Review Reports

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Thank you for the opportunity to review this manuscript. The study addresses an interesting and relatively unexplored topic in academic dentistry, particularly the use of patient-transition networks to describe the organization of multidisciplinary dental services. The manuscript is generally well structured and the methodology is described in considerable detail. However, I believe some aspects should be clarified or improved, particularly regarding the interpretation of patient transitions, the definition of the continuity indicator, and the generalizability of the findings.

Major comments

1. Interpretation of “patient transitions” versus referrals (Introduction, Methods, Results and Discussion).
One of the main issues that should be clarified throughout the manuscript is the distinction between a patient transitionand a clinical referral. The authors appropriately state in several places that the network should not be interpreted as proof of formal referrals. However, the Introduction repeatedly describes the network in terms of “referral pathways” and “referral networks” before this distinction is fully established. Since the transition matrix is constructed from consecutive service attendance rather than documented referral orders, I suggest using “patient-transition network” consistently throughout the manuscript and avoiding “referral network” unless actual referral data are available.

This is particularly important when interpreting the high PageRank of RX. The statement that RX may have a cross-cutting diagnostic role is reasonable, but this should remain clearly presented as an interpretation rather than as evidence of a coordinating or referral function.

2. Institutional Continuity Indicator (ICI) (Methods, Section 2.5; Table 1; Discussion, Section 4.2).
The ICI is an interesting descriptive measure, but its interpretation requires some additional clarification. The manuscript correctly states that it is not the Bice-Boxerman continuity index. However, because the denominator and first-visit registration practices appear to vary considerably between services, I would recommend explaining more explicitly why this ratio is useful and what exactly it measures.

For example, RX has 6,021 unique appointments and zero first visits, resulting in an ICI of 0%, while IPD has an ICI of 22.2%. These differences may primarily reflect differences in the way services register first visits rather than differences in continuity of care. The manuscript mentions this limitation, but I suggest making this point more explicit when presenting Table 1 so that readers do not interpret ICI as a quality or continuity ranking.

3. Patient-level denominator and multiservice patients (Results, Table 1 and Section 3.3).
The manuscript reports 13,026 service-specific patients and 7,260 unique institutional patient identifiers. The explanation that the former is a non-additive sum is appropriate, but this distinction could be made clearer for the reader. I suggest briefly explaining in the Results why these two numbers answer different questions and emphasizing that the network analysis is restricted to the 3,947 patients with multiservice trajectories.

This is important because the network does not represent the complete patient population of the clinic. Patients who only attended one service are not represented in the transition network, and therefore the network should not be interpreted as describing all patient pathways within the institution.

4. Construction of the transition matrix (Methods, Section 2.6).
The methodology for constructing the 18 × 18 matrix is clearly described. However, the rule whereby every service in an earlier same-day stage is connected to every different service in the subsequent stage deserves a little more explanation, as this may generate multiple transitions from a single pair of stages.

I suggest providing a short hypothetical example in the Methods or Supplementary Material. For example, if a patient has services A and B on one date and services C and D on the next date, it would be useful to explicitly show that four directed transitions (A→C, A→D, B→C and B→D) are generated. This would make the methodology much easier to reproduce.

5. Interpretation of network density and centrality (Results, Table 3; Figure 1; Discussion).
The network density is very high (0.827), with 253 of 306 possible directed pairs present. The authors appropriately mention that this affects the interpretation of betweenness and modularity. However, I think this deserves slightly more emphasis because such a dense network means that the distinction between highly and poorly connected services is relatively limited.

In particular, the high PageRank of RX should not be interpreted simply as evidence of greater organizational importance. It may partly reflect the structure of the transition data and the high frequency of imaging in dental care. The Discussion already acknowledges this, but I recommend making the interpretation more cautious.

6. Figure 1 – interpretation of the network visualization.
Figure 1 is useful and the decision to display only the 40 highest-weighted edges improves readability. However, because the complete network contains 253 directed pairs, readers may initially interpret the absence of a visible edge as the absence of a transition. The figure legend does explain this, but I suggest making the distinction even more prominent, for example by adding a sentence in the main text immediately before the figure explaining that the visualization is intentionally truncated and does not represent the complete network.

It may also be useful to provide the complete 18 × 18 transition matrix as supplementary material, if it is possible to do so while maintaining data confidentiality.

7. Generalizability (Discussion, Section 4.5).
The study is based on a single university dental clinic and the services include specific postgraduate programmes and teaching units. Consequently, the observed network structure may be highly dependent on the organization of this particular institution. The manuscript acknowledges the single-centre design, but I suggest expanding the discussion of external validity and making clear that the findings should not be directly generalized to university dental clinics with different organizational structures.

8. Data availability and reproducibility.
The Data Availability Statement indicates that the data are available on request and subject to institutional authorization. Given that reproducibility is one of the stated aims of the manuscript, I suggest clarifying whether the anonymized 18 × 18 weighted transition matrix can actually be shared as supplementary material. Since the matrix contains aggregated service-level transitions rather than patient identifiers, making it available would substantially improve the reproducibility of the network analysis, provided that this is compatible with the ethical approval and institutional data-sharing restrictions.

Minor comments

9. Terminology throughout the manuscript.
I recommend checking the manuscript for consistency in the use of “service”, “specialty”, “department”, “referral” and “transition”. These terms are not necessarily interchangeable, and consistent terminology would make the manuscript easier to follow.

10. Table 1.
Table 1 is informative, but the meaning of “Service patients” and the non-additive total of 13,026 could be made clearer directly in the table footnote. The current footnote is correct, but an additional short explanation that patients can appear in more than one service would help readers interpret the table without having to refer back to the Methods.

11. Table 3.
In Table 3, the columns “In” and “Out” appear to refer to the number of distinct incoming and outgoing service pairs, whereas “Incoming n” and “Outgoing n” refer to weighted transition counts. I suggest making this distinction explicit in the table heading or footnote, as readers may otherwise interpret these columns as equivalent measures.

12. Figure 1 and community colours.
Since modularity is very low (Q = 0.042), the community assignment is explicitly described as exploratory. I suggest ensuring that the figure caption makes this especially clear so that the different node colours are not interpreted as clinically validated groups.

Overall, I consider the manuscript potentially valuable and the methodological approach interesting. The main revisions I recommend concern clarification of the meaning and limitations of the transition network rather than a fundamental change to the study. A more consistent distinction between observed sequential service use and formal clinical referral, together with clearer explanation of the ICI and the network construction, would strengthen the manuscript considerably.

Author Response

Comment 1: One of the main issues that should be clarified throughout the manuscript is the distinction between a patient transition and a clinical referral. The authors appropriately state in several places that the network should not be interpreted as proof of formal referrals. However, the Introduction repeatedly describes the network in terms of 'referral pathways' and 'referral networks' before this distinction is fully established. Since the transition matrix is constructed from consecutive service attendance rather than documented referral orders, I suggest using 'patient-transition network' consistently throughout the manuscript and avoiding 'referral network' unless actual referral data are available. This is particularly important when interpreting the high PageRank of RX. The statement that RX may have a cross-cutting diagnostic role is reasonable, but this should remain clearly presented as an interpretation rather than as evidence of a coordinating or referral function.

Response 1: We agree. We defined patient transition at the beginning of the Introduction and revised terminology throughout the manuscript to distinguish consecutive attendance from documented referral. Statements concerning RX now describe a plausible imaging-related structural pattern and explicitly reject inference of coordination, comparative importance, or referral function.

Comment 2: The ICI is an interesting descriptive measure, but its interpretation requires some additional clarification. The manuscript correctly states that it is not the Bice-Boxerman continuity index. However, because the denominator and first-visit registration practices appear to vary considerably between services, I would recommend explaining more explicitly why this ratio is useful and what exactly it measures. For example, RX has 6,021 unique appointments and zero first visits, resulting in an ICI of 0%, while IPD has an ICI of 22.2%. These differences may primarily reflect differences in the way services register first visits rather than differences in continuity of care. The manuscript mentions this limitation, but I suggest making this point more explicit when presenting Table 1 so that readers do not interpret ICI as a quality or continuity ranking.

Response 2: We agree. Section 2.5 now defines ICI as a local first-visit-to-appointment ratio that audits registration and activity mix. The Results, Table 1 footnote, and Section 4.2 now explicitly state that values are sensitive to service-specific registration and cannot rank continuity, quality, or performance.

Comment 3: The manuscript reports 13,026 service-specific patients and 7,260 unique institutional patient identifiers. The explanation that the former is a non-additive sum is appropriate, but this distinction could be made clearer for the reader. I suggest briefly explaining in the Results why these two numbers answer different questions and emphasizing that the network analysis is restricted to the 3,947 patients with multiservice trajectories. This is important because the network does not represent the complete patient population of the clinic. Patients who only attended one service are not represented in the transition network, and therefore the network should not be interpreted as describing all patient pathways within the institution.

Response 3: We agree. Section 3.3 now distinguishes unique institutional patients from service-specific patient relationships and states explicitly that only 3,947 multiservice patients could contribute between-service transitions. Table 1 and the Conclusion repeat this distinction, and the revised text states that the network does not represent single-service pathways.

Comment 4: The methodology for constructing the 18 × 18 matrix is clearly described. However, the rule whereby every service in an earlier same-day stage is connected to every different service in the subsequent stage deserves a little more explanation, as this may generate multiple transitions from a single pair of stages. I suggest providing a short hypothetical example in the Methods or Supplementary Material. For example, if a patient has services A and B on one date and services C and D on the next date, it would be useful to explicitly show that four directed transitions (A→C, A→D, B→C and B→D) are generated. This would make the methodology much easier to reproduce.

Response 4: We agree and added the proposed A/B-to-C/D example to Section 2.6, including all four generated transitions and an explicit statement that one pair of multiservice stages may contribute multiple edges.

Comment 5: The network density is very high (0.827), with 253 of 306 possible directed pairs present. The authors appropriately mention that this affects the interpretation of betweenness and modularity. However, I think this deserves slightly more emphasis because such a dense network means that the distinction between highly and poorly connected services is relatively limited. In particular, the high PageRank of RX should not be interpreted simply as evidence of greater organizational importance. It may partly reflect the structure of the transition data and the high frequency of imaging in dental care. The Discussion already acknowledges this, but I recommend making the interpretation more cautious.

Response 5: We agree. Sections 3.4, 4.3, and 4.4.4 now emphasize that high density limits binary-connectivity discrimination and compresses betweenness. RX PageRank is framed as structural prominence likely influenced by frequent imaging, not organizational importance or quality.

Comment 6: Figure 1 is useful and the decision to display only the 40 highest-weighted edges improves readability. However, because the complete network contains 253 directed pairs, readers may initially interpret the absence of a visible edge as the absence of a transition. The figure legend does explain this, but I suggest making the distinction even more prominent, for example by adding a sentence in the main text immediately before the figure explaining that the visualization is intentionally truncated and does not represent the complete network. It may also be useful to provide the complete 18 × 18 transition matrix as supplementary material, if it is possible to do so while maintaining data confidentiality.

Response 6: We agree regarding the visualization. The main text immediately before Figure 1 and the expanded caption now state prominently that only the 40 highest-weight edges are displayed and that all 253 pairs were analyzed. We did not add the complete matrix as public supplementary material because the current institutional data-use authorization does not permit public dissemination. We clarified that the aggregated matrix and code may be shared on reasonable request, subject to authorization by Universidad Rey Juan Carlos.

Comment 7: The study is based on a single university dental clinic and the services include specific postgraduate programmes and teaching units. Consequently, the observed network structure may be highly dependent on the organization of this particular institution. The manuscript acknowledges the single-centre design, but I suggest expanding the discussion of external validity and making clear that the findings should not be directly generalized to university dental clinics with different organizational structures.

Response 7: We agree. Section 4.5 now explains how curricula, staffing, scheduling, coding, and institution-specific programs can shape topology, states that numerical metrics should not be transferred directly to other clinics, and identifies harmonized multicenter replication as the requirement for external validation.

Comment 8: The Data Availability Statement indicates that the data are available on request and subject to institutional authorization. Given that reproducibility is one of the stated aims of the manuscript, I suggest clarifying whether the anonymized 18 × 18 weighted transition matrix can actually be shared as supplementary material. Since the matrix contains aggregated service-level transitions rather than patient identifiers, making it available would substantially improve the reproducibility of the network analysis, provided that this is compatible with the ethical approval and institutional data-sharing restrictions.

Response 8: We agree that the matrix would strengthen reproducibility. It has not been deposited publicly because the present institutional authorization does not cover public dissemination. The Data Availability Statement and Section 4.6 now state specifically that the matrix, dictionary, and code may be shared on reasonable request subject to institutional approval.

Comment 9: I recommend checking the manuscript for consistency in the use of 'service', 'specialty', 'department', 'referral' and 'transition'. These terms are not necessarily interchangeable, and consistent terminology would make the manuscript easier to follow.

Response 9: We agree and completed a terminology audit. 'Service' denotes the 18 administrative clinical units, 'patient transition' denotes consecutive service attendance, and 'referral' is retained only when discussing documented referrals as an unavailable variable or when describing prior literature.

Comment 10: Table 1 is informative, but the meaning of 'Service patients' and the non-additive total of 13,026 could be made clearer directly in the table footnote. The current footnote is correct, but an additional short explanation that patients can appear in more than one service would help readers interpret the table without having to refer back to the Methods.

Response 10: We agree. The Table 1 footnote now defines service patients, explains that the same patient can be counted in several services, and contrasts the non-additive total of 13,026 with the 7,260 unique institutional patients.

Comment 11: In Table 3, the columns 'In' and 'Out' appear to refer to the number of distinct incoming and outgoing service pairs, whereas 'Incoming n' and 'Outgoing n' refer to weighted transition counts. I suggest making this distinction explicit in the table heading or footnote, as readers may otherwise interpret these columns as equivalent measures.

Response 11: We agree. The column headings are now 'In-degree', 'Out-degree', 'Weighted incoming transitions', and 'Weighted outgoing transitions', and a footnote explicitly distinguishes connected pairs from weighted counts.

Comment 12: Since modularity is very low (Q = 0.042), the community assignment is explicitly described as exploratory. I suggest ensuring that the figure caption makes this especially clear so that the different node colours are not interpreted as clinically validated groups.

Response 12: We agree. The revised caption states that node colors show only an exploratory algorithmic partition and must not be interpreted as clinically validated groups because Q = 0.042.

Reviewer 2 Report

Comments and Suggestions for Authors

Thank you for your interesting submission. Please see attached review comments.

Comments for author File: Comments.pdf

Author Response

Comment 1: While the manuscript is well written, it is riddled with jargon making the study difficult to evaluate for those who might find it most beneficial.

Response 1: We agree. We simplified the Introduction and added plain-language definitions of degree, transition counts, PageRank, betweenness, density, and modularity. We also replaced causal or managerial language with direct descriptions of what each measure can and cannot establish.

Comment 2: It is not clear who would be the target audience(s) as the information is not written at a level that would be useful to academic dental organization leadership as presented. Maybe refining the specific aims and expected outcomes would be helpful.

Response 2: We agree. Section 1.5 now identifies academic dental leadership, clinical-service managers, dental informatics teams, and researchers using routine administrative data as the intended audience. Section 1.7 now presents four explicit aims and states that the immediate outcome is descriptive screening rather than proof of improved workflow or outcomes.

Comment 3: The results presented seem to be minimal based on the amount of information produced by the data audit and study sample size. There are only 3 tables and no figures, although there is mention of a figure 1 (Line 375). It would be helpful to have some visual representation of the model being built and how this model might result in better workflows and care outcomes, if this is the intent of the analysis. (i.e. referrals, time to treatment, satisfaction, etc.)

Response 3: Thank you for identifying that Figure 1 was absent from the preliminarily edited file. We restored the patient-transition network figure and strengthened the accompanying text and caption. We also clarified that the current analysis does not estimate referrals, time to treatment, satisfaction, or clinical outcomes. The revised Discussion explains how the model can identify patterns for targeted operational audit and which additional variables are needed before workflow or outcome improvement can be evaluated.

Comment 4: While abbreviations (Line 561) are listed after the conclusion in the manuscript, it would be helpful if the names of the different dental specialty areas are translated or provided in English, again to help the reader better interpret results.

Response 4: We agree. English service names are now provided in Tables 1 and 3, and all program abbreviations in the Abbreviations section have been translated into English.

Comment 5: It would be helpful in the discussion section, to know how the results compare to what has been found using this approach, for example in other health care systems/environments. What makes the findings in this study significant for academic dentistry based on what is already known? Is the model that was developed adequate for implementation, or does it need more work to effectively be applied to improving data management in dental EHRs across areas? Having this information would make the manuscript more appealing to a broader audience.

Response 5: We agree. Section 4.4.1 now compares the edge definition used here with patient-sharing and professional networks in other healthcare settings. It explains the study's contribution to academic dentistry and explicitly characterizes the model as suitable for descriptive screening and hypothesis generation, but not yet as an implementation-ready decision tool. We list the standardized EHR fields and external validation required for broader application.

Comment 6: The conclusion seems to repeat the results but does not speak to the importance of this network analysis to how it might improve the dental workflows or data management. Additionally, how might this process/model be improved and what might be needed to further refine or implement going forward?

Response 6: We agree. The Conclusion has been rewritten to emphasize the operational value of an auditable transition map, while avoiding unsupported claims of improvement. It now specifies the data elements and multicenter validation needed to refine and implement the model.

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

Thank you for the revised version of the manuscript. The authors have carefully addressed the main concerns raised in my previous review, and I believe the manuscript has been substantially improved.

In particular, the methodological description is now clearer and more transparent. The revised manuscript provides additional information on data sources, the temporal cut-off, data deduplication, the reconstruction of patient-level trajectories, and the construction of the transition matrix. These additions make the analytical approach easier to understand and reproduce.

I also appreciate the decision to remove the previously proposed survival analyses. The explanation provided regarding the limitations of the available data for defining appropriate time origins, events, and censoring is methodologically appropriate and avoids potentially misleading analyses.

Another important improvement is the interpretation of the transition network. The manuscript now clearly distinguishes consecutive service attendances from documented clinical referrals and avoids interpreting these transitions as evidence of formal referral pathways. The Discussion and Conclusions have also been revised appropriately to avoid claims regarding improvements in workflow, waiting times, patient satisfaction, or clinical outcomes that cannot be established from the available data.

The limitations of the single-center, retrospective design and the context-dependent nature of the observed network are also now more clearly acknowledged.

Overall, I consider that the authors have satisfactorily addressed the main issues raised in my previous review. The revised manuscript is clearer, methodologically stronger, and more appropriately cautious in the interpretation of its findings. I have no further major comments

Reviewer 2 Report

Comments and Suggestions for Authors

Thank you for your revised manuscript and adequately address reviewer comments.