|---|---|---|
| Outcome variables | Low cut-off (sensitivity 80%) | High cut-off (specificity 80%) |
| OMPSQ Pain index | 75 | 97 |
| OMPSQ Function index | 76 | 86 |
| OMPSQ Work absence | 75 | 106 |
| ODI | 71 | 106 |
1.3.6
Predictive value of Oswestry Disability Index scores
Mann-Whitney U tests showed significant correlation between initial ODI scores and three outcome variables (OMPSQ functional score, OMPSQ work absence and ODI at 6 months) ( P < 0.05). The AUC for these three variables were 0.646, 0.722 and 0.669 respectively. Taken together, this showed that ODI was a poorer outcome predictor than OMPSQ. The analysis was therefore not pursued further.
1.4
Discussion
To summarise our main results, 42 to 82% of the patients recovered within 6 months, depending on the outcome variable considered. Most items on the OMPSQ questionnaire were correlated with outcome, the most significant items being depression, perceived likelihood of working in 6 months and ability to perform light work. The ROC curves and sensitivity/specificity analysis showed a reasonably good predictive value of the French version of the OMPSQ, with ROC AUC ranging from 0.73 to 0.83 for the four outcome variables. We propose to determine two cut-off scores, in order to distinguish three groups of patients for low, intermediate and high risk of chronicity. If the outcome variable chosen is the functional ODI, OMPSQ cut-off scores of 71 and 106 can be used.
Most demographic characteristics of the patients were within the range of the other validation studies of the OMPSQ : 88% patients were on paid jobs (67–100% in the literature), 65% were females (48–70%), mean age was 42 years (38–47).
Compared to other studies, pain duration at inclusion (58% patients ≤ 3 weeks, 91% < 12 weeks) was in the lowest range, with only two studies showing a shorter duration (100% < 3 weeks). This is important if the aim of the screening procedure is to predict the risk of chronicity as soon as possible in patients with (sub)acute back pain.
The OMPSQ total score at t 0 (median 85) was lowest in our study (median scores in the literature, mostly from primary care settings, ranged from 95 to 123), suggesting slightly less perceived disability in our sample. This is somewhat surprising, since our patients were recruited in the emergency unit or in an outpatient consultation of Physical Medicine and Rehabilitation of a university hospital, where one would expect more severe cases to be seen than in primary settings. This might be a consequence of the organisation of the health care system in Belgium, where it is not compulsory to meet a general practitioner before consulting a specialist.
The general evolution of back pain observed in our study was in line with other published data. For example, Linton and Boersma observed that, at 6 months follow-up, 17% of patients had more than 30 days of sick leave (18% in the present sample), 60% had not fully recovered as measured by the OMPSQ function index (58% in our data), and 48% had not recovered for the OMPSQ pain index (34% in our population).
Analysing the evolution of back pain poses the question of pertinent outcome variables. Return to work is the primary outcome of most studies, probably because it is easy to measure and is economically meaningful. Using this variable, the evolution of acute back pain is usually favourable: in a prospective study in general practice , 97% of the patients suffering from an acute episode of low back pain (LBP) had resumed work within one month. However, at one month and one year, respectively 59% and 46% of patients considered themselves as not completely recovered and 16% and 8% of patients mentioned functional limitations. This shows that return to work is a rough indicator of recovery from LBP and that functional and pain outcome variables must be used in conjunction with work ability. This also increases the range of ICF domains considered and is consistent with the biopsychosocial model used in chronic pain.
In this study, four outcome variables were defined. Three of these, corresponding to the domains cited above (pain, function, work absence), were directly derived from the OMPSQ, as in the original studies . We added the ODI to provide an external evaluation. We considered adding other measures of pain or work absence but did not do it, because the questions relating to these items in the OMPSQ are questions derived from validated tools.
Another question relating to prognostic studies is the risk of bias induced by the protocol. In this study, we took care to avoid telling the patients that we searched for factors related to the outcome. Moreover, the results of the questionnaires were not disclosed to the clinicians following the patients. Finally, the relatively high response rate and the absence of significant differences between drop-out patients and patients completing the study suggest no significant inclusion bias.
Univariate analysis of the predictive value of single OMPSQ items shows that depression, perceived likelihood of working in 6 months and ability to perform light work were best correlated with outcome. Other important items were frequency of pain episodes, stress, perceived risk of chronicity and sleep. Other studies partially overlap with these findings, identifying additional items such as previous sick leave and the belief that one should not work with the current pain level. The fact that significant items are not exactly the same in different studies is not surprising, given the diversity of patients and problems. However, the perceived likelihood of working in 6 months seems to emerge in most studies. Interestingly, the predictive value of the patient’s own expectations is salient in other questionnaires as well . This stresses the importance of addressing patient’s beliefs in back pain evaluation and treatment.
The predictive value of the OMPSQ total score observed in this study is comparable to that found in the other studies, with slightly better ROC AUC than those reviewed by Hockings et al. . Previous studies were performed in Sweden , Northern Ireland , The Netherlands , Norway and the United Kingdom , all of which belong to the German linguistic group. All these studies recruited patients from primary care settings, except Hurley et al. , which included patients from a physiotherapy department. The fact that our results, in a language from the Roman linguistic group and in a secondary care setting, are very close to those of the others studies adds further data supporting the good external validity of the OMPSQ.
However, as noted by Hockings et al. , the predictive value of the OMPSQ is moderate, suggesting that some risk factors are not taken into account. This is not surprising, since the number of potential risk factors is high. Significant risk factors may not be the same in all patients and risk factors may evolve with time in a single patient . Schultz et al. designed a predictive instrument with a slightly better ROC AUC (0.84). This questionnaire is much longer than the OMPSQ (time for completion 20–30 min instead of 5–10 min). Much shorter questionnaires have been recently proposed and preliminary validated , but their results remain to be confirmed. Indeed, short questionnaires may perform well in selected populations of patients but may have lower external validity. This is suggested by the finding that OMPSQ individual items best correlated with outcome are not the same in different studies (see above). The OMPSQ may offer the best compromise to date between brevity and accuracy.
Sensitivity and specificity values for different cut-off scores in this study are in the same range as those calculated by other authors. Linton and colleagues initially proposed to define a single cut-off score segregating patients in two groups. Proposed cut-off values ranged between 90 and 112 . However, our data suggest that it might be more accurate to distinguish three groups of patients. The same proposal has been made by Grotle et al. (cut-off scores 90 and 105) and Hill et al. (cut-off scores 90 and 112). Moreover, the cut-off scores will critically depend on clinical or research objectives: is it important not to miss any “at risk” patient and prefer high sensitivity, thus lowering the cut-off, or to spare resources and prefer high specificity, increasing the cut-off? Another critical point is the choice of the outcome variable (e.g., is the primary aim to reduce work absenteeism or to increase function?). Therefore, it is impossible to determine a single cut-off value applicable to all settings, and Hockings et al. recommended publishing detailed measures of accuracy for several cut-off points. This would allow researchers and practitioners to choose the cut-off points most relevant to their settings and objectives.
The clinical relevance of the OMPSQ has been challenged by data showing that the predictive value of OMPSQ is comparable to risk prediction by the general practitioners (GPs) , confirming the observations of Schiottz-Christensen et al. . This is not surprising, given the good general knowledge of GPs about their patients and their backgrounds . However, in Belgium, patients are allowed to self-refer directly to second-line care settings. The OMPSQ might be of special value in these cases.
Moreover, examination of separate items by the clinician might help to distinguish relevant areas of concern in a given patient . This may be of particular interest in the intermediate risk group where it can be hypothesised that most patients will have only a few specific problems (e.g., depression or fear-avoidance), which could be directly addressed. This is supported by our data showing that in the intermediate risk group, an average of 3.3 items had a score situated one standard deviation above the mean for this item (compared to 1.9 and 8.3 for the low and high risk groups, respectively). Thus, looking at individual items might help to tailor secondary prevention to the patient’s needs, sparing resources.
Another question relating to the clinical relevance of any screening method is the availability of preventive strategies. For back pain, which is very common, secondary prevention (i.e. prevention of the consequences of back pain) is important. To date, the evidence in favour of the effectiveness of treatment strategies for secondary prevention is scarce , but there are data suggesting that secondary prevention is feasible, even if the effect sizes are modest . More studies are clearly needed.
The ODI was used to provide independent information on the functional status of the patients. Indeed, the use of a functional index directly derived from the OMPSQ induces circularity, which may bias the analysis of the predictive value of the OMPSQ. The ODI has been extensively validated and is considered largely comparable to the Roland-Morris disability questionnaire . Our data show that ODI scores are correlated with the OMPSQ scores, which is not surprising since both questionnaires include questions concerning pain intensity and functional limitations. As also observed by other authors , the initial ODI scores were correlated with the outcome at 6 months. Again, this is not unexpected, because pain intensity and disability due to back pain are among the identified risk factors for chronicity. Although these factors are not strictly psychosocial, the fact that they are influenced by cognitive, emotional and behavioural factors allows considering them as bridging physical and psychological factors . However, the ODI does not include questions about psychological distress, beliefs or occupational factors. This probably explains why it is a poorer predictor of chronicity than the OMPSQ.
The main limitation of this study is the relatively small number of patients. Moreover, we did not strictly recruit consecutive cases. It may be that patients coming to the emergency unit at night or during weekends have different characteristics than patients from our sample. However, both demographic and outcome data are largely comparable to those of the original studies, suggesting reasonably good representativeness.
Another limitation is that we used a non-validated French version of the ODI. A cross-cultural validation of the ODI in French was recently published , but this was too late to be included in our protocol. However, it is unlikely that using this questionnaire would have changed the main conclusions of our study.
Several other questionnaires (e.g., specifically assessing individual risk factors such as coping, catastrophizing or psychological distress) could have been added. However, our first aim was to assess the predictive ability of the French version of the OMPSQ and not to make a theoretical analysis of prognostic factors for chronicity in LBP.
In conclusion, the present study shows a relatively good predictive validity of the french translation of the OMPSQ to assess the risk for chronicity in patients suffering from (sub)acute back pain.
Disclosure of interest
The authors declare that they have no conflicts of interest concerning this article.
Acknowledgements
Many thanks to A. Guillet for help with statistical analysis.
2
Version française
2.1
Introduction
La rachialgie est la pathologie musculo-squelettique la plus fréquente . Son impact humain et économique est important et le petit groupe de patients lombalgiques chroniques est responsable à lui seul de la plupart des dépenses dans ce domaine . En conséquence, il serait très utile de pouvoir identifier et traiter, parmi les patients souffrant de rachialgie aiguë, ceux dont le risque de passage à la chronicité est élevé.
De nombreux facteurs psychosociaux sont impliqués dans la transition de la rachialgie aiguë à la rachialgie chronique et les recommandations de bonne pratique clinique insistent sur l’importance d’évaluer ces facteurs . Ces « drapeaux orange » comprennent la détresse psychologique, les croyances autour de la pathologie, la manière de gérer la maladie ( coping ) ainsi que des facteurs socioprofessionnels et de modalités de prise en charge . Pris séparément, leur valeur pronostique est faible, soulignant l’intérêt d’une évaluation multidimensionnelle .
Plusieurs outils de dépistage existent , dont le Örebro Musculoskeletal Pain Screening Questionnaire (OMPSQ) . Ce questionnaire a l’avantage de couvrir la plupart des facteurs de risque psychosociaux (douleur, détresse psychologique, coping, croyances, perception de l’activité professionnelle, absentéisme, limitations fonctionnelles) tout en étant relativement court. Il a été validé dans plusieurs contextes cliniques et dans de nombreux pays d’Europe du Nord . Les guides de bonne pratique clinique de Nouvelle Zélande pour la prise en charge des lombalgies aiguës recommandent l’utilisation du questionnaire OMPSQ . Récemment, une revue systématique de la littérature a confirmé sa valeur prédictive modérée et recommande son utilisation, suggérant néanmoins de continuer les efforts de validation au sein de différents contextes cliniques et culturels.
Le but de cette étude était d’évaluer la valeur prédictive d’une traduction française de l’OMPSQ. Le protocole était similaire à celui utilisé pour les études initiales , afin de faciliter la comparaison des résultats entre ces études. Afin de fournir une mesure externe des capacités fonctionnelles, nous avons ajouté l’Oswestry Disability Index (ODI) .
2.2
Patients et méthodes
2.2.1
Traduction du questionnaire
Suivant les recommandations établies , une première traduction de l’anglais vers le français a été menée par trois personnes différentes (deux médecins de médecine physique et réadaptation et un psychologue expert en traitement cognitivo-comportemental de la douleur chronique). La langue maternelle de ces trois experts est le français mais ils parlent couramment l’anglais. Les trois versions ont ensuite été colligées afin de former un seul texte. Quelques problèmes mineurs furent résolus par correspondance électronique avec les Dr Linton et Boersma. Une personne de langue maternelle anglaise a ensuite réalisé une rétro-traduction du texte final du français vers l’anglais. Aucune différence significative dans le sens du texte n’a été observée en comparant les résultats avec le questionnaire original. La version définitive du questionnaire en français et les catégories de réponses sont données dans les trois premières colonnes de l’ Annexe 1 . Pour faciliter la comparaison ce tableau reprend le même format que celui de Linton et Boersma , mais les questions 3 et 4 ont été inversées (car elles ne sont pas toujours présentées dans le même ordre dans les études originales).
2.2.2
Critères d’inclusion et d’exclusion
Les patients se présentant aux urgences ou en consultation ambulatoire au sein du service de médecine physique et réadaptation des cliniques universitaires Saint-Luc (Bruxelles, Belgique), avec une douleur aiguë ou subaiguë non spécifique, localisée dans le bas du dos ou les cervicales, étaient susceptibles de rejoindre l’étude. Les médecins urgentistes n’informaient pas toujours notre équipe de la présence de patients présentant les critères d’inclusions requis, en conséquences, notre échantillon ne se compose pas strictement de patients consécutifs.
Les critères d’inclusion étaient les suivants : douleur présente depuis moins de trois mois et nombre cumulé de jours d’arrêt maladie à cause de la douleur inférieur à six mois au cours de la dernière année.
Les critères d’exclusion étaient l’incapacité à lire et à comprendre le français, la présence de « drapeaux rouges », c’est-à-dire des symptômes et signes suggérant une cause sérieuse pour la douleur .
2.2.3
Protocole
Durant la consultation initiale (t 0 ), les patients ont été informés sur l’étude et il leur était demandé de signer un formulaire de consentement éclairé. Afin d’éviter les biais, les objectifs de l’étude étaient décrits de façon assez générale (« nous voulons mieux comprendre la douleur lombaire ») sans référence aucune à l’évolution, la chronicité ou les facteurs de risque.
Les patients ayant donné leur accord pour participer à l’étude ont ensuite reçu les questionnaires, qu’ils remplissaient habituellement dans la salle d’attente. Dans certains cas (par exemple quand le patient manquait de temps pour compléter le document), le questionnaire était accompagné d’une enveloppe prépayée afin qu’il puisse nous le retourner rapidement par la poste. L’équipe médicale en charge du patient n’avait pas accès aux résultats du questionnaire et les patients étaient pris en charge comme d’habitude. Les traitements prescrits aux patients n’ont pas été enregistrés dans cette étude.
Une seconde évaluation a été réalisée six mois plus tard (t 6 ). Après un rappel téléphonique, les questionnaires ont été envoyés par voie postale avec une enveloppe prépayée pour le retour, deux semaines avant la date d’évaluation. Si aucune réponse n’avait été reçue dans ce délai, une lettre de rappel était envoyée par la poste ; si 15 jours plus tard l’équipe n’avait aucune nouvelle, l’auteur principal téléphonait personnellement au patient.
Cette étude a reçu l’approbation du comité d’éthique local.
2.2.4
Questionnaires
Un premier feuillet contenait des questions sur la langue maternelle du patient, son statut professionnel (ouvrier, employé, profession libérale), l’existence d’un éventuel accident de travail, le nombre d’années d’étude après l’âge de six ans et la localisation des douleurs (dos uniquement, dos et jambe au-dessus du genou ou dos et jambe au-dessous du genou).
Deux questionnaires ont été utilisés, tant à l’inclusion (t 0 ) que lors de l’évaluation réalisée six mois plus tard (t 6 ) : la traduction française de l’OMPSQ et la version française de l’Oswestry Disability Questionnaire (ODI) . Pour le questionnaire OMPSQ envoyé aux patients six mois après la première évaluation, l’item 6 (nombre cumulé de jours de maladie) a été modifié afin de couvrir la période concernée, c’est-à-dire « durant ces six derniers mois » (au lieu de « durant ces 12 derniers mois »). Au début de l’étude, aucune traduction française validée de l’ODI n’était disponible, nous avons donc utilisé une version française non validée disponible dans nos services.
2.2.5
Analyse des données
Les données ont été encodées dans un tableur (Microsoft Excel).
L’OMPSQ a été coté suivant les recommandations faites dans les études précédentes , c’est-à-dire en additionnant les scores pour les items 5 à 25. Les items 16 et 17 ainsi que les items 21 à 25 ont été inversés avant d’effectuer le calcul final. Pour l’item 5, le score correspondait au nombre de localisations douloureuses (de 0 à 5) multiplié par deux.
Pour l’ODI, les réponses données pour chaque item ont été additionnées et un score relatif prenant en compte les données manquantes a été calculé : score relatif = score total × 100/(5 × nombre de questions répondues) .
Quatre variables ont été définies ( Tableau 1 ), parmi celles-ci trois étaient dérivées de l’OMPSQ, comme dans les études initiales , la quatrième était une validation externe basée sur l’ODI. Les scores seuils utilisés pour la dichotomisation des patients « guéris » et « non-guéris » étaient les mêmes que dans les études initiales pour les trois variables dérivées de l’OMPSQ. Pour l’ODI, un score relatif inférieur ou égal à 20 % a été considéré normal . Les quatre variables de suivi couvrent les trois domaines de la Classification internationale du fonctionnement, du handicap et de la santé (CIF) .
| Variables de suivi | Domaine CIF | Méthode de calcul | Scores seuils pour dichotomisation | Pourcentage de patients « guéris » à t 6 | |
|---|---|---|---|---|---|
| « Guéris » | « Non-guéris » | ||||
| OMPSQ douleur | Fonctions organiques et structures anatomiques | Produit des scores pour les items 10 (intensité de la douleur) et 11 (fréquence des épisodes douloureux) | ≤ 16 | > 16 | 66 |
| OMPSQ fonction | Activités | Somme des scores pour les items 21 à 25 (avant inversion) | ≥ 45 | < 45 | 42 |
| OMPSQ absentéisme | Participation | Item 6 (nombre de jours d’absentéisme durant les 6 derniers mois) | ≤ 5 (scores correspondant à ≤ 30 jours) | > 6 (scores correspondant à > 30 jours) | 82 |
| ODI | Activités | Score relatif (voir Méthodes) | ≤ 20 % | > 20 % | 73 |
2.2.6
Statistiques
Les scores de l’OMPSQ et de l’ODI ont été considérés comme des données ordinales, car ils sont ordonnés mais les différences d’écart entre les scores ne sont pas forcément égales .
La corrélation entre les scores de l’OMPSQ et de l’ODI a été évaluée au moyen des tests de corrélation de rang τ de Kendall. La comparaison entre les scores à t 0 et t 6 a été réalisée au moyen des Tests de Wilcoxon et des tests de rang de Mann-Whitney pour les variables non paramétriques.
Le test U de Mann-Whitney a été utilisé pour la comparaison entres les scores initiaux de l’OMPSQ et de l’ODI en fonction de l’évolution des patients.
Dans l’analyse univariée, la corrélation de chacun des items des questionnaires avec la variable de suivi a été conduite à l’aide de tests- t non pairés pour la plupart des items ou le test exact de Fisher pour les variables binomiales.
Pour chaque variable de suivi, les courbes Receiver operating characteristic (ROC) et l’aire sous la courbe (AUC) ont été calculées. La sensibilité et la spécificité ont été déterminées pour différentes valeurs seuils du score total OMPSQ, pour chaque variable.
La comparaison des caractéristiques démographiques des patients sortis de l’étude avant t 6 avec les patients ayant mené l’étude à terme ont été analysées avec des tests de t non appariés pour les variables continues de distribution normale (âge, années d’éducation), le test U de Mann-Whitney pour les variables de distribution non normale ou les variables ordinales (nombre de localisations douloureuses, scores OMPSQ et ODI) et enfin le test exact de Fisher pour les variables binomiales (sexe, langue maternelle, statut professionnel, blessure liée au travail).
Le logiciel « R » a été utilisé pour les analyses statistiques .
2.3
Résultats
2.3.1
Données démographiques
Quatre-vingt-onze patients ont été inclus dans l’étude à t 0 et 73 patients (80 %) ont renvoyé les questionnaires complétés à t 6 . Comme montré dans le Tableau 2 , il n’y avait pas de différence significative dans les variables démographiques entre les patients sortis de l’étude et les patients ayant complété l’étude, ceci suggérant l’absence d’un biais d’inclusion significatif. Cependant, une plus petite proportion de patients sortis de l’étude étaient de langue maternelle française que dans le groupe de patients ayant complété l’étude. Dans le groupe ayant terminé l’étude, 47 % des patients avait une rachialgie seule, 8 % avait des douleurs cervicales et 45 % avait des douleurs à plusieurs endroits.
| Patients ayant complété l’étude | Patients perdus de vue | p | |
|---|---|---|---|
| Nombre de patients | 73 | 18 | |
| Âge (années) moyenne (écart-type) | 42,2 (10,7) | 38 (8,6) | 0,128 |
| Sexe (% de femmes) | 56 | 45 | 0,445 |
| Pourcentage de patients dont la langue maternelle est le français | 85 | 56 | * 0,037 |
| Pourcentage de patients avec un emploi rémunéré | 88 | 83 | 0,382 |
| Pourcentage de patients avec une blessure liée au travail | 16 | 28 | 0,314 |
| Pourcentage de patients dont la douleur date de moins de 3 semaines | 58 | 58 | – |
| Nombre d’années d’études après l’âge de 6 ans moyenne (SD) | 15,3 (3,5) | 14,5 (2,8) | 0,409 |
| Nombre de régions douloureuses (médiane [amplitude]) | 1 [1–5] | 1 [1–3] | 0,326 |
2.3.2
Évolution des patients
Les scores à t 0 et t 6 sont donnés dans le Tableau 3 . Les scores totaux de l’OMPSQ et de l’ODI étaient significativement plus bas à l’évaluation à six mois que lors de l’inclusion, montrant une amélioration globale. La seule variable de suivi de l’OMPSQ qui a montré une amélioration significative est l’indice fonctionnel. Les indices OMPSQ douleur et absentéisme étaient peu élevés à la visite initiale (nous avons uniquement inclu des patients avec des douleurs aiguës et subaigües) et ne montraient pas de différence significative entre t 0 et t 6 .
| Maximum théorique | t 0 | t 6 | Test de Wilcoxon p | |
|---|---|---|---|---|
| Score total OMPSQ | 210 | 85 [25–171] | 66 [12–151] | < 0,0001 |
| OMPSQ douleur | 100 | 12 [0–80] | 12 [0–80] | 0,201 |
| OMPSQ fonction | 100 | 31 [0–50] | 43 [11–50] | < 0,0001 |
| OMPSQ absentéisme | 10 | 2 [1–7] | 2 [1–9] | 0,223 |
| Score relatif ODI | 100 | 30 [0–72] | 12 [0––4] | < 0,0001 |
La proportion de patients pouvant être considérée comme guéris (dernière colonne du Tableau 1 ) était de 32 % quand toutes les variables étaient prises en compte. Vingt-cinq pour cent des patients ont montré une évolution défavorable sur une variable, 21 % sur deux variables, 13 % sur trois variables et enfin, 8 % des patients ont rapporté des problèmes persistants dans chacune des quatre variables de suivi.
2.3.3
Corrélation de l’Örebro Musculoskeletal Pain Screening Questionnaire avec l’Oswestry Disability Index
Le test de rang de Kendall a montré une corrélation faible entre le score relatif de l’ODI et le score total de l’OMPSQ (τ = 0,46 et 0,49 à t 0 et t 6 respectivement, p < 0,0001). Comme on pouvait s’y attendre, la corrélation était légèrement plus élevée en comparant l’ODI avec le sous-score fonctionnel de l’OMPSQ (τ = 0,55 à t 0 et à t 6 , p < 0,0001).
2.3.4
Corrélation des items individuels de l’Örebro Musculoskeletal Pain Screening Questionnaire avec les variables de suivi
Dans l’analyse univariée, la plupart des items du questionnaire étaient corrélés de manière significative à au moins l’une des variables ( Annexe 1 , colonne de droite). Trois items montrent une corrélation avec toutes les variables : dépression (item 14), perception par le patient de sa capacité à retravailler dans les prochains six mois (item 16) et capacité actuelle à effectuer un travail léger (item 21). Plusieurs items étaient corrélés avec trois variables de suivi : fréquence des épisodes douloureux (item 11), stress (item 13), auto-perception du risque de chronicité (item 15) et sommeil (item 25). À l’inverse, les items suivants ne montrent aucune corrélation avec les variables de suivi : le statut professionnel (item 4), le coping (item 12), la satisfaction professionnelle (item 17) et la capacité à faire les courses (item 24).
2.3.5
Valeur prédictive du score total Örebro Musculoskeletal Pain Screening Questionnaire pour évaluer la chronicité
Les tests U de Mann-Whitney comparant les scores OMPSQ entre le groupe de patients pouvant être considérés comme guéris et le groupe de patients gardant des douleurs et/ou un déficit fonctionnel marqués à six mois se sont révélés hautement significatifs pour toutes les variables ( p ≤ 0,003). Cependant, la fréquence de distribution des scores OMPSQ a montré une large superposition entre ces deux groupes, indépendamment de la variable concernée (données non montrées). L’AUC ROC, la sensibilité et la spécificité des différents scores seuils sont décrits dans le Tableau 4 . Le meilleur résultat obtenu avec l’AUC ROC concernait la prédiction de l’absentéisme au travail. Comme on pouvait s’y attendre au vu de l’importante superposition observée dans l’analyse des fréquences de distribution, il était impossible de sélectionner un score seuil unique exhibant simultanément une bonne sensibilité et une bonne spécificité. C’est pourquoi il a semblé raisonnable de proposer de définir trois catégories de patients : risque faible, risque intermédiaire et risque important. Comme proposé par Jellema et al. , le score seuil pour « risque faible » a été fixé au point correspondant à 80 % de sensibilité alors que le score seuil pour « risque important » a été fixé au point correspondant à 80 % de spécificité ( Tableau 5 ).
Related posts:
Effects of a customized biomechanical therapy on patients with medial compartment knee osteoarthritis
Comparison of gaseous cryotherapy with more traditional forms of cryotherapy following total knee arthroplasty
Reproducibility of the time to peak torque and the joint angle at peak torque on knee of young sportsmen on the isokinetic dynamometer
Comparison of energy cost in transtibial amputees using “prosthesis” and “crutches without prosthesis” for walking activities
Isokinetic muscle strengthening after acquired cerebral damage: A literature review
Ellenbogenkomplex
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
