The health of children encompasses much more than “the absence of disease” and has been described as a state of wellbeing with physical, mental and social dimensions.1 This understanding of health anticipates individual potential expressed through the life course. Childhood exposure to factors and mediators of health and development should therefore be considered with reference to medium and longer-term outcomes. Childhood residential mobility is one such factor that may be examined in terms of its association with health outcomes.

Population residential mobility rates vary internationally. Using age-standardised data, a higher proportion was reported moving in one year in New Zealand (19.6%), the United States (17.5%) and Australia (17.0%) compared with European countries such as Great Britain (10.6%), Sweden (9.5%) and Ireland (6.4%).2 These census-based statistics underestimate by not specifically accounting for individuals moving more than once in the year. Higher rates were demonstrated for children aged 1–4 years, declining until adolescence, increasing in the early twenties before decreasing again through adulthood. This pattern is observed throughout developed countries, reflecting a greater relative exposure of children and youth to relocation. Higher rates of residential mobility are associated with poverty,3 4 housing tenure, unemployment,5 family disruption and single parenting.6

Residential mobility is variably defined and quantified by researchers, using descriptors such as distance moved, reason for shift, frequency, attributes of neighbourhoods moved to or from and time since residential change. The suitability of the descriptor depends on the model being tested.7 Frequency of moves is commonly used and may investigate for a dose-effect. Distance moved may be linked to the degree of social disruption. Reasons for moving may identify other contributing factors and confounders. Time since move considers the potential for adjustment and attenuation of stress effects. Social dimensions of neighbourhood and underlying reasons for moves further describe the broader context interacting with health outcomes.

The relationship of socioeconomic deprivation and health is established.8 Underlying this empirical observation, however, is the complex interplay of factors that include, to name a few, physical housing conditions,9 economic deprivation, social disadvantage, family characteristics, neighbourhood attributes and residential mobility. Given the complex interweave of factors in this area of research, confounding challenges any analysis. Nevertheless, analysis has been attempted to consider any specific contribution of residential mobility to outcomes. The importance of social determinants remains paramount. Theoretical models broadly categorised as factors operating at individual, family and neighbourhood levels may contribute to the understanding of how residential mobility in childhood relates to health, which begs the question as to how health is conceptualised. Biological perspectives consider the physiological and psychological responses of the individual child to the stress of change.10 Relocation involving school or neighbourhood change may be experienced at an individual level as a life event stressor in which environmental change demands adaptation.11 Within operational models of life event stress the child’s cognitive appraisal of events and related explanations may moderate the child’s adaptive response and behavioural adjustment.11 12 A significant difference for children not so well addressed in the life stress literature is that usually they are not choosing to move residence. Lifespan development models also consider successful adaptation as a normal and potentially beneficial developmental process.13 Societal perspectives at the family level consider the role of family stressors including socioeconomic circumstances and disruption of family structure.14

The life course perspective considers the cumulative impact of multiple, overlapping factors operating over time to produce across the population a differential distribution of health status with a social gradient.15 The potential contribution of residential mobility during key developmental phases is considered.

From a practical point of view the relocation of children as a consequence of family break-up is common.16 Adjustment processes and capacity in the family affect the child’s experience of events. At the neighbourhood level high rates of mobility may operate by affecting perceptions of neighbourhood, social integration, support networks and social control. This concept of social capital has been described as “the social cohesion of community, sense of belonging and the level of involvement of community affairs”17 and may be protective against the health damaging consequences of poverty.18 Health outcomes may also be mediated by the impact on health service utilisation including continuity of care. Paradoxically, residential mobility may affect inclusion in studies potentially obscuring these children from research. The hypothesis examined by this review was that residential mobility experienced during childhood has an adverse effect on health outcomes after adjusting for other contributing factors including socioeconomic position.

METHODS

Search strategy

The following databases were searched in March 2005: MEDLINE (1966–2005), EMBASE (1974–2005), British Nursing Index (1994–2005), CINAHL (1982–2005), PsychINFO (1967–2005), ISI Web of Knowledge. The search terms used were “residential mobility” OR “residential instability” OR “geographic* mobility” OR “geographic relocation” OR “residential relocation” OR “residential stability”. An inclusive stance was taken towards residential mobility definitions (eg, change of address within or outside a specified region, frequency over a period of time or in lifetime) providing a clear description was offered within the study considered. Focussing on narrower definitions was considered counterproductive given the limited available studies. International relocations were excluded as these often involve major cultural and language change requiring specific analysis. Limits were set to age range “0–18 years” during which the residential mobility was measured. Given the intersectoral nature of this question, the likelihood of research existing outside major databases was recognised. The grey literature was searched using the same search terms through Internet-based search engines: Campbell collaboration and Google Scholar. References from identified studies were also checked for other relevant research.

Inclusion criteria

The authors (an academic professor and community paediatrician masters student) identified and reviewed papers for this systematic review. Titles and abstracts were screened by one reviewer (TJ) to identify studies with residential mobility analysed as an independent risk factor and with defined health outcomes. Candidate studies were then read by two reviewers (TJ, NS) and selected by criteria: (1) residential mobility during ages 0–18 years clearly defined as an independent variable; (2) health outcomes defined and measured with a reproducible method. Within each study, meaningful comparisons were sought. For example, comparisons of movers versus non-movers or frequent versus less frequent movers would allow examination for differences in dependent variables. Non-English language literature was not excluded. Exclusions were as follows: (1) studies involving only international relocation; (2) studies of migration to seek asylum; (3) studies considering only educational outcomes.

Quality assessment

The methodological quality of each selected study was independently assessed by each reviewer. A quality scoring tool was developed by consensus between the authors with reference to principles established in the literature.41 Documentation of the following was sought: aims, recruitment, setting, health measurement tools, comorbidities, demographics and objective measurement of both residential mobility and health outcomes; the inclusion of known confounders in an appropriate analysis and adequate description of “loss to follow-up” were also key criteria sought. Physical and psychological health outcomes were acceptable for inclusion if they were measured using validated and thus reproducible methods. In particular, the potential for bias threatening internal validity was considered.

Quality was graded using three descriptive categories: (1) “Sound”: clear aims, sampling well described, follow-up analyzed and confounders considered; (2) “Adequate”: moderate potential for bias identified; (3) “Suboptimal”: limited internal validity with high potential for bias.

Data extraction and analysis

Key data including study design, population sample, mobility definitions, health outcomes, confounders, biases, statistical analyses and findings were extracted using a data form. In each study the relationship between a defined residential mobility measure and health outcome was sought. All studies considered exposure to residential mobility in childhood, but were grouped by age phase during which the health-related outcome was measured (preschool, school age, adolescent and youth and adult). Studies differed in many respects with widely differing outcomes and measurement tools, and various study designs reflecting different underlying theoretical models and so meta-analysis was not considered appropriate. Effect sizes were presented as stated by the authors of the respective papers and further calculations were not undertaken.

RESULTS

Studies identified

Initial electronic database searches yielded 679 papers. Thirty-three studies were identified as potentially relevant (25 from electronic databases and eight from grey literature sources). Following full text reading of these there were 11 further exclusions, leaving 22 studies for inclusion (fig 1).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 Summary of search results.

There was adequate agreement between the reviewers in the quality grading assigned to each study (κ 0.59, 95% CI 0.31 to 0.87).

Characteristics of the studies

Twenty-two studies were included in the final analysis. Seventeen studies were identified with health outcomes in four age phases (preschool, school age, adolescence/youth and adulthood) and five studies measured health service utilisation as an outcome. Study designs included cross-sectional survey (11), cohort studies (4), longitudinal (4), case–control (1) and case series (2). Many drew on data from much larger national childhood datasets and surveys. There was a range of definitions of residential mobility used including number of moves, time since last shift and neighbourhood turnover. Similarly, there was a range of health outcomes sought in the different age phases. This variation in design and methodology is detailed in table 1.

Findings of the studies

Infant and preschool outcomes

Three studies specifically investigated early childhood outcomes (table 2). One case–control study investigating associations with sudden infant death syndrome19 identified an increased risk among infants away from their usual domicile, but no link with change of domicile compared with matched controls. A small, detailed cross-sectional US study of 70 children in the Head Start programme20 found among the movers higher levels of sibling conflict but no significant differences with behaviour, depression or cognitive competence. External validity was limited, being based on a small, highly selective, volunteer sample in the rural United States. A study of inpatients of a burns unit reported higher rates of address change but lack of adjustment for key confounders threatened internal validity.21

School age outcomes

Six studies, two longitudinal and four cross-sectional designs, investigated the association of residential mobility with school age health outcomes, focusing predominantly on behavioural problems (table 2). A large Canadian cross-sectional survey of children aged 11–12 years found that the number of lifetime moves was associated with increased behaviour problem rates by exacerbating pre-existing risk factors.22 In multivariate analysis moving three or more times was associated with increased indirect aggression, property offences and alcohol use by 12 years of age. This was a large, nationally representative sample and key confounders were addressed. Validity was, however, threatened by self-report bias relating to health outcomes, potentially reducing the effect measured. Two studies utilised data from the US 1998 National Interview Survey of Child Health investigating behavioural outcomes.23 24 Both studies depended on parental report potentially introducing recall bias. Key confounders were, however, considered and the sample was large and representative. Moving more than three times was associated with increased rates of behaviour problems23 24 and correspondingly an increased requirement for professional psychological help.24 Delay in growth or development and learning disability in frequent movers had odds ratios greater than unity but did not reach statistical significance.23 Utilising data from the second generation of the UK 1958 Birth Cohort Study, Verropoulou et al25 did not find an association with childhood aggression or anxiety, identifying other mediators of human, financial and social capital as more important. Sample attrition was relatively high at 21%, with those lost to study systematically more likely to be residentially mobile. Beyers et al26 assessed residential instability as the proportions of renter-occupiers and of households who had been resident less than 5 years. Residential instability at this neighbourhood level was not found to have a significant relationship with externalising behaviour in 11–13 year olds. Sampling was by randomised invitation, with 75% uptake in a longitudinal multi-site US Child Development Project study and selective attrition resulted in a predominantly middle income European sample limiting external validity. This longitudinal study incorporated key confounders. A study of US military families investigated behaviour problems and social adjustment identifying time since relocation as more important than the number of relocations.27 Validity was threatened by a small sample (n  =  86) and lack of detail regarding recruitment.

Adolescent and young adult health outcomes

Six identified studies investigated youth health in relation to residential mobility (table 3). Teenage premarital pregnancy was correlated with the number of residential moves.28 The contribution of moving among other factors of neighbourhood disadvantage, family structure and living arrangements was small but statistically significant given the large sample (n  =  1361, 6063 person-years). The early initiation of illicit drug use and marijuana-related problems was significantly associated with four or more geographical relocations before 16 years of age in a young adult population sample drawn from the Ontario Mental Health Supplement.29 Confounders were analysed. A long period of respondent recall may have reduced reliability in this cross-sectional design study. Increased residential moves in the preceding 5 years were associated with adolescent adjustment problems among 267 African-American female adolescents in high poverty mid-West US neighbourhoods.30 Limitations of correlational design and the use of retrospective reports for disruption variables and outcome measures argue against overstating identified associations. Stack32 also studied residential moves in adolescence, finding higher rates of premarital sexual activity reported. Report bias is likely in the accounting of this particular health outcome. In the Providence, Rhode Island Cohort Study (USA) residential mobility before 7 years of age was associated with an increased risk of depression in early teens but with early remission when encountered.31 Recall bias and lack of detail on those lost to follow-up to some extent threatened validity. Higher residential mobility was observed in a small study of inpatient psychiatric patients33 but the lack of analysis of confounders and retrospective case series design was a limitation.

Adult health outcomes

Two studies examined childhood residential mobility and adult outcomes (table 3). Juon et al35 examined mortality rates and did not find a significant association. Sample attrition is likely to have contributed systematic bias. Bures34 surveyed self-reported health status suggesting an association with general health, whereas the association with mental health reported was not statistically significant. The validation of the questionnaire used was not described.

Health service utilisation

Five studies examined residential mobility and health service utilisation (table 4). Mustard et al36 observed lower levels of continuity of medical care in a cohort study based on the US 1988 National Health Interview Survey in which mobility was measured using database maternal postcode changes. Another analysis of the same dataset by Fowler et al38 found that children aged up to 17 years who had moved more than twice in their lifetime at the point of interview were more likely to lack a regular site for healthcare and to use an emergency department for “sick care”. A small study of mothers and children in sheltered accommodation similarly found an excess use of emergency departments for medical care among more frequent movers.39 Conversely, a study in Finland found no significant association of address change with utilisation of primary health services.37 A report from the Providence Plan Group40 suggested that increased moves were associated with an increase in providers seen and reduced primary healthcare visits inferring reduced continuity. Further analysis to consider confounding was not, however, undertaken.

Sensitivity analysis

The effect on overall evidence of removing studies graded with lower quality was examined. Generally, the lower quality studies tended to find no association of outcomes with residential mobility and so their exclusion would strengthen support for the hypothesis.

DISCUSSION

This review examined childhood residential mobility and the health outcomes in those children later in their life course. Evidence was identified in particular for an association of behavioural and emotional problems in school age children with residential mobility. Twelve of the 17 health outcome studies investigated school age and adolescence and suggested increased behavioural disturbance, poorer emotional adjustment, increased teenage pregnancy rates, earlier illicit drug use, drug-related problems and teenage depression. Overall, evidence for preschool and adult outcomes was less established. There were five studies in these age groups, with methodological issues threatening validity of this evidence. Research findings on health service utilisation were mixed, with an adverse effect of mobility observed in US research, contrasting with a Finnish study finding no significant effect. This distinction may primarily reflect health service organisation differences.

Wide variation in measures of residential mobility, age groups of samples and aspects of health assessed demonstrated differences in underlying theoretical models. All the studies examined outcomes at the individual level. Residential mobility was predominantly defined at the individual level, with the exception of one study that used a neighbourhood measure.27 Confounders from a number of levels were analysed and included individual (eg, ethnicity, age, and sex), family (eg, size, parental mental health, single parenthood, family disruption measures) and community (eg, analysis by region). Socioeconomic position as a factor impacting at all levels from individual to societal is a fundamental confounder and therefore its consideration was one key measure of quality in the reading of the studies. Definitions of residential mobility could be enhanced with the inclusion of further information about the drivers or reasons for moving better to inform the conclusions regarding associated health outcomes. The quality of evidence is determined by the contributing research. Studies identified in this review were all observational, with the attendant issue of confounding. In particular, adjustment for socioeconomic status and other confounders was central to quality assessment. Some variables including family disruption and neighbourhood-level factors were accounted for in only some studies. Housing quality and crowding as factors were notably absent. When studies utilised questionnaire methods recall bias has been suggested. On first principles household moves are major events and so are associated with accurate recall but this effect is difficult to quantify further. Other research indicates the following: less recent, shorter duration moves are remembered less; coincidence with other salient life course events improves recall; asking more than one member of the household improves accuracy and interviewer assessment of response quality has predictive value.42

Publication bias was evidenced by the lack of non-English language studies despite their inclusion in the search strategy. Seventeen of the 22 studies were US-based, reflecting both the relatively high rate of residential mobility in the United States and research activity. No studies came from developing countries. The extent to which the findings from predominantly US studies may be generalised to other country settings is tempered by a number of considerations. Characteristics such as the differing relative importance of drivers of residential mobility and social attitudes to moving house may variably confound the outcomes measured.

Evidence for a causal relationship could be considered on a number of criteria. Five of the studies identified a positive dose-effect in terms of mobility and outcomes.20 26 28 30 32 Four of the studies similarly had found an effect that reached statistical significance at three or more moves.22^–24 31In these same studies a moderate strength of association was demonstrated, with odds ratios in the range 1.3–2.6. Theoretical discussions such as life-stress models or the neighbourhood social control lend plausibility to there being a causal relationship. The evidence, however, remains based on observational and predominantly cross-sectional research; alternative explanations may exist, given the likelihood of unaccounted confounders. There was a notable absence of positive health measures in the studies. Researchers have potentially neglected a continuum inherent in definitions that conceptualise health as not merely the absence of disease.

No other systematic reviews were identified. Other discussions in the literature have considered residential mobility and health. Tooley43 reviewed geographical mobility as a sociological factor understanding the child’s move as a “pivotal incident”, with both negative and positive impacts reported. Stubblefield,44 using a psychodynamic framework, emphasised the isolation affecting children who change neighbourhoods. Stokols and Shumaker45 identified problems in the literature consistent with methodological issues encountered in this review: inadequate conceptualisation of mobility; failure to delineate the psychological context and over-simplification of outcome analyses. Schmitz et al46 reviewed literature indicating that family homelessness compounds the effects of adversity. Adam47 argued that research into children’s family environments has tended to take a “snapshot” of environmental quality and that few studies take account of the “amount of change” as a variable of interest. Adam47 concluded that increased residential mobility is a predictor for social, emotional and behavioural problems after controlling for family characteristics; this concurs with the findings of our systematic review.

CONCLUSIONS

Moving house is a common experience in childhood, particularly in countries such as the United States and New Zealand, where general residential mobility rates are among the highest in the developed countries. Some children are exposed to high levels of relocation and this systematic review examined related health outcomes. After adjustment for confounders, high rates of moving were associated with increased behavioural problems during childhood and adolescence. Effects seen during childhood included increased indirect aggression, property offences and behavioural problems requiring psychological help. Outcomes in adolescence with identified association included earlier initiation of drug use and related problems, earlier onset of depression and, among adolescent girls, an increased risk of both premarital sexual behaviour and teenage pregnancy.

The contexts of moving are diverse, including parental employment factors, perceptions about neighbourhoods, housing tenure and changes in family size and structure. Factors are linked in complex intertwined ways to social and economic settings. Whereas the bulk of evidence relates to individual-based models incorporating family factors and considering individual outcomes, some support is found for interactions at a neighbourhood level exacerbating or buffering the impact of moving.

There is some limited support for a causal relationship. Nevertheless, residential mobility serves as a marker for some adverse health outcomes. Although challenging to define and measure, residential mobility is likely to be important in future research examining the “life course accumulation of disadvantage”48 or investigating the development of resilience.

Some research found an association with reduced continuity of healthcare provision. Recognising the aforesaid health associations, this argues for the re-orientation of services to identify children exposed to cumulative residential relocations and effectively provide preventive and responsive healthcare. This review specifically, and perhaps narrowly, considered health outcomes. Linkage between education and health is acknowledged in an ecological understanding of child development and wellbeing. As has been demonstrated in the educational sphere,49 by bringing to light the “mobile” population, services might be better placed to mitigate the mechanisms whereby transience contributes to social exclusion and related health effects.

Further elaboration of linkages between reasons for moving, rates of mobility and outcomes would help guide policy. At the broadest level socioeconomic gradients in health are well established but mechanisms of effect remain elusive.50 Residential mobility is linked with poverty, particularly in some populations and so, along with other related markers, forms part of a profile that is associated with poorer health outcomes and health service utilisation. This review highlights residential mobility as one of many candidate pathways between poverty and health, with the potential to inform policies that affect housing and household economics.

Residential mobility interacts at neighbourhood, family and individual levels in cumulative and compounding ways, with significance for the wellbeing of children. It is one dimension of a child’s environment that needs purposeful consideration in research design, the development of effective child health services and evaluation of policy.