Our study tested the hypotheses that perceiving neighborhood safety hazards due to crime, or living in high-crime neighborhoods would increase the risk of incident mobility disability, particularly among low-income elders. We found that perceiving safety hazards increased the risk of incident mobility disability among impoverished elders at retirement-age, and that this effect was not completely explained by baseline lifestyle behaviors and co-morbid conditions. In contrast, we found no evidence for an effect of perceived safety hazards among retirement-aged elders with incomes above the poverty line. With respect to neighborhood crime, we found: (1) strong correlations at the ecologic level between our measure of the neighborhood crime rate and the percentage of EPESE residents who believed their neighborhoods were unsafe due to crime, (2) strong associations between an individual's lack of perceived safety and residence in neighborhoods with the highest crime levels, as well as (3) significant cross-sectional associations between an individual's residence in the highest-crime neighborhoods and having multiple co-morbid conditions at baseline. Nonetheless, we found that unlike perceiving safety hazards, living in the highest-crime neighborhoods did not predict incident mobility disability in any group. Finally, we found an inverse association by which older elders above the poverty line appeared to have a decreased risk of incident mobility disability when they perceived their neighborhoods were unsafe at baseline, compared to those who did not perceive a safety hazard.
Our findings among low-income elders at retirement age (65 to 74) support observations in the literature that neighborhood conditions may have the greatest negative effects on this age group [22
]. Robert and Li evaluated data from the Americans' Changing Lives survey and the Midlife Development in the United States study, and found that relationships between community context and measures of health in aging populations (self-rated health, number of chronic conditions) are strongest at retirement age measured at age 60 to 69 in their analyses. The retirement age period may mark a vulnerable life-stage that is made more stressful by neighborhood contextual factors that breed insecurity, thereby predisposing elders to future mobility limitation. Studies of the peri-retirement period suggest that elders may experience marked changes in social identity, as well as changes in feelings of self-worth and self-esteem that could create psychosocial stress [32
]. Moreover, the peri-retirement period may also mark a change in access to an individual's financial resources, such as income [32
Our findings contrast with previous studies that have not found strong effects of perceived neighborhood safety or measured crime rates. Two studies measured perceived neighborhood safety as a separate item in relation to mobility disability incidence. Among these, neither the Alameda County Study (a prospective one-year analysis of elders aged 55 and older) nor the African American Health Study (a three-year study of middle-aged adults aged 49–65) found an association between perceived neighborhood safety and incident mobility disability [17
]. Direct measures of crime rates were not assessed in these studies, and the impact of perceived safety within low-SES groups was not a specific focus. A recent cross-sectional investigation in the US Health and Retirement Study examined the effect of government-reported Uniform Crime Report (UCR) crime rates as part of a composite measure of disadvantaged neighborhoods (county-level UCR crime rates and Black residential segregation) among elders aged 55 and older. No effect of crime/segregation on prevalent mobility disability was found, and measures of perceived neighborhood safety were not available [15
]. The positive findings in our study likely reflect the longitudinal design, the focus on age-specific effects, and the focus on low-income populations who may be most vulnerable to perceptions of neighborhood conditions. Our data support the hypothesis that the negative effect of perceived safety is strongest among retirement-aged elders who are impoverished, and who have fewer resources to buffer the effects of neighborhood conditions compared to higher income elders. Significantly, among impoverished retirement-aged elders, we did not find protective factors that may mitigate against mobility disability. While higher-income elders at retirement age enjoyed the benefits of frequent walking and long-term tenure in their baseline homes, we did not find these factors to be protective among elders who were impoverished at retirement age. In fact, long-term residential tenure tended to increase risk of developing mobility disability among these elders.
The finding that perception of safety, rather than crime rates, was related to incident mobility disability in this group generates the hypothesis that dangerous neighborhoods "get into the body" to engender mobility disability through psychosocial or psychological processes. Effects of perceived safety were robust to adjustment for cognitive impairment and depressive symptoms, both of which may increase risks for mobility disability or influence one's appraisal of neighborhood conditions. Strong correlations between living in the highest-crime neighborhoods and perceiving safety hazards among individual study participants suggest that study participants' perceptions were grounded in observable neighborhood conditions related to crime. But at the same time, perceptions more than objective measures appeared to correlate with future mobility disability. Further studies using survey data or biologic data that capture stress-related pathways that may connect perceptions of safety to disablement processes would help us understand the mechanisms underlying these associations.
With respect to the older elders in this cohort (75 and older), "inverse" associations between perceptions of safety and mobility disability should not be discounted. Previous research by Lachs et al. among EPESE participants identified an inverse cross-sectional relationship between ADL disability and criminal victimization, by which older adults who were victimized by crime were less
likely to have ADL disabilities than those who were not victimized by crime [19
]. Lachs et al. posit that elders with ADL disabilities may be less likely to travel about their neighborhoods, and therefore, may be less likely to experience unsafe conditions or be victimized by crime. It is also possible that inverse associations between perceived safety hazards or victimization among the oldest elders may indicate selection bias. This is to say, among the oldest adults, psychosocial effects of unsafe neighborhoods, or victimization from crime, may selectively lead to death as a competing risk, rather than disablement, such that survivors appear relatively healthy. Moreover, one might also reason that the oldest adults who perceive their neighborhoods are unsafe may have developed strategies for stress reduction, or for avoiding hazards, that protect them from disablement processes. Future research in other cohorts should examine effects among the oldest old to determine whether cross-over (inverse) effects can be detected in relation to neighborhood safety hazards, or other measures of neighborhood disadvantage. For example, in the Asset and Health Dynamics among the Oldest Old (AHEAD) study, chronic disease cross-over effects associated with race were seen around age 76, and ADL disability cross-over effects associated with race were seen at age 86 [35
]. Further research may identify age cross-over points with respect to neighborhood safety hazards as demonstrated in the EPESE cohort.
Taken together, our findings suggest that the effects of neighborhood safety on mobility disability operate through elders' perceptions rather than through direct measures of crime, and that such negative perceptions of safety pose a hazard primarily for retirement-aged elders who are poor. Competing risks of mortality may complicate interpreting findings among the oldest old.
Our findings should be interpreted in light of limitations in our data. First, in studying perceived safety, we cannot fully eliminate self-report bias as a contributor to our findings, where those who are developing mobility disability perceive their environments to be less safe due to crime. The prospective study design, limiting the cohort to those mobile at baseline, and adjusting for co-morbid conditions makes reverse causation unlikely as a sole explanation for our findings. Second, due to the difficulty in finding crime statistics for small areas, we obtained our measure of neighborhood crime rates from the city's newspaper. Though journalists work with local law enforcement to find source material, media reports of crime events are known to be biased toward severe and dramatic events, and only a small sample of crimes are reported [36
]. However, the strong ecologic correlation between the NHR Crime rate and the percentage of EPESE residents who felt unsafe due to crime provides some indication that NHR reported crime events may have also been observable by local residents, lending some corroborating evidence that we have classified neighborhoods appropriately. The lack of associations we report may point to the limitations of crime assessed at the census tract level as a measure of neighborhood exposure, i.e., a differential misclassification measurement error, where larger census tracts may be internally heterogeneous, and less accurately described than smaller areas. Future studies with data accurate to smaller areas than census tracts (i.e., specific blocks) may reveal associations that our data are unable to test.
An additional limitation is that we used annual household income to measure of SES, rather than an index of wealth that estimates assets and net worth, which are thought to capture socioeconomic status more completely than income in elderly populations [38
]. Moreover, because elders in the EPESE cohort have relatively low incomes, we were not able to identify an SES gradient effect of income on either mobility disability or on susceptibility to a perceived neighborhood safety hazard. Furthermore, we used a crude measure of poverty status that could not adjust for household size. Noting these limitations, we observed a strong threshold effect by which retirement-aged elders in poverty had increased risk of incident mobility disability compared with elders with incomes above poverty. To explain the effect of low income on disability, one might speculate that even elders with some wealth or assets, but low incomes, may continue to face the risk of disability, as elders with low incomes may hesitate to spend proceeds from their assets when replacement income is less available.
An additional limitation, in contrast to the existing literature on the effects of neighborhood disadvantage on health outcomes, we measured two facets of disadvantage, perceived neighborhood safety and neighborhood crime, rather than a larger multifaceted index of disadvantage with other observed conditions such as neighborhood walkability. Indices of neighborhood disadvantage may be considered more holistic appraisals of neighborhood conditions, since they assess multiple dimensions known to affect disablement, including social and built environmental characteristics [39
]. To be sure, neighborhoods with high crime or low perceived safety among residents are rarely without other aspects of disadvantage, including low-SES or impoverished neighbors, noise, traffic, and fewer health-promoting resources [18
]. On the other hand, identifying specific exposures connected to mobility disability can help target tailored interventions for at-risk communities. The issue of teasing out confounding to understand the individual contributions of built environmental characteristics and social determinants, which are likely highly collinear in most low-SES contexts, may be best approached through intervention research designed to mitigate the risk of one facet of disadvantage at a time. Our study contributes to this effort by suggesting that intervening in the psychosocial dimension of perceived safety may be useful among elders. However, our study is limited by statistical power (i.e., a small number of neighborhoods) and unable to observe relationships among neighborhood-level factors in an observational or multi-level context.
Last, a potential limitation is that observations of the EPESE cohort were made in the past, from the 1980s to 1990s. Given the small declines in disability prevalence among low-income elders, and persistent SES gradients in risks for disability between this period and the present, the experiences of low-SES elders in this cohort remain quite relevant to understanding contextual factors that contribute to mobility disability risks. We note that during this period, the New Haven EPESE cohort had few participants who were from racial/ethnic backgrounds other than non-Hispanic blacks and whites, and we do not present this data here. Additional cohort studies are needed to determine if our results generalize to other racial and ethnic groups.
These important limitations considered, our study has several strengths, including a longitudinal prospective design, a community-based random sample of elders with varied residential exposures, a high response rate (82%), and the use of previously validated measures of mobility disability and safety perception. Moreover, the analysis detected diversity in effects of socioeconomic factors and facets of neighborhood disadvantage in at least two stages of the life course, with evidence for heightened vulnerability at retirement age. Finally, our findings on the effect of a perceived safety hazard on incident mobility disability risk have implications for future public health interventions. For example, community-based interventions might survey low-SES elders at retirement age, and target social capital building, social support, or other appropriate interventions toward elders who feel unsafe in their neighborhoods as strategies to reduce risk for disablement [41
]. Whereas violence prevention programs are frequently targeted toward youth, our study suggests that interventions targeted toward improving safety perception in populations such as low-SES retirement-aged elders may reduce morbidity associated with aging.