Impact of Spatial Aggregation Level on Environmental Epidemiology Analyses: A Case Study of Combined Heat and Ozone Effects on Cardiovascular Emergencies

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This manuscript addresses an important gap in spatial epidemiology with a well-structured case study. However, major revisions require urgent attention to ensure methodological validity and reproducibility. The minor  issues are straightforward to address and would significantly enhance clarity and impact.

1. The study employs district-level spatial aggregation but draws conclusions about individual health risks (e.g., Section 4.2). This risks ecological fallacy, a well-documented limitation in spatial epidemiology. The authors must explicitly acknowledge this limitation in the Discussion (or Limitations section) and clarify that findings reflect area-level associations, not individual-level causal effects.
2. The Data Availability Statement (Line 433) cites a GitHub repository but omits the URL. This violates open-science standards and undermines reproducibility.
3. The use of the 75th percentile for EMS outcomes and 85th for temperature/ozone lacks justification (Section 2.3.2). These thresholds are common but not universally optimal; without rationale, the analysis appears arbitrary.
4. Figure 3 displays 96 district-level ORs without geographic context or significance markers. Readers cannot identify high-risk districts or spatial patterns.
5. While seasonality is addressed via moving-window binarization, other temporal confounders (e.g., day-of-week effects, long-term trends) are unaddressed. This could bias results.
6. The conclusion (Section 4.3) states findings "inform public health planning" but lacks concrete examples

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

Dear Authors,

Please consider addressing the following concerns:

1. Clarify Research Question and Novelty: The introduction provides extensive background (lines 34–54) but does not clearly state the unique contribution of this study until later (lines 93–97). The abstract and early introduction should explicitly highlight the research question and novelty
2. Strengthen Methodological Detail and Justification: The methods section (lines 127–215) needs more explanation of key choices:

• Why were districts of ~100,000 residents selected (line 131)?
• Why were the 75th and 85th percentiles chosen for binarization (lines 169–172)?
• Why was Odds Ratio selected as the primary metric (line 187) over other epidemiological measures?

3. Improve Discussion Structure and Emphasis: The discussion (lines 280–356) combines findings, literature comparisons, and methodological reflections in long paragraphs.
4. Enhance Presentation and Language: Figures and tables should be more prominent: Move Figure 2 (line 217) summarizing the analytical framework earlier in the methods section (or something similar to that). Also, please highlight Table 1 (lines 274–277) comparing spatial approaches in the main text.

1. Expand Limitations and Future Directions
2. Ensure consistent terminology (e.g., “district-level analysis” vs. “dis-aggregated analysis”).
3. Add a brief note on data harmonization and interpolation steps in the methods for clarity.
4. Consider including a short paragraph in the conclusion emphasizing practical steps for integrating geomatics into public health planning.

 

Good Luck

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

Methodological comments

1. Justification of percentile thresholds

The use of the 75th percentile for health outcomes and the 85th percentile for temperature and ozone exposure is applied consistently throughout the analysis. However, the rationale for selecting these specific thresholds is not sufficiently discussed. It is unclear whether these cut-offs are based on previous studies, methodological conventions, or empirical considerations specific to the dataset. A clearer justification is needed. In addition, a brief sensitivity analysis using alternative thresholds (e.g. 80th or 90th percentile) would help assess the robustness of the results.

2. Binary transformation of continuous variables

The binarization of both exposure and outcome variables simplifies the analytical framework but inevitably leads to a loss of information regarding exposure gradients and dose–response relationships. While this choice is acceptable given the methodological focus of the study, its implications should be discussed more explicitly. In particular, it should be clearly stated that the estimated odds ratios refer only to extreme conditions rather than to the full range of temperature and ozone variability.

3. Interpretation of odds ratios

In several sections, the interpretation of odds ratios may be read as implying a direct measure of population risk. Given the relatively frequent nature of the outcome (EMS calls), it would be appropxriate to explicitly note that odds ratios do not directly approximate relative risks in this context. A short methodological clarification would help prevent overinterpretation of effect sizes.

4. Lack of adjustment for confounding factors

The analysis does not account for potential confounders such as age distribution, sex, socioeconomic status, or co-pollutants (e.g. PM₂.₅, NO₂). Although the primary aim of the study is to examine the effect of spatial aggregatiofn rather than to estimate causal effects, the absence of confounder control should be more clearly acknowledged as a limitation. This is particularly important when discussing the epidemiological implications of the findings.

Spatial analysis and aggregation

5. Construction of population-based districts

The manuscript refers to a custom clustering algorithm used to generate districts of approximately 100,000 inhabitants, but the description remains relatively general. Additional details on the optimization criteria (e.g. population homogeneity, spatial contiguity) would improve transparency. It would also be useful to discuss whether alternative spatial configurations could lead to different results.

6. Discussion of the Modifiable Areal Unit Problem (MAUP)

The manuscript appropriately acknowledges the MAUP as a central issue, but the discussion could be expanded. In particular, distinguishing between scale effects and zoning effects would strengthen the theoretical framing. A more explicit connection between the observed results and classical MAUP concepts would enhance the meethodological contribution of the paper.

Results and interpretation

7. Public health relevance of effect sizes

The statistically significant odds ratio obtained in the cumulative analysis (OR = 1.13) is relatively modest. While this is consistent with much of the environmental epidemiology literature, the practical implications could be discussed in greater depth. For example, translating the effect into an estimated number of additional EMS calls at the regional level would help contextualize the findings for public health audiences.

8. Spatial heterogeneity of district-level results

Only a small subset of districts shows statistically significant associations. The manuscript would benefit from a more detailed exploration of tthese areas, for instance by discussing whether they share common environmental, climatic, or urban characteristics. Tzhis could help distinguish between true spatial heterogeneity and random variability

Editorial and presentation issues

9. Length and focus of the introduction

The introduction is comprehensive but somewhat lengthy. Parts of the general discussion on climate change and health could be streamlined to maintain a sharper focus on the methodological research question related to spatiaal aggregation.

10. Reproducibility and data availability

The manuscript mentions that the analysis code is available on GitHub, but no active link is provided. For transparency and reproducibility, the repository link should be included, and the conditions for accessing the underlying data should be stated more clearly.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 4 Report

Comments and Suggestions for Authors

Respected Authors,

Your manuscript investigates the impact of spatial aggregation strategies on environmental epidemiology analyses, using combined heat and ozone exposure and cardiovascular EMS events in Lombardy (Italy) as a case study. The topic is timely and relevant to geomatics and environmental health, and the manuscript is generally well written and clearly structured. The focus on spatial aggregation as a methodological determinant is appropriate for ISPRS International Journal of Geo-Information.

However, while the conceptual motivation is strong, several methodological choices require clearer justification, and some interpretive claims are overstated relative to the analytical simplicity of the Odds Ratio framework used. There are also presentation issues, minor inconsistencies, and opportunities to strengthen the manuscript’s contribution by more explicitly positioning it against existing spatial epidemiology methods.

Concerning originality and novelty, the manuscript explicitly centers spatial aggregation as a methodological variable (lines 93–111), which is rarely isolated as a primary research question in environmental epidemiology. The use of custom population-based districts rather than administrative units (lines 131–137, Figure 1B) is a notable contribution in geoinformatics. In addition, the comparison between aggregated, disaggregated, and cumulated strategies (lines 201–215) is conceptually novel and well aligned with MAUP discussions.

However, the manuscript does not sufficiently distinguish its approach from existing multi-level or hierarchical spatial analyses. (For example, lines 336–356 claim that spatial aggregation is “rarely considered,” but similar issues are addressed indirectly in small-area studies, Bayesian spatial models, and DLNM frameworks. Also, the novelty would be clearer if the authors explicitly contrasted their approach with multilevel models, spatial random effects, or Bayesian disease mapping.

My recommendation here is to explicitly clarify how this approach differs from or complements existing spatial epidemiology methods, rather than implying that spatial aggregation is generally neglected (Discussion, lines 336–356).

Concerning the significance of content, the public health relevance is clearly articulated (lines 39–59 and 100–114). Also, the Lombardy region is convincingly justified as a relevant case study (lines 117–125 and 362–365). Finally, the finding that cumulative district-level analysis yields a significant association (OR 1.13, Table 1) is potentially important for territorial health indicators. All of the above are the strengths in this case.

However, the policy implications are discussed in broad terms (lines 384–409) but remain abstract. Furthermore, it is not entirely clear how decision-makers would operationalize the “cumulated” approach differently from standard regional analyses.

My recommendation here is to add a short paragraph in the Discussion or Conclusions clarifying what concrete analytical or planning decisions would change if the proposed spatial strategy were adopted (e.g., heat-health warning systems, spatial prioritization of interventions).

Concerning the scientific soundness, there are major issues in the study design and the methodology:

Firstly, at the use of Odds Ratio without covariate adjustment, the analysis relies exclusively on unadjusted 2×2 contingency tables (lines 188–199). Also, no adjustment is made for seasonality (beyond windowed binarization), long-term trends, day-of-week effects, and potential confounders (e.g., influenza peaks, socioeconomic factors).

While the authors acknowledge this as a methodological framework study, the lack of adjustment limits causal interpretation.

Secondly, in the percentile-based threshold selection, the choice of the 75th percentile for outcomes (line 170) and the 85th percentile for exposures (lines 171–172) is not justified. Different thresholds could substantially alter OR magnitudes.

Finally, a cumulative OR interpretation is done. In the cumulative analysis (lines 211–215), districts are treated as independent observational units. This implicitly assumes independence across districts and days, which is unlikely given spatial autocorrelation.

My recommendations here are to:

• Explicitly state that ORs are descriptive associations, not causal estimates.
• Clarify that results should not be compared directly with adjusted epidemiological effect estimates.
• Provide a rationale or sensitivity analysis.
• At a minimum, acknowledge that threshold choice is arbitrary and may influence results.
• Acknowledge this assumption explicitly in the Limitations (lines 366–374).
• Clarify that the cumulative OR reflects spatial extent, not individual or population risk.

Concerning the quality of the presentation, the overall structure is logical and easy to follow. Figures 1–3 and Table 1 are relevant and generally well-integrated. The Appendix (Table A1) provides full transparency.

However, here I have to mention some specific Issues what I noticed, firstly, on unclear phrasing (e.g., in the line it is written “within specific analytical context”. This needs to be rephrased as “within specific analytical contexts”, and in lines 269–270, it is written “with an odd of recording” and should be “with the odds of recording”.

Secondly, on figure clarity, in Figure 3 (lines 261–265), the term “non-transparent areas” is unclear without seeing the color scale. Also, you must consider explicitly stating whether color, opacity, or symbol size encodes significance.

Finally, on the software and reproducibility, the authors reference a GitHub repository but provide no link (lines 222–224 and 433–434). This weakens reproducibility claims, so it should provide the repository URL or remove the claim until available.

Concerning the Discussion, there is an appropriate emphasis on methodological implications (lines 312–356). Also, the comparison with the authors' previous work (lines 329–334) is informative.

However, here I have to say that some interpretations verge on overgeneralization (lines 323–328 suggest the combined approach is “the most effective solution,” but this is demonstrated only within a single analytical framework and region. The claims about disproving correlations (lines 331–334) are also strong and may distract from the current study.

My recommendation here is to temper the language by clarifying that conclusions apply within the tested framework and context, and to avoid framing results as definitive methodological prescriptions.

Concerning the interest of the readers of ISPRS IJGI, there is a strong alignment with geomatics, GIS-based health analysis, and spatial data quality. The focus on MAUP and custom spatial units is well- suited to the journal audience. So, readers interested in spatial epidemiology, territorial indicators, and climate-health interactions will find the manuscript relevant.

My recommendation here is to emphasize geomatics contributions more explicitly in the Abstract or Conclusions (lines 26–28, 394–403), especially the role of spatial unit construction and data harmonization.

Respected Authors,

Overall, your manuscript is scientifically interesting and suitable for publication. Still, it would benefit from clearer methodological justification, more cautious interpretation of Odds Ratios, an improved discussion of limitations and comparability, and finally, minor language and presentation corrections. Addressing these points would significantly strengthen the manuscript’s rigor and impact without requiring major reanalysis.

Comments on the Quality of English Language

The manuscript needs minor language and presentation corrections

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The authors have responded to my concerns and have significantly improved the quality. Howerve, the authors must revise the writing and formatting of the paper to enhance its readability.

Author Response

We thank the reviewer for the positive evaluation of the revised version of the manuscript. We have further improved the writing and formatting of the paper to enhance its readability. We hope that the new version can now be considered fully compliant with the requirements for publication.