Is concern about young people's anti-social behaviour associated with poor health? cross-sectional evidence from residents of deprived urban neighbourhoods

Irrespective of whether one leans towards the 'realistic' or 'stereotypical' interpretation of PASB, there is evidence to show that concern about ASB and crime may have an adverse impact on people's quality of life [11]. People who worry about ASB and crime may become constrained in their use of public spaces [28, 29], or may withdraw from social life and avoid going out, especially at night [30, 31].

Several studies also suggest that residents living in disadvantaged or deprived neighbourhoods, or neighbourhoods perceived by residents to have poor reputations, are particularly likely to be concerned about crime, ASB and neighbourhood safety [19, 20, 32]. However, some studies have found that the association between deprivation and PASB prevalence appears to vary between and within disadvantaged populations: i.e. some disadvantaged neighbourhoods appear to be more resilient (i.e. have a relatively low prevalence of concerns about ASB) than others, and some residents of such neighbourhoods appear more resilient then their neighbours [2, 33, 34].

The evidence linking health with PASB also provides some mixed messages. With regards to mental health, a US study of Black and Hispanic women and children found associations between perceived crime and mental health to be weak and inconsistent [35]. Weich et al. found that associations between depression and neighbourhood graffiti were not statistically significant after adjusting for individual and household-level risk factors [36]. However, a study of adults residing in tower blocks found that fear of crime was associated with low mental health scores derived from the SF-36 (Short Form 36 questionnaire)[7]. A Scottish study found evidence that residents who perceived high levels of neighbourhood incivilities (e.g. litter and graffiti) were particularly likely to report frequent feelings of anxiety and depression [37]. Furthermore, longitudinal analysis of UK civil servants found that fear of crime was associated with poorer mental health, along with reduced physical functioning, lower social engagement and lower quality of life [38].

With regard to physical health, a literature review identified some evidence that worrying about crime and/or ASB is associated with reduced physical activity but the reviewers also identified evidence that was inconsistent with this finding [39]. Miles et al. found that neighbourhood disorder was associated with women's (but not men's) infrequent involvement in sports, whilst perceived safety was not associated with physical activity for either men or women [40]. Mason et al. found that physical activity (walking around the neighbourhood) had inconsistent associations with perceptions of poor neighbourhood safety at individual and neighbourhood levels [41]. Analysis of British Crime Survey data has found PASB to be associated with longstanding illness [20], whereas the Scottish Social Attitudes Survey found no significant association between these variables [27].

A key issue for this paper is that relatively few studies have focused particularly on associations between health and perceptions of young people's ASB. One study of walking habits amongst older people used an unflattering composite measure of 'nuisance' that combined young people hanging around with unattended dogs and dog-fouling, and found that low nuisance was associated with more walking [42]. The British Crime Survey found that self-reported long standing illness was associated with a greater likelihood of perceiving teenagers hanging around to be a neighbourhood problem (as was living in a relatively deprived area) after controlling for potential confounding variables [20].

In this study we used cross-sectional data from a survey of residents living in disadvantaged urban neighbourhoods to ask (question 1) are adults with poor health more likely to think young people's ASB is a problem in disadvantage neighbourhoods? We also invert that question and ask (question 2) do adult residents who perceive young people's ASB to be a problem in disadvantaged neighbourhoods have worse health then people who are less concerned about young people's ASB.

The findings have been adjusted to take into account demographic characteristics of residents. As PASB has also been associated in the literature with psychosocial factors such as poor neighbourhood reputation and poor social cohesion [28‐32] we have conducted further analysis to explore whether individual-level psychosocial characteristics can help explain why PASB varies within disadvantaged neighbourhoods.

A key strength of this study is that it focuses specifically on perceptions of young people's ASB in disadvantaged neighbourhoods: i.e. it focuses on the social group most frequently linked to ASB, in the kinds of neighbourhoods where previous studies have found residents are most likely to view young people's behaviour as problematic. We have measured physical and mental health using a short well-validated questionnaire (SF-12v2). We are aware of no other quantitative study that looks at both physical and mental health in relation to perceptions of young people's ASB, or that focuses on perceptions of young people's ASB in disadvantaged urban areas.

We surveyed 14 disadvantaged neighbourhoods in the city of Glasgow (UK). Two of the neighbourhoods were peripheral estates consisting mainly of houses with gardens and common entry three/four story flats (tenements). The other neighbourhoods had inner-city locations: five had a 'gardened suburb' design characterised by houses (including house-like structures containing private entry flats) with gardens; seven were mass housing estates dominated by high rise flats.

All the neighbourhoods fell well below the Scottish Index of Multiple Deprivation lowest 15% income deprivation cut-off that is used to define area poverty by the Scottish Government [44]. Around one-in-four homes were owner-occupied and the rest were rented (virtually all from government-regulated providers of social housing) [43].

Data were collected in 2006. Addresses were selected at random, although in some smaller neighbourhoods all residential addresses were selected. One adult householder (aged 16 years or over) per household was randomly sampled and, subject to the provision of informed consent, participated in a face-to-face interview. Full details of the questions asked have been published in the GoWell protocol [43]. The items relevant to the PASB analysis are described in Additional File 1.

PASB was assessed using a question adapted from the British Crime Survey asking participants if they thought 'teenagers hanging around on the street' constituted a problem in their local neighbourhood [20], defined as the local area within a 5-10 minute walk of their home. In the survey questionnaire, this was one item in a list of 17 neighbourhood problems (see Additional file 1). We were interested in more serious instances of PASB, so our analysis compared residents who identified teenagers as a 'serious problem' with those who did not (collapsing 'not a problem', 'slight problem' and 'don't know' responses into the reference category). This was our dependent variable for research question 1 (see Aims and Objectives, above).

GoWell's two primary health outcomes were the physical and the mental health scores derived from SF-12 version 2 [45]: a validated questionnaire often used in studies of residents living in disadvantaged areas [46]. SF-12v2 physical and mental health composite scores are computed using the scores of twelve questions and range from 0 to 100, where a zero score indicates the lowest level of health measured by the scales and 100 indicates the highest level of health. These scores were our dependent variable for research question 2.

We also included as secondary health outcomes two variables on health service use in the previous 12 months: number of general practitioner (GP) visits, and number of GP visits for a psychological problem. These were included to further explore whether or not attempts to reduce PASB might feasibly target frequent health service users.

Following a scoping of previous surveys and related literature we developed a list of demographic, social and psychosocial characteristics that we hypothesised could be associated with PASB [15, 20, 33, 47, 48]. Variables relevant to these characteristics were derived from items in the GoWell questionnaire, which generally took the form of structured questions with 3, 4 or 5 point Likert-type scales for self-reported responses. Most of these questions were taken from national and local surveys - these surveys were often not formally validated but have been selected because they are routinely used and provide benchmark findings from which we can compare outcomes. Some questions were developed by GoWell and piloted for ease of delivery and response. Further details about the questions can be found in Additional File 1. The other variables we included are summarised below.

Psychosocial characteristics were conceptualised as characteristics that 'bridge' individual and environmental factors by indicating how people feel about their environment[49, 50]. We included a number of neighbourhood psychosocial variables relating to social cohesion, including neighbourhood reputation, safety, tolerance, trust in neighbours, feeling of belonging to the neighbourhood, self or collective efficacy regarding local decision-making, feeling that neighbours would intervene informally to prevent ASB, how neighbourhoods affected participants' self-esteem; social contact, social support and proxies for social exposure to the neighbourhood (length of residence, speaking to neighbours, taking walks around the neighbourhood). We also included variables from questions about participants' home psychosocial environments: privacy, control, safety and self-esteem related to the home.

Using Stata/IC 11.1 [51] we ran bivariate analysis of response data for each of the independent variables. We conducted multivariate logistic regression to model associations between the demographic characteristic variables and the dependent variable for question 1 (teenager problems). We then conducted multiple regressions to model associations between the dependent variable and each of the two health variables, adjusting for demographic characteristics. To avoid over-adjustment (e.g. having one health variable adjusting for another health variable), this involved 2 separate regression models (one for SF12v2 physical health scores and the other for SF12v2 mental health scores).

To answer question 2, we used bivariate and multiple regressions with teenager problems as the independent variable (adjusting for demographic characteristics in the multiple regression models), whilst SF-12v2 physical health score and then SF12v2. mental health score were dependent variables. We explored effect sizes based on standard deviations using Cohen's suggested interpretations of Cohen's d: d = 0.2 be considered a 'small' effect size, 0.5 represents a 'medium' effect size and 0.8 a 'large' effect size [52].

Further analysis involved multiple regression models of social and psychosocial variables adjusting for the demographic and health related variables. Teenager problems was the dependent variable. Given the relatively large number of variables in the final model, we carried out a Bonferroni correction which is considered to be a conservative approach to reducing the risk of false positives from multiple comparisons [53].

In all the multiple regression models, 95% confidence intervals (CI) were based on robust standard errors which take account of the non-independence of respondents from the same area, thereby accounting for neighbourhood level effects.

Table 1 summarises our multivariate regression model showing independent associations between demographic characteristics and perceiving teenager problems. Participants from all age categories under 65 years were more likely than people from the oldest age group to perceive teenager problems (p values ranged from 0.007 for 55-64 year olds, to < 0.001 for 16-24, 40-54, and 25-39 year olds). Perceiving teenager problems was also associated with living with children (OR 1.18; 95%CI 1.00,1.39; p = 0.049) and having problems paying bills (OR 1.43; 95%CI 1.11,1.85; p = 0.005). Findings for sex, ethnic group household structure and educational qualifications were inconclusive.

Table 2 addresses our first primary research question. Teenager problems is the dependent variable and the independent variables include the SF-12 mental and physical health scores. Mental health scores were not found to be independently significantly associated with perceived teen problems after adjustment: OR 0.99 (95%CI 0.98, 1.00; p = 0.103). Lower physical health scores were found to be associated with perceiving teenager problems after adjusting for demographic characteristics: OR 0.98 (95%CI 0.97, 0.99; p = < 0.001).

Table 3 addresses research question 2. The SF-12v2 scores are the dependent variables and the independent variables include teenager problems. We found that teenager problems were associated with physical health scores (coef = -1.69; 95%CI -2.87, -0.52; p = 0.008) but there was little association with mental health scores (coef = -0.90; 95%CI -2.10, 0.31; p = 0.133) after adjusting for demographic variables. Cohen's d was 0.06 and 0.09 for SF12v2 physical and mental health scores respectively: according to Cohen's criteria these are both small effect sizes [52].

Table 4 models associations between perceptions of teenager problems and variables relating to residents' psychosocial environment, demographic characteristics, health and health service use. A Bonferroni correction was used to reduce the risk of type 1 error. People who felt unsafe going out at night, who felt low trust towards neighbours, who believed their neighbourhood did not encourage high self-esteem, reported that their neighbourhood had declined in the previous two years, lacked social support, and had less than weekly contact with relatives were all more likely to perceive teenagers hanging around to be a serious local problem (p < 0.001 before the Bonferroni correction, p < 0.05 after the correction). Similarly, participants who reported visiting a GP seven or more times in the last 12 months were significantly more likely to perceive teenager problems before and after the Bonferroni correction (p < 0.001 and p < 0.05, respectively). The only demographic variable to remain significant after the correction was age: again (similar to Table 1), the findings reported in Table 4 suggest an inverse relationship between age and perceptions that local teenagers were a problem.

A number of variables were found to have a significant association with teenager problems before the Bonferroni correction but not after: i.e. 1 to 6 GP visits during last 12 months, GP visits for a psychological issue (last 12 months), neighbourhood reputation, contact with neighbours, walk around neighbourhood, and problems paying bills.

In this study we used cross-sectional data to explore whether adults with poorer health (suggested by relatively low SF-12v2 scores) are more likely to think young people's ASB is a problem in disadvantage neighbourhoods; and whether those residents who perceive young people's ASB to be a problem have relatively poor health (compared to other residents of disadvantaged neighbourhoods). These questions are closely related and have led to similar findings.

Focusing on the findings that were adjusted for individual demographic characteristics, our study suggests that physical health may have small independent associations with perceived teenager problems, whereas independent associations with mental health are weaker. However, the means differed by less than 0.2 standard deviations suggesting that the difference was likely to be trivial, even when statistically significant. This leads us to downplay the public health significance of associations between perceptions of young people's ASB and either physical or mental health.

Our final model did suggest that frequent GP visits were more consistently associated with perceptions of teenager problems (compared to the physical and mental health scores). We assume that the well-validated SF-12v2 scores are a more robust health measure than frequency of GP visits and that, consequently, GP visits may be influenced to a greater degree by unmeasured confounding factors. However, the findings on GP use could suggest that people in deprived areas who are regularly help-seeking (for a variety of reasons) are also more likely to perceive youth ASB problems: perhaps indicating a potential link between PASB and feelings of vulnerability. Frequent health service users may therefore be a potential target population for interventions that attempt to support those most vulnerable to PASB.

We also explored the demographic characteristics of adults from disadvantaged areas who think that teenagers hanging around constitute a serious neighbourhood problem. In common with other studies[20, 54], our findings suggested an inverse age relationship whereby residents belonging to our youngest age-category (16 to 24 year olds) were more than twice as likely to perceive teenager problems compared to the oldest age group (> 64 years). These findings are interesting in that they arguably run contrary to the pattern one would expect to see if PASB was primarily driven by intergenerational intolerance characterised by older adults' negative attitudes towards young people [26].

Our findings suggests that people who perceive their psychosocial environment to be poor are more likely to be concerned about local teenagers compared to residents who rate it more favourably. For a number of psychosocial variables, this relationship appears to be stronger than the relationship between poor health and PASB. Our analysis included psychosocial characteristics that can be considered dimensions of social cohesion. Several such characteristics appeared to be independently associated with teenager problems (i.e. perceptions that the neighbourhood is unsafe, in decline, that neighbours cannot be trusted and that the neighbourhood has an adverse effect on self-esteem). This is broadly in keeping with findings on perceptions of young people's ASB based on the British Crime Survey (BCS), although BCS analysis included fewer variables and did not focus on the characteristics of disadvantaged neighbourhoods [20].

Social isolation, in terms of low social support was also found to be associated with perceiving local teenagers to be a problem. However, the findings on social contact were to some extent inconsistent in that they suggested residents with regular contact with relatives were less likely to worry about teenagers, whereas regular contact with neighbours was associated with a greater likelihood to report PASB (although this latter finding was not significant after the Bonferroni correction). The association between PASB and frequent neighbour contact is in striking contrast to arguments that increased local engagement serves to raise levels of tolerance and reduce perceptions of inter-group problems within local areas [2]. It may be that in deprived areas, everyday contacts, if they include complaints about ASB or the area in general, may serve to raise awareness of local problems.

As we asked respondents about teenagers hanging around 'on the street', one might expect neighbourhood psychosocial environments to be more relevant to people's perceptions than home environments. Perhaps this reference to 'the street' helps explain why our measure of home psychosocial environments were not found to be independently associated with PASB.

It is worth noting that this research was conducted prior to recent, well-publicised UK riots involving young people from deprived areas, and that those riots did not spread to Glasgow [18]. However, the study and its findings are clearly relevant to an issue that has both a longstanding and current place at the forefront of political and public debates. The current study contributes to existing knowledge by focusing on perceptions of young people's ASB in disadvantaged areas, and by using a validated measure of physical and mental health (SF-12v2) for its primary outcome. In contrast, previous studies have usually focused on measures of ASB that are not age-specific. The small number that do focus on young people's ASB tend to either compare deprived areas with more affluent areas (useful for identifying social inequalities in PASB but less useful for exploring variation within deprived populations), and/or do not use a validated health questionnaire.

However, we note that our independent variables were based on self-reported measures, as was the dependent variable: raising the possibility of bias from common method variance[55]. The 50% response rate is, we think, reasonable for a study of very disadvantaged neighbourhoods but raises the possibility of response bias. We also stress that cross-sectional data is useful for identifying associations but other types of study design are required to establish causal pathways between associations. This was a relatively well powered study: whilst this should be seen as a strength, it does mean the study is capable of detecting a range of associations including associations that are potentially too small to be of public health significance. Finally, the Bonferroni correction reduces the risk of type 1 errors, but has been criticised for doing so at the expense of increasing the likelihood of type 2 errors [53].

Crucially, our survey did not give participants the chance to say whether they thought that ASB problems were perpetrated by a minority or majority of local teenagers, nor did it consider the perceptions of residents who were less than 16 years old.