Based on an approach of multiple stressors, we investigated if social neighbourhood characteristics such as unemployment and residential turnover were associated with insomnia and whether these associations were modified by individual social characteristics and change of residence.
In the whole sample, we observed a statistically significant association between neighbourhood unemployment and insomnia. If neighbourhood unemployment encompasses multiple contextual stressors, this finding could hint at precarious situations in regard to psychosocial and other urban stressors influencing residents’ sleep quality in deprived neighbourhoods. Our study region is the Ruhr Area, a formerly highly industrialised area which has experienced massive deindustrialisation during the past decades. Long-term unemployment, financial strain and associated disorder and stigmatisation certainly affect both individuals and neighbourhood communities. Sleep is termed an adaptive behaviour
] and as a consequence of an uneasy and stressful neighbourhood life, a physiological mal-adaption could have a sleep-disturbing outcome expressed by difficulties falling asleep, maintaining sleep and early morning arousals. Because of their structural persistence, contextual stressors are capable of directly inducing stress bringing on insomnia. This mechanism may be mediated and modified by psychosocial-cognitive factors – resources and stressors alike – relating to residents’ perception and appraisal of their neighbourhood situation and consecutive behavioural responses. A lack of mastery and control transmitted by the social neighbourhood context could be connected to stress-related health outcomes and risk behaviour and further deteriorate sleep quality, for example
Despite its weakness, the association between months lived under high residential turnover and insomnia might still indicate the existence of neighbourhood influences which result in negative emotional states detrimental to sleep (e.g.
] for the effect of positive affect and psychological well-being). In particular, social networks and ties could fail to balance problems which emerge from an unstable social neighbourhood context. In deprived areas where population decline might have led to social fragmentation tendencies, suicide, use of drugs, disorders due to alcohol use, and assaults have been identified as leading causes of mortality
], which might give credit to the idea of residential stability being beneficial for health. We do not believe that selective migration had greatly exacerbated the effect of neighbourhood unemployment on insomnia, for correlations of residential turnover and neighbourhood unemployment were merely moderate. Regarding migration in our study cities in the Ruhr Area, though, another aspect should be taken into consideration: In some inner city areas, residential turnover is highly correlated with (younger) foreign residents and migrants who are much more affected by socioeconomic disadvantage and health risks than their German counterparts. Owing to our explorative study design tentatively showing a concurrence of higher neighbourhood unemployment and residential turnover levels, we do not presume to discard the social isolation hypothesis
adopted by others
], as mentioned in the theoretical background of this paper.
Characteristics of urban life are far too complex to be captured by neighbourhood unemployment and residential turnover alone. It is a limitation of our study that we did not know exposures inside home as well as other sleep-related environmental resources and stressors beside noise. Here, the overall null result of long-term exposure to night-time noise might, in part, result from an imprecise exposure estimation using isophone values instead of the spatially more resolved façade values and from exposure misclassification due to lack of knowledge about residential characteristics (noise protection windows, location of bedroom and living room), personal habits (time spent at home, opening of windows, etc.) and personal characteristics such as hearing loss. However, the tendency of insomniacs having endured noise for a longer cumulative time might still point to noise effects adding up to psychosocial stress associated with high neighbourhood unemployment and residential turnover.
Even if we cannot infer Ruhr Area-specific combinations of stressors from our crude contextual data basis and even if we cannot uncover specific biopsychosocial mechanisms from our analyses, it is plausible to partially ascribe insomnia as well as sleep-adverse behaviour and psychosomatic factors to neighbourhood-related living conditions.
Our approach of multiply effective stressors was further confirmed by our subgroup analyses, with each individual social characteristic suggesting a dimension of potential ‘stress’ on its own. We may assume that maintaining a high sleep quality while being exposed to an adverse urban neighbourhood context could be hindered by disadvantages in income, educational attainment and social isolation. Urban societies are shaped by market exchange and price regulations, regimes of civil rights and redistributional welfare, reciprocally social resources in social networks as well as their interdependencies. If individual social characteristics constitute a vulnerable position in relation to these realms, they consequently account for locally unequal exposures to stressors and resources, health capabilities, including psychosocial constructs like self-efficacy, attitudes towards health and collective lifestyles, and health outcomes (cf.
], cf. also with a special focus on systemic health inequalities).
The pronounced association of neighbourhood unemployment with insomnia in participants with low income supports the assumption of individual social stressors reinforcing the effect of neighbourhood deprivation on sleep and health. In participants with lower educational degrees, inconsistent associations of social neighbourhood characteristics with insomnia might not necessarily imply the irrelevance of the urban residential neighbourhood context. For instance, in less deprived neighbourhoods low education could mean being hampered in accessing available resources to one’s own health benefit or sleep hygiene practices if social thresholds are perceived as too high. In face of high neighbourhood unemployment, low education could also add to the impression of one’s own powerlessness to positively influence the neighbourhood situation. In consideration of the noise effect estimates in this subgroup, institutional rules such as legal rights for environmental protection might be difficult to interpret and apply in order to encounter physical stressors at the place of residence actively and successfully, for example. Moreover, our results give evidence of a high vulnerability of socially isolated persons who could have serious troubles to bolster up enough economic, cultural and social resources to care for a healthy diet
], to assert themselves in their daily (neighbourhood) life or to mobilise public attention to health damaging housing conditions. Helplessness and incapability to access adequate health resources might be reasonable intermediate psychosocial stressors in the relationship between neighbourhood unemployment and insomnia. At the same time, individual social characteristics did not seem to increase vulnerability to residential turnover. This finding raises the question whether multiple individual problems combined with neighbourhood unemployment outweighs potential effects of other socially unsettling processes engendered by residential turnover. Correlations between mean neighbourhood unemployment rates and months under residential turnover showed a slight tendency for an increased double neighbourhood exposure among the subgroups, however, which might have amplified the effects of higher neighbourhood unemployment levels on insomnia.
Furthermore, we studied change of residence as an effect modifier in the association of neighbourhood contextual characteristics with insomnia. Effect estimates for neighbourhood exposures gained strength in participants who had not changed residence (non-movers). By contrast, we could not detect any effects of neighbourhood contextual characteristics on insomnia in movers. First of all, this finding supports our hypothesis of time shaping the relationship between neighbourhood and health, i.e. that an enduring exposure is more likely to negatively impact sleep quality. Moreover, in participants familiar with changing their socio-spatial networks, relations and resources as a consequence of their own residential mobility, the neighbourhood context might be of less importance for well-being and health. What is more, this finding presses us to explore reasons, patterns and effects of change of residence in the future. In order to assess the ‘harmfulness’ of social segregation for residents’ social exclusion, the capabilities of the ‘moving’ households are for urban planners and sociologists decisive to know. In view of the relatively high percentage of employed movers (48.9% vs. 40.3%, results not shown in Table
), changes of residence will partially have to do with economic activity, but the motives are hidden and could be economic pressures, health problems or simply changes in need and lifestyle specific to the stage in life-course. The data did not allow us to follow past socio-spatial residential mobilities, and we could not tell whether a change of residential address meant a change of neighbourhood and exposure. This is a crucial aspect, since the material conditions of place of residence might infringe on life chances in the long term, especially for those “not (yet) in a ‘clearly weak’ socioeconomic position” (
], p. 184). Yet, we do not assume drastic up- or downward changes in the participants’ socioeconomic status. Movers had merely a bit more experience with higher neighbourhood unemployment levels than non-movers (28.1% and 25.0% vs. 26.9% and 23.1% in the third and forth level, results not shown in Table
). Maybe, age structure of the cohort accounts for this result, because residents’ socioeconomic life-courses are determined earlier in life than at the ages of 35–65. In relation to the mean age of the study population, our retrospective observation period of ten years time might be too short. Also, there was no association between individual relocations in the past and insomnia at present (Table
). Most probably, migration patterns in our sample were not driven by insomnia, though we cannot tell if this was the case with other medical conditions and leave this subject for future studies. Researchers should bear in mind then that psychosocial stress ensuing from difficult life situations, insomnia, and chronic ill-health could be well related to migration (cf.
] for the case of rural–urban migrants in China). However, studies on urban neighbourhood exposures, change of residence and insomnia are largely missing which makes it difficult to give a solid and evidence-based interpretation of the results.
At this point, we like to draw the reader’s attention to two major methodological facets, i.e. the usage of statistical units as proxies for neighbourhoods, residential history as well as the operationalisation and modelling of neighbourhood characteristics.
No doubt the statistical units we referred to as neighbourhoods will not necessarily correspond to the places involved in the social interactions and time-activity-patterns relevant for residents’ sleep- and health-related resources and capabilities. Heterogeneous in origin and boundaries, these statistical units will contain various constellations of stressors and resources. Consequently, effects of neighbourhood characteristics might be underestimated. However: “A neighbourhood need not be homogenous to affect the lives of its inhabitants. In fact, complete homogeneity within in area precludes the study of contextual effects altogether” (
] p. 112). In a cross-sectional neighbourhood study with HNR study data
], effects of neighbourhood unemployment on coronary calcification did not alter when a smaller statistical unit was used.
Alongside the need to consider life-course trajectories, epidemiology has become interested in residential history approaches and modelling cumulative exposures
]. We tried to capture this aspect by relying on a ten year exposure history based on administrative data. The public source of data acquisition proved to be quite trustworthy for two cities: Firstly, residential history data for the participants was almost complete, with an effective rate of 95.6 per cent (1680 out of 1757) and 99.0 per cent (1640 out of 1656) for participants from Mülheim and Essen respectively. Secondly, outcome and socioeconomic characteristics of the study sample with a complete residential history data did not substantially deviate from the whole study population including participants with incomplete residential information. Thirdly, the residential mobility is obviously typical for the age groups covered in our analysis. In a study dealing with social inequalities and residential choice in another city in Germany, Cologne, ca. 24 per cent of the participants of the same age had changed their home address in a ten years time (
], special analysis). So, the residential history data enabled us to consider exposure duration while investigating the potential effect of moves and to recognise possible social differences underlying individual changes of residence, e.g. in movers, non-movers, and in those participants with a disrupted residential history. This latter subsample excluded from our analysis turned out to be a specific group in the population which should be thought of when interpreting results from cross-sectional studies: Incompleteness seemed to signal rather dynamic life stages, as if these participants had not been settled yet, even after moving into the study region.
Modelling of long term neighbourhood exposures has become rather sophisticated, using multiple-year measures
], indices of cumulative area-based socioeconomic environments
], neighbourhood trajectories and hierarchical latent growth curve modelling
], for instance. Lack of retrospective socioeconomic neighbourhood indicators prevented us from setting up a true time series and adopting such complex approaches. Yet, we managed to build a socioeconomic neighbourhood status variable overcoming exposure misclassifications owing to changes of residence. Correlations of neighbourhood unemployment and residential turnover as well as the expertise of the municipal departments in charge of statistics and urban research revealed relatively stable city structures throughout the 1990s, which compensates in part the flaw in socioeconomic data availability. Furthermore, we developed a dynamic measure of residential turnover in the neighbourhood, though a bimodal distribution of exposure to residential turnover and small observation numbers in monthly exposure strata thwarted differentiating exposure durations in more detail. In principle, our measure of residential turnover is rather unique: As far as we know, residential (in-)stability has been mostly analysed in cross-sectional studies and is frequently operationalised as the percentage of residents having been resident in a neighbourhood for a certain time span, e.g. for five years of time. Although implying a perspective into the past, this previously used measure does not allow for long-term analysis relying on several points in time. Such a summary measure is necessary to quantify the cumulative impact of social neighbourhood contextual characteristics. Our study would have demanded a larger sample, however, in order to scoop out the dynamic of this measurement approach.
From 2000 onwards, German cities have been growing more and more fragmented, but public socio-spatial monitoring has been much improved, partly as a reaction to this development of social segregation. Prospectively, downward spirals wearing away neighbourhood resources may become more and more traceable as selective population dynamics are developing at a quicker pace. For this reason, our study continues to assess neighbourhood contextual exposures during follow-up, creating a longitudinal design at both individual and neighbourhood level, testing more precise neighbourhood exposure accumulation measures, and exploring the relationship between neighbourhood unemployment and residential turnover in an enlarged database. In this way, we might be able to identify critical neighbourhood windows in the ongoing process of aging, because growing economic and physical inactivity in daily life are thought to enhance susceptibilities.