Multilevel analysis concerning the relationship between social vulnerability and the healthy use of leisure time in children and adolescents in Argentina : A national population-based study

Leisure time is a human right and has to be considered part Background: of any health promotion initiative aimed at children and adolescents. The objective of this study was to analyze the relationship between social vulnerability and the healthy use of leisure time in children and adolescents in urban contexts of Argentina, in 2012. A cross-sectional and analytical study using data from the Methods: Module on Activities of Girls, Boys and Adolescents of the Annual Urban Household Survey was carried out. In this survey, a self-administered instrument was applied to 25,915 individuals aged from 5 to 17. A Social Vulnerability Index (SVI) was developed. Association was estimated by multilevel logistic regression. Children and adolescents use most of their leisure time to carry Results: out school activities (90.1%) with art activities having the lowest percentage (21.8%). In the multilevel models on the relationship between a high SVI and non-performance of socialization activities, the OR was 1.99 (p = 0.002, 95% CI: 1.28-3.12). The association between high SVI and non-use of ICT gave an OR of 14.17 (p < 0.001, 95% CI: 5.13-39.17), and between high SVI and non-use of internet, an OR of 21.89 (p < 0.001, 95% CI: 7.50-63.88). A high SVI negatively impacts on some healthy activities of Conclusions: leisure time for children and adolescents in Argentina. The SVI could be a useful tool to guide health promotion initiatives in this population. 1 2,3 4 5


Introduction
Leisure time allows children and adolescents to participate in a diverse range of activities that contribute to developing their identity, improve their self-regulation and express their interests. Several studies show that extracurricular activities benefit the positive development of the child-adolescent population [1][2][3] .
In recent decades, Argentina has made notable progress in expanding the rights of children and adolescents through the adoption and promulgation of various laws, and the adoption, and ratification of the Rights of the Child. Despite the progress made in the legal and institutional framework in 2011-2012, it was observed that 30.1% of children in Argentina were affected by multidimensional poverty 4 .
Social vulnerability in children and adolescents is a central element in the definition of social protection policies 5 that seek to improve the quality of life in children and adolescents 6 and promote healthy free-time uses 7 . Today, leisure time is considered a right. This is claimed by the Convention on the Rights of the Child 8 , in which it is understood as a time for rest and leisure 7 , and it is considered a necessary element in the approach to health promotion 9 ; but it is also one of the main social determinants of health 10 .
Previous studies indicate that the use of leisure time is related to social vulnerability in children and adolescents [11][12][13] . In these, it is observed that as the social stratum diminishes, the low performance of extracurricular recreational activities (sports, art and cultural) increases and the probability of not having access to this type of incentives increases 14 . The social inequality gap in socialization opportunities is significant and clearly regressive for children and adolescents living in poverty in Argentina 14 . On the other hand, gender socialization may be an important mechanism to both understand and counteract observed differences in participation in free-time activities among genders 12,14,15 .
The scarce evidence documented in the field explored in the Argentina and Latin America highlights the importance of the production of knowledge of the impact of social vulnerability on the healthy use of leisure time in children and adolescents in order to contribute to the quality of life and health of this group from a holistic and integral perspective.
The main objective of this study was to analyze the relationship between social vulnerability and the healthy use of leisure time of children and adolescents in urban contexts in Argentina in 2012.

Study design and sample
This was a cross-sectional, analytical study of a secondary database of the Module on Activities of Girls, Boys and Adolescents (MANNyA) 16 that was included as part of the Annual Urban Household Survey (EAHU, by its Spanish initials) during 2012. The MANNyA was carried out on the basis of an inter-institutional initiative between the Ministry of Labor, Employment and Social Security, the National Secretariat for Children, Youth and Family, the National Commission for the Eradication of Child Labor, the National Institute of Statistics and Censuses (INDEC, by its Spanish initials) and the Provincial Statistical Offices.
The sample consisted of children and adolescents aged 5 to 17 from urban conglomerates in Argentina. It was a probabilistic, stratified and multi-stage sample of 34,487 households, with a total of 25,915 children and adolescents surveyed 16 .

Data collection instruments
The data from the MANNyA secondary database 16 , the basis of this analysis, was collected using a self-administered instrument. The databases of MANNyA 2012 are public and can be obtained in the web site of INDEC.
The instrument included closed-ended questions aimed at obtaining what children and adolescents usually do in their leisure time.
Result variable "Healthy use of leisure time", defined as the time for rest and leisure activities, play and recreational activities suitable for the age. It also implies the right to participate freely in cultural life and the arts 7 . In the framework of the analyzed module, leisure time activities are defined as everything that the child does before or after school, or on the weekends 16 .
The dimensions of analysis for this study were: -School activities (homework or studying for school); -Sports/physical recreation (football, swimming, cycling/ horseback riding, etc.); -Art (painting, theatre, music, dance, language or art-related workshop or course); -Socialization (going out with friends to the cinema, to the square, to the cyber cafe, etc.); -Use of Information and Communication Technologies (ICT; computer or netbook); and -Internet use Explanatory variable "Social vulnerability", defined as the situations of insecurity and defenselessness experienced by communities, families, and individuals in their livelihood conditions, as a consequence of the impact caused by any socio-economic event. In addition, the management of resources and the strategies used by them to cope with the effects of this event were considered 17 . Today, social vulnerability is considered one of the main social determinants of health 10 , where most health problems can be attributed to people's socio-economic conditions.
For the construction of this variable, a Social Vulnerability Index (SVI) was prepared (Table 1) based on the data from the MANNyA secondary database 16 . The dimensions and weighting values of the conceptualization of social vulnerability were based on previous publications 17, 18 .
The categories included material assets, such as employment and housing, and non-material assets, such as those related to human capital (access to the health system and educational system of the head of household).
The construction of the SVI was based on the selection of dimensions represented by different categories which, depending on the risk situation, were defined as «moderate» or «critical».
Since each of the selected categories may have different levels of intensity, it was decided to define differential weights within them 18 .
Next, the SVI of each household where the child or adolescent is living was categorized as follows: without SVI was assigned to those cases with a value of 0.00; low or moderate SVI was assigned to those cases with values between >0.00 and ≤0.45; high SVI was assigned to those cases with values greater than 0.45.

Statistical analysis
The variables of interest were analyzed descriptively using position and dispersion measurements and frequency distribution.
Given the hierarchical structure of the data (individuals grouped in regions of the country), a multilevel logistic regression analysis was conducted to explore the relationship between the social vulnerability and the healthy use of leisure time. The models considered socio-economic and socio-demographic variables at two levels: an individual level, relating to the child/ adolescent (performance of domestic and economic activities, gender, age, school attendance, health coverage, illiteracy), and a household level (level of education of the head of household). An empty model (Model 0) was performed, and the level 2 variance (household) and the intraclass correlation coefficient (ICC) were calculated. Education -Households whose head has an average of <7 years of schooling -Households whose head has an average of >7 and <12 years of schooling 0.25 0.10 * It represents the ratio between the total number of people in the household and the total number of rooms or spaces available in the household (excluding bathroom(s) and kitchen(s) (Instituto Nacional de Estadísticas y Censos). **The predominant materials of the constituent components of the dwelling (floors, walls, and roofs) are assessed and categorized in relation to their strength, resistance, and capacity for thermal, waterproof, and sound insulation. The dwellings are classified as: Calmat I and II: the dwelling was made from resistant and solid materials used in floors and roofs, although it may/ may not incorporate insulating or finishing elements in at least one of these. Mosaic/tile/wood/ceramic/carpet floors were considered. The roof had a membrane/asphalt singles and did have a ceiling/interior cladding.
Calmat III: the dwelling was made from resistant and solid materials used in floors and roofs, but it lacks elements of insulation or finishing in all of these. Cement floor/fixed brick floor was considered. Roof with tile/slab roof without any cover, slate/clay tile, and no ceiling/interior cladding.
Calmat IV and V: the dwelling was made of non-resistant materials in at least one of the constituent components. Loose brick/ earth floor and roofs of sheet metal without cover, fiber cement/plastic sheet, corrugated roofing sheet, cane/board/straw with mud/ straw alone were considered and there was no ceiling/interior cladding. ***In the weighting structure, greater importance has been given to the occupation dimension (dependency burden of income earners), given that the association of the population with the labor market becomes a key factor in social vulnerability, and monetary income can change the situation of social inclusion/exclusion more immediately.
Firstly, a univariate analysis was carried out and, from the variables that resulted in statistical significance, a multivariate model of random intersection was constructed by introducing one variable at a time. The standard errors were calculated taking into account the cluster effect of each region of the country. The variable-addition models were compared using the likelihood ratio test. The proportion of the variance at level 2 explained (PVE) by the different models was calculated as PVE= (V) × 100.
The statistical package Stata® v14.2 (Stata Corporation, College Station, Texas, USA) was used for all the analyses, and a p<0.05 value was considered statistically significant.

Ethical considerations
The study was carried out on the basis of a secondary analysis of the MANNyA database 16  The database is currently public and open-access, and is de-identified by the responsible public body 19 . Thus, this study did not require an evaluation by an ethics committee and, in addition, qualifies for the status of being exempt from obtaining informed consent.

Descriptive phase
As shown in Table 2, the sample consisted of 52.6% males and 63.5% adolescents. In total, 32.2% of the respondents were from Gran Buenos Aires and 30.9% from the Pampas region. It is also observed that 95.5% were attending school at the time of the survey and 58.1% had some kind of health coverage.  Among those with a low or moderate SVI (87.1%), the distribution of the performance of leisure time activities was the same as that     In the group composed of those with a high SVI (2.1%), the distribution of the performance of leisure time activities changes as compared to the other two groups of SVI. While school activities (83.1%) and art activities (11.4%) continues to be the most and least popular choice, respectively, sports/ recreational activities are in second place (48.1%), replacing ICTs, which have moved into third place (45.1%). In turn, socialization activities (36.3%) were more prevalent than Internet use (33.7%).

Analytical phase
In the multilevel models, statistically significant associations between some leisure time use activities and social vulnerability are seen (Table 5).
In the multilevel model, the non-performance of school activities was statistically associated with low/moderate SVI, with an OR of 1.165 (p=0.023).
In analyzing the relationship between high SVI and the risk of not performing sports/recreational activities, it was observed that the OR was 1.292, although this association was not significant (p=0.088).
With respect to the relationship between high SVI and the non-performance of art activities, an OR of 2.232 was found; that is, a higher risk of not performing these activities, being statistically significant (p<0.001). This was similarly reflected in the relationship between SVI in the highest stratum and the socialization activities (OR 1.999 and p=0.002).
With respect to the relationship between high SVI and the non-use of ICTs, a statistically significant association was observed, in which the OR was 14.171; that is, high SVI resulted to be a risk factor (p<0.001). On the other hand, it could be observed that the non-use of the Internet during leisure time in the group with high SVI yielded an OR of 21.887, being significant (p<0.001).
Based on the aforementioned, in the final models (Model 3), which considered individual and contextual variables, a clear upward gradient in the likelihood of unhealthy use of leisure time is observed when there are increased SVI levels in the cases of art, socialization, ICT, and the Internet.

Discussion
Our results highlighted the impact of a high social vulnerability index on the reduced performance of some healthy leisure time activities such as art, socialization and Internet use, among children and adolescents in Argentina.
A new look at the concepts of leisure time in Latin America has to be constructed, rescuing the profound knowledge and practices already elaborated in Latin American territories 20 . However, reflection on this relationship requires various perspectives and levels of analysis. In this sense, it is necessary to highlight the influence of the public social protection policies in guaranteeing the right to leisure time in childhood and adolescence.
Some risk factors from the social sphere where biographies of children and adolescents are developed may have an impact on the ways in which they perform in society. The fact that a significant part of the risk to their health and quality of life occurs in this context was highlighted in the field of social epidemiology 21 .
As Feito 22 suggests that vulnerability has a dimension of susceptibility to harm, conditioned by intrinsic and extrinsic factors, anchored in the radical fragility of the human being, but undoubtedly largely attributable to social and environmental elements. In this sense, our study showed that a significant part of the population under analysis had some degree of social vulnerability, which makes us consider this indicator as one social determinant of health 10 .
Regarding the healthy use of leisure time in the population under analysis, it was observed that some activities were more prevalent than others; generally, art, sports/recreational, and socialization activities were less frequent. This information is important because, as other authors have shown 23,24 , these activities should be proposed as strategies to promote the health of children and adolescents at social risk 25 , since they contribute to the psychosocial and physical state of children and adolescents, while at the same time promoting life skills. This fact is relevant since they could also be the activities to be promoted through different public health policies and programs in the country.
In addition, some activities such as sports and recreation were performed only by half of the sample. This leads to the need to promote these activities in order to have a positive impact on some aspects related to the physical and mental health and nutrition of children and adolescents in a complex epidemiological context, characterized by the increase in chronic diseases, low physical activity, and malnutrition 26,27 .
Another aspect to highlight is the high prevalence of school activities during leisure time. This can be explained by the implementation of public policies of great impact in the years prior to this survey, such as the Universal Child Allowance (AUH), in force since 2009 28 and considered a key element for staying at school.
In this sense, the debate on the psychosocial development in childhood and adolescence should be focused not only on school activities, but also on the possibility of choosing and performing multiple leisure time activities freely; activities that entail an enjoyment linked to them. In this sense, we agree with what Fredriksson and colleagues stated in a recent publication, in which they suggest that increasing the participation of young people in leisure activities, especially those from more socially vulnerable environments, can help to reduce social inequalities in health. In this sense, it is necessary to promote varied activities, both structured and unstructured, during free time 29 . Another aspect to consider is that of gender differences in the performance of activities during leisure time. For example, it is observed that "sexual division in the activities between men and women is already established in childhood and adolescence" 30 . Our results showed that men tended to do more sports activities and women more art activities.
With respect to some social determinants of health, such as literacy or school attendance, in all the activities the prevalence was found to be higher in the low and without SVI subgroups, highlighting the weight of these determinants. In this sense, the ecological study conducted by Viner et al. on the health of adolescents suggested that the most effective health interventions are probably those addressing structural changes, e.g. access to education 31 .
Furthermore, it should be noted that the use of digital-free time, mainly focused on the use of the Internet and ICTs, has become increasingly present in the activities of children and adolescents and has become an agent of socialization 32 .
However, our results show that the use of the Internet is significantly conditioned by social vulnerability. In this sense, it was observed that, in the group with the highest level of vulnerability, there is a 14 times higher risk of not using the Internet. This is relevant and shows how the condition of vulnerability becomes a determinant barrier when accessing certain goods and services, especially if progress is to be made in reducing digital divides 33 .
Our findings on the use of ICTs were similarly, albeit less strongly, reflected in the use of the Internet; that is, it was observed that there was a risk of not using ICTs in the groups with high levels of social vulnerability. It should be noted that two years before this survey, Argentina implemented the "Connecting Equality" policy 34 , a federal broad-scope program. This is important because the effectiveness of the digital divide reduction policies, especially among the most vulnerable groups, could present certain barriers to access 35 .
There were also marked differences with respect to art and socialization practices, where it was found that the higher the social vulnerability index, the higher the risk of not performing these activities. However, this result highlights the importance of promoting this type of activity in children and adolescents, since, according to the evidence and based on the findings by Wald 36 , the changes perceived as a result of participating in art workshops are closely linked to feelings of well-being, the development of personal capacities and the strengthening of group relations, which may be linked in a broad sense to the paradigm of health promotion.
Furthermore, the subjective processes of socialization in human health acquires a central role. In this line, the World Health Organization 37 hierarchized and gave relevance to these aspects within the concept of quality of life. Several points can be highlighted in our results and our contributions to the knowledge of the situation of leisure time use in children and adolescents.
Firstly, it was found that in 2012 there was a social inequality gap, as measured by the SVI. On this point, and going back to the theoretical framework that underpinned this proposal, it is necessary to reflect on the weakening of social networks, unfavorable economic scenarios and the differential impact of targeted social protection policies at national level 38 .
Secondly, the negative effect of social vulnerability on some specific activities was highlighted, such as those related to art, socialization processes and access to goods and services such as the Internet. From here, it is possible to postulate that this would not be fully contributing to the achievement of the Rights of the Child 8 , which claim the free participation in cultural life and the arts 7 .
Thirdly, it was observed that, in a large part of the activities of healthy leisure time use, the presence of social vulnerability conditions its full performance; in this sense, it could be believed that these activities are hampered or compete with other activities e.g. those that are domestic and economic 30 .
In a study carried out in Ecuador, Mexico and Peru, CEPAL and UNICEF stressed that work, both paid and unpaid, is another activity that occupies an important part of adolescents' time; this reality is not in line with the fundamental rights of this group 30 . Thus, we face a double challenge concerning the effective right to the healthy use of leisure time. On the one hand, the aim is to promote leisure and welfare activities for children and adolescents 30 ; sports, art, and socialization activities. On the other hand, it is necessary to consider the digital divide and access to differential goods and services, which could be considered as socialization agents 32 in this group.
Another weakness of the proposal is that the research question was addressed based on the analysis of a secondary data source, so it is possible that some aspects related to the construction of the SVI were left out. However, the methodology for the construction of this index is flexible and there is not a sole theoretical frame of reference. Lastly, another limitation was the data collection date (2012), where some aspects may have been modified so far.
However, this study has many important strengths. Among them, the large sample size stands out. It allowed for multivariate analyses and adjustments by multiple confounders and the national representativeness of the sample, since the survey of the secondary database was coupled to the Annual Urban Household Survey (EAHU) by following the application of a rigorous probabilistic sampling, thus ensuring the accuracy of the data obtained and the scope of the entire urban population of the country 16 .

Conclusions
In conclusion, the presence of social vulnerability has an unfavorable impact on the performance of some healthy leisure time activities, such as art, socialization and use of the Internet in the group of children and adolescents in Argentina. The obtained findings lead us to highlight some strategic moves concerning the field of health social determinants where the SVI could be a useful tool to guide health promotion initiatives in the population of children and adolescents.

Grant information
The author(s) declared that no grants were involved in supporting this work.

General comments:
This manuscript studies the relationship between social vulnerability index and healthy use of leisure time. The manuscript can be considered as a case study, as it focuses only on the Argentinian population. Some information is missing in order to understand better what authors would like to show, and they also need to improve the interpretation of the results obtained at the discussion section. Please, find my specific comments that can help authors.

Specific comments: Abstract:
Even it is known what ICT means, I suggest to add what it means in words.

Introduction:
This section can be considered somehow weak. Concerning schools, it is missing information about how school schedule is in Argentina, for example, how long children are staying at school daily; is it the same schedule for primary, secondary and high school? Do they have all afternoon free, or some of them? Do they have more than one turn (morning turn and afternoon turn? It could be interesting to know about public/private schools, because sometimes their schedule is different, and if some activities are included into the daily schedule. In fact, the survey asks about the timetable they have at school, but has not been considered. Updated data from other countries (not only Latin American) are missing, just to compare and discuss later about them. It could be of interest to conceptualize what health is, and how leisure time can be healthy, not only physically (aerobic, strength, body composition, etc), but also from a functional, psychological and social point of view.

Methods:
It could be interesting to explain more about how data collection was done.
Searching the original survey [spanish] it is understood that it is not intended to know about leisure time (only one block of questions, of a total of 28, asked about it), but to know about their working activities. Perhaps, because of that, this information is not enough to interpret when analysing data. It is suggested, as well, to annex the survey, even it is in Spanish.
If it is a self-administered survey, all the questions were the same for all the children and adolescents (5-17 years old)? If there were the same, for a 5 year old child it can be difficult to answer some (or all) the questions, so it is needed to know who helped them to answer. Moreover, authors presented a number of participants that did not know how to read and write, so who answered? In fact, the item 29 of the survey asks about the adult's participation when answering, but, was a parent, older brother/sister, relative, baby sit…? Was there an age cut-off point considered when receiving support for answering? Did the survey taker participate, as well? All these aspects make the methods confusing for getting the data. Was it really a selfadministered survey or there was a survey taker? As the data studied were obtained from the 2012, co-authors should state if there was or not a later survey that explores leisure time.  (Table 3). Why Region has been included, since the group of variables shown here are individual variables? See comment referred on item 3.6. (statistical analysis). In Table 3, there is no mention related to weighted data, why? 4.3. Table 5: a) In general, odds ratios showed higher magnitudes than expected. Probably, as referred above, it is a result of using odds ratios instead of prevalence ratios. Moreover, the large confidence intervals denote lack of precision, due to the strata dilution (high SVI category). On the other hand, multicollinearity between SVI and other contextual variables shown on the footnote may contribute to that. It might be a good strategy to join moderate + high strata as one. b) ICC values are highest than expected and they increase from model 0 to model 3. How does it is plausible, since after adjusting by individual variables, ICC should decrease because the influence of the last one variables? c) By the way, those contextual variables has not been described in the sections 'Methods' and/or descriptive 'Results', as well as some individual variables like household activities of the child/adolescents, etc… d) One strategy to evaluate multicollinearity is to run matrix correlations among the independent variables. e) It is recommended to add the model quality adjusted indexes (Hosmer & Lemeshow χ 2 in the case of logistic regression models or likelihood ratios if prevalence ratios are used).

Discussion:
Most paragraphs on this Section are not in consonance with the study purposes and results. Some of them are only related to the descriptive analysis (Ex: paragraphs 6 e 7). There were neither specific discussion related to the multilevel analysis nor comparison between such results and those found in the literature. Some paragraphs are extremely far from the supposed aims of the present study (Ex:, paragraphs 2, 5, 8, 10,11, 12). In paragraph 4 the authors reported: … "in this sense, our study showed that a significant part of the population under analysis, had some degree of social vulnerability, which makes us consider this indicator as one social determinant of health". Based on the text highlighted above, it should be emphasized that just because most of the population (around 90%) were classified as moderate or high SVI, this variable probably does not keep good discrimination power, so that it should not be considered as a good determinant of health. There is a statement that is not found in the respective Table 5 (in paragraph 12, the authors reported an odds ratio = 14. This value was not found in Table 5, and its interpretation is misleading, since odds ratio is not risk). Strengths and limitations reported in paragraphs 21 and 22 were poorly described. The main limitation is not the study base (secondary) but mainly related to the representatives what was not reported by authors (missing data in the sample process could lead to selection bias), and the 'self-administered instrument' without mention about who replied the questions (parents, children, adolescents) could lead to measurement bias. No statement related to potential selection or measurement bias was reported by authors. The conclusions are equally misleading. Taking into consideration the scope of this section, the authors did not present a consistent discussion, and, so, the section should be completely rewritten.

References:
As already reported by Reviewers, no relevant references related to the analytical phase (multilevel analysis) were found. Among 38, 15 References are related to International Institutions, or similar. There should be specific references related to the study aims and the main findings.

If applicable, is the statistical analysis and its interpretation appropriate? Partly
Are all the source data underlying the results available to ensure full reproducibility?

Are the conclusions drawn adequately supported by the results? No
Competing Interests: No competing interests were disclosed.

Reviewer Expertise: Epidemiology; Statistical methods applied to Epidemiology
We confirm that we have read this submission and believe that we have an appropriate level of expertise to confirm that it is of an acceptable scientific standard, however we have significant reservations, as outlined above. Our thanks to you and to the reviewers for the extremely helpful comments that have been provided to improve our paper. We've addressed all the reviewers' as well as the editorial comments. They are at the end of this letter.

Comments on this article
Best regards, The Authors. The manuscript can be considered as a case study, as it focuses only on the Argentinian population. Some information is missing in order to understand better what authors would like to show, and they also need to improve the interpretation of the results obtained at the discussion section. Please, find my specific comments that can help authors.
Answer: Thank you for the observation. If it is a self-administered survey, all the questions were the same for all the children and adolescents (5-17 years old)? If there were the same, for a 5 year old child it can be difficult to answer some (or all) the questions, so it is needed to know who helped them to answer. Moreover, authors presented a number of participants that did not know how to read and write, so who answered? In fact, the item 29 of the survey asks about the adult's participation when answering, but, was a parent, older brother/sister, relative, baby sit…? Was there an age cut-off point considered when receiving support for answering? Did the survey taker participate, as well? All these aspects make the methods confusing for getting the data. Was it really a self administered survey or there was a survey taker? Answer: The data collection was made by a trained survey taker. We added more information in the data collection instruments section.

General Comments:
The study aims to investigate the association between Social Vulnerability Index (SVI) and healthy use of leisure time (outcome). The authors applied multilevel analysis including in the models SVI and individual variables potentially related to the outcome. In general, the manuscript is reasonably well written (exception the section Discussion), and some specific questions, listed below, claim to be replied. Answer: Thank you for the observations. What References related to multilevel analysis were used? Answer: All proposed changes were considered; new multilevel models are generated and more information is added in the data analysis section described below: Following the recommendation of the reviewer; the IVS variable recategorized, considering new cut-off points where moderate and high stratum joined. Then the models were made, and it was observed that the OR decreased and the CI were reduced, so it was decided to leave OR in all analyzes and not reasons of prevalence. In these new multilevel models, two levels are left: individual and contextual level. For the second level, the variable of "Decile group of total family income in the region" was used, which is explained in the methods section. Although models tested taking the third level of the region, the performance of said models (ICC) was not adequate and did not add a further explanation to the phenomenon studied. So, we decided not to include the third level.
The ICC values were added in all the proposed models, taking only a second level. As described in the previous paragraph.
In the method section, the formula PVE = V x 100 is further developed, and references are added. 4. Results: 4.1.  (Table 3). Why Region has been included, since the group of variables shown here are individual variables? See comment referred on item 3.6. (statistical analysis). In Table 3, there is no mention related to weighted data, why? Answer: The "SE" is eliminated in tables 3 and 4. The results of table 3 are reviewed (can read and write). It is added in the table footer that the data is weighted. 4.3. Table 5: a) In general, odds ratios showed higher magnitudes than expected. Probably, as referred above, it is a result of using odds ratios instead of prevalence ratios. Moreover, the large confidence intervals denote lack of precision, due to the strata dilution (high SVI category). On the other hand, multicollinearity between SVI and other contextual variables shown on the footnote may contribute to that. It might be a good strategy to join moderate + high strata as one. b) ICC values are highest than expected and they increase from model 0 to model 3. How does it is plausible, since after adjusting by individual variables, ICC should decrease because the influence of the last one variables? c) By the way, those contextual variables has not been described in the sections 'Methods' and/or descriptive 'Results', as well as some individual variables like household activities of the child/adolescents, etc… d) One strategy to evaluate multicollinearity is to run matrix correlations among the independent variables. e) It is recommended to add the model quality adjusted indexes (Hosmer & Lemeshow χ in the case of logistic regression models or likelihood ratios if prevalence ratios are used).
Answer: All proposed changes were considered; new multilevel models are generated, and more information is added in the data analysis section described below: Following the reviewers' recommendations, the IVS variable recategorized, considering new cut-off points where the moderate and high stratum joined. Then the models were carried out, and it observed that the OR decreased and the CI reduced, so we decided to leave OR in all analyzes and not reasons of prevalence. Multicollinearity was analyzed, and we decided to remove from the individual variables: "can read and write" and "child's health coverage". The new variable of the individual level "selfconsumption activities" was added, which is explained in the methods section. The ICC recalculated in all the models, where it observed that they are reduced, and are lower values concerning the models previously presented. The ICC values were added in all the proposed models (Table 5). Contextual and individual variables were included in methods. The quality-adjusted index of the models were added. We have performed AIC-BIC statistics to judge the goodness of fit shown in each table (Table 5). In these new multilevel models, two levels are left: individual and contextual level. For the second level, the variable of "Decile group of total family income in the region" was used, which is explained in the methods section. Although models tested taking the third level of the region, the performance of said models (ICC) was not adequate and did not add a further explanation to the phenomenon studied. So, we decided not to include the third level.

We have added a PCV explanation that stands for Proportional Change in Variance (V N-1 -V N-2)/ V N-1) where VN-1 is the DGHIR (cluster) variance in the empty model and VN-2 is the DGHIR variance in the model including individual characteristics.
In the method section, the formula PVE = V x 100 is further developed and references added. 5. Discussion: Most paragraphs on this Section are not in consonance with the study purposes and results. Some of them are only related to the descriptive analysis (Ex: paragraphs 6 e 7). There were neither specific discussion related to the multilevel analysis nor comparison between such results and those found in the literature. Some paragraphs are extremely far from the supposed aims of the present study (Ex:, paragraphs 2, 5, 8, 10,11, 12). In paragraph 4 the authors reported: … "in this sense, our study showed that a significant part of the population under analysis, had some degree of social vulnerability, which makes us consider this indicator as one social determinant of health". Based on the text highlighted above, it should be emphasized that just because most of the population (around 90%) were classified as moderate or high SVI, this variable probably does not keep good discrimination power, so that it should moderate or high SVI, this variable probably does not keep good discrimination power, so that it should not be considered as a good determinant of health. There is a statement that is not found in the respective Table 5 (in paragraph 12, the authors reported an odds ratio = 14. This value was not found in Table 5, and its interpretation is misleading, since odds ratio is not risk). Strengths and limitations reported in paragraphs 21 and 22 were poorly described. The main limitation is not the study base (secondary) but mainly related to the representatives what was not reported by authors (missing data in the sample process could lead to selection bias), and the 'self-administered instrument' without mention about who replied the questions (parents, children, adolescents) could lead to measurement bias. No statement related to potential selection or measurement bias was reported by authors. The conclusions are equally misleading. Taking into consideration the scope of this section, the authors did not present a consistent discussion, and, so, the section should be completely re-written.
The benefits of publishing with F1000Research: Your article is published within days, with no editorial bias • You can publish traditional articles, null/negative results, case reports, data notes and more • The peer review process is transparent and collaborative • Your article is indexed in PubMed after passing peer review • Dedicated customer support at every stage • For pre-submission enquiries, contact research@f1000.com