Prevalence of depression in China during the early stage of the COVID-19 pandemic: a cross-sectional study in an online survey sample

Objectives We aimed to determine (1) the prevalence of depression during the COVID-19 pandemic among Chinese adults and (2) how depression prevalence varied by province and sociodemographic characteristics. Design Cross-sectional study. Setting National online survey in China. Participants We conducted a cross-sectional online survey among adults registered with the survey company KuRunData from 8 May 2020 to 8 June 2020. We aimed to recruit 300–360 adults per province (n=14 493), with a similar distribution by sex and rural-urban residency as the general population within each of these provinces. Primary outcome Participants completed the Patient Health Questionaire-9 (PHQ-9). We calculated the prevalence of depression (defined as a PHQ-9 score ≥10) nationally and separately for each province. Analysis Covariate-unadjusted and covariate-adjusted logistic regression models were used to examine how the prevalence of depression varied by adults’ sociodemographic characteristics. All analyses used survey sampling weights. Results The survey was initiated by 14 493 participants, with 10 000 completing all survey questions and included in the analysis. The prevalence of depression in the national sample was 6.3% (95% CI 5.7% to 6.8%). A higher odds of depression was associated with living in an urban area (OR 1.50; 95% CI 1.18 to 1.90) and working as a nurse (OR 3.06; 95% CI 1.41 to 6.66). A lower odds of depression was associated with participants who had accurate knowledge of COVID-19 transmission prevention actions (OR 0.71; 95% CI 0.51 to 0.98), the knowledge that saliva is a main transmission route (OR 0.80; 95% CI 0.64 to 0.99) and awareness of COVID-19 symptoms (OR, 0.82; 95% CI 0.68 to 1.00). Conclusion Around one in 20 adults in our online survey sample had a PHQ-9 score suggestive of depression. Interventions and policies to prevent and treat depression during the COVID-19 pandemic in China may be particularly needed for nurses and those living in urban areas.


Introduction
Page 5 line 6 "As of June 2021", do you mean early, middle or late June 2021? Ps check the citation for consistence.
Page 5 line 52-55 "More recently, literature has emerged that offers contradictory findings about the association between depression and these sociodemographic characteristics 17 19" Ps rewrite this sentence for clarification.

Study Design and Sampling
Have you consult a statistician whether the use of sampling is appropriate? Is your sample representative of your population? Ps attach your Institute Ethics approval for this study.

Page 7 line 47-48
What is the rationale to use the Patient Health Questionnaire-9 (PHQ-9)? Why not use other Depression scale? Table 1. Education Level shall be converted to western style so that readers from different country can understand. consideration include sociodemographic characteristics, whether one works as certain type of health care providers, COVID-19 awareness, and acquaintance with COVID-19 posi-tive cases. The work claims to be the rst to address how the prevalence of depression varies by regions, sociodemographic characteristics, and other things that a ect mental health in China.
For the study design, the authors collaborated with an information technology company in China and conducted a crosssection survey in 2020 and sampled 10,000 adults from 31 provinces. The COVID-19 awareness and the depression level, together with personal demo-graphic information, were measured from a questionnaire. A Patient Health Questionnaire-9 (PHQ-9) score larger than 4 was considered depression. The COVID-19 awareness consisted of questions such as perceived risk, COVID-19 symptoms, awareness of vaccine availability, knowledge of transmission prevention actions etc.
Through the study, the authors found that (1) depression was more common among urban residents, nurses and those who has a family member diagnosed with COVID-19; (2) knowledge of COVID-19 transmission and healthcare-seeking behavior can e ectively reduce the risk of depression. I think the paper provides intuitive conclusions. Through the data analysis, we get more support and details on the e ect of di erent factors under consideration on depression.
I only have some minor comments listed below.
2. Page 7, line 48, \a score 4". Why do we use 4 as the threshold? Is it for a more balanced data in logistic regression? Or do we have literature convention or argument from mental health studies? In some other places of the paper, the de nition becomes > 4 (e.g. Page 9, line 48). In Table 1, no depression is de ned as PHQ-9 4 and depression is de ned as PHQ-9 4. Please be consistent with how to handle the case with PHQ-9 = 4.  [13][14][15][16][17][18][19][20][21][22]. People in urban areas have higher education and greater access to COVID information. How come their depression is higher or positively correlated with urban dummy variable? The authors argued more knowledge on COVID should reduce the depression. They seem to contradict here. Is it possible that the higher depression may not be explained as something related with COVID, but just by the fact that working in a city is more stressful?
9. Page 17, line 31, \more vulnerable" ! \are more vulnerable". 10. Page 17, line 34, \contracting" ! \contacting". 11. Page 17, line 45, \initial stage"? The data was collected in May 2020. Is this still considered as initial stage? By this time, the peak outbreak of the pandemic in China has passed.

2) Update description of vaccines (being developed)
3) Description of China having returned to 'normal life' whereas other countries have not does not tally with objective indicators such as the COVID-19 Stringency Index where China is currently ranked as one of the countries with the most stringent restrictions globally (https://ourworldindata.org/grapher/covid-stringencyindex). If China had returned fully to normal life it would also beg the question as to why it is necessary to conduct this study. Please update the introduction to reflect how China compares to other nations on indicators such as the Stringency Index. Method: 4) Please include more information on how participants were recruited / why people register with KuRunData (e.g. is this an existing online internet panel for market research etc.).
5) The typical cut-off used on the PHQ-9 is greater than or equal to 10. A cut-off of greater than or equal to 4 is unusual. Even mild depressive symptoms are typically guaged using a score of greater than or equal to 5 on this measure. Unless there is a strong rationale not to do so I suggest using the more conventional cut-off of ≥10 which the original  paper has shown to have a sensitivity of 88% and a specificity of 88% for major depression. The use of the 4 cut-off partly explains why the rate of depression appears to be very high in the current results.

Results
6) Please provide more information on how the sampling weights were calculated. The method section indicates that "The sampling weights were calculated by the inverse of the probability of age, residence type, and sex of selecting participants." Yet, looking at Discussion: 9) The discussion begins "This is the first study that quantified the prevalence of depression symptoms and its variation by province and sociodemographic characteristics in China". Where in the paper was the prevalence of depression quantified by province?
10) The PHQ-9 cut-off described in the method was ≥4 and this was reported as greater than 4 in the discussion. Please provide the correct description. 11) Discussion of depression rates identified in other studies needs to be better contextualised. For instance, are there other nationally representative studies that have used the PHQ-9 during COVID-19 and reported on the same cut-offs? The rate of depression identified will depend on the measure and cut-off used which needs to be noted.

Response to Reviewer 1
We thank Reviewer 1 for the valuable and constructive comments. We have carefully incorporated the reviewer's advice and as a result our paper has improved substantially. Below, we explain in detail how we have taken each of the comments into account in the revision of our manuscript. We show the reviewer's comments in italics and reply to them in standard font.
[Comment-1] Page 5 line 6 "As of June 2021", do you mean early, middle or late June 2021?
Ps check the citation for consistence.
[Answer] Thank you very much for this comment. For this comment, we carried out modifications on the manuscript, as shown below: (1) The authors added the word "middle" into the sentence and updated the reference (Page 5, paragraph 1): "The KurunData is an online internet panel for market research, which recruits participants and provides access to the questionnaire via Wechat mini. To assure the representativeness of the data, we utilized stratified sampling by sex, residence type and province. We calculated the size of each stratum based on a sequential method (Table S1)  "We utilized the PHQ-9 depression scale as a screening tool. There are several reasons for using this tool. At only nine items, it is suitable for a large-scale population survey and easy to complete by the general population as a quick depression assessment. The nine diagnostic symptom criteria of the PHQ-9 correspond to the DSM-IV major depressive disorder (MDD) criteria and can facilitate follow-up review of symptoms and the diagnosis process 3 . The PHQ-9 was also selected as it has good internal consistency 4 5 , reliability, and validity 6 , 7 with a Cronbach's alpha coefficient between 0.80 and 0.90."

[Comment-4] Measurements
Since data on knowledge of COVID-19 was published, this part can be shortened.
[Answer] The authors shortened the description of COVID-19 awareness published previously. (Page 7, paragraph 3) "As a part of investigation for willingness to participate in COVID-19 vaccination research, the questionnaire of our study included three parts:1) COVID-19 awareness (perceived risk of death from COVID-19, knowledge of the transmission of COVID-19, awareness of recommended healthcareseeking behavior); 2) sociodemographic information (sex, age, highest education level, ethnicity, residence type, health care providers, annual household income, personal history of COVID-19 diagnosis , positive acquaintance COVID-19 diagnosis); and 3) the Patient Health Questionnaire-9 (PHQ-9; a depression screening tool 8 ).
We utilized the PHQ-9 depression scale as a screening tool. There are several reasons for using this tool. At only 9 items, it is suitable for a large-scale population survey and easy to be completed by the general population as a quick depression assessment. The nine diagnostic symptom criteria of the PHQ-9 correspond to the DSM-IV major depressive disorder (MDD) criteria and can facilitate follow-up review of symptoms and the diagnosis process 9 . The PHQ-9 was also selected as it has good internal consistency 10 11 , reliability, and validity 12 13 , with a Cronbach's alpha coefficient between 0.80 and 0.90. We defined a score of 10 or greater as depression; COVID-19 awareness consisted of nine items which were previously published 14 . Among these nine items, for categorical outcomes, data are expressed as binary or categorical (range: 0 -100% [Answer] Thank you for your comment. The authors prefer to use the term "prevalence of depression" as suggested by the editors above. Prevalence is the proportion of a specific population with a particular disease during a given time, and incidence refers to the rate of new cases of a disease occurring in a specific population over a particular period of time. We did not know whether the participants with depression represent new cases, so we can only calculate the prevalence. "This study also found that nurses had a higher odds of depression, consistent with previous studies 15 16 . One possible explanation was that as front-line healthcare workers, nurses had higher risk to be infected by longer contact with patients than doctors, and worked longer hours than usual 17 , which might make them become more frustrated."  [Answer] Thank you for this comment. The education levels in Table 1 have been revised. (Page 10~12)  1 Weighted using survey sampling weights. 2 As per the 2020 China Statistical Yearbook.
For dichotomous outcomes, data are expressed as a percentage with the correct response (95% confidence interval). For continuous outcomes, data are expressed as median (interquartile range).  "Effective vaccines are critical to the containment of the COVID-19 pandemic; therefore, more and more vaccines are in clinical and pre-clinical development. Over 56 confirmed effective candidate vaccines for COVID-19 are being produced in America, Europe, China, and Australasia. As of middle January 2022, 139 vaccines are in clinical development, and 194 vaccines are in preclinical development preclinical phases 23 . Moreover, about 59.9% of the world population has uptake at least one dose of a COVID-19 vaccine, and 9.68 billion doses have been administered worldwide, as well as 32.92 million doses are administered each day 24  [Answer] Thank you for this helpful comment. The authors have followed your suggestion and have revised the description with the following sentence. (Page 5, paragraph 1) "Although China is currently ranked as one of the countries with the most stringent restrictions, scoring the sixth highest in Stringency Index globally 25 , asymptomatic infection and mutated variants continue to pose an unprecedented threat."

[Comment-4]
4) Please include more information on how participants were recruited / why people register with KuRunData (e.g. is this an existing online internet panel for market research etc.).
[Answer] Thank you for this suggestion. The authors have expanded the explanation of how participants were recruited in the methods section and this online market research tool. (Page 7, paragraph 1) "The KurunData is an online internet panel for market research, which recruits participants and provides access to the questionnaire via Wechat mini. To assure the representativeness of the data, we utilized stratified sampling by sex, residence type and province. We calculated the size of each stratum based on a sequential method (Table S1) to reflect the province's population composition by sex and urban and rural residence (according to the China Statistical Yearbook 2020 1 ). When sufficient participants were recruited for a given stratum, we halted recruitment for this stratum. Participants were paid 5 yuan (US$0.77) for completing the questionnaire. All participants completed informed written consent. Finally, 14493 participants initiated the online survey, with 10000 participants completing the survey."

[Comment-5]
5) The typical cut-off used on the PHQ-9 is greater than or equal to 10. A cut-off of greater than or equal to 4 is unusual. Even mild depressive symptoms are typically guaged using a score of greater than or equal to 5 on this measure. Unless there is a strong rationale not to do so I suggest using the more conventional cut-off of ≥10 which the original  paper has shown to have a sensitivity of 88% and a specificity of 88% for major depression. The use of the 4 cut-off partly explains why the rate of depression appears to be very high in the current results.
[Answer] Thank you for this helpful suggestion. The authors have utilized a cut-off of ≥10, as recommended by this reviewer and updated Table 1 and Table 2 (Page 10-13). The results and discussion sections were also updated. (Page 9-17) "The majority of participants were male (50.8%), had at least a high school education (73.0%), lived in urban areas (59.4%), and had an annual household income >90,000 RMB (~US $13,000) (51.6%). Han ethnicity accounted for the vast majority of participants (93.8%). Except for 9% of participants aged under 20 years, the percentage of other age groups ranged from 16.5% to 19.0%. Health care providers accounted for 4.1% of participants, with the largest category being community health workers (1.6%). 0.08% of participants had been diagnosed with COVID-19 and 0.34% of participants knew someone diagnosed with COVID-19. 6.0% of participants reported being previously diagnosed with depression. The correct rate of COVID-19 awareness ranged from 11.1% to 96.3%, knowledge of COVID-19 transmission prevention actions was lowest (11.1% [95% CI: 10.5% -11.7%]) and knowledge about perceived risk of COVID-19 death among people with other diseases was highest (96.3% [95% CI: 96.0% -96.7%]) (See Table 1.)

Variation of depression prevalence by province
Overall, the depression prevalence was 6.3% (5.7%-6.8%). As seen in Figure 1, the prevalence of depression in Henan was the highest (9.4% [6.6%-12.7%]) and in Hainan was the lowest (3.7% [1.9%-6.1%]). The prevalence of COVID-19 in Hubei province (1,148 per million) tended to be higher than other areas in China during the same time period, and Tibet (0.28 per million) saw the lowest prevalence (Table S2). While the prevalence of COVID-19 cases in other provinces tended to be similar, Hubei and Tibet were the exception (Figure 2).

Variation of depression prevalence by sociodemographic characteristics within provinces
The results of covariate-unadjusted and covariate-adjusted logistic regression analysis of associated factors with depression are shown in Table 2. After adjusting for covariates, urban residents and nurses had higher odds of depression than rural residents (OR, 1,50; 95% CI, 1.18 -1.90) and other health care providers (OR, 3.06; 95% CI, 1.41-6.66). Participants who had correct knowledge of COVID-19 transmission prevention actions (OR, 0.71; 95% CI, 0.51 -0.98) and knowledge of saliva as the main transmission route (OR, 0.80; 95% CI, 0.64 -0.99) had lower odds of depression, as did participants who had accurate awareness of COVID-19 symptoms (OR, 0.82; 95% CI, 0.68 -1.00).
"  1 Weighted using survey sampling weights. 2 As per the 2020 China Statistical Yearbook.

Discussion
In our online survey sample, 6.3% of adults had depression as defined by a PHQ-9 ≥10. We found a higher prevalence of depression among certain population groups, including urban residents and nurses. Furthermore, we found that knowledge and awareness of COVID-19 were associated with lower odds of depression. Under the strong assumption that these associations might be causal, this finding could indicate a "protective effect" for mental health and thus indicate the importance of effective communication and education about COVID-19 amid the pandemic.
The correlation between urban residence and depression was positive in our study, which is consistent with several previous studies 26 27 . A possible interpretation of this finding is that while the virus could be transmitted more quickly in urban areas with a higher density population 28 , those in urban areas tend to have higher education levels and greater access to the latest report on the COVID-19 pandemic 26 . Another explanation is that depression was more common in the urban than in rural areas in China before the COVID-19 pandemic. Moreover, the social distancing restrictions allowed less travel in cities than in rural areas, potentially contributing to the higher prevalence of depression in urban than in rural areas. This study also found that nurses had higher odds of depression, consistent with previous studies 15 16 . One possible explanation was that as front-line healthcare workers, nurses had higher risk to be infected by longer contact with patients than doctors, and worked longer hours than usual 17 , which might make them become more frustrated. the peak of the COVID-19 epidemic, which ranged from 8.3 to 48.3% 29 31-33 . One possible explanation is that our study occurred in May when the epidemic had decreased in severity, and the prevalence of depression may have diminished. Another possible reason for a lower depression prevalence found in our study may be that with the continuous strict quarantine policy, the government provided timely mental health service in response to COVID-19 34 .
The prevalence of COVID-19 in 31 provinces tended to be similar except for Hubei, which had the highest prevalence ( Figure 2). However, the proportion of the population by province with depression (PHQ-9≥10) did not seem to be significantly associated with the prevalence of COVID-19 cases.
Moreover, from the covariate-adjusted logistic regression results, we found that depression was not associated with the prevalence of COVID-19 cases confirmed by province, which also suggested that the prevalence of depression is not associated with the severity of the regional epidemic during the COVID-19 outbreak. One explanation is that resilience plays a protective role in mitigating the impact of stress and trauma on depressive symptoms and the consequences associated with depressive symptoms 35 . During the initial phase of the COVID-19 outbreak in China, the Chinese government enforced a rigid social distancing policy through social media and strongly promoted hand washing, surface disinfection, and the use of protective masks 36 . In Wuhan, as the most affected city, people with mild and asymptomatic infection received care in Fangcang shelter hospitals, for facility-based isolation, treatment, and monitoring 36 , an effective method to control the epidemic 37 38 . Fangcang shelter hospitals also provided mental health counseling services and social support to help patients recover during isolation 39 , which may also reduce the incidence of depression in hard-hit areas. "We calculated the size of each stratum based on a sequential method (Table S1) to reflect the province's population composition by gender and urban and rural residence (according to the China Statistical Yearbook 2020 1 ) and assure the sufficient samples collected in each stratum. Once we have recruited enough samples for a stratum, we would stop recruiting samples for this stratum.
"The sampling weights were calculated from the 2019 population census and the sampling quotas, accounting for some feature of the survey, including oversampling for sex, residence type and province. Specifically, sampling weights are the inverse of the probability of selecting participants with some specific residence type (urban or rural), sex (male or female), in some specific province among the population. We used sampling weights adjusted for the survey design to calculate statistics. For binary and categorical response options of knowledge about COVID-19, we calculated the percentage of participants with correct responses."

[Comment-7]
A table such as Table 1 including the weighted prevalence of depression by sociodemographic characteristics would be worth including.
[Answer] Thank you for this suggestion. The authors added the weighted prevalence of depression by sociodemographic characteristics in Table 1. (Page 10-12) "  1 Weighted using survey sampling weights. 2 As per the 2020 China Statistical Yearbook.
For dichotomous outcomes, data are expressed as a percentage with the correct response (95% confidence interval). For continuous outcomes, data are expressed as median (interquartile range).

[Comment-8]
8) It is not clear how you are finding such large odds ratios for family and neighbor confirmed COVID ( "and having a family member (OR, 23 [Answer] Thanks for this comment. The authors updated all results using a score of 10 or greater as depression symptoms definition in Table 2 and gave details of the prevalence of depression in these groups for positive acquaintance COVID-19 diagnosis. The cases of neighbor and friend groups were close to 0, which may be producing extreme ORs, we combined these two groups in Table 2 as suggested. (Page 12-13) " 9) The discussion begins "This is the first study that quantified the prevalence of depression symptoms and its variation by province and sociodemographic characteristics in China". Where in the paper was the prevalence of depression quantified by province?
[Answer] The description of the prevalence of depression quantified by province appears in the results section. (Page 10, paragraph 1) and Table 1 (Page 10-12).

"Variation of depression prevalence by province
As seen in Figure 1, the prevalence of depression (PHQ-9 score ≥10) in southern China tended to be greater than in northern China. The overall prevalence of depression was 6.0% (95% CI: 5.6%-6.5%).
The prevalence of COVID-19 in Hubei province (1,148 per million) tended to be higher than other areas in China, and Tibet (0.28 per million) was the lowest province. While the prevalence of COVID-19 cases from other provinces tended to be the same except for Hubei and Tibet (Figure 2)." "  1 Weighted using survey sampling weights. 2 As per the 2020 China Statistical Yearbook.
For dichotomous outcomes, data are expressed as a percentage with the correct response (95% confidence interval). For continuous outcomes, data are expressed as median (interquartile range). 10) The PHQ-9 cut-off described in the method was ≥4 and this was reported as greater than 4 in the discussion. Please provide the correct description.
[Answer] Thanks for this comment. The authors corrected the description about the PHQ-9 cut-off in the discussion section. (Page 14, paragraph 1) "In our online survey sample, 6.3% of adults had depression as defined by a PHQ-9 ≥10."

[Comment-11]
11) Discussion of depression rates identified in other studies needs to be better contextualised. For instance, are there other nationally representative studies that have used the PHQ-9 during COVID-19 and reported on the same cut-offs? The rate of depression identified will depend on the measure and cut-off used which needs to be noted.
[Answer] Thanks for this comment. The authors added the following text to the Discussion section. (Page 14, paragraph 3) "Several studies have described the prevalence of depression during the COVID-19 pandemic in the general population. Among studies using the PHQ-9 scale with the same cut-off value(≥10), the prevalence of depression observed in our study (6.3%) was lower than a national study among 56,679 participants conducted February 28 to March 11, 2020 in China (10.8%) 29 as well as a study of 1,470 individuals among the general population during the COVID-19 outbreak in the United States from March 31 to April 13, 2020 (27.8%) 42 43 . Most of the other studies were conducted in February 2020─at the peak of the COVID-19 epidemic, which ranged from 8.3 to 48.3% 29 31-33 44-46 . One possible explanation is that our study occurred in May when the epidemic had decreased in severity, and the prevalence of depression may have diminished. Another possible reason for a lower depression prevalence found in our study may be that with the continuous strict quarantine policy, the government provided timely mental health service in response to COVID-19 34 ."
[Answer] Thank you for this comment. The spelling has been corrected in the Methods section. (Page 7, paragraph 1) "Participants were paid 5 yuan (US$0.77) for completing the questionnaire."  [Answer] Thank you for this comment. The authors have revised the cut-off to ≥10 as the depression symptom definition and updated Table 1 and Table 2 as well (Page 10-13). The results and discussion sections were also updated to reflect this change. (Page 9-17). The PHQ-9 cut-off was revised in the methods (Page 8, paragraph 1) and discussion section (Page 14, paragraph 1). "We defined a score of 10 or greater as depression;" "In our online survey sample, 6.3% of adults had depression as defined by a PHQ-9 ≥10." [Answer] Thank you for this comment. The authors have rephased this sentence for clarity as suggested. (Page 8, paragraph 2) "The prevalence of depression was stratified by participants' sociodemographic characteristics and overall knowledge about COVID-19." [Comment-4]4. Page 8, line 43, "an odds ratio" →"odds ratios".
[Answer] Thank you for this suggestion. The authors have modified the following text in the methods section. (Page 8, paragraph 2) "We used covariate-unadjusted and covariate-adjusted logistic regression with a binary indicator for each province (province-level fixed effects) and obtained odds ratios (OR)." [Comment-5]5. The paper has no clear reference to Table 1. Please add a sentence in the main text about something like \Please refer to Table 1 on summary statistics of ...".
[Answer] The authors added the following sentence in the results section. (Page 9, paragraph 3) "…(See Table 1 [Answer] Thank you for this question. Since depression is defined as PHQ-9 ≥ 10, the following text has been added to the Discussion section. (Page 14, paragraph 3) "Several studies have described the prevalence of depression during the COVID-19 pandemic in the general population. Among studies using the PHQ-9 scale with the same cut-off value(≥10), the prevalence of depression observed in our study (6.3%) was lower than a national study among 56,679 participants conducted February 28 to March 11, 2020 in China (10.8%) 29 as well as a study of 1,470 individuals among the general population during the COVID-19 outbreak in the United States from March 31 to April 13, 2020 (27.8%) 42 43 . Most of the other studies were conducted in February 2020─at the peak of the COVID-19 epidemic, which ranged from 8.3 to 48.3% 29 31-33 44-46 . One possible explanation is that our study occurred in May when the epidemic had decreased in severity, and the prevalence of depression may have diminished. Another possible reason for a lower depression prevalence found in our study may be that with the continuous strict quarantine policy, the government provided timely mental health service in response to COVID-19 34 ." "This finding suggests that, to reduce the prevalence of depression, effective communication and education of COVID-19 preventive measures and recommended healthcare-seeking behaviors are urgently needed."  13-22. People in urban areas have higher education and greater access."to COVID information. How come their depression is higher or positively correlated with urban dummy variable? The authors argued more knowledge on COVID should reduce the depression. They seem to contradict here. Is it possible that the higher depression may not be explained as something related with COVID, but just by the fact that working in a city is more stressful?
[Answer] Thank you for this comment, the authors agree. The previous wording may have been unclear. We improved the following text in the discussion section. (Page 14, paragraph 2) "A possible interpretation of this finding is that while the virus could be transmitted more quickly in urban areas with a higher density population 28 , those in urban areas tend to have higher education levels and greater access to the latest report on the COVID-19 pandemic 26 . Another explanation is that depression was more common in the urban than in rural areas in China before the COVID-19 pandemic." [Comment-9] 9. Page 17, line 31, "more vulnerable" → "are more vulnerable". 10. Page 17, line 34, "contracting" → "contacting".
[Answer] Thank you for these suggestions. Since depression is defined in the revised manuscript as PHQ-9 ≥ 10, the results have been revised. No association was found between those with a family member or neighbor with confirmed COVID-19 and depression. Accordingly these sentences have been removed.
[Comment-10] 11. Page 17, line 45, \initial stage"? The data was collected in May 2020. Is this still considered as initial stage? By this time, the peak outbreak of the pandemic in China has passed.
[Answer] The authors improved the following text in the limitations section. (Page 16, paragraph 3) "This is the first study that investigated the prevalence of depression by province during the early stage of the COVID-19 pandemic in China;" [Comment-11]12. Page 18, lines 4-6. The study design was cross sectional, so the authors "cannot make temporal conclusions". But the paper does compare the prevalence of depression with previous studies in lines [10][11][12][13][14][15][16][17][18]Page 15. Could you explain what the meaning of "temporal conclusions" is here?
[Answer] Thank you for your comment. This phrasing has been removed as suggested.
[Answer] Thank you for these corrections. The authors have improved the following text in the Limitations section. (Page 17, paragraph 1

and Page 17 paragraph 2)
"Although we used stratified sampling to increase the representativeness of the data, it is still difficult to avoid response bias as potential participants with depression might be either less or more interested in taking part in the survey." Accurate knowledge of COVID-19 transmission and awareness of COVID-19 symptoms were associated with lower odds of depression."