This study examined the associations of sociodemographic and socioeconomic circumstances with sleep duration and insomnia-related symptoms among nationally representative sample of Finnish adults. The main findings were as follows:
1) Childhood socioeconomic position was mostly unassociated with adult sleep.
2) After full adjustments, the associations between sleep and sociodemographic or socioeconomic factors were attenuated, except for marital status, household income and employment status which remained associated with sleep duration and insomnia-related symptoms. Sleep is shorter and insomnia-related symptoms are more prevalent among the divorced and separated as compared to married adults. However, income and employment status were the most consistent determinants of short and long sleep duration and insomnia-related symptoms. Thus, those with low household income levels, the unemployed, and disability retirees were the most likely to report poor sleep.
3) Having small or adolescent children was associated with better sleep.
4) A clear gradient was observed in many associations regarding occasional and frequent insomnia-related symptoms. Correspondingly, disadvantaged social position was mostly related to short and long sleep duration, and the strength of the association increased toward the extreme ends of the sleep duration distribution.
Poor sleep in adulthood may reflect childhood circumstances, chronic problems, and adversity. Accordingly, recent studies have shown that several problems including economic difficulties in childhood are associated with adult sleep quality  and insomnia-related symptoms . Contrary to our findings, the association between low parental education and difficulties falling asleep remained in an earlier study, when current socioeconomic position as measured by education and occupation had been accounted for . However, no association was found for sleep maintenance. As our measure did not distinguish between different insomnia symptoms, it is difficult to directly compare the findings. However, in line with this study, parental education has not been associated with insomnia-related symptoms . While lower socioeconomic position in childhood has been linked with chronic diseases and mortality on adulthood,  most examined diseases such as coronary heart diseases may take decades to develop. Thus, this may explain the lack of an association between childhood socioeconomic position and sleep, as insomnia symptoms can emerge in adulthood, and in addition to changes in health status, they can be attributable to strenuous working conditions, and different life situations, for example. In all, due to the crudeness of our measure on parental education, more comprehensive data about childhood socioeconomic position is needed to confirm its significance to adult sleep. It is also possible that otherwise disadvantaged childhood circumstances could be important to sleep in adults.
One might have assumed that parents of young children have more sleep problems and sleep less. Unexpectedly, we found that having children was associated with fewer insomnia-related symptoms than not having children. The age range examined did not allow us to separate infants from toddlers or older children. Thus it is possible the requested age (under 7-years) was too inclusive and differences between parents of infants and other children remained undetected. Nonetheless, similar inverse association between number of children in a household and frequent insomnia-related symptoms has also been reported in a British cohort . Even though the models were adjusted for age, it is possible that this cannot fully account for the fact that insomnia-related symptoms were rare among younger adults (with children) and much more prevalent among older adults in our cohort as well as in the British cohort . However, similar association was found when we restricted the analyses to younger (30-45 years old) participants (data not shown). Furthermore, it was repeated among both mothers and fathers. Nonetheless, while it is practically evident that infants disrupt especially their mothers’ sleep, teenaged children may cause worry and anxiety by keeping their mothers awake waiting for their children to return home, for example . It is also possible that parents report their sleep as it were without possible disruptions by their children. As our stratified analyses showed differences in sleep between participants with and without children, which were not explained by age or gender, the reasons for these associations require further scrutiny.
In line with previous studies [1–6], sleep duration and insomnia-related symptoms varied according to gender, age, and marital status. Thus, sleep was shorter and insomnia-related symptoms were more prevalent among older adults and among single, divorced, and widowed adults. However, healthy elderly people have been reported to sleep as well as their younger counterparts [50, 51]. Thus, sleep duration and insomnia-related symptoms should not only be seen as a function of age; they are likely to be attributable to other causal factors . Overall, these findings highlight the need of increasing our understanding about the importance of sleep for ageing people and those living alone. Additionally, even though the higher prevalence of insomnia-related symptoms among women is in line with previous evidence, the sociodemographic patterning of sleep was similar between genders as judged by the lack of interactions. Since we focused on multiple social determinants of both sleep duration and insomnia-related symptoms more detailed examination of gender differences was both unfeasible and beyond the scope of this study.
Although residential area was not associated with insomnia-related symptoms, those living in densely populated or rural municipalities tended to sleep more often ≥10 hours than those living in urban towns. Studies of residential area and sleep duration are sparse, but this association might be related to different age and occupational structures among residential areas in Finland. For example, people living in rural areas tend to be older, and more often farmers. In line with our findings concerning an association between long sleep duration and living outside urban cities, a previous study reported that both average sleep duration and subjective sleep need were slightly higher among those living in rural areas as compared to those living in urban areas .
Despite clear sociodemographic differences in sleep, deviation from population mean sleep duration, and insomnia-related symptoms varied most consistently with socioeconomic circumstances. The associations between poorer sleep and low income were in line with previous evidence, although the associations were particularly strong in our cohort [1, 3]. Unemployment has also been shown to be associated with sleep duration and insomnia-related symptoms in several studies [1, 2]. However, the associations between sleep and retirement are complex. Thus, although sleep may tend to improve after old age retirement, for example due to the removal of work-related stressors, sleep among disability retirees is poor before and after retirement [52, 53]. Accordingly, we found associations only between disability retirement and occasional and frequent insomnia-related symptoms, whereas the sleep quality of old age retirees did not differ from the employed participants. Finally, further material resources, such as housing tenure and economic difficulties, may contribute to poorer sleep and partly account for sleep inequalities according to income and employment status [3, 7, 54, 55]. As data concerning broader material circumstances were unavailable for this study further scrutiny is needed to corroborate these findings.
This study showed that after mutual adjustment, many associations among sociodemographic and socioeconomic circumstances and sleep are attenuated. This suggests that part of the effects of other determinants is mediated through other determinants. Since we were able to include a broad range of sociodemographic variables, the results show which of the associations remain when the effects of other determinants have been taken into account. Thus this wider approach provided a more detailed understanding on the production of socioeconomic inequalities in sleep and highlighted the importance to consider multiple socioeconomic circumstances simultaneously. After adjustments, the results showed the importance of e.g. income and employment status to sleep over and above the effects of other indicators. This examination of multiple socioeconomic circumstances further confirms that the socioeconomic indicators are not interchangeable, but each indicator has a specific nature and reflects particular socioeconomic circumstances across lifecourse [56–59]. Most previous studies have focused on one or a few indicators as determinants of sleep, but our results suggest the importance of considering multiple indicators simultaneously. Among the adult Finnish population, income and employment status are key socioeconomic determinants of sleep, and are likely to explain the associations between low education and poor sleep, for example. Because most of the associations remained after adjustment for health status, this highlights the role of social factors in poor sleep. Adjusting for health might also bias the estimates if sleep, among other factors, mediates the association among sociodemographic and socioeconomic circumstances and poor health, as suggested earlier [3, 60]. As our aim was to focus on a range of sociodemographic determinants of both insomnia-related symptoms and sleep duration, inclusion of a full array of explanatory factors was beyond the scope of this study. However, in previous studies [3, 7], several potential explanations for the found socioeconomic inequalities in sleep have been considered. For example, the associations between unemployment, low income and poor sleep could be partly accounted for by worries and stress in addition to poorer health status . Education, in turn, reflects health-related values, behaviours, and attitudes which can be assumed to include sleeping habits as well. High educated are assumed to have better knowledge on the means to improve sleep, importance of sleep to health, and they may also more actively seek help and treatment to their insomnia or sleep deprivation [3, 61]. Albeit adjusting for such potential explanatory variables has resulted in attenuation of the associations between sociodemographics and sleep, most of the associations remained.
Finally, our statistical models revealed a clear gradient in many associations, which is in line with previous evidence linking the highest morbidity and mortality risks to the extreme ends of sleep duration distribution . Furthermore, health risks indicated by work disability, for example, are higher for frequent as compared to occasional insomnia symptoms [18, 19, 62]. In order to promote better health and well-being, our results also highlight the importance to focus on the milder insomnia-related symptoms as well as the more serious ones. As such, insomnia symptoms are highly prevalent in the population and early detection and better identification of such symptoms and risk groups could have a notable effect on public health.
Strengths and limitations
Several limitations need to be acknowledged. First, this study was cross-sectional, which makes it impossible to infer on causality of observed associations. Two-way associations between sleep and socioeconomic position are conceivable. Disadvantaged position is likely to be a determinant of sleep due to financial strain and related stress for example. It is also possible that poor sleep, as part of a medical condition that is severe enough to affect the global functioning in the long-term, leads to disability retirement [18, 62]. However, as social welfare and health services are relatively good in Finland as compared to many other countries, and all population groups have access to health care, consequences of poor sleep do not necessarily imply varying health cost by socioeconomic position as much as in some other countries. Although we took into account health status in this study, the association between sleep and health is complex, and it is difficult to interpret the effects of health adjustments, and separate primary insomnia from comorbid conditions.
Second, we only had a single item measure for insomnia-related symptoms. The item has, however, been shown to have relevant psychometric properties (associations with other sleep-related variables and outcome variables) in several our previous studies [34, 40, 63–66]. Such single items have also been shown to have important predictive value for various physical and mental health outcomes [33, 67, 68]. Although our sleep measures were not validated, validity of similar self-reported items has been assessed in several previous studies [69, 70]. Since insomnia-related symptoms tend to be persistent , a further limitation of this study was that data about childhood sleep duration and sleep-related problems were not included. However, childhood socioeconomic position could be taken into account, and the main focus of this study was on the associations among current sociodemographic and socioeconomic circumstances, sleep duration, and insomnia-related symptoms among the adult Finnish population.
Third, retrospective data about parental socioeconomic position was used. The validity of such retrospective reports may be questionable and may vary according to the age of the respondent. A review of the validity of retrospective responses showed that such reports can be used, although responses concerning adverse conditions are likely to be substantially biased [72, 73]. As our measure is not focused on interpretation, such as experiencing financial difficulty, its validity and reliability is likely to be better. Since our key sociodemographic and socioeconomic determinants of sleep were based on current position, these results are less prone to bias. Furthermore, all indicators asked about concrete details of sociodemographic and socioeconomic circumstances that did not involve judgments about personal situation, perceived conditions, or experiences of socioeconomic disadvantage that are more difficult to interpret.
Fourth, since income and employment status, in particular unemployment, had the most consistent associations with sleep duration and insomnia-related symptoms, this raises a question about the mechanisms and role of economic difficulties. Previous studies suggest that economic difficulties exist at all income levels, even among affluent employed populations [74, 75], and have adverse effects on sleep , and other behaviours  after other socioeconomic circumstances are taken into account. Thus, it is possible that the associations found in this study are also explained by greater economic difficulties and related financial and other psychosocial strain.
Fifth, the number of cases with complete data for all sociodemographic and socioeconomic circumstances and sleep varied slightly among our models. However, complete case analyses produced similar results to those reported in our study, suggesting that the estimates are valid. Further control analysis including missing responses as a separate category also produced similar results (data not shown). Thus, we preferred to retain the full sample and use all data available for each analysis without redundant exclusions.
Sixth, ethnicity or race were not assessed as part of the sociodemographic framework as in many other studies [1, 2, 48]. However, ethnic groups are very small in the population (less than 1%) and the data can be considered very homogeneous with this respect.
Seventh, it might be assumed that the use of hypnotic drugs interferes with the reported associations among sleep duration, insomnia-related symptoms and sociodemographic and socioeconomic circumstances. For example, if hypnotic drug use reduced insomnia-related symptoms and lengthened sleep, and if such medication was disproportionally distributed among socioeconomic groups, this would distort our examination of the associations among sleep and sociodemographic and socioeconomic circumstances. Nonetheless, we conducted control analyses adjusting for hypnotic use (data not shown). The contribution hypnotic drugs made to the associations among sociodemographic and socioeconomic circumstances and sleep was small. Prevalence of hypnotic drug use was 7% and it only partly captures insomnia-related symptoms in these data. Additionally, alcohol may be used as sleep aid, but it also adversely affects sleep maintenance . However, the associations are complex [77–79]. In our sensitivity analyses, adjustment for alcohol had negligible contribution to the examined associations. Further examination of the associations between sleep and alcohol drinking patterns was out of the scope of this study.
Eighth, it should be noted that as these data were collected a decade ago, it is possible that changes occurred in e.g. use of electric media could adversely affect sleep and limit generalizability of the findings to the current situation [80–82]. However, it is of note that our focus was on lifecourse sociodemographic determinants of insomnia-related symptoms and sleep duration among adult Finns. These determinants and patterns of the associations are unlikely to be largely affected by the changes in the use of electronic media. If such media usage is disproportionally distributed across the examined socioeconomic groups, this might suggest that our estimates for e.g. high educated high income participants with potentially more exposure to electronic media are conservative or that the inequalities might be narrowed. Further research is needed to elaborate these issues.
The strength of this study is that we used a large amount of nationally representative data about the adult population. The results can, therefore, be generalized at the population level in Finland. Generalizability to countries with different socioeconomic structure should be cautious. Moreover, since the data are representative of the general Finnish population, these results provide novel evidence about the distribution of sleep duration among population subgroups and point key groups for insomnia-related symptoms. We also used weighted analyses to improve the generalizability of the findings. A further strength is the availability of a variety of sociodemographic and socioeconomic circumstances over life course. We had data about childhood socioeconomic position and we were therefore able to show that the associations among current socioeconomic circumstances are independent of childhood socioeconomic position. Additionally, we were able to focus on two key characteristics of sleep: quantity and quality.