Article Text

Download PDFPDF

Original article
Sick leave in workers with arm, neck and/or shoulder complaints; defining occurrence and discriminative trajectories over a 2-year time period
  1. A Feleus1,
  2. H S Miedema1,
  3. S M A Bierma-Zeinstra2,3,
  4. T Hoekstra4,5,
  5. B W Koes2,
  6. A Burdorf6
  1. 1Research Center Innovations in Care, Rotterdam University, Rotterdam, The Netherlands
  2. 2Department of General Practice, Erasmus MC, Rotterdam, The Netherlands
  3. 3Department of Orthopaedic Surgery, Erasmus MC, Rotterdam, The Netherlands
  4. 4Department of Health Sciences, VU University, Amsterdam, The Netherlands
  5. 5Department of Epidemiology and Biostatistics, VU University Medical Center, the EMGO Institute for Health and Care Research, Amsterdam, The Netherlands
  6. 6Department of Public Health, Erasmus MC, Rotterdam, The Netherlands
  1. Correspondence to Dr A Feleus, Research Center Innovations in Care, room RS 02.123, Rotterdam University, Rochussenstraat 198, 3015 EK Rotterdam; P. O. Box 25035, Rotterdam 3001 HA, The Netherlands; a.feleus{at}


Objectives Impediments due to complaints of non-traumatic arm, neck and/or shoulder (CANS) during work often leads to consultation in primary care. This study examines the occurrence of sick leave among workers with new CANS, and evaluates sick leave trajectories and their characteristics.

Methods This prospective 2-year cohort study included workers with a new CANS presenting in general practice. Participants filled out postal questionnaires on sick leave at 6-monthly intervals. Latent class growth mixture modelling was used to identify distinct trajectories of sick leave. Multinomial regression analyses identified characteristics of the subgroups.

Results During follow-up, of the 533 participants 190 reported at least one episode of sick leave due to CANS. Three sick leave trajectories were distinguished: (1) ‘low-risk’ trajectory (n=366), with a constant low probability over time; (2) ‘intermediate risk’ trajectory, with a high probability at first consultation followed by a steep decrease in probability of sick leave (n=122); (3) ‘high-risk’ trajectory (n=45), with a constant high probability of sick leave. Compared to the ‘low-risk’ trajectory, the other trajectories were characterised by more functional limitations, less specific diagnoses, more work-related symptoms and low coworker support. Specific for the ‘high-risk’ subgroup were more recurrent symptoms, more musculoskeletal comorbidity, high score on somatisation and low score on job demands.

Conclusions Three trajectories of sick leave were distinguished, graded from favourable to unfavourable. Several complaint-related and work-related factors and somatisation contributed modestly to identify an unfavourable trajectory of sick leave when presenting in primary care with CANS.

Statistics from


Complaints of arm, neck and/or shoulder, not caused by trauma or systemic disorder (also referred to as CANS), constitute an important health issue.1 Reported 12-month prevalence rates in general working age populations range from 12% in the USA to 36% in the Netherlands and 44–52% in the UK.2–6 In the Netherlands, 19% of the respondents reported chronic symptoms.2 In about 30–45% of all cases, people feel hindered and consult a general practitioner (GP).7–9 In a population in primary care, about 77% of the patients with CANS experienced symptoms in the upper back, neck and/or shoulder, 25% in the elbow and/or arm and 19% in the wrist and/or hand.10 In 42% of the cases, symptoms were reported in more than one region.10

Among workers, these symptoms frequently result in productivity loss, including sick leave.11–14 In a diverse population of Finnish workers seeking medical advice due to incipient CANS, 56% of the workers reported productivity loss.11 In a Dutch population of computer workers with CANS, productivity loss was involved in 26% of all cases.12

Sick leave was reported in 19% of the workers with chronic CANS in the Netherlands, in which 39% with a duration ≥4 weeks.13 In addition, a cohort study in the USA reported recurrence of sick leave in 26% of the cases, within 5 years, after the first episode.14

Of all registered sick leave days in the Netherlands in 2012, 11.2% was due to CANS.15 In the UK, the average number of work days lost per affected person is estimated at 14.7 days in 2015.16 Yearly costs of CANS in the Netherlands are estimated at 2.1 billion euros of which most costs are due to sick leave (920 million euros).17

Workers with limitations due to CANS generally consult their GP at a much earlier stage than the occupational physician, even though 71% report their symptoms to be influenced by their work.10 Therefore, the potential consequences of these symptoms for their workability is an important item when presenting in primary care. Furthermore, workers may return for consultations due to prolonged symptoms2 or recurrence of symptoms over time.18 ,19

In working individuals with shoulder pain, diversity in the course of symptoms was pointed out.19 To gain insight into functional limitations due to CANS over time, we studied the disability trajectories in a cohort of patients presenting with CANS in primary care over a 2 year time period, including workers and non-workers.20 Although one expects disability and sick leave to be related, various other factors may be important in taking sick leave and may influence its course over time.

So far, some studies21–23 have reported on the predictors of sick leave in workers with CANS presented in primary care. In these studies, the symptom variables ‘higher severity of symptoms at baseline’, ‘previous musculoskeletal trauma’ and a ‘perceived cause of strain or overuse during regular activities’ were predictive of sick leave at 6-months follow-up. Of the psychosocial factors, ‘coexisting psychological symptoms’ and a ‘higher score on somatisation’ were reported as predictors of sick leave.22 ,23 Furthermore, the work characteristics ‘self-reported work-relatedness of symptoms’, ‘heavy physical work’ and ‘low decision authority’ were determinants of sickness absence at the 6-month follow-up.23

Studies at the workplace have reported high intensity of pain, heavy physical work, low level of adjustment latitude and work pressure, to be predictive of, or associated with, sickness absence in musculoskeletal symptoms.24–26 Thus, these studies show that a combination of symptoms, physical and psychosocial variables, can be predictive of sick leave.

However, no information is available to help identify subgroups of workers with CANS, at high risk of sick leave over time. This knowledge may be used to compose more personalised treatment in order to limit the loss of workability.

Therefore, we performed a secondary analysis of cohort data, including workers presenting with CANS in primary care over a 2-year time period.

The present study aimed to:

  1. describe the occurrence of sick leave due to non-traumatic arm, neck and/or shoulder symptoms in a heterogeneous population in primary care, up to 2 years after their first consultation with the GP;

  2. identify trajectories of sick leave patterns among individuals with non-traumatic symptoms of arm, neck and/or shoulder; and

  3. evaluate personal, complaint-related psychological, social and work-related factors associated with these trajectories.


Design and setting

The CANS study is an observational prospective cohort study, with a 2-year follow-up, conducted in 21 general practices in the southwest region of the Netherlands. The Medical Ethical Committee of Erasmus Medical Centre approved the study. From September 2001 to December 2002, 36 GPs recruited consulters with a new complaint or a new episode of complaints of the neck, upper back, shoulder, upper arm, elbow, forearm, wrist or hand.

Data were collected by means of five self-administered questionnaires at baseline and every 6 months thereafter during follow-up. The GPs provided care as usual without implementation or promotion of any diagnostic or therapeutic and each participant provided written informed consent.

Additional information on the procedure, follow-up regarding non-recovery of complaints and management within the first 6 months, and on disability trajectories for the full cohort (including non-workers) are published elsewhere.10 ,20 ,27 The present secondary analysis determines the trajectories for the outcome of sick leave over a follow-up period of 2 years. The Strobe checklist was utilised to prepare this paper.28


Patients had to be aged 18–64 years and able to complete Dutch language written questionnaires. An episode of CANS was considered ‘new’ if they had not visited their GP for the same complaint during the preceding 6 months. Patients were excluded when complaints could be explained by a trauma, fracture, malignancy, amputation, prosthesis, congenital defect or a previously diagnosed existing systemic disorder and/or generalised neurological disorder, or when they reported to have recovered at the time of filling in the baseline questionnaire.

Patients who reported to be currently (self-)employed were eligible for the present study.


During the first consultation, patients received study information, a consent form and the baseline questionnaire from their GP. A fax was sent by the GP to the investigators with a patient number, age, gender, diagnosis and expected prognosis of the symptom. After the research team received the completed informed consent and the first questionnaire within 8 weeks, inclusion criteria were verified in the computerised medical records. At every consecutive 6 months after the first consultation a self-administered questionnaire was sent from the research centre, with a total of five questionnaires. This way, data on possible indicators and sick leave were collected.


Sick leave due to CANS was measured at consecutive 6-month periods, from the first consultation with their GP up to 2 years after, by the questions: (1) ‘Were you absent from work in the past 6 months due to CANS?’ Additional information on duration was collected by the question: (2) ‘What was the total number of days absent due to those symptoms in the past 6 months?’ Response categories were: 0 working days; 1–5 working days; 6–10 working days; or >10 working days.29

Possible indicators of sick leave

Possible indicators of sick leave over the course of 2 years were selected from a theoretical biopsychosocial perspective. The same indicators were studied as were reported previously with regard to non-recovery at 6 months.10 These include the following individual, complaint, lifestyle, psychological, social and work characteristics:

  1. Personal characteristics: age (years), gender and level of education (low: no education, primary school or lower vocational school; medium: lower or higher general secondary school level or middle vocational school; high: higher vocational school or university) were included.

  2. Complaint characteristics: duration of the complaint (0–3 months, or more than 3 months), region with most complaints during the last week (neck–shoulder–upper arm, elbow–forearm–wrist–hand, both), trauma arm, neck or shoulder in the past (yes/no), musculoskeletal comorbidity, non-musculoskeletal comorbidity and recurrence were assessed.

    The diagnosis, as registered by the treating GP, was dichotomised by the researcher into specific or non-specific according to the CANS model.1 A diagnosis was categorised as specific when it could be attributed to a specific medically objectifiable disorder (see online supplementary appendix 3). When the GP indicated more than one diagnosis, specific diagnosis was given priority.

    Furthermore, the severity of symptoms during last week (0–10) was measured with the numeric rating scale, and disability was measured with the Disability of Arm Shoulder and Hand (DASH) questionnaire (0–100), including the question ‘feeling limited in performing ones job or daily activities because of CANS’ (item 23).30 For perceived health, the categories of the first question of the SF-12 were recorded as ‘good’ (‘excellent’, ‘very good’ and ‘good’) and ‘poor’ (‘fair’ or ‘poor’).31

  3. Lifestyle characteristics: the body mass index (kg/m2) was calculated from reported height and weight (<25, 25–30 (overweight), >30 (obese)), and workers were coded as active in sports when participating at least 1 h a week (yes/no).

  4. Psychological and social characteristics: somatisation and distress were both measured with the Four Dimensional Symptom Questionnaire (4DSQ).32 Social support was measured with the Social Support Scale (SOS), a Dutch version of the Social Support Questionnaire (SSQ).33 Catastrophising was measured with a subscale of the Dutch adaptation of the Coping Strategy Questionnaire.34 Kinesiophobia was measured using the shortened version of the Tampa scale without the four reversed items.35 Of all psychosocial variables, higher scores indicate more of the measured characteristic. Health locus of control was assessed by one simple question ‘Do you believe you can influence your health through your behaviour?’ scored on a 4-point Likert scale. The scores ‘considerable’ or ‘to a large extent’ were considered as ‘yes’.

  5. Physical work characteristics: full-time work (working 36 hours a week or more) or part-time work (working <36 hours a week), <5 years working in the current job, sick leave due to complaints of arm, neck or shoulder in the past 6 months and work-relatedness of complaints were included. Complaints were defined as work-related if participants with a paid job confirmed one of the three following questions: (a) Do the complaints return or worsen during the activities at work? (b) Have you adapted or reduced your activities at work because of your complaints? (c) Do the complaints diminish after several days of being off work?

    Physical load at work was measured with the physical workload questionnaire (PWQ), a validated short version of the Dutch Musculoskeletal Questionnaire. The items were scored on a 4-point Likert scale ranging from 1 ‘seldom’ to 4 ‘always’. Two sum scores were calculated: ‘heavy physical workload’ and ‘long-lasting postures and repetitive movements’.36

  6. Psychosocial work characteristics: the psychosocial factors at work were measured with the Dutch translation of the core of Job Content Questionnaire (JCQ),37 including quantitative job demands, skill discretion, decision authority, supervisor support and coworker support. Job insecurity was measured with the item ‘My job security is good’.

Data analysis

Demographics will be presented in percentages, means and SD or median and ranges. Data on sick leave are presented for each time period. For all follow-up moments, the proportion of non-responders was calculated. Backward logistic analysis served to detect selective response in non-complete cases versus complete cases.

Analysis of the study data consisted of two steps:

Step 1: Latent class growth mixture modelling (LCGM) was used to identify possible subpopulations (ie, classes) of clinically meaningful groups of workers who are similar in their presence or absence due to CANS during the follow-up.38–40 Each trajectory class has its own growth parameters (intercept, slope), representing its sick leave trajectory. Within-class variation in either, or both, of the growth factors was accounted for.41 The model was built stepwise.

The most optimal model was determined by a combination of:

(1) indices of fit: the Bayesian information criteria (BIC), the Vuong-Lo-Mendell-Rubin likelihood ratio test (LRT) and the Bootstrap likelihood ratio test (BLRT); (2) posterior probabilities for the sick leave trajectory to which the individual was assigned to compared with the probability for any other trajectory, to check for distinctions between the trajectory classes;38 ,39 ,42 (3) interpretability of the model: by clinical meaningfulness of the groups, the trajectory shapes for similarity, number of individuals in each trajectory class and the number of estimated parameters41 (see online supplementary appendix 2 for details).

Step 2: Characteristics of the different trajectories were explored using multivariable multinomial regression analyses with the trajectory class as outcome measure. For possible relevant characteristics with clinically relevant classifications or validated classes, the existing cut-offs were used. For continuous variables, the linearity of the relation between this variable and trajectory class was examined. In case of non-linearity of continuous variables, or lack of clinically relevant classifications/validated classes for ordinal variables, their scores were split based on the median score of the total population at baseline.

OR expresses the association between the characteristic measured at baseline and the trajectory. Variables that were univariably associated with membership of one of the sick leave classes (p<0.05) were selected for multivariable multinomial analysis by a step-backward procedure (backward Wald). First, analyses per domain were conducted. Contributing variables (p<0.05) were selected for the final multivariable analyses. The percentage of explained variance (Nagelkerke's R2) and percentage correctly predicted were calculated to give an indication of the fit of the final models.

In case of multicollinearity, the variable with the highest association was retained for further analyses.

Mplus V.6.1 was used for LCGM (Mplus V.6.1 ed 1998–2010). Non-response analysis, description of the course of sick leave, analysis of the characteristics of class membership and multinomial regression analyses were performed using SPSS software V.20 (SPSS, Chicago, Illinois, USA).


Sample characteristics and occurrence of sick leave

Of the 798 patients who consulted their GP with a new CANS and aged 18–64 years, 682 (86%) returned the completed questionnaire and informed consent.10

In total, 533 patients were (self-)employed, and therefore eligible for the present study (for diagnosis see online supplementary appendix 3). Table 1 lists the response at the follow-up moments that ranges from 85.0% to 72.0% over time, with a total of 311 (58.3%) complete cases.

Table 1

Response at the four periods of follow-up

In general, differences between complete cases and non-complete cases were small. However, being a non-responder at one or more time-points was associated with: having symptoms located at neck, shoulder or upper arm, a trauma of arm, neck or shoulder in the past, and a higher score on somatisation and catastrophising.

The mean age of the respondents was 42 years and 47% were men. Of this population, 58% worked full time (≥36 hours/week) and 87% of the workers reported to feel limited in performing their jobs or daily activities because of CANS (item 23, DASH).30 In total, 71% reported their symptoms to be work related.

Around the time of the first consultation with their GP (up to 6 months prior to baseline), 127 workers (23.8%) availed sick leave, any length, due to CANS. During the 2-year follow-up period, the percentage of workers reporting sick leave decreased from 18.5% during the period baseline to 6 months, to 8.6% during the period 18–24 months follow-up.

At baseline, 4.9% of the workers reported sick leave due to CANS for ≥10 working days, which increased to 8.2% in the first 6 months after the first consultation and decreased to 2.3% at the 2-year follow-up. In total, 190 (35.6%) workers reported sick leave due to CANS during one or more of the 6-month follow-up periods (table 2).

Table 2

Availed sick leave due to CANS

Trajectories of sick leave

Based on fit, distinctions between the trajectory classes and clinical relevance of the model, we chose the three-class linear model with pooled variance of intercept and slope as the optimal solution (see online supplementary appendix 4).

All three estimated trajectories of sick leave show considerable similarity with the observed trajectories (figure 1). The trajectory of the largest group including 68.7% (n=366) of the sample represents workers who have a low probability for sick leave from baseline to the 2-year follow-up (‘low risk’). In figure 1, the line in the middle represents the trajectory of 22.9% (n=122) of the workers who start with a sick leave probability of around 0.65 at baseline, which shows a steep decrease over time (‘intermediate risk’). The third trajectory represents a small but clinically relevant group comprising 8.4% (n=45) of the workers with CANS with a repeatedly high probability of sick leave, about 0.65, during follow-up (‘high risk’).

Figure 1

Mean probabilities of sick leave in workers with CANS in primary care per trajectory in the three-class model. The observed and estimated trajectories are presented for the three classes.

Characteristics of the three classes

Table 3 presents descriptive data of the participants and results of the univariable multinomial regression analysis. The individual and complaint characteristics (ie, gender, duration of complaints at the time of first consultation, body region with the most symptoms and having non-musculoskeletal comorbidity) are equally distributed between the three trajectory classes. Furthermore, there were no differences in the distribution of low health locus of control or body mass index.

Table 3

Characteristics of the sick leave trajectories and results of univariate multinomial regression analysis

Regarding the work variables, ‘working full time’ and ‘years working in current job’ were also equally distributed across the trajectory classes. The majority of workers reported that their complaints are work related, with higher scores among workers with ‘intermediate risk’ and the ‘high-risk’ trajectory.

Multinomial regression analysis showed that in the ‘intermediate risk’ and ‘high-risk’ trajectories workers had more functional limitations, less specific diagnoses, more work-related symptoms and low coworker support was more frequently reported compared with the ‘low-risk’ trajectory’ (table 4).

Table 4

Results of multivariable multinomial regression analysis for characteristics of sick leave; the final model

Specific for workers with ‘intermediate risk’ trajectory was that they more frequently reported poor perceived general health and high physical load. Workers in the ‘high-risk’ trajectory more frequently reported having a recurrent complaint, musculoskeletal comorbidity, high somatisation and low quantitative job demands.

Overall, information on the variables in the final regression model classified 71% of the workers correctly in the trajectory classes, as distinguished by the LCGM analyses. Per trajectory class, the percentages of workers that could be correctly classified in the ‘low-risk’ trajectory, ‘intermediate risk’ and ‘high-risk’ trajectory were 92%, 26% and 23%, respectively (Nagelkerke R2=0.36).

No correlations higher than 0.28 were found between the variables that remained in the multivariable model.


Summary of occurrence of sick leave and trajectories

In this population of workers who consulted their GP with CANS, the prevalence of sick leave at baseline was 23.8% which declined to 8.6% at 2-year follow-up. In total, 190 workers (35.6%) availed sick leave due to CANS during one or more of the measurement time periods.

The three distinct trajectories that emerged can be graded from favourable to unfavourable: (1) the large ‘low-risk’-trajectory (68.7%) including workers with a continuous low probability of taking sick leave during follow-up; (2) the workers in the ‘intermediate risk’ trajectory (22.9%) with, initially, a relatively high probability of taking sick leave which quickly decreased over time; and (3) the ‘high-risk’ trajectory (8.4%), a small but clinically relevant group with a repeatedly high probability of taking sick leave.

Prognostic indicators compared to other studies

When comparing the presented results with those of our study on disability trajectories (workers plus non-workers, n=682),20 also three trajectories were identified: fast recovery (67.6%), modest recovery (23.6%) and continuous high disability (8.8%). Despite the similarities in number and distribution of the trajectories, only the variables: musculoskeletal comorbidity, poor general health and high score on somatisation, predicted unfavourable trajectories in both sick leave and disability.20 In addition, we also found several work factors to be predictive of an unfavourable trajectory in sick leave that could not be included in the full cohort. The predictor ‘low coworker support’, however, seems in line with ‘less social support’ reported in disability trajectories.20

Compared to others who studied predictors of sick leave at a specific follow-up time (3, 6 or 12 months),21–23 ,25 there are similarities with their findings. In the present study, the symptoms characteristic ‘more functional limitations’ was more frequently reported in the unfavourable trajectories. Although the OR of this continuous variable is close to one and the mean difference in DASH score with the low risk group (12 points, resulting in an OR=1.51) is smaller than the smallest detectable change (16 points),43 the finding seems in line with others.22 ,23 ,25 Studies on workers with shoulder pain,22 neck–shoulder pain25 and CANS23 reported that variables related to impediments such as ‘higher severity of symptoms’ and ‘high pain intensity’ were predictive of sick leave.

Our finding that having a ‘non-specific diagnosis’ was predictive of an unfavourable sick leave trajectory was in contrast to a study in physical therapy.23

Furthermore, the work variables ‘work-related symptoms’ and ‘low coworker support’ were also more frequently present, with highest associations for the high-risk trajectory. Only ‘work-related symptoms’ have been reported by others.23

Factors specific for the ‘intermediate risk’ trajectory were ‘poor perceived general health’ and ‘high physical load’. The latter was also reported by others.23–26

Distinct factors for the ‘high-risk group’ were having a recurrent symptom, musculoskeletal comorbidity, high somatisation and low quantitative job demands. Only high score on somatisation was also reported elsewhere.23

Despite some overlap in predictors with other prognostic studies on sick leave due to CANS, it is difficult to compare the results of the studies due to differences in the possible predictors that were investigated. Although they were included in our study as well, the variables ‘previous musculoskeletal trauma’22 and ‘low decision authority’,23 and low level of adjustment latitude,24–26 which were predictive in other studies, were not retained in our final model.

Strengths and limitations

The present study has some limitations. First, at the different follow-up periods loss to follow-up ranged from 15% to 28% of the total cohort of workers. Nevertheless, for a follow-up study with a large initial cohort, these data seem acceptable. Furthermore, the LCGM analysis provides estimates for missing data during follow-up and non-response analysis yielded only small differences. However, the complete cases are slightly less likely to have neck, shoulder or upper arm symptoms, having had a trauma of arm, neck or shoulder in the past, and are more likely to have higher scores on somatisation and catastrophising.

A second possible limitation of the present study is that we measured sickness absence based on self-reported questionnaires, whereas sickness data from company records might be considered more accurate.44 However, assessed reliability of data on sickness absence using company records as reference suggested that, for overall sickness absence and sickness absence due to back pain, the questionnaire data on prevalence, frequency and duration compared relatively well with the company's absence records. For the frequency and duration of sickness absence due to low-back pain in the past 6 months, the κ values for registered sickness absence were 0.61 and 0.65, respectively.29

In addition, results of a large cohort study of municipal employees in Sweden shows good agreement between self-reported and registered information on sickness absence (no sick leave days, 1–7 days and >28 days), with a recall period of 12 months.45 Because our recall period measuring the occurrence of sick leave (yes/no; >10 working days) was 6 months, we expect that the self-reported information in this study is in good agreement with the registered sickness absence.

To further explore the grading in the three trajectories, additional analyses were performed regarding the duration of sick leave. Sick leave periods lasting >10 working days during one or more of the time periods was reported by 2% in the ‘low risk’ trajectory, compared to 30% in the ‘intermediate risk’ and 62% in the ‘high-risk’ trajectory. This confirms the increase in severity in the different trajectories. Whether only the workers in the ‘intermediate risk’ trajectory have a better natural course because of less musculoskeletal comorbidity and less recurrent complaints, or are able to make the relevant adjustments in their work to regain/continue their working tasks, cannot be concluded from our data.

Information on the variables in the final model seems mostly helpful to support identification of workers with a low risk of (recurrent) taking sick leave (92%). However, the variables combined, assigned only a small portion of workers in the ‘intermediate risk’ (26%) and ‘high-risk’ trajectories (22%) correctly. To check whether this was due to imperfect classification by the LCGM analyses because of incomplete cases (all cases with at least baseline scores on sick leave were included), additional LCGM analyses were performed for those with at least two measurements (n=488), and for the complete cases (n=311). However, both analyses resulted in similar trajectories with similar distributions. This implies that other variables that were not included in our study may be associated with the sick leave trajectories. In addition, the variability in working tasks and contexts between the participants, and the consequences for the ability to continue ones work, is complex and may be difficult to capture in a questionnaire. Information from qualitative studies may have additional value to unravel the contributing factors of sick leave probability. Two studies may contribute to possible additional factors. One study explored experiences of workers with CANS,46 and reported on the factors: insufficient awareness of possibilities to influence and manage their symptoms, inadequate communication with supervisors and lack of adaptations at the workplace that made it difficult for them to continue working. The other study focused on success factors to continue working in workers with chronic musculoskeletal symptoms.47 They reported the factor: support, specified as support in a change of behaviour related to: raising adjustment latitude, changing pain-coping strategies, organising modifications and conditions at work, finding access to healthcare services and asking for support.47 Although we included support of supervisor and coworker, and autonomy in our study, the items regarding these domains of the Job Content Questionnaire,37 do not explicitly address the above mentioned problems and solutions.46 ,47

Implications for research and/or practice

Although all the found predictors only modestly contribute to identify an unfavourable trajectory, including those variables in the management of CANS, they might still contribute to limit loss of workability. Furthermore, in line with our findings, the majority of these variables are also recommended to address in history-taking in a multidisciplinary guideline on non-specific CANS, in which also contact with an occupational physician is advised when work-related symptoms continue for more than 2 weeks.48 However, how to combine all these specific aspects into an effective approach, to prevent absenteeism or to promote a sustainable return to work is still unclear.49–51 The finding that workers with a specific diagnosis more frequently have a favourable course may be because more guidelines are available, and at the workplace it is more clear what to aim for.

Looking at the various predictors requires a multifaceted approach. At the workplace, our findings and those of the qualitative studies46 ,47 combined, suggest addressing physical and cognitive strategies tailored to the individual needs, and using a participatory approach to support commitment from the stakeholders and set priorities.52–55


This study showed that most workers with CANS who present in primary care have symptoms related to their working activities and that, in one-third of the cases, these symptoms result in one or more episodes of sick leave during follow-up. Three trajectories of sick leave were distinguished, graded from favourable to unfavourable. Several symptom variables, work factors and scores on somatisation modestly contribute to correct classification when they present in primary care.

Therefore, more studies on this topic are needed, taking into account the complexity of sick leave as outcome measure. Including qualitative research may help to identify factors to develop a more comprehensive model.

What this paper adds

  • The percentage of workers who availed sick leave at each 6-month period from baseline up to the 2-year follow-up, started at 23.8% around the time of the first consultation with their GP and dropped to 8.6% during the period 18–24-month follow-up.

  • Long-lasting sick leave (more than 10 working days) was reported in 8.2% to 2.3% of the participants during the four 6-month follow-up periods.

  • During the 2-year follow-up period, 190 (35.6%) workers availed sick leave during one or more of the 6-month follow-up periods which, in one-third of the patients, lasted more than 10 working days.

  • Three distinct trajectories regarding sick leave were found, which can be graded from favourable to unfavourable.

  • Several symptom-related and work-related factors and somatisation contributed modestly to identify an unfavourable trajectory of sick leave when presenting in primary care with CANS.


The authors thank the general practitioners and the patients for their valuable contribution to this study.



  • Contributors SMAB-Z and BWK are responsible for the initial idea to conduct a prospective cohort study in primary care, to study the course and outcome of patients with CANS. SMAB-Z, BWK, AF, HSM and AB were responsible for the study design, choice of measures and content of the questionnaires. AF conducted data collection under the supervision of SMAB-Z, BWK and HSM. AF, SMAB-Z and BWK were involved in the organisation of the network of GPs that participated in this study. AF, HSM, AB and BWK were responsible for the planning of the analyses of the 2-year follow-up data. TH is a statistician and was consulted for specific advice regarding the use of LCGM in Mplus and interpretation of the results. AF performed all the analyses for this study and drafted the manuscript. All authors were involved in discussions about the study results, commented on drafts of this manuscript, and all approved the final version. All authors are accountable for all aspects of the work.

  • Funding The study was supported by internal funding from Erasmus MC (Revolving Fund) Rotterdam, and by the Rotterdam University of Applied Sciences.

  • Competing interests None declared.

  • Patient consent Obtained.

  • Ethics approval The Medical Ethical Committee of the Erasmus Medical Centre approved the study, and written informed consent was obtained from each participant.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.