Response to reviewers’ comments

Manuscript ID PONE-D-20-35162 entitled "Food insecurity and mental health of women during COVID-19: Evidence from a developing country"

We thank the reviewers and the editor for their constructive comments. We believe the comments have improved our work. Please find our response to the individual comments below. Where appropriate, we addressed the queries from THE reviewers together, and we have mentioned that in our responses. As we addressed the comments and revised the paper, the line numbers have changed in this revised version of the paper. Therefore, we mentioned page number corresponding to all changes.

Response to the comments from the editor

RESPONSE:

1. PLOS ONE style guide has been followed to format the revised version of the manuscript.

2. The corresponding author has made the declaration of competing interests on behalf of all authors.

3. De-identified data have now been shared for scientific scrutiny.

Reviewer #1

This paper reports the results of a rigorous longitudinal study of women’s mental health and food insecurity status during a COVID lockdown in rural Bangladesh. The topic is obviously very timely and deals with a group that is largely underrepresented in the research on this topic. The data collection and analysis, such as they are, are presented in sufficient detail and appear to have been conducted robustly. My critiques have to do with measurement approach and contextual framing of the paper more generally.

Comment 1. First, the introduction is very brief and provides little context or justification for the study. There is a lot of literature on household management of food insecurity, and particularly how women manage household food shortages, from a wide range of LMIC settings. Drawing on some of that literature would help situate the present study. It might also provide some foundation for a justification (currently missing from the paper) for working exclusively with women. The authors cite some literature around food insecurity and mental health in the discussion (e.g. page 15-16) that might actually be more effective if included in the introduction to help frame and justify the paper—especially if they also add some discussion of the aforementioned literature on women and FI in LMICs.

RESPONSE:

We have now expanded the introduction section including relevant literature on food insecurity, stress and mental health, and knowledge and attitude.

ADDED:

“Such health crises were also observed to usher in food insecurity via disruptions in economic activities. [REF]”(p4)

“In varying degrees, a lockdown was widely adopted to curb the spread of coronavirus. Millions living in developing countries found themselves treading-off between keeping safe from the virus, and poverty and hunger.” [REF](p4)

… “Bangladesh, where this study was conducted, had an extended lockdown, from 26 March to 30 May 2020. [REF] It was estimated that about 52% of the Bangladeshi population was at the risk of becoming poor,and about five million people could fall into poverty due to COVID-19 pandemic. [REF] The economic disruption caused by COVID-19 and the lockdown that followed was likely to impact Bangladesh's food security scenario, with 31% of its population either moderately or severely food insecure even before the pandemic. [REF] Consequently, with the rise of poverty and food insecurity, the health effects, particularly on women and children, such as anemia and prevalence of thinness among women of reproductive age, malnutrition, stunting and wasting of children may potentially follow suit. There has been ample evidence on the influence of food insecurity on the mental health of individuals. In a study that included data from 149 countries, Jones (2017) found that food insecurity has a dose-response relationship with poor mental health status regardless of socioeconomic and demographic characteristics from 11 regions, indicating that the relationship holds across cultural contexts. [REF] Food insecure individuals were found to have higher odds of experiencing adverse psychosocial conditions such as sadness, worry, stress, anger, and lower odds of experiencing positive psychosocial conditions such as enjoyment, feeling well-rested, being treated with respect. In another study that included data from 147 countries, Frongillo et al. (2019) found that prevalence of food insecurity was strongly negatively associated with subjective wellbeing measured using daily experience index with variation among developed and developing countries likely due to other relevant factors. [REF] Thus, food insecurity can potentially have adverse effect not only on individual’s physical health but also on mental health [REF] ” (p4-5)

… “The variation of magnitude by which sex alters the association between food insecurity and mental health status is unclear, with literature demonstrating poorer mental health status among women than men or no effect. [REF] However, the rationale for conducting this study with women has been driven by evidence that shows Bangladeshi women have low nutritional status and poorer mental health than men, particularly in rural areas, [REF] and women’s pivotal role in managing household food shortage in developing countries including Bangladesh. [REF]

As the COVID-19 pandemic set in, information has been abundant, often not evidence-based and spurious, misleading people to adopt harmful practices and trivialise the risks of the disease. [REF] Knowledge and attitude have profound significance in influencing compliance to healthful practice, such as maintaining hand hygiene in developing countries. [REF] Additionally, a lack of knowledge about a specific physical condition or disease can harbour misconceptions and stigma about the disease and hinder help-seeking. Lack of awareness at the community level impede social acceptance, fuel stigma, induce fear in individuals, deter disease disclosure, and seek care. [REF] Paradoxically, knowing about a disease and its potential after effect could induce paralysing fear and denial such an extent that it could interfere with reception of health messages, hinder seeking care and care provisions, and encourage people to disregard the existence of the disease altogether. [REF] Understanding knowledge and attitude, including stigma, is critical in assessing and testing likely patients and maintaining the practice of COVID-19 preventive measures. In Bangladesh, women, particularly in rural areas, lack access to authentic health information [REF] due to their limited access to information in general [REF] and lower access to technology. [REF] Women are predominantly caregivers for children and elderly family members at home. Thus, exploring women's knowledge and attitude regarding COVID-19 can provide important insight into infection prevention and containment in women, children and the elderly in rural Bangladesh.

Mental stress may occur due to emotional or physical stressors. [REF] There are various ways stress and its impact have been conceptualised and studied from epidemiological, psychological, and biological perspectives, employing different methodologies. [REF] Despite these differences, there is substantial evidence that mental stress, through various pathways, can affect individuals mental and physical health, and thus, overall wellbeing [REF] Our study is based on the evidence on food insecurity during economic crises, health crisrs such as epidemics and COVID-19 pandemic, and the evidence on mental health outcomes related to food insecurity. Economic crises and shocks may lead to food insecurity due to supply shortage, food price volatility, and income loss compromising people's purchasing power. [REF] Health crises can also lead to food insecurity as observed during and after previous epidemics such as HIV/AIDS, ebola influenza, scurvy and cassava virus. [REF] The economic corollaries of COVID-19 have had a significant adverse effect on food security scenario, [REF] particularly in the developing countries. [REF] Given the influence of food insecurity on mental health, [REF] it is critical to examine the association between food insecurity and women’s mental health in the context of COVID-19 in Bangladesh where food insecurity is common, and women have poor mental health outcomes. In this study, we present evidence from rural areas of Bangladesh, where we examined the association between food insecurity and women's mental health, that is their perceived stress level, during COVID-19. Taking food insecurity as a potential stressor during the lockdown, we hypothesise that more food insecure women will experience a higher stress level.” (p6-7)

Comment 2. Second, the authors state briefly that the perceived stress scale was adapted for use in Bangladesh, but not so for the FIES. The citations associated with PSS and FIES are the original versions, developed in English using Western populations. Cultural and linguistic adaptation of measures of subjective states (stress, food insecurity) is very important for the integrity of the data collected—so the authors should spend at least a few sentences explaining how these measures were adapted for use in a Bangladeshi population. Who translated the instruments? Were they pilot tested before use? How did the authors ensure that participants understood and could respond fully to each question? Etc?

RESPONSE:

ADDED:

“PSS scale was chosen based on previous study findings and applicability in the context of Bangladesh. However, the previously validated versions were validated and used in special populations, predominantly in urban areas. [REF] FIES was previously used to conduct a multi-country study. [REF] PSS and FIES scales were adapted for this study considering the cultural and linguistic aspects of stress and food insecurity experiences. The adaptation process was completed in three steps. Firstly, the first author (TR) translated the original PSS scale in Bengali. The translated version was then shared with the co-authors (MGH and AI). The first stage of translation was finalised after discussion among the authors. FIES was translated in Bengali following the same steps. Being born and raised in the study region, AI provided critical guidance at this stage about the perception of stress and food insecurity in study locations.

At the second step, AI shared the PSS and FIES with research collaborators based in the study locations. The group of researcher collaborators included both women and men from the communities participating in this study. This group considered each question of PSS and FIES carefully as to what they might mean to the study participants given their sociodemographic characteristics. The group suggested some changes in wording. In the third step, TR prepared second drafts of the questionnaire, incorporating changes suggested in the previous step. Subsequently, the translated scales were finalised in upon discussion among all authors. Pilot testing was not feasible due to lockdown and the time-critical nature of the study. However, the community members' involvement in the scale adaptation process helped prepare the scales for study context, taking into account the cultural perceptions and verbal expressions of stress and food insecurity.” (p9-10)

Comment 3. Third, there is almost no discussion of where COVID knowledge and attitude questions of this study came from. These are interesting questions, many of which deal with culturally important questions of stigma or fate. Who wrote them, and how were they decided?

RESPONSE:

ADDED:

“The knowledge and attitude questions were developed by the first author (TR) consulting the World Health Organization (WHO), [REF] Médecins Sans Frontières (MSF), and the government of Bangladesh [REF] online resources including daily updates and reports and manuals. The questions were finalised and approved after repeated discussion among all authors. Moreover, the authors have experience working in social development and health with rural communities across Bangladesh and thus have relevant insight into knowledge and attitudes around health issues.” (p10)

Comment 4. Fourth, why does this paper report knowledge and attitude levels around COVID, anyway? These don’t seem to be part of the paper’s main concern (food insecurity and mental health during lockdown) and might be better left to another report. A separate paper dealing in detail with knowledge and attitudes toward COVID (especially the stigma elements) would be a good contribution in its own right. If the authors want to keep these data in the present paper, then there should be some discussion of how knowledge and attitude toward COVID might be related to the key variables—mental health and food insecurity—that form the heart of this report. How, for instance, did mental health vary along with COVID knowledge and attitudes? Or food insecurity?

RESPONSE:

We have considered the reviewer’s advice seriously. After careful consideration we decided to keep this part of the evidence in this paper. Following reviwer’s suggestions, however, we have now discussed to link how knowledge and attitudes are related with mental health and food security in introduction, results and discussion sections. We have measured the effect of the level of knowledge and attitude on participants’ stress level and evaluated the effect of wave 2 food security status on the participants' knowledge and attitude level. Please see our response to comment 1.

ADDITIONALLY ADDED:

“Effect of Wave 2 FIES score on KLC-19, ALC-19 and PSS score at the Individual and Village level

At the individual level, the mean value of PSS, KLC-19 and ALC-19 score were 19.39 (SD= ± 4.41), 0.63 (SD= ± 0.16) and 0.75 (SD= ± 0.15) respectively. At the village level, the mean values of the PSS, KLC-19 and ALC-19 score were almost similar to individual level; 19.52 (SD= ± 2.29), 0.63 (SD= ± 0.09) and 0.75 (SD= ± 0.09).

At the individual level, when evaluating the effect of wave 2 FIES score on PSS, KLC-19, and ALC-19, PSS score showed a significant increase of 1.87 (95% CI: 1.61; 2.13; p-value: 0.000) for each unit increase of wave 2 FIES score, indicating that worsening the level of food security status increasing the stress level of the participants. The effect of KLC-19 and ALC-19 scores on PSS score were statistically non-significant. While measuring the effect of wave 2 FIES score on KLC-19 and ALC-19 score, a statistically significant effect was observed between wave 2 FIES score and KLC-19 (coef: -0.03; 95% CI: -0.04; -0.02; p-value: 0.000), indicating that worsening of food security status was negatively associated with the knowledge level of the participants on COVID-19. Findings from the village level analyses were also almost similar to the individual level (Table 4).

Table 4: Effect of wave 2 FIES score on PSS, KLC-19 and ALC-19 scores

List of Scores Individual level Village level

PSS score 1.87 (1.61; 2.13; p=0.000 1.34 (0.46; 2.22); p=0.003

KLC-19 score -0.03 (-0.04; -0.02); p=0.000 -0.03 (-0.06; 0.00); p=0.052

ALC-19 score 0.00 (-0.01; 0.00); p=0.350 -0.01 (-0.04; 0.02); 0.418

Effect of the change in FIES score (from wave 1 to wave 2) on KLC-19, ALC-19 and PSS score at the Individual and Village level

At the individual level, a significant negative association was observed between the PSS score and change in food security status (Coefficient: -1.15, 95% CI: -1.30; -0.99, p-value: <0.001), indicating that worsening the food security status over the pandemic period increasing the stress level of the participants. The effect of the change in FIES score on KLC-19 and ALC-19 were non-significant. The analysis results at the village level were also similar to the individual level analysis (Table 5).

Table 5: Effect of change in FIES score from wave 1 to wave 2 on PSS, KLC-19 and ALC-19 scores

List of Scores Individual level Village level

PSS score -1.15 (-1.31; -0.99); p=0.00 -1.02 (-1.52; -0.53); p=0.00

KLC-19 score 0.00 (0.00; 0.01); p=0.233 0.00 (-0.02; 0.03); p=0.722

ALC-19 score 0.00 (0.01; 0.00); p=0.115 0.00 (-0.02; 0.02); p=0.704

” (p19-20)

Comment 5. Fifth, given all the stresses associated with living through a COVID lockdown, how did the authors account for potential confounding variables in the food insecurity-stress relationship they’ve documented here? That is, how can they be confident that their measures reflect an actual direct effect of food insecurity on stress levels, instead of a mediated effect resulting from other lockdown stressors, such as loss of income (which they briefly mention), concerns about children’s schooling, juggling new roles and dynamics in the home environment when everyone is shut in, possible domestic violence increases, etc (which they do not mention)? There needs to be some specific discussion of the potential confounders not accounted for in this study.

RESPONSE:

The evidence we presented here should be interpreted as correlational and identifying causal effects and magnitude of those causal estimate would require underlying variables to be controlled for, or pre-covid related measures. In the absence of such data, we provide evidence which should be interpreted as association between independent variables of interest and the outcome variables.

ADDED:

“Although we found a statistically significant effect of change in food security on stress levels, we were unable to adjust for some potential confounders such as the burden of household chores during the lockdown, children’s education and domestic violence due to unavailability of data. These factors might have contributed to higher stress level. Collecting this information over the phone was not feasible due to the sensitive nature of the information which might contribute to the deterioration of the situation. We also had to limit the survey time over the phone to not take too much of participants’ time, and hence could not collect many related valuable data despite our best efforts. Therefore, the presence of other stressors as indicated by Frongillo et al. (2019) needs to be considered while interpreting this evidence. [REF]” (p27-29)

Minor editorial requests:

Comment 6. I noted some significant standard English errors throughout, such as missing or sometimes overused definite articles, subject-verb agreement, and odd prepositions. PLOS does not provide copy editing services that will correct these, to my knowledge, so the authors need to do a careful proof-read before resubmission.

RESPONSE:

Proof-reading has been completed. Changed that happened in this process are in track-changes in the unclean version.

Comment 7. Can the authors explain why they chose mixed effects modeling? Just a sentence or two justifying it as part of a strategy to link the present sample’s trends as part of larger statistical data would suffice.

RESPONSE:

Based on the comments from reviewer 2, we have conducted our analysis from both the individual and village level. In the revised version, we have conducted univariable linear regression only. Please see the revised version of the statistical analysis section.

ADDED:

“Participant characteristics were summarised using descriptive statistics. Frequency and percentages were used for dichotomous variables. Mean and standard deviation or median and interquartile range were used for continuous variables, based on the distribution of the data. A descriptive analysis, from both the individual and village level, knowledge and attitude level and PSS scores were also performed using similar descriptive statistics.

To measure the effect of food security status (measured via FIES score during wave 2) on participant’s stress level (measured via PSS score during wave 2), a univariable linear regression model was used. Two separate univariable linear regression models were used to measure the effect of knowledge and attitude level on COVID-19 (measured during wave 2) on PSS score. The effect of wave 2 food security status on participants' knowledge and attitude level of participants were also measured using two univariable linear regression models. Finally, to measure the effect of change in food security status (measured via subtracting the wave 2 FIES score from wave 1 FIES score) on the PSS, knowledge and attitude scales, three separate univariable linear regression models were used. All the regression models were conducted using both the individual and village level data using robust standard error. All statistical analyses were conducted using Stata 14.0 version, and p-value <0.05 was considered as statistically significant level.” (p11-12)

Comment 8. Provide international monetary conversion for BDT when discussing average income, please.

RESPONSE:

ADDED.

“The median monthly income was $82.15 ($67.65-$115.99) (Based on the rate at OANDA on 15/01/2021).” (p13)

Reviewer #2

Comment 1. [83] The only citation here for the link between food insecurity and mental well-being is from ten years ago. At a minimum, in addition, the authors should cite the two articles by Frongillo et al. and the one by Jones on the relationship between food insecurity and subjective well-being that were done using the Gallup World Poll data (and FIES). The second Frongillo et al. article investigates how the relationship depends on the degree of food insecurity (and why) which may be relevant to the findings in the current article. doi:10.3945/jn.116.243642, doi.org/10.1093/jn/nxy261, doi.org/10.1016/j.amepre.2017.04.008

RESPONSE:

ADDED. PLEASE SEE THE INTRODUCTION SECTION (p4, 5-7).

Comment 2. [84-86] The statement of aims is weak. “Investigate” and “examine” are activities and not scientific aims. So what were the scientific aims and what were the hypotheses?

RESPONSE:

CHANGED TO

“Taking food insecurity as a potential stressor during the lockdown, we hypothesise that more food insecure women will experience a higher level of stress. In this study, we also measured awareness about coronavirus among women and their compliance with the health directives, and their association with food security and mental health.” (p7)

Comment 3. [118-120] “We attempt”? Either the authors measured knowledge, attitudes, and income loss or they did not. Furthermore, the three tables that display the items used need to be referenced here.

RESPONSE:

CHANGED TO

“we measured” (p9)

“In this research, we measured …” (p22)

Comment 4. [129-131] Categorizing changes in food insecurity into negative, zero, and positive is fine for descriptive purposes but not for formal analysis.

RESPONSE:

Thank you very much for your valuable and thoughtful guidance. We have now revised our result section of the manuscript and have considered categorisation based on food security status only in the descriptive analysis result part. We added additional information on household head occupation status. We also described the proportion of participants with mild, moderate and severe food insecurity security status during the both wave 1 and wave separately. An additional supplementary table describing the participant characteristics based on participants’ change in food security status over the pandemic period has been added.

Comment 5. [142] The authors are strongly encouraged to re-do the analysis using a two-level model (villages and individuals) regressing, for example, PSS on food insecurity at visit 1, food insecurity at visit 2, and other covariates. The coefficient for food insecurity at visit 2 is then interpreted as quantifying the association of the change in food insecurity with PSS (since control is made for food insecurity at visit 1). The analysis that the authors have done assumes that change means the same regardless of what the value of food insecurity was at visit 1, which is not likely to be true. There is no reason to make this assumption when it is easy to model without the assumption. (See articles by McKenzie appended for more information if desired.)

RESPONSE:

Thank you very much for your valuable guidance. We have now revised our result section thoroughly and also revised our analysis. We have conducted our analysis from both the individual and village level. In the revised version we have conducted linear regression only. We measured the effect of wave 2 food security and on the participants’ stress level during the pandemic period as PSS is only measured during wave 2, and hence we could not use ANCOVA as suggested by Mckenzie. The change of the FIES in wave 2 is relative to wave 1. Please see the revised version of the statistical analysis section. The following lines have been added in the result section. Please see our response to comment 7 below where we described the changes made in the results section in more detail.

Comment 6. [144-146] The analysis using the categories of food insecurity is arbitrary and should be deleted. Only 7% reported positive change. This means that Tables 4 and 6 can be deleted and instead a table should be created with the results from the suggested analysis using food insecurity as a continuous variable at each visit.

RESPONSE:

Thank you very much. We have now deleted the table 4 and table 5 from the revised version of the result and have removed the formal analysis results based on categorised food insecurity variable.

Comment 7. [163-168] The primary exposure variable is food insecurity, and no descriptive information has been presented about it. We need to know the mean and SD at visit 1, visit 2, and the change.

RESPONSE:

ADDED:

“A total of 2402 female responders across 357 villages were included as participants of the study. The average age of the participants was 38 years (SD ± 8 years), and among them, 993 (42%) participants were above 40 years of age. The median number of family members was 4 (3-5). Majority of the household head were daily labourer, 789 (33%) and small businessman 564 (24%) followed by farming on own land 413 (17%), farming on other people’s land 246 (10%), private job 107 (4%), Government job 77 (3%) and others 206 (9%). The median formal education year of the participants was 8 (8-10) years. About 91% of participants (2096 out of 2,402) can read at least, wherein, 1896 (88%) participants mentioned that they could read and write. The median monthly income was $82.15 ($67.65-$115.99) (Based on the rate at OANDA on 15/01/2021). During the wave 1 majority of the participants 1604 (67%) had a FIES score of 2 (Moderate food insecurity) followed by 418 (17%) with the score 1 (Mild food insecurity), 197 (8%) with the score of 3 (Severe food insecurity), and 183 (8%) with the score of 0 (No food insecurity). On the other hand, during the wave 2, 1336 (56%) participant had a FIES score of 2, the number and rate for score 0, 1, and 3 were 507 (21%), 380 (16%), and 179 (7%) respectively. Also, 829 (35%) participants reported that their food security status worsens (a negative change in FIES score from wave 1 to wave 2) over the pandemic period. Fig 1 shows the difference in various food security status between two waves and a significant difference in the proportions of participants was observed in the food secure, mild insecure, and moderate insecure groups. Supplement 1 described the participant characteristics based on the change in food security status over the pandemic period.

Supplement 1: Description of participant status based on change in food security status (Improved, No Change, Worsen)

Improved No Change Worse p-value

Age in yers

- Mean±SD 38.08±8.31 37.96±8.33 38.83±8.09 0.08

Number of Family Members

- Median (Q1-Q3) 4 (4-5) 4 (3-5) 4 (3-5) 0.66

Level of Education, n (%)

- Illiterate

- Can Read Only

- Can Read and Write

32 (7)

24 (5)

432 (88)

66 (7)

53 (6)

808 (87)

46 (6)

34 (5)

656 (89)

0.79

Occupation of the Household Head, n (%)

- Day Laborour

- Farming on Own Land

- Farming on Other’s Land

- Small Businessman

- Private Job

- Government Job

- Others

186 (35)

78 (15)

69 (13)

114 (21)

26 (5)

15 (3)

47 (9)

400 (39)

190 (18)

110 (11)

117 (17)

41 (4)

23 (2)

97 (9)

203 (24)

145 (17)

67 (8)

273 (33)

40 (5)

39 (5)

62 (7)

0.00

Monthly Household Income, BDT

- Median (Q1-Q3)

7000 (5833-10,000)

7500 (5833-10,000)

7125 (5833-10,333)

0.26

” (p12-13)

Comment 8. [231] The terms food secure, mild insecure, and moderate insecure are used here but have not been explained in the Method section.

RESPONSE:

we have added additional information in the revised version of the result section of the manuscript (response to comment 7 above).

Comment 9. [297] Delete “promisingly”. Such a term has no place in a scientific manuscript.

RESPONSE:

DELETED (p22)

Comment 10. [334] This study is observational, so no causal language is appropriate. The authors can say that “worsening food insecurity was associated with…”

RESPONSE:

CHANGED TO

“food security status was found to be associated with a higher perceived stress level” (p26)

Comment 11. Table 4. How can there be a change in PSS score when PSS was only assessed at visit 2? The term change can only be used when something has been measured at least twice so change over time can be observed.

McKenzie, D. (2012). Beyond baseline and follow-up: The case for more T in experiments. Journal of Development Economics, 99:210–221. [http://doi:10.1016/j.jdeveco.2012.01.002]http://doi:10.1016/j.jdeveco.2012.01.002

McKenzie, D. (2015). Another reason to prefer Ancova: dealing with changes in measurement between baseline and follow-up. Developmental Impact (blog). Washington, DC: World Bank. https://blogs.worldbank.org/impactevaluations/another-reason-prefer-ancova-dealing-changes-measurement-between-baseline-and-follow

RESPONSE:

Table 4 and Table 5 from the revised version of the result have now been changed. (p20)

Reviewer #3

Comment 1. Overall

The article presents important findings, and has two time points of data collection, which for that reason is the main reason to not recommend rejection. However, the paper needs major improvements in the description of the work, the methods and the results to meet the publication standards for PLOS One. Grammar editing will be important before the paper is published, suggest a careful review- several places where plural/singular state of verbs were wrong, and a few other things.

RESPONSE:

Thank you. We have revised the paper significantly. We have also read the paper thorougholy and completed Proof-reading.

Comment 2. Abstract:

Could be improved by some copy editing. It is unclear what ‘panel’ means in line 42.

In the conclusion, the authors note that there are implications. Provide a sentence of two about what the implications are here.

RESPONSE:

CHANGED TO

“Data were collected at two time-points (waves 1 and 2) from 2402 women, one per household, participating in a larger study during extended COVID-19 lockdown in the rural areas of the southwest region of Bangladesh.” (p2)

…“ The results indicate a higher stress level, a potential contributor to poor mental health, as food insecurity deteriorated. Policy initiatives in ameliorating immediate food insecurity during crises and improving long-term wellbeing, and expanding the reach of mental health support are warranted.” (p3)

Comment 3. Introduction:

Your introduction needs a more robust literature review that allow you to frame your questions and objectives, and currently very short. The paper’s lack of a theoretical grounding weakens its ability to contribute to broader conversations, and to contribute to theory on food security and wellbeing. Having this could allow you to have sound hypotheses to test. What have others have found regarding food security and perceived stress? What has been the differences between the findings between low-, middle-, and high-income countries? And between vulnerable groups (i.e. women, farmers, children, etc.)? Suggest an introduction that allows the reader to understand your research questions, objective, and theoretical grounding, and then present a literature review on the topic, and section that explains why you chose Bangladesh. It is important for the reader to understand the broader social, geographical, political, and cultural context in which this study was carried out. Furthermore, given that your study is based upon data using FIES and PSS, you could expand your literature review on these two scales.

RESPONSE:

We have now expanded the introduction section including relevant literature on food insecurity, stress and mental health, and knowledge and attitude. Please see the introduction section (p4, 5-7).

Comment 4. Methods:

There is additional clarity necessary for this work to meet the standards of PLOS One and for reproducibility. The authors need to better explain the survey protocols, and the literature that informed it. Though the study focuses on understanding change, not much of the results show what was found on the first round of surveys. It is also not clear whether PSS data was also collected in wave 1? (line 117). Doing so, would be most important to have a baseline to compare pre and post COVID.

Furthermore, given that the paper lacks a theoretical grounding, so it is hard to understand the variables chosen for the models. How do these models compare to what is typically included in understanding PSS and food security? Why do the authors include assessment of attitudes and knowledge- is this demonstrated in the literature to have an impact on the outcomes? What is the hypothesis driving this inclusion? Including the survey questions (i.e. scale questions) in a supplementary form would be helpful. Finally, could the authors say something about how the respondent demographics compare to the overall population of the region or Bangladesh? It’s important to know whether the population is biased. For example, do we know if any of these women were pregnant or postpartum? Did the number of charges/children that they had make any difference in their mental health? What about increased domestic violence?

RESPONSE:

PSS data were collected only at the wave 2 (visit 2) of the study. A couple of considerations shaped the decision of collecting pss data at that point only. Firstly, the loss of income of the poor and looming food insecurity became a priority at the beginning of the nationwide covid-19 lockdown. Our goal was primarily informing policy makers about the urgently needed food support for the poor. At this point the likely duration of the lockdown was unknown as it continually extended before finally coming to an end on 30 may 2020. Secondly, as the lockdown had been extended, our observations on food insecurity, and concern about mental health particularly for women due to lockdown and resulting effects, encouraged us to explore how the covid-19 induced circumstances impacted level of perceived stress in women in rural areas.

We have now expanded (i.e. On food security, knowledge and attitudes, and mental health) the introduction section. Please note our response to comment 1 of reviewer #1, p4, 5-7.

In our manuscript, in the method section we have commented on the representativeness of our sample. (p8)

Our surveys were conducted over the phone. Given the context of the study, we had to omit questions that would have added valuable evidence due to time constraint of the participants. Our participants are women who are primarily responsible to taking care of their family. During the pandemic they might be facing additional care responsibilities as children were not attending school and people were urged to stay in. Therefore, we had to be mindful of not adding to the burden. We have now addressed this issue in the strengths and limitations subsection (p28-29).

Comment 5. Results:

More information about the model parameters and outputs would be helpful to the reader. What variance was associated with the random effects? How many groups were there? Also, the results focus heavily on knowledge and attitudes, but that isn’t the main framing of the paper, so suggest introducing these concepts more in the introduction or, increasing the clarity of results on the food insecurity and other topics.

RESPONSE:

We have now revised our entire result section of the manuscript based on all of our reviewers’ valuable advice and guidance. Please see the result section on the revised version of the manuscript. (p12-13, 18-19)

Comment 6. Discussion

This section is the strongest. Much of the literature discussed in this section should be better foreshadowed in the introduction. Though you mention that the paper has policy implications, your discussion could be expanded to better discuss how these findings inform future strategies that allow emergency management to take into consideration mental wellbeing. Furthermore, how do your results compare with others?

In your discussion or limitation section, explore potential factors that could have caused both increased stress and food insecurity. Are there other environmental, political or social changes that occurred over the time of the study which could have contributed to increased stress among the population prone to food insecurity? i.e. poorer families may have increased stress due to financial challenges, this would also make them prone to food insecurity. This should be explored in your discussion & limitations.

RESPONSE:

Unfortunately, we were unable to explore the effect of several social, environmental and political factors because of limited access to collecting data during the pandemic situation. Therefore, we could also not adjust some potential factors such as household environment factors while locked into the house, domestic and social violence, and others. As a result, we were also unable to adjust those potential factors in our formal analysis an dwe have now addressed this in our manuscript. (p28)

ADDED:

“Household food insecurity during crises may put women living in rural areas under stress. Addressing food insecurity is critical in ensuring the physical health of the people and their mental health. Collaborative actions are warranted from the relevant government and non-government agencies to initiate public policy dialogue around ensuring food security for the poor during crises such as the COVID-19 pandemic and long-term. Distribution of food during crises needs urgent consideration, and equitable distribution must be ensured so that the support reaches the poor. Mental health support needs to be more available in rural areas, in a contextually appropriate form. Given that food insecurity is common in Bangladesh and women’s critical role in managing household food shortage, community-based awareness-raising activities on the significance of mental health in overall wellbeing, the importance of help-seeking, and available support services could be beneficial.” (p29-30)

Comment 7. Conclusion

There needs to be a conclusion section.

RESPONSE:

ADDED. PLEASE SEE p29-30.

Comment 8. General comments:

The paper uses the concepts of “psychological wellbeing”, “mental health”, and “mental wellbeing”. The authors should better define the concepts, and no use them interchangeably.

RESPONSE:

This has now been clarified in the introduction, and necessary changes have been made throughout the manuscript.

Attachments

Attachment

Submitted filename: Response to Reviewers.docx

Response to reviewers’ comments

Manuscript ID PONE-D-20-35162R1 entitled "Food insecurity and mental health of women during COVID-19: Evidence from a developing country"

We thank the reviewers and the editor for their constructive comments. We believe the comments have improved our work. Please find our response to the individual comments below. As we addressed the comments and revised the paper, the line numbers have changed in this revised version of the paper. Therefore, we mentioned page number corresponding to all changes in the subsequent pages.

Response to the comments from the editor

RESPONSE:

Reviewer #3

Comment. Discussion and Results

The authors have implemented many changes to their manuscript, which has certainly improved it's clarity and context. The addition of other literature and more methodological details was critical to publish this paper. For the discussion section, I was not implying that the authors have the data itself to explore social, environmental and political factors, but rather that they authors could integrate other data in this space. For example, were there monsoons or major climate disruptions during this time that could have also affected these outcomes? Were there any political unrest?

While the authors have mostly addressed the original comments, I am concerned about their new results and modelling section, which is not clear. Reviewer 2 suggested a two-level model, or a hierarchical model. I would have interpreted that to mean a model looking at food security outcomes at Time 1 and Time 2, with two levels of random effects- individual and village. Instead, the authors did something very different, and I don't understand what they did. The seem to have calculated the mean of different scores at the individual and whole village level. Then, it looks like they ran a simple univariate model predicting different scores at the individual and village level, and the effect of change t the two levels. I'm not sure whether the "effects" they are reporting are random effects? Or some other variable? Also, I think it would be important to look at all of these different potential contributors to food security in one model, so a hierarchical random effects model that includes the PSS, KLC-19, ALC-19, and some covariate controls such as income, with a random effect at the individual and village level. This would be the most robust way to examine food security outcomes, rather than individually. Regardless of whether Reviewer 2 thinks that this model has been done correctly, it at least needs further explanation. It is not clear what the coefficients are that are being reported in the results (random effects?) or how the model was built, whether it was truly hierarchical. Model fit statistics are also not present, which minimizes the ability to understand these results.

RESPONSE:

Thank you once again for your comments and for clarifying the comments further for us. We have now clarified the changes made to the analysis in the previous submission and the results in the light of those changes (p12 & 19), and addressed the potential influences major climatic events or political unrest might have on our findings in the Discussion section (p22). We explained the changes made to the Result section first and then the changes made to the Discussion section.

In the first version of the manuscript we implemented the hierarchical model with two levels of random effects. Based on the reviewer 2 comments, we changed the analysis and revised the result. However, in this revised version we also added the results from the two levels of random effect and the overall result remains same.

ADDED TO RESULTS:

Added information in the method section:

“Three mixed-effects linear regression modelling was used to assess the relationship between changes in FIES scores and knowledge level, attitude status, and PSS score. Each model included individual and village level random effects to address the correlation of the outcomes. The models were also adjusted for the age, level of education and monthly income of the participants.” (p12)

Added information in the result section:

“When performing a mixed effect linear regression model, a significant negative association was observed between the PSS score and change in food security status (Coefficient: -1.20, 95% CI: -1.37; -1.03, p-value: <0.001). On the other hand, the effect of change in food security status over KLC-19 was non-significant (Coefficient: 0.05, 95% CI: -0.05; 0.16, p-value: 0.302) but significant for the ALC-19 score (Coefficient: -0.06, 95% CI: -0.11; -0.01, p-value: 0.029).” (p19)

ADDED TO DISCUSSION:

“Climatic conditions such as cyclones or floods which happen frequently in Bangladesh are important factors to consider for examining the food security of households. [REF] This study sample was drawn from a region in Bangladesh which experienced cyclone ‘Amphan’ just after the first wave of survey. [REF] However, wave 2 survey was conducted about three weeks after the cyclone. While the livelihood, food security and mental health situation could be affected by the disaster we believe that the results reported in the paper are not significantly affected by cyclone ‘Amphan’. There are several reasons. First, people in this region frequently face natural disasters and very much adaptive to changes caused by cyclone. [REF]. Second, only limited number of the sample we surveyed during wave 2 was directly affected by this cyclone. Still there are reasons to expect that mental health of women are affected by the cyclone. To the extent cyclone Amphan had a lasting effect, our results would be subject to bias, and one should interpret the results with caution. Note also that our results are not affected by any political and social unrest since it was relatively a peaceful time in Bangladesh politics.” (p22)

Attachments

Attachment

Submitted filename: Response to Reviewers.docx

Response to reviewers’ comments

Manuscript ID PONE-D-20-35162R2 entitled "Food insecurity and mental health of women during COVID-19: Evidence from a developing country"

We thank the reviewer and the editor for their constructive comments. We believe the comments have improved our work. Please find our response to the individual comments below.

Reviewer #3

Comment

On the hierarchical model, there is still a lack of information that is necessary to publish this paper. It is not standard to just report the coefficients and confidence intervals in a multivariate hierarchical model for just one variable of interest, as is done in Table 5. At a minimum, the authors should have supplementary materials that include the coefficients, p values, confidence intervals, and standard errors for all independent variables and controls, as well as the random effects. Otherwise, this does not adhere to reporting of statistical results. Furthermore, the authors did not address my comment, which was to create a single model looking at the relationship of all these different stress and other factors on food security status. Why are they isolating the effect of these three variables into single models? What is the justification, rather than including them all into a single model to predict food insecurity?

RESPONSE:

In the very first version of the manuscript, we implemented the following statistical method of analysis:

Three mixed-effects linear regression modelling was used to assess the relationship between changes in FIES scores and knowledge level, attitude status, and PSS score. Another mixed-effects linear regression model was used to evaluate the relationship between PSS score and categorized food security status change (negative change, no change and positive change). A mixed-effects ordinal logistic regression was also conducted to see the relationship between ordinally categorised PSS score and categorised food security status. Each model included village-level fixed effects to adjust for the correlation of outcomes within the village; the models also adjusted for participants’ age and educational level.

In the first round of review, we received the following comment from Reviewer 2:

The authors are strongly encouraged to re-do the analysis using a two-level model (villages and individuals) regressing, for example, PSS on food insecurity at visit 1, food insecurity at visit 2, and other covariates. The coefficient for food insecurity at visit 2 is then interpreted as quantifying the association of the change in food insecurity with PSS (since control is made for food insecurity at visit 1). The analysis that the authors have done assumes that change means the same regardless of what the value of food insecurity was at visit 1, which is not likely to be true. There is no reason to make this assumption when it is easy to model without the assumption. (See articles by McKenzie appended for more information if desired.)

In the same round, the comments from Reviewer 3 were as follows:

More information about the model parameters and outputs would be helpful to the reader. What variance was associated with the random effects? How many groups were there? Also, the results focus heavily on knowledge and attitudes, but that isn’t the main framing of the paper, so suggest introducing these concepts more in the introduction or, increasing the clarity of results on the food insecurity and other topics.

Following the suggestion of Reviewer 2, we re-analysed the data. As we re-analysed the data and removed the mixed models based on Reviewer 2, we were not able to address the above comment of Reviewer 3 at that point.

We have now addressed the potential influences major climatic events or political unrest might have on our findings in the Statistical Analysis section (p11), Result section (p18-19), and Supplement 4 and 5.

ADDED:

“In addition, three mixed-effects linear regression modelling were used to assess the relationship between wave 2 FIES scores and knowledge level, attitude status, and PSS score. Each model included individual family and village level random effects to address the correlation of the outcomes. The models were also adjusted for the age, level of education; monthly income and the respective scores of the participants.” (p11)

“Table 4: Effect of wave 2 FIES score on PSS, KLC-19 and ALC-19 scores

List of Scores Individual level Model# Village level Model# Mixed Model$^@

PSS score 1.87 (1.61; 2.13); p=<0.001* 1.34 (0.46; 2.22); p=0.003* 1.80 (1.48; 2.11); p=<0.001*

KLC-19 score -0.03 (-0.04; -0.02); p=<0.001* -0.03 (-0.06; 0.00); p=0.052 -0.46 (-0.62; -0.29); p=<0.001*

ALC-19 score 0.00 (-0.01; 0.00); p=0.350 -0.01 (-0.04; 0.02); p=0.418 -0.04 (-0.14; 0.07); p=0.495

#Univariable linear regression models were applied

$Mixed-effects linear regression modelling was used

^Models were adjusted for the age, level of education, monthly income and respective scores of the participants

@Each model included individual family and village level random effects

*p-value <0.05 considered as significant

When performing a mixed effect linear regression model, a significant positive association was observed between the PSS score and change in food security status (Coefficient: 1.80, 95% CI: -1.42; -0.98, p-value: <0.001). On the other hand, the effect of change in food security status over KLC-19 was significant (Coefficient: -0.46, 95% CI: -0.62; -0.29, p-value < 0.001) but non-significant for the ALC-19 score (Coefficient: -0.03, 95% CI: -0.14; 0.07, p-value: 0.495). A detailed description of the mixed effect model results has been described in supplement 4.” (p18)

“Table 5: Effect of change in FIES score from wave 1 to wave 2 on PSS, KLC-19 and ALC-19 scores

List of Scores Individual level Model# Village level Model# Mixed Model$^@

PSS score -1.15 (-1.31; -0.99); p=<0.001* -1.02 (-1.52; -0.53); p=<0.001* -1.20 (-1.42; -0.98); p=<0.001*

KLC-19 score 0.00 (0.00; 0.01); p=0.233 0.00 (-0.02; 0.03); p=0.722 0.05 (-0.09; 0.19); p=0.482

ALC-19 score 0.00 (0.01; 0.00); p=0.115 0.00 (-0.02; 0.02); p=0.704 -0.03 (-0.12; 0.05); p=0.445

#Univariable linear regression models were applied

$Mixed-effects linear regression modelling was used

^Models were adjusted for the age, level of education, monthly income and respective scores of the participants

@Each model included individual family and village level random effects

*p-value <0.05 considered as significant

…A detailed description of the mixed effect model results has been described in supplement 5.” (p19)

Please note that we did not attempt to examine the effect or stress, knowledge, and attitude on food security in this paper, or predict food insecurity; rather we attempted to examine the effect of food security status on the level of perceived stress. We also attempted to evaluate the effect of food security status on the COVID-19 related knowledge and attitude level of the participants. (p 7-8)

Thank you once again for your comment.

Attachments

Attachment

Submitted filename: Response to Reviewer.docx