Review Reports

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors 1. The most substantive methodological concern relates to the construction of age-specific incidence rates. Throughout the manuscript, incidence is calculated using the total resident population as denominator without adjustment for vaccination status or underlying susceptibility. This approach implicitly treats the entire population, including individuals who are vaccinated and therefore largely protected against infection, as uniformly at risk. In the context of pertussis, where primary vaccination coverage in infants is relatively high and booster coverage among adolescents is substantially lower (reported as approximately 75–79% for 16-year-olds in Tuscany), this assumption is not epidemiologically neutral. It means that the denominators for different age strata contain markedly different proportions of susceptible individuals, rendering direct comparisons of age-specific incidence rates methodologically problematic.

 

The consequence is non-trivial. The 10–14 year age group, identified in this manuscript as carrying the highest cumulative incidence (217.61 per 100,000), has suboptimal booster coverage and therefore a larger proportion of genuinely susceptible individuals in its denominator than, say, the 0–4 year group, where primary vaccination rates are higher. Cross-age comparisons of population-based incidence rates under these conditions do not isolate differential susceptibility from differential exposure or case ascertainment; they conflate all three. The authors therefore are encouraged to calculate or discuss incidence rates restricted to the unvaccinated or incompletely vaccinated stratum within each age group, where vaccination records permit, or at minimum to explicitly acknowledge this limitation and its direction of bias in the text. The recently published analysis by Moa and colleagues on the burden of pertussis in Australian adults, which similarly grapples with the intersection of population-based incidence estimates and vaccination status, may provide a useful methodological reference (https://doi.org/10.3390/vaccines13101029)   2. A central analytical challenge for any passive surveillance study of this kind is the assumption that case ascertainment probability remains constant over time. This assumption is implausible here. The dramatic increase from 10 cases in 2023 to 598 cases in 2024 is interpreted primarily as a true resurgence of pertussis, and this may well be correct. However, the ECDC Communicable Disease Threats Report reintroduced pertussis as a monitored threat from 2023 onwards after nearly a decade of absence, and a dedicated Rapid Risk Assessment was published in May 2024. The authors themselves acknowledge that heightened awareness among clinicians may have increased diagnostic sensitivity and reporting completeness in 2024. These institutional and behavioural changes in the surveillance system constitute a time-varying confounder of unknown magnitude that cannot be separated from the true underlying trend using passive surveillance data alone.
This limitation is particularly consequential for two specific claims: first, the inference about changes in the age distribution of cases, which could reflect differential improvements in case detection across age groups rather than genuine epidemiological shifts; and second, the interpretation of the declining hospitalization proportion in 2024 (10.4% versus 60% in 2023). The latter is almost certainly influenced by surveillance inflation of the denominator, as milder cases become detectable through heightened diagnostic vigilance, the total number of reported cases increases disproportionately relative to severe cases, mechanically reducing the hospitalization fraction without any necessary change in disease severity. The authors should discuss this numerator-denominator artifact explicitly and avoid presenting the decline in hospitalization proportion as straightforward evidence of attenuated clinical severity. The severity of underreporting in Italian surveillance, estimated at approximately 141-fold among adolescents and over 3,000-fold among adults, provides essential context for interpreting apparent changes in age-specific incidence [please cite https://doi.org/10.3390/jcm12051732]. Active surveillance data from comparable settings, such as the community-based study conducted in Yiwu, China, which estimated pertussis incidence at 108.3 per 100,000 compared to far lower passive estimates, illustrates the systematic gap between passive and active surveillance systems and the magnitude of true burden likely obscured by notification-only approaches [cite https://www.mdpi.com/2076-2607/12/11/2186 ].

3. The manuscript calculates a "cumulative incidence" across the six-year observation period using the average resident population as denominator (reported as 18.19 per 100,000). This is not a standard epidemiological measure. Cumulative incidence, properly defined, estimates the probability of disease occurrence in a defined population over a specific period under the assumption that all individuals are followed from the start of the period and are event-free at baseline. Its calculation requires a closed cohort or a well-specified open cohort with person-time denominators. The approach used here, dividing total cases over six years by an average annual population count, yields a period incidence rate rather than a true cumulative incidence, and conflates the very different epidemiological conditions of the pandemic suppression period (2020–2023, 32 total cases) with the 2024 resurgence (598 cases). The resulting figure (18.19 per 100,000) is dominated by a single calendar year and is unlikely to be informative for surveillance purposes. The authors should either adopt standard incidence rate terminology, present annual rates individually, or, if a multi-year summary measure is desired, calculate an age-standardised average annual incidence rate with appropriate denominators and a confidence interval. 

4.  The authors note that a "high proportion of notifications with incomplete information on hospitalization status ('NA')" limited severity analysis, but neither the proportion of missing values nor any assessment of the likely missingness mechanism is provided. If missing hospitalization records are concentrated in particular years, age groups, or local health authorities, their omission could systematically bias estimates of hospitalization rates and their apparent temporal trends. The authors should quantify the extent of missing data for this variable, stratified by year and local health authority if possible, and discuss whether a missing-not-at-random mechanism is plausible given the surveillance context. At minimum, a sensitivity analysis presenting hospitalization rates under complete-case assumptions and under a plausible worst-case imputation should be reported.

5.  Confidence intervals are not reported for any of the incidence estimates, including those presented in Tables 1 and 2 and in the supplementary material. For a surveillance study with small absolute case counts in certain strata (notably 2021–2023, where zero to ten cases per year were recorded), point estimates without uncertainty bounds are unreliable bases for comparison. Poisson-based confidence intervals should be provided for all rates, with exact methods applied where counts are small.
The geographical analysis attributes the majority of cases (64%) to the Central Tuscany Local Health Authority and offers population density as one explanatory factor. While this is plausible, the authors acknowledge the alternative explanation of differential reporting capacity. These two explanations have entirely different public health implications. The manuscript would benefit from a more systematic treatment of this ambiguity, for example by examining whether the proportion of laboratory-confirmed cases (if available) differs across local health authorities, which would serve as a partial proxy for surveillance quality.
The use of the phrase "significant differences" in relation to the annual variation in incidence (Methods section) is potentially misleading in the absence of formal statistical testing. If no inferential tests were performed, which is methodologically defensible for a descriptive surveillance study, the language should be revised to "substantial" or "notable" differences throughout.

Author Response

Dear Reviewer,
Thank you for your valuable suggestions and for the time you dedicated to evaluating our manuscript. Your comments have been extremely helpful in improving the quality and clarity of our work.

We provide a point-by-point response to the comments raised in attached file

Best regards

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The topic is relevant to the general public, healthcare workers, and policymakers. The title reflects the content of the manuscript. Although the abstract is an adequate summary of the manuscript, there are no p-values that can be used to determine whether the differences are statistically significant or not. The introduction orients the readers to the topic. However, there is no justification on the choice of Tuscany as a study site. Additionally, no knowledge gap was identified. The materials and methods should have discussed how statistical significance of the differences was determined. The results also lack test to determine statistical significance of the findings. The authors compare their findings with previous literature and provide plausible explanations to their findings. The conclusions are based on the findings.

MAJOR REVISIONS

Abstract

1. There is no way readers are able to know whether the differences being presented are statistically significant or not. Although only cases were included in the study, a non-parametric one-sample Chi-square test could have been used to determine if the differences were statistically significant.
2. The authors state, ‘The 2024 resurgence likely reflects waning immunity, pandemic-related vaccination disruptions, and reduced pathogen circulation.’ I am not sure how reduced pathogen circulation can cause a resurgence.

 

Introduction

3. It is not clear why Tuscany was chosen.
4. What knowledge gap were the authors intending to fill?

 

Materials and Methods

5. The authors did not clearly explain how they determined whether the differences were statistically significant or not. Just stating the different percentages without determining if the differences were statistically significant or not is not that informative to readers. I encourage the authors to address this.

 

Results

6. The authors should provide some statistical test to show if any differences being discussed were statistically significant or not.

MINOR REVISIONS

1. Write LHU in full in line 79.
2. Write PREMAL in full in line 82.
3. Write ECDC in full in line 220.
4. Provide citations in lines 279-282.

 

Author Response

Dear Reviewer,
Thank you for your valuable suggestions and for the time you dedicated to evaluating our manuscript. Your comments have been extremely helpful in improving the quality and clarity of our work.

We provide a point-by-point response to the comments raised in attached file.

Best regards

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The authors have responded to all five points raised in the first round of review, and several substantive improvements are evident. The addition of confidence intervals across tabulated estimates, the revised description of the PREMAL surveillance system, and the correction of inferential language represent meaningful contributions to methodological transparency. Nevertheless, a number of responses remain procedurally compliant rather than analytically complete, and at least one central methodological issue: the misclassification of the summary incidence measure, has not been resolved with sufficient rigour. Minor concerns also persist regarding the treatment of missing hospitalization data and the interpretive framing of geographical heterogeneity. The manuscript is not yet acceptable in its current form and requires targeted revision before publication.

Point-by-Point Assessment

• R1 — Age-specific incidence denominators and vaccination status. The authors acknowledge the limitation appropriately and have expanded the PREMAL description to note that vaccination status is not captured in the surveillance data. This is an acceptable response given the constraint that individual-level vaccination records cannot be linked. The revised Methods text provided in the cover letter is clear and proportionate. This comment is considered satisfactorily addressed, provided the expanded PREMAL description appears in the published manuscript exactly as quoted.
• R2 — Time-varying case ascertainment as a confounder of temporal trends. The authors accept the substantive point and report having strengthened the Discussion to acknowledge differential surveillance sensitivity across time. This is welcome.
• R3 — Misuse of the term "cumulative incidence." This remains the most significant unresolved issue. The authors state that they "calculated incidence rates, presented the annual rates individually, and reported the period incidence," and have added a mean annual incidence rate with confidence intervals. This is an improvement, but the response does not confirm whether the term "cumulative incidence" has been replaced throughout the manuscript, or whether the metric previously reported as "cumulative incidence 2019–2024" (18.19 per 100,000) continues to appear under that label. Inspection of the revised manuscript tables (Supplementary Table 1 and Tables 1–2) suggests that the column header "Cumulative incidence 2019–2024" is retained, as is the associated figure of 18.19 per 100,000 in the abstract. This is terminologically incorrect. A quantity derived by dividing total cases over a multi-year period by an average annual population count is a period incidence rate or, at best, a mean annual rate, not a cumulative incidence in any epidemiologically defensible sense. Cumulative incidence requires a defined cohort, event-free at baseline, followed for the duration of the period, and is bounded between 0 and 1. The retained terminology will mislead readers and is inconsistent with standard epidemiological usage. The authors must either rename this measure throughout (including in the abstract, tables, and supplementary material), or replace it entirely with the mean annual incidence rate now correctly calculated and presented. This revision is mandatory.
• R4 — Missing hospitalization data and confidence intervals. The addition of confidence intervals across all tabulated estimates is appropriate and directly responsive to the reviewer's request; this element is satisfactorily addressed. The treatment of missing hospitalization data, however, remains inadequate. The authors assert that PREMAL is standardised across Local Health Authorities and that data collection is uniform, which they appear to offer as reassurance that the missing data are inconsequential. This conflates standardisation of data collection procedures with completeness of data reporting — two distinct properties of any surveillance system. Standardised forms do not preclude differential missingness if reporting behaviour varies systematically by year, LHA, age group, or clinical severity. I explicitly requested quantification of the proportion of missing values stratified by year and LHA, as well as a discussion of whether a missing-not-at-random mechanism is plausible. Neither has been provided. At minimum, the proportion of cases with "NA" hospitalization status should be reported in a supplementary table stratified by year and LHA, and the Discussion should acknowledge that if missing records are concentrated in 2024 (the year with the largest case load), apparent hospitalization rates for that year may be understated. A sensitivity analysis under a plausible worst-case assumption, for example, treating all NA cases as hospitalised, would further strengthen the analysis.
• R5 — Inferential language and statistical significance. The authors confirm that "significant differences" has been replaced with appropriate descriptive language throughout. This is satisfactorily addressed and requires no further action.

Additional Concern Not Previously Raised

The confidence intervals reported for several strata in Tables 1 and 2 include negative lower bounds (e.g., mean annual incidence for the 15–19 year group: –0.10 to 4.29; for the 20–24 year group: –0.38 to 0.58). Negative incidence rates are not epidemiologically interpretable. This arises from the application of normal approximation methods to Poisson counts that are too small for this approximation to be valid. Where case counts are sparse (fewer than five events per stratum per year), exact Poisson confidence intervals, or, if preferred, score-based intervals, should be substituted. The presence of negative lower bounds in published tables constitutes a reporting error that should be corrected prior to acceptance.

Author Response

See file. Thank you

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The authors have addressed all my comments, and that has improved the manuscript. However, the confidence intervals added are not included in the narrative. The authors should interpret the confidence intervals.

Author Response

See file. Thank you

Author Response File: Author Response.pdf

Round 3

Reviewer 1 Report

Comments and Suggestions for Authors

The authors have responded conscientiously to all three mandatory revisions specified in the second round, and the manuscript is substantially improved as a result. The correction of confidence interval methods, the terminological revision throughout the manuscript, and the important clarification regarding hospitalization data completeness represent meaningful advances in methodological transparency. One residual textual inconsistency, a direct consequence of the R4 clarification, remains present in the body of the manuscript and must be corrected before acceptance. Two minor points are noted below for completeness.

1. The revision has been implemented correct period incidence consistently throughout the abstract, all main tables, supplementary tables, and the body text. Table 1 now correctly presents a "Period incidence rate 2019–2024" column alongside the mean annual incidence rate, and the abstract has been updated accordingly. This mandatory revision is satisfactorily resolved.

2. The application of exact Poisson methods for strata with five or fewer events has eliminated all inadmissible negative lower bounds across Tables 1, 2, and Supplementary Table 1. The methodological clarification has been appropriately incorporated into the Materials and Methods section. This mandatory revision is satisfactorily resolved.

3. The authors' clarification that PREMAL contains no missing values for hospitalization status and that the original reference to "NA" pertained to disease severity variables introduced only in 2024 resolves the substantive concern. The rephrased sentence in the Discussion (lines 255–258) correctly redirects the missing-data caveat to clinical severity rather than hospitalization. This clarification is ok, and the mandatory revision is satisfactorily resolved in its substantive dimension.

Residual Inconsistency Requiring Correction

One textual error introduced by the R4 correction has not been propagated to all relevant locations in the manuscript. The Discussion section (lines 327–330) retains the following sentence in its recommendations paragraph: "The incomplete information on hospitalization status ('NA') limited the assessment of disease severity and hospital burden." This sentence directly contradicts the authors' own clarification, confirmed in the cover letter, that no missing values exist for the hospitalization variable. The reference to hospitalization status and the parenthetical "(NA)" must be removed or revised in this paragraph to align with both the corrected sentence at lines 255–258 and the authors' stated position. Failure to correct this passage will create an internal inconsistency in the published article that could mislead readers.

Minor Points

1. The abbreviation list continues to define "NA: Not applicable," a designation that the authors have clarified does not apply to hospitalization data in this dataset. Given that the only substantive use of the NA notation in the original manuscript has now been reattributed to a different variable (disease severity), this abbreviation entry should either be updated to reflect its corrected referent or removed if it no longer serves a necessary function in the revised text.

2. The period incidence rate (18.19/100,000 over six years) remains reported as the headline figure in the abstract, despite the authors now also presenting the mean annual incidence rate, which is the more interpretable summary measure given the extreme between-year heterogeneity of the data. Consideration should be given to foregrounding the mean annual rate in the abstract and relegating the period rate to a secondary position, as the latter is dominated by a single calendar year and conveys limited epidemiological information in isolation. This point is offered as a suggestion rather than a requirement.

My Final Recommendation

Minor revision. Acceptance is contingent solely on correction of the residual inconsistency at lines 327–330 of the Discussion. No further methodological review is required following this correction.

Author Response

Thank you, See file 

Author Response File: Author Response.pdf