Author's response to reviews Title:A multilevel analysis to explain self-reported adverse health effects and adaptation to heat: A cross-sectional survey in the deprived areas of 9 Canadian cities Authors: Diane Bélanger (diane.belanger@ete.inrs.ca) Belkacem Abdous (belkacem.abdous@fmed.ulaval.ca) Pierre Valois (pierre.valois@fse.ulaval.ca) Pierre Gosselin (pierre.gosselin@inspq.qc.ca) Elhadji A Laouan Sidi (Elhadji.Anassour-Laouan-Sidi@crchudequebec.ulaval.ca) Version:4Date:1 December 2015 Author's response to reviews: see over
Reviewer 1 Received Nov. 02, 2015 Modifications I am not convinced by the statements made about the perceptual approach to exposure assessment. Relatively unobtrusive objective measures/estimates of heat exposure (as well as internal heat generation for exertional heat stress, and heat strain) are currently available. A stronger argument needs to be made for why objective exposure assessment was not considered as part of this study. The authors did not reply to this questions. Even if they could not assess healt/exposure assessment they have to discuss potential biases realted to missclassification. The reviewer may not be convinced, but the argument we presented is real. Moreover, this is not a grant proposal review, the survey is completed. We may add that in countries such as Canada where hot and humid conditions are rare events, numbering only 10-15 days per summer for the region under study, it can become a logistic nightmare to simultaneously measure temperature and humidity in a high number of dwellings and conduct interviews for 3485 people in 9 cities over 14 random days. There are also several pitfalls with so-called objective exposure assessment, as some authors highlight. We suggest that the reviewer reads those arguments in Ormandy, D. and V. Ezratty, Health and Thermal Comfort: From Who Guidance to Housing Strategies. Energy Policy, 2012. 49: p. 116-121. In short, these are the discrepancies of ambient temperature within a room (vertical and horizontal gradients), at different times of the day, between different rooms of the house and between rooms that are occupied and used for different reasons. With this in mind, they describe the perception approach as a proxy for thermal comfort, which relies on occupant feedback. Studies such as the No changes.
WHO Large Analysis and Review of European Housing and Health Status have explored and used this technique of self-reporting in a survey on housing which included indoor temperature and health-related questions, but no HOBO measurements of indoor temperature. So we don t see the need for any misclassification bias discussion, as it s possibly worse with indoor temperature measurements and no gold standard exists to assess the direction of such a bias. As the reviewer may never be convinced by our detailed arguments (previous round and this one), we ask the editor to make a final decision here. Reviewer 2 Received Nov. 02, 2015 1. In addressing original Major Compulsory Revisions Comment #2, authors cite a previous publication of theirs using the same dataset to confirm validity of the approach presented in the current manuscript. Independently collected data and analysis is required to assess validity of an approach. Additional references should be added to support this statement. 2. Original minor essential comment #3 a, b, and d, were not addressed sufficiently. Further clarification of the methods, in particular MCA methods was not provided. Although a new publication on the same dataset is cited, it is necessary to provide at least details of the dimensions used in the present analysis (and how it is different from published analysis) for the reader to be able to Several references have been added. Our own publications have been deleted. The Analysis section now includes a detailed explanation about the differences between previous papers and this one. We also described the ascending regression procedure. See line 54 See lines 139-140 and lines 164-176 and lines 183-187
understand and interpret presented results. d. Were results from bivariate analyses used to pick variables to include in multivariate analyses? If so, methods should state what bivariate analysis metrics were used to determine inclusion in multivariate model. Variables not included should be noted We disagree with this request, as this is not a course in model building, but rather a paper that should be concise as requested by the journal and previous comments from other reviewers. No change 3. The response for Original minor essential comment #3 c suggests that individual level, building level, and DA level covariates are not independent (e.g. individual-level variables are used to create building level variables). It is unclear why this is justified and how it could result in meaningful/interpretable results. 4. Original minor essential comment #1 An abstract is still not part of manuscript pdf. An abstract in html was found on the website. It is unclear what the main conclusion from the analysis is from this abstract because only numbers of indicators are described. Specific indicators the authors feel are important in differentiating perceived impacts versus perceived adaptation should be highlighted as it relates to development of better surveillance and public health interventions. The statistical explanation and references have been added. As the reviewer didn t receive the abstract (my apologies for BMC Public Health negligence), we paste the conclusion of the new abstract here, as it answers this request about specific indicators of importance: This 3-level analysis shows the differential importance of the characteristics of residents, buildings and their surroundings on self-reported adverse health impacts and on adaptation (other than air conditioning) under hot and humid summer conditions. It also identifies indicators specific to impacts or adaptation. People with negative health impacts from heat rely more on adaptation strategies while low physical activity and good dwelling/building insulation lead to lower adaptation. See lines 196-199 Conclusion of abstract improved (pasted in column beside).
5. Based on several references to previous publications using this same dataset, authors should detail the specific findings from this multilevel analysis that are different from the previously published manuscripts. For example, were different indicators found when using the multi-level approach versus the original multivariate and weighted logistic regression model in http://www.mdpi.com/1660-4601/11/11/11028? How are variables using in the adaptation index treated differently in this publication versus http://link.springer.com/article/10.1007 %2Fs10584-015-1420-4? These distinctions should be highlighted in Discussion and abstract. Quality of written English:Needs some language corrections before being published Better neighbourhood walkability favors adaptations other than air conditioning. Thus, adaptation to heat in these neighbourhoods seems reactive rather than preventative. These first multilevel insights pave the way for the development of a theoretical framework of the process from heat exposure to impacts and adaptation for research, surveillance and public health interventions at all relevant levels. We have added a paragraph in the analysis section to detail the differences between the multilevel analysis used in the present article and GEE used in two of the three articles published with the same database. That said, the specific contribution of the buildinglevel and the DA-level in the present article is already reported in the first paragraphs of the discussion. We have now added some precisions in the text for the other differences between the articles already published and the present one The article has already been reviewed by a native English speaker (Canada). lines 164-176 Lines 321-339 Lines 377-380 Lines 398-400 Line 410