Review Reports

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

 I recommend its publication following minor revisions addressing the following points.

1. Clarification of the conceptual distinction between incidence and sickness absence duration

One of the most interesting contributions of the study is the distinction between factors associated with the occurrence of commuting accidents and factors associated with their consequences after the accident has occurred. This distinction is already present in the manuscript and emerges clearly in relation to age.

However, I would recommend making this conceptual point slightly more explicit at the end of the Introduction. In particular, the authors could briefly state that accident incidence and sickness absence duration represent two complementary but different dimensions of occupational health burden.

This would help the reader understand from the outset why both types of analyses are included in the study.

2. Clarification regarding the incidence-rate population

The incidence-rate analyses shown in Figures 1 and 2 correspond to all economic sectors in Spain, whereas the main analytical sample is restricted to health and social care workers. The authors correctly acknowledge this limitation later in the manuscript, but I suggest making this distinction more explicit when the incidence results are first presented.

A short sentence in Section 3.2 indicating that these rates provide national contextual information rather than sector-specific estimates would avoid possible misinterpretation.

3. Further justification of sickness absence duration as an outcome

The manuscript appropriately considers sickness absence duration as an indicator of accident consequences. Nevertheless, a brief additional explanation in the Methods or Introduction would strengthen the conceptual basis for this outcome.

The authors could clarify that sickness absence duration should not be interpreted as a direct measure of injury severity alone, since it may also reflect recovery, functional impairment, occupational demands, clinical management, and return-to-work processes.

This point is partly discussed later in the limitations section, but introducing it earlier would improve the interpretation of the outcome variable.

4. Interpretation of the traffic-related accident finding

An interesting result is that the crude mean sickness absence duration is very similar for traffic and non-traffic accidents, whereas traffic involvement is associated with a substantially longer expected duration after multivariable adjustment (EMR = 1.287).

The Discussion already addresses this finding appropriately. Nevertheless, I suggest slightly emphasising that this difference between crude and adjusted estimates likely reflects the different age and anatomical injury profiles of traffic and non-traffic accidents.

This is an instructive result and highlights the added value of the multivariable model.

5. Clinical and occupational interpretation of injured body regions

The strong associations observed for shoulder/arm and wrist injuries are among the most practically relevant findings of the study. These injuries are not necessarily the most frequent, but they are associated with substantially longer sickness absence.

The authors could add one or two sentences discussing why upper-limb injuries may have a particularly important impact on return to work in healthcare and social care occupations, where manual handling, patient care, repetitive tasks, and upper-limb functional requirements are common.

This would strengthen the link between the statistical findings and occupational health practice.

6. Minor clarification concerning the 545-day maximum

The manuscript clearly explains that sickness absence duration was bounded at 545 days because of the Spanish temporary incapacity system. This is useful information.

I suggest slightly rephrasing the explanation to make completely clear that this was an administrative ceiling in the original data rather than an analytical truncation introduced by the researchers.

This distinction may be relevant for international readers unfamiliar with the Spanish social protection system.

7. Preventive implications

The Discussion contains several useful preventive recommendations. I suggest distinguishing slightly more clearly between recommendations directly supported by the present findings and those derived from previous literature, particularly those concerning fatigue, shift work, sleep, and working schedules.

Since these variables were not available in the database, wording such as “may”, “could”, or “should be evaluated in future studies” would preserve the appropriate observational interpretation of the results.

8. Strengths of the study

I encourage the authors to retain and perhaps slightly emphasise the strengths already identified in Section 4.1. In particular, the combination of:

• a nationwide administrative dataset;
• a very large sample;
• a long observation period;
• analysis of both accident patterns and sickness absence consequences; and
• multivariable modelling adapted to an overdispersed count outcome

constitutes an important contribution of the study.

The sample size is particularly noteworthy and allows detailed stratification by age, sex, commuting direction, traffic involvement, and injured body region.

9. Minor editorial comments

The presentation of Tables 2–5 could also be checked during final formatting to ensure that headings, group labels, and statistical significance columns are immediately understandable to readers.

 

This is a relevant and well-conducted study addressing an occupational health issue that has received relatively limited attention in healthcare workers. The large national sample is a major strength, and the analytical strategy is appropriate for the study objectives and distribution of the outcome.

The manuscript provides useful evidence that the factors associated with the frequency of commuting accidents are not necessarily those associated with prolonged sickness absence. In particular, the findings concerning age, commuting direction, traffic involvement, and anatomical location of injury have potential relevance for occupational health surveillance and preventive strategies.

The points raised above are primarily intended to improve clarity and interpretation and do not require substantial changes to the analytical approach.

 

 

Comments on the Quality of English Language

A careful final language review would be advisable to ensure consistent terminology throughout the manuscript, particularly regarding the terms “sickness absence duration”, “days lost”, and “lost working days”. I recommend using “sickness absence duration” consistently when referring to the study outcome and retaining the administrative terminology only when describing the original variable in the database.

Author Response

We thank the reviewer for the positive assessment of our manuscript and the constructive recommendations. We have addressed each point below. Following the Academic Editor’s additional methodological recommendations, we also expanded the adjustment set and adopted a zero-truncated negative binomial model with a non-linear age term. Accordingly, some estimates have been updated since the original submission.

1. Clarification of the conceptual distinction between incidence and sickness absence duration

Response: Thank you for highlighting this distinction. We have revised the final paragraph of the Introduction to state explicitly that accident incidence and sickness absence duration represent complementary but distinct dimensions of occupational health burden. Incidence describes accident occurrence relative to the population at risk, whereas recorded sickness absence duration describes absence following an accident among affected workers. We also clarify that the populations underlying the incidence figures and the individual-level duration analyses are identified separately.

2. Clarification regarding the incidence-rate population

Response: Thank you for identifying this potential source of misinterpretation. The incidence figures have been revised since the original submission. Figure 1 now compares incidence in health and social care activities with incidence across all economic sectors, while Figure 2 presents health and social care sector rates by age group and sex.

We have clarified these populations at the beginning of Section 3.2 and in the figure captions. The Methods specify that the denominators correspond to the population represented by each series, with additional age and sex stratification for Figure 2. The all-sector series provides national context; the sector-specific series and the individual-level duration analyses are explicitly distinguished.

3. Further justification of sickness absence duration as an outcome

Response: We have strengthened the explanation in Section 2.3. Recorded sickness absence duration is defined as an indicator of absence-related burden rather than a direct measure of injury severity or verified time to recovery or return to work. We explicitly acknowledge that duration may also reflect functional limitations, occupational demands, treatment and rehabilitation, administrative procedures, and return-to-work arrangements. This interpretation is also reflected in the Introduction and limitations.

4. Interpretation of the traffic-related accident finding

Response: We have expanded the Discussion of the difference between crude and adjusted findings. Although the overall crude means were similar for traffic-related and non-traffic-related accidents, traffic-related accidents had longer mean durations within every reported age group. The groups also differed in their age and anatomical injury-site distributions. These compositional differences are consistent with the contrast between the crude and adjusted comparisons.

In the revised expanded model, the marginally standardised adjusted mean ratio is 1.271 (95% CI: 1.248–1.294), replacing the original estimate of 1.287. Additional adjustment for injury type and recorded severity attenuated this ratio to 1.201 (95% CI: 1.180–1.222). We clarify that these results do not quantify the separate contribution of age or injury profiles and do not establish a causal effect of traffic involvement.

5. Clinical and occupational interpretation of injured body regions

Response: We have expanded the Discussion of shoulder, arm and other upper-limb injuries and wrist injuries in relation to the manual and functional demands of healthcare and social care occupations. We discuss patient assistance and transfers, personal care, equipment handling, and repetitive or precise manual tasks as activities potentially affected by limitations in shoulder mobility, grip strength, or wrist function.

We explicitly identify this explanation as a possible interpretation, since individual functional limitations, task demands, rehabilitation, and workplace accommodations were not measured. The updated adjusted mean ratios are 1.888 (95% CI: 1.751–2.037) for shoulder, arm and other upper-limb sites and 1.773 (95% CI: 1.638–1.919) for wrist injuries, relative to head injuries.

6. Minor clarification concerning the 545-day maximum

Response: Section 2.3 now states that 545 days was the maximum recorded value in the supplied dataset and that no additional analytical upper limit was imposed by the researchers. We explain its correspondence with the Spanish temporary-incapacity exhaustion threshold while distinguishing this threshold from verified recovery or return to work.

The extract does not establish whether records at this value represent completed episodes, administrative closure, or ongoing absence. We therefore acknowledge the uncertainty concerning their endpoint rather than assuming a verified censoring mechanism. We report that 34 accidents had a recorded duration of 545 days and examine their exclusion in a sensitivity analysis.

7. Preventive implications

Response: We have revised the preventive implications to distinguish the observed associations from possible interventions informed by previous literature and occupational practice. Fatigue management, scheduling changes, safer travel arrangements, rehabilitation, and temporary task modifications are presented as potential approaches requiring assessment and prospective evaluation.

We explicitly acknowledge that fatigue, shift work, sleep, journey duration, transport mode, and domestic responsibilities were not measured. Their contribution to the observed associations remains hypothetical, and the effectiveness of the suggested interventions was not evaluated in this study.

8. Strengths of the study

Response: Thank you for recognising these strengths. We have retained the description of the nationwide administrative source, the large sample of 139,412 registered accidents, and the 13-year reporting period in Section 4.1. The sample supports detailed stratification by worker and accident characteristics and permits examination of recorded absence-related burden alongside accident patterns.

The revised Methods and Results also document the expanded zero-truncated NB2 model, which accommodates positive recorded durations and overdispersion, together with diagnostic checks and sensitivity analyses. We balance these strengths with an explicit acknowledgement of the remaining limitations of distributional fit.

9. Minor editorial comments

Response: We have reviewed the headings and labels in Tables 2–5. These now identify age groups in years, accident counts and percentages, and mean recorded sickness absence duration in days. Commuting-direction and traffic-involvement labels have been clarified, and the statistical significance columns are labelled “p-value”. The accompanying notes explain the percentage denominators and identify the duration comparisons as unadjusted.

We appreciate the reviewer’s recommendations, which have helped improve the conceptual clarity, occupational interpretation, and presentation of the manuscript.

Reviewer 2 Report

Comments and Suggestions for Authors

Dear Authors,

in your abstract and methodology it is important to specify that the data used in the longitudinal analysis are secondary data. 

The explanations in the subsection ”3.5 Factors Associated with Sickness absence Duration: Negative Binomial Regression” are incorrect. 

the analyses of data need to follow step by step the statistical rules of  Negative Binomial Regression https://www.scalestatistics.com/negative-binomial-regression   The subchapter 3.5 Factors Associated with Sickness Absence Duration: Negative Binomial Regression must be rewritten concerning the statistical rules. 

Negative Binomial Regression is a prediction method. This method is used for prediction, not just association. 

Please consult a resource that explains how to interpret the data.

Synchronize all parts of the manuscript with the correct explanation (abstract, methodology, analyses, discussion, conclusions). 

Please cite the source of the data below each table. 

 

 

Author Response

We thank the reviewer for the comments concerning data provenance, the statistical analysis, interpretation, and consistency of reporting. We have substantially revised the regression analysis and its presentation, as detailed below.

1. Identification of secondary data

Response: The Abstract and Section 2.1 now explicitly state that the study uses secondary administrative occupational accident data provided by the Spanish Ministry of Labour and Social Economy and originating from the Delt@ reporting system. We clarify that the records were collected for administrative purposes and that no participants were recruited or contacted. Although the records span several years, the individual-level analysis concerns registered accidents rather than longitudinal follow-up of identified workers.

2. Revision of the negative binomial analysis and Section 3.5

Response: We have rewritten Section 2.4 and Section 3.5, now entitled “Adjusted Associations with Recorded Sickness Absence Duration”, to describe the model specification, assessment, and interpretation explicitly.

Because the analytical sample contains only positive recorded durations, the revised primary analysis uses a zero-truncated negative binomial model with the NB2 parameterisation and a log link. The model includes an expanded adjustment set, and age is represented by a restricted cubic spline rather than a single linear term.

We now report the estimated dispersion parameter, log likelihood, AIC, BIC, and comparisons with ordinary negative binomial and zero-truncated Poisson models. The dispersion estimate was 1.063 (95% CI: 1.054–1.073). We also describe numerical convergence, the information matrix, Pearson residual statistics, calibration, randomized quantile residuals, and influence diagnostics. Sensitivity analyses and additional diagnostic results are provided in Supplementary Tables S1–S3 and Figure S1.

We acknowledge that the model does not fully reproduce the duration distribution, particularly very short absences and parts of the upper tail. These discrepancies and their implications are explicitly discussed rather than assuming complete distributional fit.

3. Purpose of the model and interpretation of estimates

Response: We agree that negative binomial regression can generate predicted expected counts. Its use is also compatible with estimating adjusted associations through contrasts of model-based expected outcomes. We have clarified that the objective of this study is to estimate adjusted associations with recorded sickness absence duration, rather than to develop and validate an individual prediction tool.

For the zero-truncated model, we calculate expected positive durations as μ/[1 − P(Y = 0)]. Adjusted mean ratios are obtained by marginal standardisation over the observed distribution of the remaining covariates. They are therefore not interpreted automatically as exponentiated coefficients of the underlying negative binomial distribution.

Tables 6 and 7 report adjusted mean ratios and adjusted mean durations in days, respectively, with 95% confidence intervals. We explicitly distinguish these estimates from causal effects, verified recovery times, and externally validated individual predictions.

4. Consistency across manuscript sections

Response: We have updated the Abstract, Methods, Results, Discussion, and Conclusions to reflect the revised model and interpretation. The age association is now described as non-linear, and the sex association is qualified according to the adjustment set: no detectable difference was observed in the primary model, whereas a modest difference emerged after additional adjustment for injury type and recorded severity. The revised text consistently refers to recorded sickness absence duration and adjusted mean comparisons.

5. Identification of data sources beneath tables

Response: We have added explicit source statements beneath the analytical tables, identifying the occupational accident microdata provided by the Spanish Ministry of Labour and Social Economy. The coding tables in Appendix A identify the official Delt@ reporting guide as their source. We distinguish tables containing authors’ calculations from those documenting administrative coding.

We appreciate the reviewer’s recommendations, which have helped improve the transparency and consistency of the statistical reporting.

Reviewer 3 Report

Comments and Suggestions for Authors

Dear Juan and team,

Thank you for the chance to review your work. While reading this paper, I kept returning to one issue, it moves between two different populations. As I understood, the accident records come from health and social care activities, whereas the incidence figures refer to workers across all economic sectors in Spain. I bring this up because of the paper’s main interpretations, especially those concerning age and sex, draw on the national figures. It is true that the study has a useful dataset and an important question, but the current version needs a sharper analytic boundary and a closer look at the model used for sickness-absence duration.   

I have raised some points that need urgent attention before this work can be ready for publication. Please pay attention to these points, to avoid unnecessary rounds of reviews.  

1. The incidence figures should not be taken as findings from the health and social care sector. Figures 1 and 2 are based on all-sector national rates, while the analytic sample is restricted to CNAE divisions 86, 87, and 88. As a result, the figures cannot show whether incidence falls with age among health and social care workers. The conclusion currently comes very close to making that claim. If sector-specific denominators can be obtained, the figures should be recalculated for the study population. If they cannot, I would keep the figures only as national context and state this plainly in the text, captions, discussion, and conclusion. The reason why the rate series starts in 2013, despite the study period beginning in 2011, also needs to be explained. 
2. Please pay special attention to these suggestions here, as I could not fully reconstruct how the final sample was obtained. The manuscript reports 139,412 included accidents, yet it does not show how many records were initially available or how many were removed at each stage. A short flow diagram would solve this. It should include relapses, inconsistent commuting-direction codes, missing values, and the final number used in the regression. The classification of direction also deserves more detail. Since “to work” and “from work” are derived from values 0 and 99 in the hours-worked variable, readers need an official source confirming that this coding is valid for commuting accidents. A supplementary codebook for the ESAW categories would be very helpful, especially for traffic involvement and the grouping of injured body parts. 
3. On the other hand, your outcome requires a model that acknowledges how it was recorded. The sample excludes accidents without lost workdays. Therefore, sickness-absence duration begins at one day by design. The outcome also ends at 545 days because of the Spanish temporary-incapacity framework. Note that a standard negative binomial regression does not directly address this lower truncation or the possibility that cases at 545 days represent unresolved absence. Here, you can consider a zero-truncated negative binomial model and explore a time-to-return-to-work analysis as a sensitivity check. If 545 days marks administrative censoring, these observations should be handled accordingly. Please report the number of records at this limit, the estimated overdispersion, model-fit statistics, and diagnostic checks. 
4. If the study spans thirteen years, including the period of COVID-19 mobility restrictions, why was calendar year not included in the model. The risk environment, commuting patterns, reporting practices, and return-to-work procedures may have changed considerably over this period. The model would also benefit from adjustment for activity division, region, occupational category, employment characteristics, and injury diagnosis or severity, wherever these variables are available. Body part does not fully capture injury severity. Age is central to your work, though its effect is assumed to rise at the same pace every year. A flexible age term or age categories would allow readers to see whether the association is actually linear. As I see it, adjusted predicted durations at ages such as 30, 40, 50, and 60 would be easier to understand than the single yearly EMR.  
5. The authors correctly show that absence duration is strongly skewed. In a dataset of this size, repeated t-tests will identify very small differences as statistically significant. I would reduce the attention given to p-values in Tables 2–4 and focus on the size of the differences. Confidence intervals for mean differences, median comparisons, or quantile-based analyses would give the reader a better sense of what is practically meaningful. Some of these detailed stratified tables could be moved to supplementary material. The main Results section would then have more space to explain the adjusted findings.  
6. The longer absence after accidents occurring on the journey from work may be related to fatigue, shift work, sleep loss, time pressure, or longer journeys. These are plausible possibilities; however, the register contains no information on them. The same applies to transport mode, domestic responsibilities, and the “double presence” framework. See that these ideas are useful, but the current dataset cannot test them. Moreover, the recommendations concerning scheduling, recovery time, and flexible work hours should be retained as policy considerations, with wording that makes clear they have not been evaluated in this study. 
7. Ultimately, I found Figure 3 difficult to interpret. The red shading has no legend, scale, or numerical values, so the reader cannot tell how the figure relates to the percentages in Table 5. A heatmap or a bar chart with labelled percentages and denominators would be more informative. The wording for anatomical categories should also be consistent throughout the manuscript; for example, “trunk and thoracic region”, “torso and thoracic region” and “multiple body parts/sites” should not alternate.  
8. The author-contributions statement needs copyediting. Also, a STROBE checklist, an appendix detailing the coding decisions, and information on access to the analysis code would strengthen the reporting.  

I hope these suggestions guide you to a revised version ready for publication. 

Author Response

We thank the reviewer for the detailed methodological assessment and recommendations. These comments have led to substantial revisions of the analytical framework, regression model, interpretation, and reporting. We address each point below.

1. Population underlying the incidence figures and study-period coverage

Response: We agree that all-sector national incidence rates cannot establish incidence patterns within health and social care activities. We have revised the incidence figures: Figure 1 now compares health and social care activities with all economic sectors during 2011–2023, while Figure 2 presents sector-specific rates by age group and sex in 2023. The Methods identify the corresponding populations and denominators, and the figure captions explicitly distinguish the sector-specific estimates from the national comparison.

The revised series now begins in 2011, removing the previous discrepancy between the incidence reporting window and the individual-level study period.

We have also revised the final paragraph of the Introduction to distinguish incidence from recorded sickness absence duration. The Conclusions focus on adjusted duration associations among registered accident cases and no longer assert a general decline in accident incidence with age. In particular, the sector-specific age patterns differ between men and women and do not support a universal decreasing age gradient.

2. Sample selection and variable coding

Response: Section 2.2 now reports the successive selection counts. The source files contained 7,695,970 occupational accident and relapse records. Restriction to CNAE divisions 86–88 yielded 746,556 records, including 144,938 commuting records. Exclusion of 5,526 relapses left 139,412 accidents, all of which were included in the primary regression.

No records were excluded for inconsistent commuting-direction codes or missing values. However, missing contract information in 78 records was retained as a separate unknown category, rather than treated as a reason for case exclusion. Detailed diagnostic information was unavailable for 40,045 records; this variable was omitted from the adjustment set without excluding those accidents.

Supplementary Table S5 provides the sequential counts and reasons for exclusion in tabular form. We have used this table rather than a separate graphical flow diagram to document the selection process.

The official Delt@ reporting guide is now cited to support the use of codes 0 and 99 for journeys to work and return journeys, respectively. We clarify that these are administrative direction codes rather than actual hours worked. Appendix A and Supplementary Table S4 document traffic involvement, anatomical groupings, other covariates, reference categories, and the treatment of missing or unspecified values.

3. Outcome support, the 545-day maximum, and model adequacy

Response: We have replaced the ordinary negative binomial model as the primary analysis with a zero-truncated NB2 model, whose likelihood is conditional on positive recorded duration. This explicitly accommodates the minimum of one day imposed by the inclusion criteria.

We now state that 545 days was the maximum recorded value in the supplied dataset and that no additional analytical upper limit was imposed by the researchers. There were 34 records at this value, representing 0.0244% of the sample. Although this maximum coincides with the Spanish temporary-incapacity exhaustion threshold, the supplied duration field does not distinguish completed episodes, administrative closure, or ongoing absence. We therefore cannot establish the censoring status of these observations.

A time-to-return-to-work analysis could not be performed because a verified return-to-work event and censoring status were unavailable. We explain this limitation explicitly and interpret the outcome as recorded administrative absence duration. Exclusion of the 34 endpoint records was examined as a sensitivity analysis, without presenting it as a censoring analysis.

Section 3.5 reports the dispersion estimate of 1.063 (95% CI: 1.054–1.073), log likelihood, AIC, BIC, and comparisons with ordinary NB2 and zero-truncated Poisson models. The Methods and Supplementary Materials document convergence, information-matrix checks, Pearson residual statistics, calibration, randomized quantile residuals, influence diagnostics, and sensitivity analyses. We also acknowledge the remaining discrepancies in distributional fit.

4. Additional adjustment and the age association

Response: The expanded primary model now includes accident year (AÑO), activity division, workplace province, three-digit occupational category, and employment-contract code, in addition to the original worker and accident characteristics. These covariates are represented categorically. An alternative adjustment using recorded accident year was examined in sensitivity analysis.

Injury type and recorded severity were included in an additional adjustment model. Detailed diagnosis was not included because 40,045 records lacked usable information. We therefore distinguish anatomical location from injury type and severity rather than assuming that body part adequately captures clinical severity.

Age is now represented by a four-knot restricted cubic spline. The test of departure from linearity was significant (robust Wald χ² = 57.09, 2 degrees of freedom; p < 0.001). We have consequently removed the interpretation of one constant percentage increase per additional year.

Table 7 reports adjusted mean durations and 95% confidence intervals at ages 25, 35, 45, 55, and 65, with age 45 as the comparison reference. Figure 4 shows the adjusted curve across ages 20–70. These estimates provide the more interpretable absolute-duration presentation suggested by the reviewer. Year associations are explicitly not interpreted as causal effects of COVID-19 or mobility restrictions.

5. Magnitude, precision, and practical interpretation

Response: We agree that statistical significance alone is insufficient in this large sample. We have expanded the presentation of effect magnitudes through marginally standardised adjusted mean ratios with 95% confidence intervals in Table 6 and adjusted mean durations in days with 95% confidence intervals in Table 7.

We also report absolute differences in the descriptive Results. Across age groups, return-journey accidents had unadjusted mean durations 4.45–8.05 days longer than journeys to work, while traffic-related accidents had durations 4.25–7.37 days longer than non-traffic-related accidents. These descriptive differences are distinguished from the adjusted estimates.

We retained Tables 2–5 in the main manuscript because they help explain case composition and the contrast between crude and adjusted findings. The t-tests remain descriptive comparisons; we do not interpret small p-values as evidence of practical importance. We have not added quantile regression or confidence intervals for the crude mean differences; the principal inferential results are the adjusted estimates and their confidence intervals.

6. Unmeasured mechanisms and preventive implications

Response: We have revised the Discussion to present fatigue, shift work, sleep loss, time pressure, journey characteristics, transport mode, and domestic responsibilities as possible interpretations informed by previous literature. We explicitly state that these factors were not measured and that their contribution cannot be inferred from commuting direction, sex, or traffic involvement alone.

The preventive recommendations have also been revised. Scheduling changes, fatigue management, safer travel arrangements, rehabilitation, and temporary task modifications are presented as potential approaches requiring assessment and prospective evaluation. Their effectiveness was not tested in this study.

7. Figure 3 and anatomical terminology

Response: Figure 3 has been replaced by a combined anatomical overview and labelled heatmap. The figure includes a common percentage colour scale, group denominators, and the percentage and absolute count for each anatomical category. Its caption clarifies that the percentages describe distributions of registered accident cases rather than workforce incidence rates or injury severity. Table 5 has been retained.

The anatomical coding groups are explicitly defined in Appendix A, and the revised model table uses labels reflecting their complete code composition, including other or multiple sites where applicable.

8. Author contributions and reporting transparency

Response: We have revised the author-contributions statement and provided a completed STROBE reporting checklist in Supplementary Table S7. Appendix A and Supplementary Table S4 describe the coding decisions, while Tables S5 and S6 provide selection counts, missing-data information, and distributions of selected additional covariates.

Analysis code is not included in this submission. The manuscript and Supplementary Materials instead provide the model specification, spline knots, adjustment variables, reference categories, estimation and standardisation procedures, diagnostic checks, and sensitivity analyses. Conditions for requesting access to the administrative microdata are explained in the Data Availability Statement.

We appreciate the reviewer’s careful assessment, which has helped strengthen the analytical boundaries, methodological transparency, and interpretation of the study.

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The changes implemented are appropriate and sufficiently improve the clarity, coherence, and overall quality of the manuscript

Author Response

We thank the reviewer for this positive assessment and for the constructive feedback throughout the review process, which has helped us improve the clarity, coherence, and overall quality of the manuscript. We are pleased that the revisions have satisfactorily addressed the reviewer’s concerns.

Reviewer 2 Report

Comments and Suggestions for Authors

The methodology description is OK now. 

The SPSS program use must be include in Abstract also. 

"Descriptive and bivariate analyses were conducted using IBM SPSS Statistics". 

Author Response

We thank the reviewer for their careful review and constructive feedback, and we are pleased that the methodology description is now satisfactory. As requested, we have added the following sentence at the end of the Methods section of the abstract: “Descriptive and bivariate analyses were conducted using IBM SPSS Statistics”.

Reviewer 3 Report

Comments and Suggestions for Authors

Dear Juan and colleagues,

Thank you for your careful response to my comments. 

Having reviewed the revised manuscript, I am satisfied that the issues raised have been adequately addressed, and I consider the work suitable for publication.

I look forward to seeing this study contribute to knowledge in this field!  

Warmest! 

Author Response

We sincerely thank you for your careful review, constructive feedback, and kind words. We are pleased that the revisions have adequately addressed your concerns and that you consider the manuscript suitable for publication. We greatly appreciate your time and contribution to improving our work, as well as your encouraging remarks about its potential contribution to the field.