4. Discussion
The results of this study highlight the association between lockdown and poor sleep quality in the Italian population. These results suggest that lockdown might impact sleep quality and that this association could be mediated by crucial determinants of health, such as physical activity and diet, and to a lesser extent, smoking and alcohol consumption, the presence of financial problems, and symptoms of psychological distress. Indeed, anxiety was widespread among the Italian population during lockdown, as were feelings of uncertainty, fear, and loneliness [9,10]. This impact on sleep quality was seen especially among females, those with low education level, and those who experienced financial problems.
Since the beginning of the pandemic, observational, mostly cross-sectional, studies have been conducted around the world to study lifestyle changes in populations subjected to restrictions. While sleep quality has been frequently investigated, results have not always been consistent. In China, Wang and colleagues found that 75% of 2289 individuals isolated at home rated their sleep as very good [11], but a smaller prospective study found that 37% of a sample of young Chinese adults reported a worsening of their sleep quality during the pandemic [12]. Studies conducted in other countries have demonstrated that lockdown could either lead to an increase in the amount of time spent in bed [13,14], with good sleep efficiency [13] or associated with sleep disorders [15,16] and disruption of one’s habitual circadian rhythm [14]. This topic has also been investigated in Europe, where 55% of Spanish adults changed their sleep pattern after the beginning of the lockdown [17], and individuals who were physically active prior to the SARS-CoV-2 spread reported major sleep problems [18]. In Italy, cross-sectional investigations have suggested that while the number of hours spent in bed might have increased during lockdown [2], many individuals reported insomnia (43%) [19] and poor sleep quality (57%) [9]. Similar percentages of sleep deterioration (48%) were also shown in the UK [20], and a longitudinal study conducted in Italy confirmed the detrimental effects of the 40-day lockdown on all the parameters of the Sleep Quality Index [21], except for sleep duration [22]. Our cohort seems to confirm that sleep quality deteriorated during lockdown, as among those individuals who experienced a change in their sleep quality, 85.8% defined their sleep as poor. This proportion of individuals accounted for 46% of the entire cohort investigated, which is strikingly similar to the proportion of those who reported a worsening of sleep quality in the above-mentioned Italian studies.
Our results corroborate the findings of previous research that the worsening of sleep quality during lockdown was more pronounced among females [9,23,24]. Poor sleep quality was strongly associated with the presence of many financial problems, although this group in our cohort accounted for a very small proportion of participants (2.1%). However, poor sleep quality was not associated with the interruption of work. One possible explanation for these apparently inconsistent data could be because of the economic measures promptly introduced by the Italian Government to counter the economic effects of the forced closure of most businesses. Similarly, a large survey conducted in the UK found that poor sleep quality during lockdown was associated with perceived financial problems, but not with unemployment [23].
In this study, 35.1% of individuals reported a worsening of their level of physical activity, which proved to be a determinant of poor sleep quality in the multivariate analysis. This result was not surprising, as a moderate effect of regular exercise on overall sleep quality is well documented [25,26,27], and the association between less exercise and sleep disorders was confirmed during lockdown in a sample of German individuals [16].
In addition, 17.6% of individuals declared a change for the worse in their eating habits, which was associated with poor sleep quality. This proportion of individuals who adopted a less healthy diet is lower than the 35.8% estimated by Di Renzo and colleagues in a cross-sectional survey, which involved 3533 Italians throughout the country [2]. In this regard, it should be noted that 18.5% of our sample introduced changes in their diet that we categorized as mixed, because they were neither completely positive nor completely negative. It is plausible that some of these mixed changes, upon further investigation, could be reclassified as worse. In fact, the association between sleep quality and this type of mixed change in eating habits went in the same direction as that shown by the worsening of one’s diet. Still, regardless of the accuracy of the categorization for this predictive variable, it seems certain that several individuals’ diets worsened and that this was associated with poorer sleep. This also seems to be true outside of Italy, as data collected from more than 1000 individuals in three different continents, including Europe, which found that unhealthy diet increased by 10% during the period of social isolation, and that worse eating habits were significantly associated with a worsening in sleep quality [15].
Finally, our results confirm the association between poor sleep quality and symptoms of psychological distress. The positive relationship between anxiety and poor sleep quality has been extensively investigated and was confirmed during lockdown [9,20,26]. While the cross-sectional design of our study does not allow us to verify whether the symptoms of psychological distress that we documented were related to the restrictions imposed to curtail the spread of the SARS-CoV-2 virus, a causal relationship is plausible, as also suggested by the literature [9,10,15,16,20,28]. The convenience sample investigated was recruited in the Province of Reggio Emilia, one of the most affected Italian provinces during the early phase of the pandemic. Considering that, at the time of the survey, little or nothing was known about the disease and its prognosis, it is plausible that feelings of fear and uncertainty were more widespread in the populations living in areas with a high prevalence of infection [9,29]. Added to this were the concerns arising from the forced temporary closure of many business and industries in one of the most productive areas of Italy and of Europe. The COVID-19 pandemic has called into question the stability of the Italian National Health Service, also triggering uncertainties regarding the continuity of care for thousands of individuals, often elderly and with comorbidities, in one of the longest-lived countries in the world. Therefore, the association between poor sleep quality and feelings of fear and uncertainty was expected.
Less obvious was the relationship between poor quality sleep and feelings of loneliness as, based on the univariate analysis conducted, living alone was not a predictor of poor sleep quality. Thus, loneliness was perceived regardless of the presence of other household members; the reasons underlying this feeling must be sought in other causalities. We have no explanation for this, but it is possible that forced living with others in close quarters for an extended period may have worn down the quality of relationships between individuals, or loneliness may have been due to the forced separation from important, vulnerable family members. Moreover, the sense of loneliness could also have derived from the need to limit interpersonal contact to a minimum, a measure that, while necessary to reduce the contagion, is in stark contrast with Italian culture.

Limitations and Strengths
The results of this study should be interpreted with caution as they derive from a cross-sectional design, which makes causal inferencing challenging. Moreover, the survey sampling was based on an online invitation, which does not allow for generalization because the population that does not use the Internet was not explored.
Indeed, a recruitment bias emerged in our sample, as females and working-age individuals with a high education level are overrepresented compared to the general Italian population. While the overrepresentation of females is a feature common to other similar studies [2,9,16,20,28,30], it is likely that the biases observed in age and education reflect a high level of adhesion of participants who are also healthcare professionals or administrative staff of the LHA or of the major municipalities of the Province of Reggio Emilia, which disseminated the survey through their social media and websites.
Moreover, the outcome of interest and data related to predictive variables were collected through self-reporting rather than through a valid survey or clinical assessment. This was because, as we wanted to capture the changes that occurred during lockdown through a cross-sectional design, we chose specific questions to comply with this purpose.
Finally, we decided to observe the association between each change occurring during the lockdown, adjusting only for those variables that certainly could not change during the study period and cannot be intermediate effectors of the other putative exposure. However, this approach does not allow us to rule out that some of the observed associations were not independent and could have been due to confounding or could have been mediated by other putative exposures.
Despite these limitations, this study had a relatively high acceptance in our community. The sampling strategy allowed us to collect data from quite a large sample, which in fact numerically mirrors that investigated by other similar study designs conducted to explore the same theme [9,15,28,30]. Further, this was the only sampling method feasible during lockdown.