Children's Mental and Behavioral Health, Schooling, Sociodemographic and Socioeconomic Characteristics During School Closure in France Due To COVID-19: The SAPRIS Project.

Background: COVID-19 limitation strategies have led to widespread school closures around the world. The present study reports children’s mental health and associated factors during the COVID-19 school closure in France in the spring of 2020. Methods: We conducted a cross-sectional analysis using data from the SAPRIS project set up during the COVID-19 pandemic in France. Using multinomial logistic regression models, we estimated associations between children’s mental health, children’s health behaviors, schooling, and sociodemographic and socioeconomic characteristics of the children’s families. Results: The sample consisted of 5702 children aged 8 to 9 years, including 50.2% girls. In multivariate logistic regression models, children’s sleeping diculties were associated with children’s abnormal hyperactivity-inattention (adjusted Odds Ratio (aOR) 2.05; 95% Condence Interval 1.70-2.47) and emotional symptoms (aOR 5.34; 95% CI 4.16-6.86). Factors specically associated with abnormal hyperactivity/inattention were: male sex (aOR 2.29; 95% CI 1.90-2.76), access to specialized care prior to the pandemic and its suspension during school closure (aOR 1.51; 95% CI 1.21-1.88), abnormal emotional symptoms (aOR 4.06; 95% CI 3.11-5.29), being unschooled or schooled with assistance before lockdown (aOR 2.13; 95% CI 1.43-3.17), and tutoring with diculties or absence of a tutor (aOR 3.25; 95% CI 2.64-3.99; aOR 2.47; 95% CI 1.48-4.11, respectively). Factors associated with children’s emotional symptoms were the following: being born pre-term (aOR 1.34; 95% CI 1.03-1.73), COVID-19 cases among household members (aOR 1.72; 95% CI 1.08-2.73), abnormal symptoms of hyperactivity/inattention (aOR 4.18; 95% CI 3.27-5.34) and modest income (aOR 1.45; 95% CI 1.07-1.96; aOR 1.36; 95% CI 1.01-1.84). Conclusions: Multiple characteristics were associated with elevated levels of symptoms of hyperactivity-inattention and emotional symptoms in children during the period of school closure due to COVID-19. Further studies are needed to help policymakers to balance the pros and cons of closing schools, taking into consideration the educational and psychological consequences for children.


Introduction
The emergence of COVID-19 in 2020 led many countries to implement strict sanitary measures, many of which resulted in substantial changes in children's lifestyles for several months. On March 31, 2020, 170 countries temporarily closed schools in an attempt to limit the spread of the new coronavirus responsible for this disease (1). In France, the rst lockdown lasted from March 16 to May 11, 2020, and schools and universities, which offer teaching to 14 million students, closed from the beginning of the lockdown until June 22, 2020 (2). School closures, the absence of usual sports and leisure activities, and a reduction in social contacts may have impacted children's mental health in an unprecedented manner (3)(4)(5)(6)(7)(8)(9). Children may have been exposed to stressors generated by the media, the fear of contagion, the lack of social interactions with individuals outside the immediate family, and by family members affected with COVID-19 (10,11). Moreover, the COVID-19 crisis may have resulted in socioeconomic di culties (e.g., nancial loss and changes in employment status) and impacted the mental health of adults, which in turn may have had indirect effects on mental health in children (10,12,13). The potential bene ts of closing schools to curb the spread of COVID-19 need to be weighed against the effects on children's mental health (14).
Although there is no consensus on the effectiveness of school closures to minimize the spread of COVID-19 (14)(15)(16), deleterious short-term effects on children's mental health have been observed. A Chinese study of 7143 college students found that 24.9% experienced anxiety related to the COVID-19 outbreak (3). Another Chinese cross-sectional study using an online questionnaire on 359 children and 3254 adolescents showed that 22.3% suffered from depressive symptoms during the outbreak (17). In addition, an internet-based survey evaluated the behavioral problems of 1264 children using the Strengths and Di culties Questionnaire (SDQ), in two elementary schools in China during the COVID-19 outbreak (18).
The prevalence was 6.3% for hyperactivity/inattention and 4.7% for emotional symptoms (18). However, to date the majority of studies on this topic have been carried out in China where the epidemic has been of shorter duration than in other parts of the world (3,4,17,(19)(20)(21). In France, a survey that used the SDQ on 432 parents showed that 24.7% of their children had hyperactivity/inattention symptoms and that 7.1% had emotional di culties during the rst lockdown in the spring of 2020 (13). This study based on a small sample of children identi ed signi cant correlates of these symptoms with parental mental health or nancial di culties, as well as the child's sleeping di culties or high screen time use. Therefore, social inequalities related to children's health were exacerbated by this unprecedented situation (5).
These studies had several limitations like a small sample size or the non-inclusion of other factors that might affect children's mental health, such as chronic illness and number of hours of study at home per day (22). Furthermore, while a signi cant number of studies were conducted on the indirect effects of the COVID-19 pandemic on mental health, only a few focused on children (12,23). More research is therefore needed to quantify the impact of socioeconomic disparities and inequalities that are exacerbated by school closure and that of the lockdown on children's mental health (24). The present study sought to assess the factors associated with the mental health of children aged 8 to 9 years-old during the school closure in the rst lockdown in spring of 2020 in France. We focused on children's hyperactivity/inattention and emotional symptoms, and on a range of factors such as health behaviors (e.g., sleeping di culties), schooling (e.g., tutoring, time devoted to schoolwork), and sociodemographic and socioeconomic characteristics of children's families (e.g., nancial di culties, type of housing, occupational category).

Methods
In the early weeks of the COVID-19 crisis, the SAPRIS survey was set up to study the health, social interactions and socioeconomic characteristics of the general population during the crisis in France (25). The project was based on the collection of data using the same questionnaire on the participants of ve large ongoing French cohorts, three of adults (Constances, E3N-E4N and NutriNet-Santé) and two of children (ELFE/EPIPAGE-2). For the present study, we used data collected on the 5702 children aged 8-9 years old in 2020 and participating in the ELFE and EPIPAGE-2 population-based birth cohorts that focus on child health and include late pre-term and term (ELFE) and very and extremely premature children (EPIPAGE-2) born in 2011 (26-28). More details on ELFE and EPIPAGE 2 are available elsewhere (26-28).
Data for the present study were reported by parents during the rst lockdown period (from April 16 to May 4, 2020) or immediately after (from May 5 to June 21, 2020), thus encompassing the whole period of school closure (29). To make the target populations completely disjointed and complementary, 48 children born before 35 weeks in ELFE were excluded. All methods were performed in accordance with the relevant guidelines and regulations. Ethical approval and written informed consent were obtained from each participant before enrolment in the original cohort. The study was approved by the Inserm ethics evaluation committee (n° 2020.04.24 bis_ 20.04. 22.74247, 2020 April 27), and the CNIL ((n° 920193, 2020 April 30). According to French law, the present nested survey did not require speci c additional written consent from the participant. Representatives of the participants tested and validated the questionnaires, but they did not contribute to other aspects related to the design, conduct, reporting or dissemination of the research.

Children's mental health
Symptoms of hyperactivity/inattention and emotional symptoms were ascertained by two subscales of the SDQ, a widely used measure of children's mental health which has satisfactory psychometric properties (30,31). These symptoms concerned the 15 days prior to the questionnaire and were reported by parents on a 3-point scale (not true, somewhat true, and certainly true; ranges from 0-2) to indicate the extent to which each item applied to their children (32,33). The following ve items were used to assess symptoms of hyperactivity/inattention: "Restless, overactive, cannot stay still for long"; "Constantly dgeting or squirming"; "Easily distracted, concentration wanders"; "Thinks things out before acting"; and "Sees tasks through to the end, good attention span" (32). The ve items used to assess the emotional symptoms were: "Complains of headache/stomach ache"; "Many worries, often seems worried"; "Often unhappy, down-hearted or tearful"; "Nervous or clingy in new situations, easily loses con dence"; and "Many fears, easily scared" (32). From the parents' responses to the ve items of each subscale, we calculated scores that ranged from 0 to 10 with cut-offs of the French version of the SDQ (34). Concerning hyperactivity/inattention, a score ≤5 is considered normal, equal to 6 as boundary state, and >6 as abnormal. For emotional symptoms, a score ≤3 is considered normal, equal to 4 as boundary state, and >4 as abnormal.

Covariates
Children's health behavior Health behaviors included the following: sleeping di culties (e.g., di culty falling asleep, waking up frequently or too early at night without being able to go back to sleep) since the beginning of lockdown (yes: new occurrence, increase, or stability, no: decrease, disappearance or none); time per day spent by child on the following activities: (a) reading, drawing, board games; (b) screen for recreation, (c) social network, (d) physical activity (sports or walks outside; indoor and outdoor physical activity).
Children's schooling Educational characteristics included the following: school situation before lockdown (normal; with assistance or unschooled); average time devoted to schoolwork per day (none or less than one hour, 1-3 hours, > 3 hours); tutoring di culty (tutoring with di culty, tutoring without di culty, no tutoring); average screen time for educational reasons per day. Tutoring refers to the presence of someone who can help the child with homework (e.g., parents, siblings, grandparents etc.). Screen time for educational reasons refers to time spent on television or other screens to follow school or educational programs.
Sociodemographic and socioeconomic characteristics of children's families Socioeconomic characteristics included the following: the parents' occupational category (executive, intermediate and executive, intermediate and employee, independent, laborer, 2 inactive or only one employee/laborer); change in parents' work situation (change at least for one parent, no change for either parent); parents' distance working (neither working, one teleworking and the other not working, at least one working outside, both teleworking); perceived nancial situation (a uent and constant income; a uent and declining income; modest and constant income; modest and declining income). Sociodemographic information concerning housing conditions included the following: type of housing (rural house, urban house, at with balcony/garden, at without balcony/garden, other); region classi ed according to the prevalence of COVID-19 at the beginning of the epidemic (High: Ile de France and Grand Est vs Medium: Bourgogne-Franche-Comté, Auvergne-Rhône-Alpes and Hauts-de-France vs Low: other regions); number of rooms per inhabitant and if the child lived with both parents or not.

Others
Additional data included the following: sex; chronic disease (yes, no); born pre-term (yes, no); children's access to specialized care prior to the COVID-19 epidemic (e.g., physiotherapist, speech therapist, psychologist, rehabilitation) and its continuation during school closures (yes and pursuit, yes but no pursuit, no); COVID-19 cases among household members (yes, no).

Statistical analysis
We rst described child and family sociodemographic characteristics, (i.e., weighted means with standard deviations for continuous variables; frequency with weighted percentages for categorical variables). Descriptive analyses were conducted after correcting the sample data to be representative of children born in France in 2011 (35). For this purpose, a weighting coe cient was calculated. For more information, see the online supplement.
Before performing multiple imputation, we removed variables with more than 30% missing data (screen time and physical activity) and observations with missing data for both outcomes (hyperactivity/inattention and emotional symptoms). Classi cation and Regression Tree methods (CART, imputation method for mixed data: both Continuous and Categorical) (36) for multiple imputation were used to handle missing data. Secondly, to test the association between each covariate and mental health outcomes (i.e., hyperactivity/inattention and emotional symptoms), we used multinomial logistic regressions adjusted on prematurity to estimate Odds Ratios (OR) and 95% con dence intervals (CI). Thirdly, to perform relevant covariate selection, we applied a penalized regression method, Elastic-Net, combining bootstrapping with multiple imputed data (37)(38)(39). We used the variable inclusion probability with a 50% threshold to select variables associated with children's mental health. Finally, we used multinomial logistic regression to estimate adjusted OR and 95% CI for the association between variables selected by Elastic-Net, prematurity and both mental health outcomes.
Data were analyzed using the nnet package in R (version 3.6.1) with multinomial logistic regressions speci ed using the multinom function. The mice package and its cart method were used to perform multiple imputation. Finally, the caret, glmnet, doParallel and dplyr packages were applied to implement the Elastic-Net method on imputed data.

Sample characteristics
The study included 5702 children (Table 1). In total, 2808 (50.2%) children were females, 1161 (19.4%) had access to specialized care prior to the COVID-19 epidemic, 289 (4.5%) had chronic disease, and 1122 (2.9%) were premature children. COVID-19 cases were reported in 235 (4.4%) households. Concerning the children's mental health, hyperactivity/inattention scores were elevated for 672 (13.6%) and moderate for 452 (8.2%) children, respectively. Emotional di culties scores were elevated or moderate for 404 (7.5%) and 288 (5.8%) children, respectively. In terms of children's health behaviors, 39.9% had sleeping di culties. Children spent an average of 2 hours and 20 min reading, drawing, and playing board games per day, and 2 hours and 45 min doing physical activities per day. Regarding screen time, children spent an average of 3 hours and 10 min per day for recreation, and 5 min on social networks. Only 188 (2.8%) children were unschooled or in school with assistance before the lockdown, 2764 (62.8%) spent between 1-3 hours per day on schoolwork and an average of 50 minutes on educational programs per day. 4202 (96.7%) children had a tutor at home to help them with their homework. Table 2, the pandemic situation led to a change in work situation in 1483 (43.9%) households, working from home was implemented in 2371 households (32.6%) and 1813 (41.6%) perceived nancial di culties. Children lived mostly in a house (67.4%) and with their two parents (84.5%). 9.0% and 17.3% of children lived in regions with high or moderate prevalence of COVID, respectively.

Factors associated with hyperactivity/inattention and/or emotional symptoms
The factors associated with hyperactivity/inattention and/or emotional symptoms when adjusting for prematurity are presented in Tables 3 and 4. The factors associated with an increased risk of both hyperactivity/inattention and emotional symptoms in univariate analysis were the following: having chronic disease, being born pre-term, having access to specialized care prior to the COVID-19 epidemic, having di culties in sleeping, being schooled with assistance or unschooled before the lockdown, having di culties with tutoring or not having a tutor, low parents' dominant socio-professional category, presence of change in the parents' work situation, having modest and declining outcome during the lockdown, and not living with both parents. The factor associated with a lower risk of both hyperactivity/inattention and emotional symptoms adjusted on prematurity was having at least one parent who worked outside.
Factors exclusively associated with an increased risk of hyperactivity/inattention adjusted on prematurity were the following: being a boy; having emotional symptoms. Factors associated with a decreased risk of symptoms of hyperactivity/inattention were the following: spending 1-3 hours or more than 3 hours per day devoted to schoolwork, having parents working remotely, and living in an urban environment.
In contrast, factors only associated with an increased risk of emotional symptoms adjusted on prematurity were the following: presence of COVID-19 cases in the household, hyperactivity/inattention symptoms, having modest and constant income, and living in an urban at without a balcony/garden. The only factor associated with a lower risk of emotional symptoms was being a boy.
Finally, the number of rooms per inhabitant and the French regions most affected by the COVID-19 epidemic were not associated with children's mental health.

Factors associated with children's mental health after Elastic Net selection
The factors most strongly associated with children's mental health after using Elastic Net selection are presented in Tables 5 and 6. Elastic Net selected 10 factors as most associated with hyperactivity/inattention in children: sex, access to specialized care prior to the COVID-19 epidemic, having emotional symptoms, having sleeping di culties, being schooled with assistance or unschooled before the lockdown, devoting one hour or less to school work per day, having di culties with tutoring or not having a tutor, low parents' dominant socio-professional category, living in an urban house, and not living with both parents (Table 5).
After adjusting for these 10 factors most associated with hyperactivity/inattention symptoms and prematurity, boys remained at greater risk for abnormal hyperactivity-inattention than girls (aOR 2.29; 95% CI 1.90-2.76). Children who had regular care before the COVID-19 epidemic and could not continue the sessions during the lockdown had an increased risk of abnormal hyperactivity/inattention symptoms (aOR 1.51; 95% CI 1.21-1.88).
Concerning children's mental health, compared to normal children, children with abnormal emotional symptoms or at boundary state were associated with higher risk of abnormal hyperactivity/inattention symptoms (aOR 4.06; 95% CI 3.11-5.29 and aOR 2.14; 95% CI 1.54-2.97, respectively). Regarding the children's health behavior, risk of abnormal hyperactivity/inattention symptoms or at boundary state appeared to be higher for children with sleeping di culties (OR 2.05; 95% CI 1.70-2.47, OR 1.71; 95% CI 1.39-2.10) than in those without sleeping di culties (Table 5).
Regarding children's schooling, being schooled with assistance or unschooled was associated with a higher risk of abnormal hyperactivity/inattention (aOR 2.13; 95% CI 1.43-3.17). Spending 1-3 hours per day devoted to schoolwork or more than 3 hours seemed to be associated with less risk of abnormal hyperactivity/inattention symptoms than spending no time or less than 1 hour (aOR 0.55; 95% CI 0.34-0.91 and aOR 0.56; 95% CI 0.33-0.92, respectively). In this multivariate analysis, having di culties with tutoring or not having a tutor were still associated with a higher risk of abnormal hyperactivity/inattention (aOR 3.25; 95% CI 2.64-3.99 and aOR 2.47; 95% CI 1.48-4.11, respectively).
For sociodemographic and socioeconomic characteristics, the low parents' occupational category (laborer) was still associated with a higher risk of abnormal hyperactivity/inattention as compared to the executive category (aOR 1.66; 95% CI 1.01-2.71). Compared to living in a rural house, living in an urban house was associated with less risk of abnormal hyperactivity/inattention symptoms (aOR 0.79; 95% CI 0.64-0.97).
Concerning the factors most associated with emotional symptoms, Elastic Net selected the following seven variables: being born pre-term, COVID-19 cases in the household, hyperactivity-inattention symptoms, sleeping di culties, parents' occupational category, change in parents' work situation, and nancial situation ( Table 6). The results described below are adjusted on these seven variables.

Main ndings and interpretations
Our study showed high rates of hyperactivity/inattention and emotional symptoms during school closure due to COVID-19 in France. It also helped in identifying risk factors of hyperactivity/inattention and/or emotional symptoms in children. Di culties in sleeping and parents' occupational category were the only factors independently associated in the multivariate models with both dimensions of mental disorder. Factors speci cally associated with hyperactivity/inattention were the following: sex, presence of regular care and its pursuit during school closure, emotional symptoms, school situation, presence of tutoring and di culties with it, and type of housing. Factors related to emotional symptoms were the following: being born pre-term, the presence of COVID-19 cases in the household, hyperactivity/inattention and nancial di culties.
Children's sex, chronic disease, specialized care and being born pre-term In the context of the COVID-19 crisis, our results indicate a higher risk of hyperactivity/inattention among boys than girls, and a higher risk of emotional symptoms among girls than boys. Findings by others were similar outside the context of the COVID-19 pandemic (40,41) and our ndings are consistent with those of recent studies on it (4,42,43). A cross-sectional online survey of 8079 Chinese adolescents aged 12-18 affected by the COVID-19 pandemic revealed that female gender was the most important risk factor for depressive and anxiety symptoms (4).
After adjustment on covariates, we found that children who had access to specialized care prior to the COVID-19 pandemic and who were unable to continue follow up during school closures had a higher risk of abnormal hyperactivity-inattention. Being born pre-term was associated with a higher risk of abnormal emotional symptoms during the COVID-19 pandemic. These results are coherent with epidemiological studies conducted outside the COVID-19 context which found that pre-and perinatal factors (e.g. prematurity, chronic disease) are associated with Attention-De cit/Hyperactivity Disorder (ADHD) (44-46) and emotional problems (47,48), although the de nite causes remain unknown. While being born preterm is a known risk factor for hyperactivity/inattention disorder (49,50), it is the association between prematurity and emotional symptoms that is emphasized by our ndings.

COVID-19 cases in the household
In our multivariate analyses, the presence of COVID-19 cases in the household seemed to be a risk factor of abnormal emotional symptoms in children. This was one of the strongest associations according to Elastic Net. These results are in line with a large cross-sectional online survey of 44 447 college students conducted in China (51). Compared with students who reported not having infected or suspected cases in family members and relatives, those who reported having con rmed and suspected cases in family members and relatives had a higher risk of depressive symptoms (OR 4.06; 95% CI 1.62-10.19 and OR 2.11; 95% CI 1.11-4.00, respectively) (51). Interestingly, several studies evidenced an association between emotional symptoms and COVID-19 epidemic-related factors such as potential exposure to the virus, fear of infection (i.e., self-and/or family members), and loss of loved ones (3,52,53). Mental stress, anxiety, depressive symptoms, insomnia and fear are triggered by the pandemic itself, mandatory preventive measures, and COVID-19-related circumstances (52-54).

Children's mental health
Our ndings showed that during the school closure in France, hyperactivity/inattention and emotional symptoms were associated. Elastic Net found this to be one of the strongest associations. Our results con rm those of another study conducted outside the epidemic context which showed a stronger correlation between hyperactivity/inattention and emotional symptoms in the SDQ (33).

Children's health behavior: sleep di culties
Our study is consistent with recent reports on the association between di culties in sleeping and children's mental health during the COVID-19 pandemic (13), in particular with hyperactivity/inattention and emotional symptoms. Di culty in sleeping is one of the most predictive variables of hyperactivity/inattention in children, as well as emotional symptoms. Our ndings con rm those of previous studies conducted outside the COVID-19 pandemic (55)(56)(57). Hyperactivity/inattention and emotional symptoms are both characteristics of children with ADHD. A meta-analysis of 16 studies comparing sleep in ADHD children versus controls found that the former were signi cantly more impaired than the latter on most subjective and some objective sleep measures (57). The relationship between ADHD and di culties in sleeping is bidirectional (56,58-60). ADHD may cause sleep problems and sleep problems may cause or mimic ADHD (55,56). Moreover, a reciprocal relationship between sleep quality and anxiety problems seems very likely, although the speci c mechanisms that contribute to these sleep disturbances remain unclear (61-64).
Children's schooling: school situation, time devoted to schoolwork, and tutoring In multivariate analyses, being unschooled or schooled with assistance before the lockdown was associated with a higher risk of abnormal hyperactivity/inattention. Spending 1-3 hours or more than 3 hours per day on schoolwork seemed to be associated with a lower risk of abnormal hyperactivity/inattention symptoms than spending no time or less than 1 hour. Furthermore, children without a tutor and those with tutoring di culties were at greater risk for abnormal hyperactivityinattention compared with those without tutoring di culties. Our results are consistent with those of a previous study, which found that prolonged school closure regardless of the epidemic context can cause signi cant mental health problems in children, especially among economically disadvantaged groups (65). While schools are essential for children's academic education, they also play an important role in addressing their physiological and mental health needs (65,66). However, according to Elastic Net, these schooling variables are not strongly associated with emotional symptoms.

Sociodemographic and socioeconomic characteristics of children's families
Our multivariate analyses found an increased risk of abnormal emotional symptoms in children whose families experienced nancial di culties. Children of parents in a low socio-professional category were also the most affected in terms of hyperactivity/inattention. Similar results were recently reported in a French community-based sample of 432 parents and their children during the COVID-19 lockdown (13). The authors showed that family socioeconomic factors such as nancial di culties and unemployed parents were signi cantly associated with a higher risk of children's mental health problems, such as hyperactivity/inattention and emotional symptoms (13). Stability of family income was signi cantly associated with children's psychological di culties during the COVID 19 pandemic (3), due to economic decline and the stress that it causes (67,68).

Housing
Our adjusted analysis showed that, compared to living in a rural house, living in an urban house was associated with a lower risk of hyperactivity/inattention during school closure in France. The type of housing was not associated with emotional symptoms in multivariate analysis after an Elastic Net selection. These results are contrary to most studies showing a lower risk of ADHD in children living in rural rather than urban areas (69,70). A large web-based survey of 8177 Italian students found that poor housing (i.e., living in apartments <60 m 2 with poor views and poor indoor quality) was associated with an increased risk of depressive symptoms during lockdown (71). In addition, an Iranian study which evaluated people's preferences and priorities to choose healthy homes after the COVID-19 pandemic revealed that the most critical priorities for residents were natural light, having a view, acoustic protection, and an open or semi-open space (72). Our con icting results may be explained by the greater proximity to other children in urban than in rural areas.
Contrary to ndings in the literature (17), our adjusted analysis did not nd a signi cant association between living in one of the regions of France most affected by COVID 19 and children's mental health. Finally, living with both parents was found to be protective for hyperactivity/inattention but not for emotional symptoms. Living with both parents has been shown to be a protective factor for children's mental health in the context of the COVID-19 pandemic, since a Chinese study of 7 143 college students found that living with parents was a protective factor against anxiety (OR 0.75; 95% CI 0.60-0.95) (3). A recent study on the mental health of children aged 9 to 18 in France also showed the protective effect of a two-parent family on the psychological distress of children during the COVID-19 pandemic (73).

Strengths and limitations
This study has some limitations. First, owing to too many missing data regarding screen time and physical activity, we were unable to estimate the associations between these variables and the presence of hyperactivity/inattention or emotional symptoms in children. The results could be different or even more marked in the general population. Second, only a few parents reported their alcohol and smoking consumption as well as their mental disorders, which prevented us from taking these variables into account in our analyses. Third, we did not have any objective measures of sleep quality but only a parent's self-reported measure of di culty sleeping. Fourth, we had no data on the consumption of media information related to the pandemic in minutes per day and quality of information received. Therefore, we were not able to study the association between these environmental factors with children's mental health, even though signi cant associations have already been reported (43,52,74). Fifth, our data were collected during a period of school closure in the rst COVID-19 wave in France, so we could not analyze data on children's mental health prior to the lockdown. As a result, we were unable to assess the impact of school closure on children's mental health independently from that of the COVID-19 pandemic. Sixth, we could not re-evaluate the mental health of the children, yet it is possible that the reopening of schools in France has resulted in health bene ts for them. Future study designs should include re-assessment of mental health and compare children's health scores between countries that implemented vs. did not implement school closure as a preventive measure. Future research should also investigate potential environmental factors that might in uence children's mental health, such as parental history of mental illness, consumption of media information related to COVID-19, and domestic violence. In addition, the causal relationship between environmental factors related to COVID-19 and children's mental health should be explored.
The study also has some strengths. First, it includes the large number of parent respondents during school closures in France at the beginning of the COVID 19 pandemic. It also includes pre-term children, a population more at risk for hyperactivity/inattention, so it provides more power for studying risk and protective factors in a lockdown situation. Second, we were able to control a wide range of possible covariates from sociodemographic and socioeconomic characteristics of children's families to children's schooling. To our knowledge, this is the rst study to consider a set of variables related to children's housing and schooling. Third, using validated measures of hyperactivity/inattention and emotional symptoms (SDQ) with satisfactory psychometric properties is a strength and novel aspect of this study (30,33). Overall, we believe that our ndings contribute to the evolving debate concerning school closure and children's mental health. We suggest that school closure due to COVID-19 could increase the risk of mental health problems and exacerbate health inequalities (68). Our ndings provide important guidance for the development of children's psychological support strategies in France.

Implications
The current focus on measures to contain COVID-19 transmission around the world may distract attention from emerging mental health issues (54,75). The closure of schools due to COVID-19 and the lockdown could lead to an increase in mental disorders in children, especially among vulnerable populations. Efforts must be made to address the pandemic situation while considering children's mental health and to adapt the delivery of care to meet the demands of COVID-19 (68). Our ndings have important clinical and policy implications. In this global public health emergency, identifying vulnerable groups of children and environmental risk factors of children's mental health is a critical step in developing and adapting mental health services and tools (76). Health authorities and governments should implement services that improve the mental health of the population, particularly that of vulnerable children. Emerging digital applications such as telehealth, social media, and mobile health could serve to promote children's mental health (77).

Conclusion
This study conducted in France explored several risk factors of hyperactivity/inattention and emotional symptoms in children during the school closure due to the COVID-19 pandemic. Some factors that may increase the risk of children developing mental health problems are described: sex, pre-existing health disorders, COVID cases in the household, di culties in sleeping and in schooling, and sociodemographic and socioeconomic disparities. Policymakers need to balance the pros and cons of reopening school, taking into consideration the educational and psychological consequences for children. Therefore, it is essential to implement appropriate protective measures to manage the unprecedented repercussions of school closure on children's mental health while avoiding an increase in socio-economic inequalities. The authors warmly thank all the volunteers of the ELFE, and EPIPAGE2 cohorts.
We thank the staff of the SAPRIS study group that have worked with dedication and engagement to collect and manage the data used for this study and to ensure continuing communication with the cohort participants.

Competing interests
Mrs Maeva Monnier has been funded by the region of New-Aquitaine (AMI Flash Recherche et Innovations COVID). The other authors declare no competing interests related to this paper.

Fundings
Role of the funding source Sponsor and funding sources played no role in the study design, data collection, analysis, interpretation or drafting of the study. MM and CG had full access to all data in the study.

Data availability
In regards to data availability, data of the study are protected under the protection of health data regulation set by the French National Commission on Informatics and Liberty (Commission Nationale de l'Informatique et des Libertés, CNIL). The data can be available upon reasonable request after a consultation with the steering committee of the Sapris study. The French law forbids us to provide free access to Sapris data; access could however be given by the steering committee after legal veri cation of the use of the data. ²In the absence of a spouse, the respondent who has undergone a change in work situation during the lockdown refers to «change at least for one parent» category and the one who has not undergone any change to «no change for either parent» category.
³In the absence of a spouse, the respondent who is not working is associated with the «neither working» category; the one who teleworks refers to «both teleworking» category; and the respondent who works outside to «at least one working outside» category.