Evaluating the health effect of a Social Housing programme, Minha Casa Minha Vida, using the 100 million Brazilian Cohort: a natural experiment study protocol

Introduction Social housing programmes have been shown to influence health, but their effects on cardiovascular mortality and incidence of infectious diseases, such as leprosy and tuberculosis, are unknown. We will use individual administrative data to evaluate the effect of the Brazilian housing programme Minha Casa Minha Vida (MCMV) on cardiovascular disease (CVD) mortality and incidence of leprosy and tuberculosis. Methods and analysis We will link the baseline of the 100 Million Brazilian Cohort (2001–2015), which includes information on socioeconomic and demographic variables, to the MCMV (2009–2015), CVD mortality (2007–2015), leprosy (2007–2015) and tuberculosis (2007–2015) registries. We will define our exposed population as individuals who signed the contract to receive a house from MCMV, and our non-exposed group will be comparable individuals within the cohort who have not signed a contract for a house at that time. We will estimate the effect of MCMV on health outcomes using different propensity score approaches to control for observed confounders. Follow-up time of individuals will begin at the date of exposure ascertainment and will end at the time a specific outcome occurs, date of death or end of follow-up (31 December 2015). In addition, we will conduct stratified analyses by the follow-up time, age group, race/ethnicity, gender and socioeconomic position. Ethics and dissemination The study was approved by the ethic committees from Instituto Gonçalo Muniz-Oswaldo Cruz Foundation and University of Glasgow Medical, Veterinary and Life Sciences College. Data analysis will be carried out using an anonymised dataset, accessed by researchers in a secure computational environment according to the Centre for Integration of Data and Health Knowledge procedures. Study findings will be published in high quality peer-reviewed research journals and will also be disseminated to policy makers through stakeholder events and policy briefs.


GENERAL COMMENTS
This is an interesting paper outlining a protocol for a study on the impact of a social housing intervention in Brazil. The study uses secondary data analysis of the novel merger of datasets from a large cohort study, a major social housing intervention and health datasets available in the country. The data analysis protocol appears robust, with appropriate plans for the identification of comparable individuals using appropriate data on location and socioeconomic status.
Here you are claiming to be the "first study to evaluate the effect of a social housing programme on health outcomes at national level". I don't dispute that what you are planning is important and that it is novel, but I wonder about this statement. There is work using the SAIL database in Wales on certain interventions that have been made on housingsome of which largely fall in the area of improving social housinge.g. the Housing Regeneration and Health Study. Admittedly this study focuses on improving existing social housing as the intervention, rather than the creation of new housing. There are also studies in Australia (e.g. Bentley et al, 2018) on mental health and social housing. A systematic review of the literature from 1887 to 2007 shows some studies that cover the provision of new housing (Thomson et al, 2009). While what you are doing is a major advance, I wonder if you might focus on scale and the low-to mid-income focus in terms of the novelty of the study on lines 10-16.
2. You suggest there is a lack of patient and public involvement in the research (Page 5, lines 42-47). I am surprised you are not including any public involvement in the research processfor dissemination if not for research design. Do your stakeholders mentioned in line 39 of the Ethics summary not include the public or representative groups? If there is truly no public engagement then it might be worth putting something in lines 42-47 on why this is (perhaps funding constraints?) and it might be worth thinking of attempting to obtain further funding to support such engagement. It is all too easy to analyse data on households without engaging with the people in them, which can give a richer understanding of the reasons why the relationships are as they come out or of what questions could be posed in the research. 3. P7 -I am wondering if you might consider increasing the consideration of potential short term negative impacts of the programmee.g. if relocation leads to increased stress, fragmentation of social networks (which you do mention briefly) and the wider consequences for households of such a move. I think you may also be being optimistic that reduced housing costs would lead to disposable income being spent on healthy food.
Minor comments 1. Page 3 Line 16 -meaning of "intersectionality" not clear to me. "Intersectoral" or "interdisciplinary" might be better words.
2. I think you are overstating weakness on longer term outcomes for bullet 5 (Page 3, line 31)to have up to 8 years is not exactly limited (often studies are shorter in their focus than this in terms of follow-up). I would consider rephrasing this as a strength.
3. Page 4 Line 37 -add "and infectious disease, such as leprosy and tuberculosis" 4. Page 5, line 24 insert "and" at end of line individuals in other housing tenures using propensity score approaches. The proposed project will make efficient use of existing cohort and registry data. Importantly, because of the large sample size, 114 million people or 55% of the Brazilian population, the experiences of subgroups defined to capture the complexity of identity (intersectionality) can be explored. There is a lack of information on important behavioural risk factors (e.g. smoking), however, these factors (while correlated with social housing) are unlikely to determine access to the program. Authors acknowledge that the maximum follow-up is 8 years, which is a short time window to examine the chronic pathway under consideration. Pathways are explicitly described in logic models, however, these models (unlike directed acyclic graphs) do not identify sources of confounding, the direction of relationships, mediators and modifiers which would be useful for framing the approach more clearly (particularly if complex sources of heterogeneity are to be explored). Sensible robustness checks and sensitivity analyses are proposed.

This is an interesting paper outlining a protocol for a study on the impact of a social housing intervention in Brazil. The study uses secondary data analysis of the novel merger of datasets from a large cohort study, a major social housing intervention and health datasets available in the country.
The data analysis protocol appears robust, with appropriate plans for the identification of comparable individuals using appropriate data on location and socioeconomic status.
We thank you for your observation. We do not have permission to contact individual MCMV recipients, but we are working with democratically elected public representatives as part of our policy stakeholders. We compromise to disseminate the findings to the public. We hope in the future to undertake qualitative research which will provide a more nuanced understanding of the mechanisms through which any observed impacts operate and if we are able to contact individual recipients, we will also seek to involve them in the research itself too.
As you note, we have been engaging closely with policy stakeholders and the protocol has been devised as a joint effort with the National Housing Secretariat from the Ministry of Regional Development, in order to guarantee that the findings would answer relevant policy questions. CIDACS staff are, in synergy with these key stakeholders and decision-makers, providing the methodological rigor needed to assure sound results. These will, in turn, be able to be incorporated in the National Housing Plan, which is currently under development in Brazil. The National Housing Secretariat from the Ministry of Regional Development will not interfere in the analysis and results from studies planned in this protocol. We updated the information related to this topic in the protocol (Please see section "Methods and analysis/Ethics summary, Line 163-175 Page 5-6): "Patient and public involvement: This research was done without public involvement. Public were not invited to comment on the study design and were not consulted to develop public relevant outcomes or interpret the results, since we use an administrative and deidentified dataset and do not have permission to contact individuals. Study findings will be discussed with managers and specialists from the National Housing Secretariat from the Ministry of Regional Development and the published results will be disseminated to the public through the mass media. This study is a joint effort with the National Housing Secretariat from the Ministry of Regional Development, in order to guarantee that the findings would answer relevant policy questions. CIDACS staff are, in synergy with these key stakeholders and decision-makers, providing the methodological rigor needed to assure sound results. Findings will be incorporated into the National Housing Plan which is currently under development in Brazil. The National Housing Secretariat from the Ministry of Regional Development will not interfere in the analysis and results from studies planned in this protocol." 3. P7 -I am wondering if you might consider increasing the consideration of potential short term negative impacts of the programmee.g. if relocation leads to increased stress, fragmentation of social networks (which you do mention briefly) and the wider consequences for households of such a move. I think you may also be being optimistic that reduced housing costs would lead to disposable income being spent on healthy food.
We acknowledge the possibility of short term negative impacts of the programme on health and wider outcomes. In order to support our initial hypothesis -receiving a new house will improve health and wellness of vulnerable families -, we focus our logic model on possible positive effects of new houses on health, but we mentioned possible negative outcomes related to the loss of social networks (pp6-7) Since the literature does not provide a consensus related to the short-term effect of relocation on health outcomes, we highlight both options in our logic model (Please see section "Methods and analysis/Logic models Line 229-231, page 7). We have now added an additional sentence to highlight these potential negative impacts within the text too: "In addition, reducing the cost of housing also provides greater access to resources which can be spent on healthier food and health care, leading to better control of cardiometabolic risk factors [31,32]. In contrast, it is possible that greater availability of income could lead to greater consumption of unhealthy products (such as ultra-processed food, tobacco and alcohol) which could in turn increase cardiovascular risk (Martins & Monteiro, 2016;Sperandio et al 2017
In line 16 we use "intersectionality" in the sense proposed by Crenshaw (1990), which has been used in studies on social epidemiology to capture how the inequalities of race, gender, and class among others, as well as their interaction (e.g. race x gender x class) influence health outcomes. In the Brazilian context, it is important to consider the intersectionality framework to improve understanding of the distribution of health outcomes, and to analyze health disparities among social groups. The large sample size of our data (114 million people or 55% of the Brazilian population) will allow us to apply this framework and to explore health disparities among social groups, considering race/ethnicity, gender and class, as well as considering the intersection between those multiple social markers of health inequalities (Please see section "Strengths and limitations" Line 84-85, page 3).
2. I think you are overstating weakness on longer term outcomes for bullet 5 (Page 3, line 31)to have up to 8 years is not exactly limited (often studies are shorter in their focus than this in terms of follow-up). I would consider rephrasing this as a strength.
We have rephrased this issue. We understand that 8 years is not exactly limited, but since we are working with chronic diseases, such as CVD mortality, and a neglected disease with a long term of incubation, such as leprosy, we choose to be more cautious in our initial approach, would ultimately require many more DAGs and given that we have pre-defined covariates available which are not time-varying, we feel that DAGs would actually be less useful than logic models to inform the analysis. By drawing on logic models, we are better able to illustrate mechanisms through which the intervention might operate and key moderators. Furthermore, the logic models have informed the development of our analysis plans and we would therefore prefer to retain them in preference of DAGs, since they more accurately reflect the theoretical underpinning of our analyses.