Author's response to reviews Title:Medically Unexplained Symptoms and the risk of loss of labor market participation - A prospective study in the Danish population Authors: Katja Loengaard (nfa.klo@gmail.com) Jakob B Bjorner (jbj@nrcwe.dk) Per K Fink (perklfi@rm.dk) Hermann Burr (burr.hermann@baua.bund.de) Reiner Rugulies (rer@nrcwe.dk) Version:2Date:3 July 2015 Author's response to reviews: see over
Revision of manuscript # 2013322820134308 Medically Unexplained Symptoms and the risk of loss of labor market participation - A prospective study in the Danish population Dear Dr. Pafitis, National Research Centre for the Working Environment Lersø Parkallé 105 DK-2100 Copenhagen Denmark Thank you very much for inviting us to revise and re-submit our manuscript. We are delighted to do this. We found the reviewers comments very helpful and have revised the manuscript accordingly. Please find below our detailed point-by-point response to the reviewers comments. In the manuscript, we have highlighted the changes in yellow. 3 June 2015 File no.: Sagsnr Ref.: Navn Direct phone: (+45) 39 16 52 18 e-mail: rer@nrcwe.dk Yours sincerely, Reiner Rugulies, PhD, MPH, MSc Professor of Psychosocial Work Environment and Health National Research Centre for the Working Environment Copenhagen, Denmark Phone: (+45) 39 16 52 00 Fax: (+45) 39 16 52 01 e-mail: nrcwe@nrcwe.dk www.nrcwe.dk VAT no. DK15 41 37 00 EAN-no.: 5798000399518
Point-by-point response to Referee comments to manuscript MS 2013322820134308 - Medically Unexplained Symptoms and the risk of loss of labor market participation - A prospective study in the Danish population RESPONSE TO REFEREE 1 (Chris Jensen) 1) Little is known about how medically unexplained symptoms (MUS) affects labor market participation. The paper's main finding is that employees with MUS (and no chronic disease) have a higher risk of LTSA and unemployment than healthy employees. The paper provides an important contribution to the epidemiologic literature on this topic. AUTHORS RESPONSE: Thank you very much for this positive assessment of the relevance of our paper. We very much appreciate this. Major compulsory revisions 2) Page 5: The study cohort comprises individuals from 18 year through 59 years. The authors argue that the upper boundary was chosen because they wanted to focus on participants of working age at baseline, before the possibility of early retirement. This is a reasonable argument. However, could 58 or 59 years old participants wishing to withdraw from working life use sick leave and/or unemployment as an exit pathway, i.e. receive sickness or unemployment benefit until they become eligible to early retirement benefit? To avoid this potential problem, the author could either lover the age limit to e.g. 55 years, or include age as a categorical variable (including a category for 55-59 years old) thus avoiding to reduce the sample size. It would be informative to present additional results from analyses broken down by age-groups. AUTHORS RESPONSE: We followed the advice and conducted a sensitivity analysis that excluded the 1,294participants who were 55 years or older. Results did not change substantially for long-term sickness absence and unemployment. However, for disability pensioning the RR in the high-symptom-no-chronic (MUS) group increased from 2.06 (95% CI: 0.77-5.52) to 3.06 (95% CI: 1.08-8.62). We report the results from this sensitivity analysis in the result section under the new header Sensitivity analyses. 3) Page 6: The comparison of the MUS group with especially the Low-symptomchronic group is interesting, because this comparison tries to answer whether employees with MUS have worse labor market prospects than employees with chronic diseases. In this respect the lumping together of presumably non-disabling chronic diseases as hay fever and diabetes with potentially very disabling diseases as smokers lung and psychiatric disorders may mask important differences (and the reason why e.g. LTSA apparently does not differ significantly between the MUS group and the Lowsymptom-chronic group). Therefore, it would be relevant and interesting to see how the labor market outcomes of MUS group differ from the labor market outcomes of more detailed (groups of) chronic diseases. 2
AUTHORS RESPONSE: This is an important comment but also a complicated issue. In response to the reviewer s request, we now conduct a sensitivity analysis that no longer regards hay fever, skin diseases and allergy as chronic diseases, because they are presumably less disabling than the other included chronic diseases. We did not exclude diabetes, although the reviewer listed diabetes as an example for a nondisabling disease, because a recent Danish population-based study showed that diabetes is actually highly disabling (Cleal et al. European Journal of Public Health 2015, Published Online First 9 February 2015). The results from this sensitivity analysis are reported in the result section under the new header Sensitivity analyses There was no substantial difference to the main analysis with regard to long-term sickness absence and unemployment. For disability pensioning, however, re-defining chronic disease resulted for the high-symptom-nochronic group in a RR of 3.29 (95% CI: 1.67-6.47) that was statistically significant, whereas it was not statistically significant in the main analyses. The RR for disability pensioning of the two chronic disease groups that were already statistically significant in the main analyses, further increased in the sensitivity analysis. The sensitivity analysis is now presented in the result section and the results are addressed in the discussion. 4) Page 6: As discussed by the authors, the definition of MUS is the most important challenge in the present paper. It is defined from an instrument that was developed in an un-published paper. It should be described in some detail in the methods section or as an appendix to the present paper. Which symptoms were assessed? AUTHORS RESPONSE: We agree with the Reviewer (and with Reviewer #2 who had a similar request, see response #12 below) that more details on the MUS-POP instrument are needed. In response, we have expanded the method section that now includes a more detailed description of the development of the MUS-POP sale and that now lists all 15 symptoms that were assessed with the MUS-POP scale. 5) Table 2: The study of the association between MUS and subsequent labor market outcomes is based on people who were working at baseline. As shown in Table 2, people in the High-symptom-chronic group are more frequently permanently out of work at baseline than the Low-symptom-chronic group, and this group is more frequently permanently out of work at baseline than both the MUS group and especially the Healthy group. This analysis is clearly a strength of the paper as a background for the prospective analyses, but the authors do not argue why the analysis is performed nor do they discuss the consequences this selection might have for the estimated coefficients in the prospective analyses. AUTHORS RESPONSE: Thank you for this suggestion. We have now added to the beginning of the discussion section (in the sub-section Summary of results ) a paragraph that relates the findings from table 2 and table 3 to each other. We address that at baseline the three groups who had either a high symptom score, a chronic disease or both had a higher prevalence of being out of work than the healthy group (although in the case of the high-symptom-no-chronic group only in 3
the crude analysis). Thus, our cohort for the follow-up analysis consisted of employees who had found a way to keep labor market attachment despite their health conditions at baseline. One could have assumed that this group would not be at increased risk for loss of labor market during follow-up. Clearly, this was not the case. We conclude therefore that although our sample consisted of employees who had managed to keep labor market attachment at baseline despite their health condition, they were still at increased risk of losing labor market attachment during follow-up. 6) Abstract, Conclusion: Can the findings be given causal interpretation? It is stated that MUS seem to have a negative effect. If so, causality should be discussed more clearly in the Discussion. AUTHORS RESPONSE: Causal inference in observational epidemiological studies has to be drawn with caution, because of the possibility of confounding by unmeasured factors. We now address this in the beginning of the sub-section Methodological considerations in the discussion section. 7) Page 14, line 326. The response rate of 62.5% should be discussed. It may be argued that it was considerably lower because people responding in telephone interviews were not included. I would consider them as non-responders in the present study. AUTHORS RESPONSE: We would have preferred to include all participants, however we decided against this, because of convincing evidence from DWECS that response patterns to key variables deviated substantial depending on type of data collection. This is now discussed in the end of the section Methodological considerations. 8) Conclusion: Related to this discussion, another issue concerns how practitioners can use the findings. The authors conclude (last sentence of the conclusion) We recommend that these possible consequences are taken into consideration by health and social security professionals who are working with people suffering from MUS. What could they do? AUTHORS RESPONSE: Possible action by health professionals and in particular social security professionals include informing employees with MUS about their increased risk of loss of labor market participation and to consider employees with MUS for prevention and re-integration programs aimed at individuals at high risk of loss of labor market participation. We now elaborate on this in the section Implication. Minor essential revisions 9) Page 10, line 225: low symptom no chronic group should probably be low symptom chronic group? AUTHORS RESPONSE: Thank you very much for spotting this mistake. This has been corrected. 4
REFEREE 2 (ERIK L. Werner) 10) This is a well written very interesting paper with a highly relevant research question. My major concern regards the validity of the methods but overall I find these questions well accounted for in the discussion section. AUTHORS RESPONSE: Thank you very much. We very much appreciate this positive assessment. 11) 1. Is the question posed by the authors well defined? The research question and aim of the study is well defined both in the title and in the background section 12) 2. Are the methods appropriate and well described? The methods are well described. A combination of survey data and register data is a challenging task. How well are the specific diagnoses of a chronic disease defined both by the people themselves, and by the doctors making the premises for the register data? What is the definition of a chronic disease? Is the categorization in fact on specific / unspecific diagnoses? It would be helpful to have examples on chronic and no chronic disease. The paper refers to an instrument for classification of symptoms (MUS-POP); it would be helpful to have this instrument presented, for example as an appendix. AUTHORS RESPONSE: We now provide examples of chronic and non-chronic diseases. With regard to the MUS-POP instrument we agree that more information on the instrument is needed (see also response #4 to Reviewer 1 above). We have expanded the method section that now includes a more detailed description of the development of the MUS-POP sale and that now lists all 15 symptoms that were assessed with the MUS-POP scale. 13) 3. Are the data sound? The results are well presented in tables 14) 4. Do the figures appear to be genuine, i.e. without evidence of manipulation? Yes 5
15) 5. Does the manuscript adhere to the relevant standards for reporting and data deposition? Yes 16) 6. Are the discussion and conclusions well balanced and adequately supported by the data? The discussion seems well balanced, particularly with regard to the high uncertainty of assessment of the patient groups are the MUS category really MUS, and are those categorized as chronic disease purely well-defined conditions? Although as stated in the introduction, this study being the first prospective study on MUS and long term sickness absence, the results seem to fit in with previous research on the topic. 17) 7. Are limitations of the work clearly stated? Yes 18) 8. Do the authors clearly acknowledge any work upon which they are building, both published and unpublished? Yes but it would be helpful to have access to previous work and instruments from the same research group. AUTHORS RESPONSE: We agree and we have added a more detailed description of the MUS-POP instrument to the method section (see also response #4 to Reviewer 1 and response #12 to Reviewer 2). Further, we now refer to the Ph.D. dissertation by Katja Loengaard that includes the most comprehensive description of earlier work and we state that this dissertation can be can be requested from the corresponding author. 19) 9. Do the title and abstract accurately convey what has been found? Yes 20) 10. Is the writing acceptable? Yes 21) In conclusion, I find this paper highly relevant and the study very well conducted. There are some major concerns regarding the validity of the classification of patients but these seem well accounted for in the discussion section. AUTHORS RESPONSE: We very much appreciate this positive assessment by the reviewer 6