To our knowledge, this study is the first to identify adolescent perception subgroups during lockdown, using an unsupervised approach based on many variables. Three perception clusters can be distinguished. A cluster composed of almost two-fifths of the adolescents had good mental health and did not feel stressed during lockdown. A cluster representing one-fifth of the adolescents was mainly unhappy, felt stressed, showed signs of depression, suffered from overcrowded living conditions, and experienced worsening family relationships during lockdown. A third group with moderate well-being and stress was more supported by family and worried about the health of their relatives. These three clusters differed with regard to their behaviors during lockdown. Compared with the unstressed cluster, the group of unhappy adolescents were more affected by COVID-19, had difficulty doing school activities, spent more time on social networks and less time on video games; they also more often reported decreased sleep time and a deterioration in their diet. The cluster of adolescents with moderate well-being, including more girls, spent more time on social networks, was more physically active, slept less, and more often reported eating fruit and vegetables, and drinking alcohol.

Overall, the adolescents in this study appeared to be healthy, mildly anxious, and to have a well-being score (average WEMWBS score = 49.0 [SD = 8.3]) close to that observed in English and Scottish teenage school students (48.8 [SD = 6.8]) [25]. However, the use of a cluster method allowed us to identify different subgroups of adolescents who did not experience lockdown in the same way, in line with previous results in French adults [26]. A significant portion of the adolescents (39%) seemed resilient and were only minimally affected by lockdown. They experienced high well-being and low stress, perhaps due to greater social support from family, less deterioration in the family’s financial situation, and less deterioration in their school situation [27]. In contrast, another group of adolescents (19%) reported a moderate to high overall stress level and symptoms of depression: unhappiness, loss of interest, low optimism, low self-esteem, and loss of energy, as found in other studies [5]. This could be explained in part through lack of support from their entourage, increased family tensions, and the social distancing inherent in lockdown [6, 8, 28]. In fact, these adolescents suffered from loneliness, overcrowded households, and more often experienced a deterioration in relationships with their relatives. The final subgroup of adolescents were anxious, especially about COVID-19 (higher level of worry about their health or the health of their relatives, higher coronavirus-related stress), but they were better supported; they more often improved their relationships with their family, and were more often understood and helped by their parents and teachers. This could be explained in part by the predominantly female composition of this subgroup. Indeed, girls are known to be generally more anxious and to have more prosocial interactions than boys [13, 29].

During lockdown, adolescents generally increased their sedentary activities (78%), especially the time spent on screens: watching movies, videos, television; using social networks; playing video games; working remotely or practicing online sports. Consistent with other studies, adolescents from the PARIS birth cohort reported an average of 6 h and 24 min spent on screens per day, including almost 2 h on social networks and 2 h playing video games [14, 19]. Similar to our results, Albrecht et al. reported an increased sleep time and a sleep duration of 9 h on school days and 9.75 h on free days for Swiss adolescents during lockdown, which is in line with sleep recommendations for adolescents [15, 30]. With regard to physical activity, most adolescents reported a decrease; however, 27% increased their indoor sports activities, which is slightly higher than Ng et al., who described a 20% increase in physical activity [31]. Our results may be due to the high SES of PARIS families. Indeed, it has been shown that high parental education and high family income are associated with physical activity [32]. Furthermore, as described in previous studies and related to the increase in sedentary activities, 10% of adolescents declared that their diet had deteriorated, and 49% had increased their consumption of snacks, chips, candies, or pastries [33, 34]. Alcohol and tobacco use appeared to be lower than in other studies [35, 36], possibly due to the high SES of PARIS families, being locked down with parents, a decrease in social activities, and underreporting [37, 38]. Finally, about half of the adolescents had difficulty doing school activities, reporting a loss of motivation, difficulties concentrating, or the unavailability of teachers. Remote learning is known to increase distraction and decrease the interaction with teachers for adolescents [39]. This phenomenon may have been exacerbated in adolescents who had learning difficulties prior to lockdown.

This study shows that depressed, anxious, stressed, and lonely adolescents spent more time on social networks during lockdown. Indeed, compared with cluster “happy” of adolescents least affected by lockdown, the most unhappy, stressed, and lonely adolescents (cluster ”unhappy”) and, to a lesser extent, the anxious adolescents in cluster “intermediate” spent more time on social networks and less time playing video games. As noted by Cauberghe et al., this could be explained by the need of anxious or lonely adolescents to use social networks to cope with the situation and to keep in touch with their loved ones [18]. Unfortunately, as shown in several studies, the use of social networks did not seem to increase the feeling of well-being [40]. Similarly, adolescents in clusters “unhappy” and “intermediate” more often experienced decreased sleep time compared with cluster “happy”. This confirms results from several studies showing an association between decreased well-being in adolescents and decreased sleep time during lockdown [13, 15].

Furthermore, our results indicated that adolescents in cluster “unhappy” with symptoms of depression, stress, and overcrowded living conditions were more likely to be possible COVID-19 cases, despite their relatively low coronavirus-related stress. As suggested by Ren et al., a stricter lockdown of adolescents with COVID-19 may have led to a deterioration in their well-being [7]. In the meantime, cluster “unhappy” was associated with a poorer diet. This finding is supported by existing literature showing that stress eating increased due to the negative impacts of lockdown [17]. The association with difficulty doing school activities could be attributed to a poor social environment and greater overcrowding in the household, resulting in concentration difficulties. These adolescents had less support from their parents and teachers, which could impact their transition to remote learning.

Finally, compared with cluster ”happy”, adolescents in cluster “intermediate” were mainly girls, more often increased their level of physical activity, and consumed more fruit, vegetables, and alcohol. Higher anxiety levels have been previously shown in girls [8, 29]. Moreover, adolescents in cluster “intermediate” appeared to be more concerned about their health and adopted a healthier lifestyle. They increased their consumption of fruit and vegetables, which could also be explained by a greater presence of relatives at a time when families were more eager to cook. Nevertheless, they consumed more alcohol than other clusters, perhaps due to greater social drinking with family or through video calls with friends.

This study has multiple strengths. First, to our knowledge, this is one of the first studies to investigate in such detail both perceptions and behaviors of adolescents during lockdown. Second, the use of the unsupervised clustering method is innovative and allowed us to highlight different subgroups of adolescent perceptions never published before. The use of online questionnaires at the end of lockdown made it possible to cover the whole period of the first lockdown. Despite the unavailability of baseline levels, questions were asked about improvement and deterioration in perceptions and behaviors since the beginning of lockdown. Nevertheless, the cross-sectional design of the study makes it impossible to draw conclusions about the direction of the associations between perception profiles and behaviors. Moreover, this study focuses on Parisian families mainly of high SES and living in an urban environment; therefore, these findings cannot be generalized to all French families. Lastly, lower income families less participated in the survey which could result in underestimating the “unhappy” cluster size.