Background Early life adversities may play a role in the associations observed between neighbourhood contextual factors and health behaviours.
Methods We examined whether self-reported adverse experiences in childhood (parental divorce, long-term financial difficulties, serious conflicts, serious/chronic illness or alcohol problem in the family, and frequent fear of a family member) explain the association between adulthood neighbourhood disadvantage and co-occurrence of behavioural risk factors (smoking, moderate/heavy alcohol use, physical inactivity). Study population consisted of 31 271 public sector employees from Finland. The cross-sectional associations were analysed using two-level cumulative logistic regression models.
Results Childhood adverse experiences were associated with the sum of risk factors (cumulative OR 1.32 (95% CI 1.25 to 1.40) among those reporting 3–6 vs 0 adversities). Adverse experiences did not attenuate the association between neighbourhood disadvantage and risk factors; this cumulative OR was 1.52 (95% CI 1.43 to 1.62) in the highest versus lowest quartile of neighbourhood disadvantage when not including adversities, and 1.50 (95% CI 1.40 to 1.60) when adjusted for childhood adversities. In adversity-stratified analyses those reporting 3–6 adversities had 1.60-fold (95% CI 1.42 to 1.80) likelihood of risk factors if living in the neighbourhood of the highest disadvantage, while in those with fewer adversities this likelihood was 1.09–1.34-fold (95% CI 0.98 to 1.53) (p interaction 0.07).
Conclusions Childhood adverse experiences and adulthood neighbourhood disadvantage were associated with behavioural risk factors. Childhood experiences did not explain associations between neighbourhood disadvantage and the risk factors. However, those with more adverse experiences may be susceptible for the socioeconomic conditions of neighbourhoods.
- Health Behaviour
Statistics from Altmetric.com
In recent years associations have been reported between neighbourhood socioeconomic status and several health risk behaviours that largely contribute to the burden of diseases.1 The behaviours include alcohol consumption,2 ,3 smoking2 ,4 and physical inactivity5 ,6 as well as clustering of these risk behaviours.7 ,8 Neighbourhood associations for these behaviours have been found in studies using register and cohort data when controlling for the current individual-level socioeconomic status. One randomised control trial has also provided evidence that living in a socioeconomically better neighbourhood could decrease the risk of extreme obesity and diabetes in women,9 as well as improve physical and mental health.10
In addition to the individuals’ current living environment and state of socioeconomic well-being the conditions to which they were exposed in childhood may have far-reaching effects on health. Adverse experiences in childhood (eg, long-term financial difficulties, parental divorce, alcohol problem of a family member) have been linked to measures of adult risk factors such as negative life events11 or heavy alcohol consumption,12 and socioeconomic childhood adversities have been linked to substance dependence,13 ,14 exceeding guidelines of ‘sensible’ alcohol consumption in men,15 and being a current smoker.16 ,17 A pathway from childhood financial adversity through socioeconomic position in adulthood, and smoking to poor lung function has also been established,18 and there are other pathways that could link childhood adversities, neighbourhood disadvantage and health behaviours.
Because adverse experiences and socioeconomic disadvantage in childhood are a common prior cause of adult neighbourhood and health behaviours, it is plausible that the childhood exposures confound the associations between neighbourhood disadvantage and health behaviours. For example, it is possible that due to continuity in social disadvantage19 those who have experienced adversities in childhood tend to become residents of disadvantaged neighbourhoods, and they also have poorer health behaviours.14 Furthermore, if neighbourhood disadvantage and adverse experiences in childhood were interacted, the neighbourhood associations could be contingent on adversities. However, the possible confounding and/or modifying role of childhood adversities in the associations between neighbourhoods and health behaviours have not been examined. This information would help in directing research either towards neighbourhood factors or childhood stressors in order to reduce the burden of diseases attributed to poor health behaviours in adulthood.
Therefore, we sought to determine whether the association between neighbourhood disadvantage and co-occurrence of behavioural risk factors (smoking, moderate to heavy alcohol use and physical inactivity), that we have previously reported,7 is confounded by adverse childhood experiences, and whether the associations for neighbourhoods are contingent on childhood adversities. The study population consisted of over 31 200 Finnish Public Sector (FPS) study participants who responded to a study survey in 2008 including questions about health behaviours and childhood adversity.
Participants were selected from the FPS study's register cohort that includes employees working in 10 towns and 6 hospital districts for more than 6 months in any year between 1991 and 2005 (from city mayors and doctors to semiskilled cleaners).20 Employers’ records have been used to identify the eligible employees for nested survey cohorts for which questionnaire surveys have been mailed/emailed every 4 years, starting from year 2000. For this study, we used cross-sectional questionnaire data from 2008 when information about childhood adversity was requested from the participants who were employed by the towns (response rate 71%). Information of all needed variables was available for 31 271 participants for whom the coordinates of residential address were obtained from the Population Register Center (22 214 were excluded due to missing information in any of the variables). The sex distributions of the included and excluded participants were similar: 77.6% women (included) and 77.2% women (excluded). Mean age of the included was 48.1 years and of the excluded 47.8 years. Distribution of the level of education among the included was: 39% high, 45% intermediate, 16% low; and among the excluded: 32% high, 48% intermediate, 20% low. The included did not differ markedly from the excluded in regard of neighbourhood disadvantage: 25% of the included vs 31% of the excluded lived in areas of the lowest quartile of disadvantage, and 25% of the included vs 24% of the excluded lived in areas of the highest quartile of disadvantage. The Ethics Committee of the Helsinki and Uusimaa Hospital District has approved the study.
Adverse childhood experiences
The occurrence of adverse childhood experiences was assessed on the basis of six survey questions modified from the Survey of living conditions collected by Statistics Finland.21 Respondents were asked whether they had experienced the following adversities in their childhood: divorce/separation of the parents, long-term financial difficulties in the family, serious conflicts in the family, frequent fear of a family member, serious or chronic illness of a family member and alcohol problem of a family member (response categories for each item: no; yes; or cannot say). The reliability of the summary measure of these adversities has been previously found to be good, the κ-values of responses with a 5-year interval ranging from 0.56 to 0.90,22 and shown to longitudinally predict, for example, coronary heart disease23 (particularly chronic illness in the family), asthma24 (particularly financial difficulties and chronic illness in the family) and depression.11 The distribution of the number of adversities by increasing number was: 39%, 25%, 14%, 9%, 7%, 4% and 1%. In the analyses a summary variable of the ‘yes’ answers (ie, 0, 1, 2 or 3–6 adversities) was used, as in previous studies using the same adversities and where an increased risk of incident coronary heart disease (CHD), asthma and depression was found among those reporting three or more childhood adversities.11 ,23 ,24 Other cut points (eg, ≥4) have also been presented in prior literature when using different definitions for adversities,25 and therefore we run a sensitivity analysis using ≥4 as the upper cut point.
Data on the characteristics of residential neighbourhoods was derived from Statistics Finland's grid database.26 This database contains demographic area-level information that is based on the total population of Finland at the time of data collection in 2008–2009. We calculated neighbourhood disadvantage score for each 250×250 m map grid (a neighbourhood) using information on median household income (coded inversely), education attainment (proportion of those with low education), and unemployment rate, with lower scores indicating lower disadvantage. In the neighbourhoods of the highest versus lowest disadvantage quartile the median annual income ranged from €27 649 to €67 993, unemployment rate ranged from 2.3% to 14.1%, and the proportion of low educated people from 36.7% to 14.2%. For each of the three variables, we derived a standardised z-score (mean=0, SD=1). The summary disadvantage score was then calculated by taking the mean value across all z-scores.7 Missing data for the 250×250 m neighbourhoods (ie, information on income and education was confidential if <10 households within a square), were replaced using information from the eight surrounding neighbourhoods (n=264). The means of the indicator variables in the surrounding neighbourhoods were used for calculating the summary score for neighbourhood disadvantage.
We determined the existence of three behavioural risk factors based on the questionnaire responses. Smoking status was measured with two questions: (1) ‘Do you smoke or have you previously smoked regularly, that is, daily or nearly daily?’ and (2) ‘If you have smoked, do you still smoke regularly?’, and dichotomised as ‘current smoker versus other’. We also requested habitual alcohol consumption that was used to determine moderate to heavy alcohol use. For men, >24 drinks (each 12 g of pure alcohol) per week, and for women >16 drinks per week was used as the lower limit of moderate/heavy use (yes vs no).27 Physical activity was measured by the Metabolic Equivalent Task (MET) index and was expressed as the summed score of MET hours per day. Reporting less than two MET hours per day (corresponds to 30 min brisk walking28) indicated physical inactivity (yes vs no). The main outcome variable was a sum of these three dichotomised behaviour-related risk factors: (1) being a current smoker, (2) moderate/heavy alcohol use and (3) leisure-time physical inactivity, and this summary variable had four classes: 0, 1, 2 and 3 risks.
Marital status (living alone vs married/cohabiting) was also inquired in the surveys. Information on age, sex and occupational title was obtained from the employers’ registers. As in our earlier studies7 ,29 we used the Classification of Occupations by Statistics Finland,30 an established classification system, to classify individuals into three occupational positions: the high=upper grade non-manual workers, intermediate=lower grade non-manual workers and the low=manual workers. Other markers of the participants’ socioeconomic status were the level of education (high=university degree, intermediate=high school or vocational school, low=comprehensive school), obtained from Statistics Finland, and housing tenure that indicates household wealth (owner vs other), from the Population Register Center. Information on the number of inhabitants per neighbourhood was obtained from the Statistics Finland's grid database, and used as a proxy for the degree of urbanisation in the neighbourhood.
The odds of family adversities in childhood by adult neighbourhood disadvantage was examined with age and sex adjusted binomial logistic regression models (GLIMMIX procedure of SAS V.9.2), and the results are presented as ORs with 95% CI. To examine whether neighbourhood disadvantage and childhood adversity associate with co-occurrence of behavioural health risk factors, and whether childhood adversity attenuates the association between neighbourhood disadvantage and the risk factors, two-level cumulative (ordinal) logistic regression models (GLIMMIX procedure of SAS V.9.2) were used.
‘Model 1’ was used to estimate the association between neighbourhood disadvantage and co-occurrence of behavioural risk factors, and the association between childhood adversity and the risk factors adjusting for age and sex. In age and sex adjusted ‘Model 2’ both exposures were included simultaneously, and ‘Model 3’ was further adjusted for the other covariates; marital status, individual socioeconomic characteristics (occupational position, education and housing tenure) and number of inhabitants in the neighbourhood. Results are presented as cumulative ORs (CORs, ie, the average of three logistic comparisons: ≥1 vs <1, ≥2 vs <2 and 3 vs <3 risk factors) with 95% CI (p value for the Score Test for the Proportional Odds Assumption was 0.74 indicating that the proportional assumption holds). To examine if there were differences in the associations between neighbourhoods and clustering of risk behaviours according to the level of childhood adversities, we calculated the CORs of risk sum with a model including an interaction term (risk factors ∼ neighbourhood disadvantage+adversity+neighbourhood disadvantage×adversity), using the least deprivation-no adversity group as the reference. As sensitivity analyses we examined the possible confounding by individual adversity items (yes/no), and the associations for individual risk factors separately. Because of the skewed sex distribution of the data, we also stratified analyses by sex.
The mean age of the study population was 48.1 (SD 9.1) years and the proportion of women was 77.6%. Descriptive statistics of the population by childhood adversities are presented in table 1. The neighbourhoods in which the participants lived had on average 258 (1–2521) inhabitants. Participants with no childhood adversities were more prevalent in the least (42.7%) versus most (35.7%) disadvantaged neighbourhoods, whereas those with three to six adversities were more prevalent in the most (23.5%) vs least (19.0%) disadvantaged neighbourhoods (table 1). Prevalence of unadjusted co-occurrence of the three behavioural risk factors among those reporting three to six adversities was 0.9%, whereas the prevalence of no risk factors was 54.4%. Among those with zero adversities, the corresponding percentages were 0.7% and 60.9%, respectively. By neighbourhood disadvantage, the prevalence of three risk factors was 1.3% in the most disadvantage quartile, whereas the prevalence of no risk factors was 52.0% (table 2). In the age and sex adjusted logistic regression models the ORs of individual adversity items increased with increasing neighbourhood disadvantage (table 3) suggesting those with adversities can end up living in more disadvantaged neighbourhoods. Results using the fully adjusted model were more consistent for women than men but overall very similar (see supplementary table 1).
Neighbourhood disadvantage was associated with the co-occurrence of risk factors with an age and sex adjusted COR of 1.52 (95% CI 1.43 to 1.62 for the highest vs lowest quartile of disadvantage). When additionally adjusted for childhood adversity, the association remained similar (COR 1.50; 95% CI 1.40 to 1.60), but further adjustments for the other individual-level and area-level covariates (Model 3) attenuated this association (table 4). The results remained the same when we used an adversity measure with the highest exposure group ≥4 (COR for Q4 1.22; 95% CI 1.14 to 1.31). Associations were slightly stronger among men than women, but the patterns of associations were similar for both sexes (see supplementary table 2). None of the individual adversity items confounded the association between neighbourhood disadvantage and co-occurrence of risk factors. These age, sex and adversity adjusted CORs for risk factors ranged from 1.50 to 1.52 (highest vs lowest quartile of disadvantage, data not shown). There was a positive association also between childhood adversity and the sum of behavioural risk factors in the age and sex adjusted model (COR 1.34; 95% CI 1.26 to 1.42, for three to six adversities vs zero adversities). After adjustment for neighbourhood disadvantage the association remained, but when all other covariates were added the association attenuated slightly (COR: 1.25; 95% CI 1.18 to 1.33).
Since there appeared to be a statistical interaction between adversity and neighbourhood disadvantage (p=0.07), we estimated the associations between neighbourhoods and risk behaviour clustering according to the level of childhood adversities. In the group of no adversities there was a non-significant increase in the association between disadvantage and risk factors, whereas among those with adversities, increasing levels of neighbourhood disadvantage linearly associated with the co-occurrence of behavioural risk factors (figure 1). One quartile increase in neighbourhood disadvantage increased the cumulative odds of risk factors in the ‘one adversity’ group by 7%, in the ‘two adversities’ group by 5%, and in the ‘three to six adversities’ group by 10%. Compared with those with no adversities and living in the most affluent neighbourhoods, those with no adversities but living in the most disadvantaged neighbourhood had COR of 1.09 (95% CI 0.98 to 1.21) for the co-occurrence of risk factors, while in those exposed to three to six adversities the corresponding CORs were 1.15 (95% CI 1.02 to 1.31) when living in the most affluent neighbourhood, and 1.60 (95% CI 1.42 to 1.80) when living in the most disadvantaged neighbourhood (figure 1) (for all effect estimates by adversity groups and by sex see etable 3). The patterns for co-occurrence of risk factors by neighbourhood and childhood adversities were similar when using cut point ≥4 for the adversities (see supplementary figure 1).
When we considered each of the individual risk factors separately, we found that neighbourhood disadvantage associated with smoking and physical inactivity, but not with moderate/heavy alcohol use (see etables 4–6). Childhood adversity associated with all individual risk factors, but it did not confound any of the associations between neighbourhood disadvantage and individual behavioural risk factors.
Neighbourhood disadvantage and adverse experiences in childhood were associated with co-occurrence of behavioural health risk factors in adulthood even when controlling for each other. These findings suggest that the evidence for an association between neighbourhood socioeconomic factors with health behaviours is not confounded by early life exposures, at least in these Finnish data. We found evidence of effect modification of the association between neighbourhood disadvantage and health behaviour according to childhood circumstances: the deleterious effect of living in a distressed neighbourhood was amplified by adverse experiences in childhood. This finding suggests another, understudied mechanism through which childhood experiences may affect public health.
Only a few studies have examined the role of childhood experiences in the neighbourhood associations. One study reported that family socioeconomic status was not a confounder, not even a significant covariate, in the analyses of associations between neighbourhood disadvantage and age-crime curves in youth.31 Another study suggested that ‘supportive parenting’ was mediating the association between neighbourhood socioeconomic status and children's antisocial behaviour.32 However, neither of these studies examined health behaviours in adults as we did. In a US study, childhood exposure was measured as self-reported physical, sexual and emotional abuse, and this kind of child maltreatment was found to modify the association between living in a disordered neighbourhood and binge drinking.33 This agrees with the current findings, but as in that study the childhood exposure was abuse, and the study population was dominated by African Americans,33 the results are not directly comparable with ours.
Our findings indicate that adverse childhood experiences and current neighbourhood disadvantage have comparable associations with the co-occurrence of behavioural risk factors in adulthood. The long-term health effects of childhood experiences may be related to differences in health-promoting parenting as financial difficulties or parental distress may affect children's health behaviors34 that may have long-lasting effects on health. One pathway suggested for the neighbourhood disadvantage—health behaviour association is that the initiation and maintenance of healthy behaviours is more difficult for people living in disadvantaged vs wealthy neighbourhoods because in their daily lives they are exposed to behaviours such as smoking or public drinking.35 A common pathway for both observed associations can be through social stressors. In childhood, exposure to social class related stressors may affect biological systems and thus health in the long-term36; in the neighbourhood level, high crime rates, for example, may increase smoking prevalence37 and low safety may decrease physical activity.38
Our study population consisted of FPS employees and it was female-dominated, mainly of Caucasian ethnicity, and included employed persons from a welfare state, which limits the generalisability of the findings. Thus, these results should be confirmed in other populations and in countries where the variation in the disadvantage between neighbourhoods is larger and people have lower level of social security through life. Another limitation is that we lacked detailed information on residential history of the participants. Neighbourhood conditions in childhood could be very influential in the formation of lifestyle habits—for example, smoking initiation, which occurs in most people before the age of 18 years.39 Additionally, we used cross-sectional self-reported data on childhood adversities and health behaviours both of which are prone to recall bias and under-reporting.40–42 However, in large cohort studies such as ours objective measures of health behaviours is seldom if ever available. If either one or both measures were under-reported, our results may have been underestimations of the true associations. When using retrospective measures there is also a chance of differential misclassification error that may limit the validity of the data. We had no information of severe adversities such as abuse, which may have led to exposure misclassification. In addition, although there is some evidence that the reliability of the self-reported measure of childhood adversity is good,22 only prospective studies beginning from childhood could assess the validity of self-reported adversity. However, we had large number of participants reporting childhood experiences, accurate data on objectively measured current area-level disadvantage, and adequate control for the current individual-level socioeconomic status.
In conclusion, these findings suggest that the associations between neighbourhood disadvantage and health risk behaviours are not confounded by exposure to adverse experiences in childhood in a context of Scandinavian welfare society. However, those not having experienced adversities may be more resilient to the effects of neighbourhood deprivation than those exposed to several childhood adversities, a group which seemed to be vulnerable to the adverse neighbourhoods’ socioeconomic effects.
What is already known on this subject
Childhood adverse experiences have been linked to adverse health outcomes in adulthood.
Poor neighbourhood conditions in adulthood have also been associated with adverse health in adulthood.
What this study adds
When simultaneously analysed, childhood adverse experiences and poor socioeconomic status of adult neighbourhood were associated with co-occurrence of poor health behaviours in adulthood.
Although childhood experiences did not remove the association between neighbourhood socioeconomic status and poor health behaviours, those exposed to several adverse experiences may be more vulnerable to the effects of poor neighbourhood conditions in adulthood.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Data supplement 1 - Online tables
Contributors JIH, JV and SVS designed the study, directed its implementation, including quality assurance and control. JIH performed the analyses, did the literature review and drafted the manuscript. JP, JV and SVS designed the study's analytic strategy and critically reviewed the manuscript. MK and IK helped to prepare the Materials and Methods and the Discussion sections of the article and critically reviewed the manuscript. All authors have approved the submission of this manuscript version.
Funding This work was supported by the EU Era-Age2 programme (Academy of Finland grant 264944) and the participating organisations and by the Bupa Foundation, UK, and the National Institute on Aging (R01AG034454-01).
Competing interests None.
Ethics approval The Ethics Committee of the Helsinki and Uusimaa Hospital District.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement The study approval by The Ethics Committee of the Helsinki and Uusimaa Hospital District does not allow distribution of the data.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.