Objective: To assess health improvement and differential changes in health across various sociodemographic groups in neighbourhood renewal areas.
Design and setting: A longitudinal survey of 10 390 residents in New Deal for Communities (NDC) areas and 977 residents in comparator areas in England.
Measures and methods: Changes on several outcomes across five domains (health, unemployment, education, crime and the physical environment) were assessed by sex, age, educational and ethnic group.
Results: Small overall improvements were seen on all domains in NDC areas but similar improvements were also seen in comparator areas. In NDC areas, higher educational groups were more likely to stop smoking, less likely to develop a limiting long-term illness, more likely to find employment and more likely to participate in education or training (p for trend <0.05). Older people and women were less likely to find employment and experienced smaller increases in income. These patterns were generally mirrored in comparator areas, although the education gradient in participation in education or training was less steep in NDC areas.
Conclusions: Evidence from two-year follow-up does not support an NDC effect, either overall or for particular population groups. Residents with lower education experienced the least favourable health profiles at baseline and the smallest improvements. Programme leaders should consider how to encourage participation among those with the lowest education. A shallower social gradient in participation in education and training in NDC areas and a lack of gradient in income, crime and environmental outcomes indicate that some aspects of the programme may be reaching all sections of the community.
Statistics from Altmetric.com
Reducing inequalities in health is a key priority for the UK government.1 To this end, public service agreements have been announced to address inequalities in life expectancy by level of area deprivation by 2010.1 Neighbourhood renewal programmes—initiatives aimed at improving conditions in the most deprived areas—are one way of tackling these inequalities. Area-based initiatives are thought to be popular among policy makers because they: (1) are assumed to be an efficient means of targeting the most deprived individuals; (2) provide a context for involving local people in identifying local problems and delivering solutions; and (3) recognise area effects.2 Robust evidence relating to the impact of area-based interventions on health and health inequalities is, however, scarce.3–5
There are three distinct goals of policies aimed at reducing health inequalities.6 The first goal is to improve the health of the worst off only. Interventions could be deemed effective if the health of the worst off had improved in absolute terms, irrespective of whether other population groups had made even greater progress. The second goal is to narrow the health gap between the most disadvantaged and the better off groups, by the worse off groups making better progress in health relative to the rest of the population. The third goal is to tackle the overall gradient in health across the whole population. Interventions would have to demonstrate differential changes in health across socioeconomic groups over time. This typology is made more complex when applied to area-based initiatives for which the socioeconomic characteristics of both areas and residents must be considered. It is theoretically possible that an area-based initiative is effective in raising the average health of the deprived intervention area (at a slower, equal or faster pace than nationally) while at the same time contributing to widening inequalities between people within that area. In other words, a reduction in socioeconomic inequalities between areas does not necessarily equate with a reduction in inequalities between socioeconomic groups within the areas.
Existing studies of area-based initiatives suggest that residents with greater financial and human capital may be more able to access information about new services, may be better placed to utilise those services and may have a more positive reaction to new services.7 8 A suggested explanation is that recipients with the greatest socioeconomic disadvantage, who have experienced instability, change and disadvantage, may be overwhelmed by the introduction of the programme. In addition, interventions may not be designed for and targeted to the most deprived groups.
The present study investigates the impact of the New Deal for Communities (NDC) initiative on inequalities in health and the determinants of health. Policy documents suggest that the NDC initiative in England has been primarily concerned with achieving absolute improvements in the targeted deprived areas alone and, to a lesser extent, with reducing the gap between the NDC areas and the rest of the population.9 The interest of this study is not only in whether there has been an absolute improvement in health in the NDC areas, but crucially whether there have been differential changes in health across demographic or socioeconomic groups over time within NDC areas. Have some groups benefited more than others from the extra resources provided in intervention areas? Have intervention areas been able to target resources to residents who had the greatest socioeconomic disadvantage and poorest health at baseline? Three related questions were addressed: (1) Have there been overall improvements in health or its determinants in the deprived areas targeted by the NDC initiative? (2) Have there been differential changes in health or its determinants for different socioeconomic, ethnic, gender and age groups within NDC areas? (3) To what extent do any differential changes mirror what is happening in similarly deprived non-intervention areas?
The NDC programme
The NDC programme is an area-based initiative that aims to improve conditions in some of the most deprived neighbourhoods in England and reduce the gap between them and the rest of the country. There are 39 NDC areas, each with a budget of approximately £50 million with which to address five specific outcome areas (health, unemployment, education, crime and the physical environment) over 10 years. In order to be considered for NDC funding, community partnerships involving local residents, local authorities, public service providers, community and voluntary organisations and businesses had to prepare a proposal for regeneration. Proposals were assessed on the basis of the deprivation score for the community and were stratified by region. Out of the 39 NDC areas, 29 fall within the 10% most deprived wards in England (measured by the Index of Multiple Deprivation 2000) and a further eight are within the 20% most deprived wards. More details on the NDC programme are given elsewhere.9
The variety of projects being implemented in each NDC area is too wide to describe here (see http://extra.shu.ac.uk/ndc/ndc_reports.htm) but examples include the provision of a new facility providing workspace for new and existing small businesses, refurbishment of poor quality housing using local workers and the introduction of a drug and alcohol outreach service for under 18 year olds, with costs ranging from a few thousand to over one million pounds. Project case studies, partnership reports on process and management and individual survey data have been collected to aid evaluation. We conceptualised the intervention as holistic (a complex intervention based on action across the five domains) rather than project based (a specific, single intervention on a given demographic group or domain) and analyses were based on individual survey data.
In 2002, MORI/NOP undertook a survey of 500 residents aged 16 years and over in each of the 39 NDC areas.10 Contact addresses were revisited two years later providing longitudinal data for 10 390 residents (73% of baseline respondents). Simultaneously, a survey of residents in comparator areas, matched on deprivation score and local authority, was undertaken to provide longitudinal data for 977 residents (72% of baseline respondents). Only residents who remained in the NDC or comparator area were included in the analysis reported here. The possible biases introduced by residential mobility are discussed later.
The survey covered health and health behaviours, quality of life, employment and other socioeconomic characteristics, demographic characteristics, housing, community life and perceptions of crime in 2002 and 2004. This study focused on changes in health and the determinants of health using the measures summarised in table 1.11 For the present study, outcomes were selected a priori to capture short-term impacts that most comprehensively summarised change across the five domains.
Socioeconomic position is commonly captured by education, income or occupation. Here we use education, which structures occupation and income,12 and was measured by highest educational attainment and split into six bands (none up to national vocational qualification (NVQ)5 or equivalent). As a guide, NVQ1 is up to general certificate of secondary education (GCSE) level and NVQ4 is up to degree level. As a result of small numbers in comparator areas, education was further collapsed into three bands (none, NVQ1–2 and NVQ3–5). Reported ethnic background was categorised as white or white mixed/south Asian (Indian, Pakistani, Bangladeshi, other Asian background)/black (Caribbean, African, other black background)/other (Chinese, other background).
Continuous change variables were normally distributed. Quitting smoking, new long-term limiting illness, new employment and participation in education/training were analysed as binary variables.
Changes in health and the determinants of health were assessed by sex, age, educational and ethnic group to explore the possibility that NDC initiatives may have differential success across demographic and socioeconomic groups. Minority ethnic groups had a younger population profile than white individuals. Those aged 25–54 years had greater educational attainment than younger and older participants. Models for educational and ethnic group were adjusted for age to account for the varying age structures.
Two-level regression models were used to account for the clustering of residents within areas using MLwin.13 Change in health (or other outcome) was modelled by sex, age, ethnic and educational group. Variables indicating NDC/comparator area and the interaction between NDC/comparator area and sociodemographic group were added (to assess whether differential changes mirror what is happening in similar deprived areas). For presentation, a stratified analysis was used to describe change in health by sociodemographic group separately for NDC and comparator areas. All estimates were weighted for the probability of the selection of individuals within households into the survey.
The sample is summarised in table 2. Women were overrepresented, approximately 20% were from non-white ethnic backgrounds and over a third had no formal qualifications. The sex, age and ethnicity profiles of residents in comparator areas were similar to those of residents in intervention areas (table 2). The comparator areas had a slightly lower proportion of residents with no educational qualifications (33% versus 39%), however, which may indicate that the areas were slightly less deprived than their NDC counterparts. As expected, residents from the lowest educational group had the poorest baseline outcomes across all health and non-health domains (data available on request).
The results from the multilevel regression models are shown in table 3 (for health), table 4 (education and employment) and table 5 (crime and environment). Overall, statistically significant improvements were seen in NDC areas for income, fear of crime and satisfaction with the local area. These improvements were mirrored in comparator areas. With the exception of satisfaction with the local area, no statistically significant differences in overall change in NDC versus comparator areas were found. In other words, evidence does not support an overall NDC effect at the two-year follow-up.
The overall figures mask some differences according to the demographic and socioeconomic group. In NDC areas, the likelihood of quitting smoking, finding work and participating in education or training increased steadily with increasing education and the likelihood of developing new long-term limiting illness decreased with increasing education (fig 1). The changes in health and its determinants according to socioeconomic or demographic group in NDC areas largely mirrored changes seen in the comparator areas. Statistical interactions between NDC status and sociodemographic group were not significant, with two exceptions. In both comparator and NDC areas, the likelihood of participation in education or training in the past 12 months increased with increasing education, but this association was less steep in NDC areas. Also, an age-related decline in self-rated health was seen in comparator areas but not in NDC areas.
Differential improvement was also seen by gender (for example, women saw smaller increases in income and were less likely to find work), age (for new limiting illness, income, participation in education or training, fear of property crime and satisfaction with local area) and ethnicity (for example, black residents saw smaller improvements in psychological well-being compared with white residents, and south Asian residents were more likely to quit smoking). As highlighted above, these changes were not significantly different to those seen in comparator areas.
Significant variation between NDC areas was found on several domains. As a proportion of the total variation, however, differences between NDC areas were small (less than 2% on all domains). This indicates that differences in health and its determinants between demographic and socioeconomic groups within areas were more apparent than differences between areas.
Programme leaders should consider how to inform and include the least educated groups and design projects with a view to encouraging uptake among those with the lowest education. Further data collection sweeps are needed to assess progress towards this goal.
What is already known on this subject
Area-based interventions are a key element of policies to reduce health inequalities in England.
In the 39 NDC areas, small improvements in health and the determinants of health were seen between 2002 and 2004. These improvements were mirrored in similarly deprived comparator areas.
In summary, there was evidence of improvement in NDC areas between 2002 and 2004 across several domains. Greater improvements were seen for some socioeconomic and demographic groups. Both the overall and differential improvements were, however, mirrored in comparator areas. The evidence does not therefore support an “NDC effect” for the whole population or for different types of resident.
Turning in more detail to the first research question of whether there have been overall improvements in health and the determinants of health in NDC areas, this study found improvements in psychological well-being, income, fear of crime and satisfaction with the local area. More than 10% of residents quit smoking and nearly 10% of adults who were not working at baseline were in work at the 2004 follow-up. Although statistically significant, the improvements were modest in magnitude. This is in line with expectations given the short follow-up period. The NDC programme was announced in 1998 but it took time for projects to be implemented and bedded down and this may contribute to the lack of change noted on some outcomes. Our estimates relating to the first research question are broadly similar to those reported by the NDC National Evaluation Team, although they recorded small improvements across all domains for a wider range of outcomes.9
What this study adds
Over the two-year follow-up, there were changes in inequalities in health between socioeconomic groups within the areas. In both NDC and comparator areas residents with the lowest educational attainment and poorest health at baseline experienced the smallest improvements. Health and socioeconomic inequalities among residents in NDC areas were no smaller at follow-up than inequalities in comparator areas.
Local programmes may require more specific targeting to those with the least resources. Longer-term follow-up is needed to confirm trajectories of change in intervention and comparator areas.
On the second research question, there was some evidence of differential impact within NDC areas. Women and older people, for example, do not appear to have benefited from new employment opportunities, nor from the rising incomes of those in paid employment. There was a graded relationship between educational attainment and improvement on the health, employment and education domains. There were already large differences by education for long-term limiting illness, smoking and income and these differences widened over the two-year follow-up.
These changes within NDC areas reflect a wider trend unfolding in other deprived areas. There were no consistent differences between NDC and comparator areas in the pattern of health-related outcomes for different demographic groups. In other words, and in answer to the third research question, robust evidence of an NDC effect was not found, either overall or in terms of differential impacts, over and above the developments in the comparator areas. The finding of a weaker association between the take-up of training and baseline education levels in NDC areas is, however, intriguing and could indicate that inequalities are growing less fast in NDC areas. The usual caveat about multiple significance testing applies here. “Contamination” of initiatives from intervention to comparator areas may contribute to the general lack of NDC effect, although intervention and comparator areas were chosen to be non-contiguous to reduce this possibility. There was considerable overlap of area-based initiatives in NDC areas14 and it is likely that interventions were underway in some of the similarly deprived comparator areas.
Investigation of the reasons for the differential improvement by educational group was beyond the scope of this study. These results are based on analysis of change over two years and it is possible that lower educational groups are simply slower to take up new services and resources and that, over time, the socioeconomic differences will diminish. Several possible plausible barriers to take-up can, however, be identified including inadequate information about services, socioeconomic differences in the concepts of health and disagreement about the importance of the service, language differences and the costs of participation.15–19 The potential for health promotion activities to increase social inequalities in health has been noted20–22 and community development approaches to health promotion have been proposed as one way of building more appropriate interventions.23 Inequalities across the whole socioeconomic spectrum were not assessed here, but the findings indicate some widening of inequalities between residents of the deprived intervention areas, and they highlight the need to continue to pay careful attention to who is benefiting from area-based initiatives. It is worth noting that there was no evidence of an education gradient in change on the crime and environment domains. The crime and environment initiatives may provide transferable lessons on reaching all sections of the community.
By their very nature, area-based initiatives introduce some form of geographical clustering into the data and one strength of this study was to use multilevel analysis to account for this. Methodological and other limitations of the study must be considered. Multiple testing is a possible explanation for the significant findings seen. Rather than focusing on one-off statistically significant estimates, we should look for consistent patterns (such as that of differential change according to educational group). People who moved in or out of the areas were not included here. An NDC effect may be greater for stayers who have had greatest exposure to the intervention. To the extent that the intervention facilitates moving out (for those that want to) by expanding life choices, excluding movers will underestimate the NDC effect. Future analysis should compare the health and other outcomes of movers and stayers. Although a holistic approach that investigates NDC status as a binary variable is suitable for the quantitative analyses undertaken here, it is unable to consider variation in programme priorities or implementation. For example, the present analyses did not distinguish NDC areas that aimed to target particular sociodemographic groups. Future work may review programme documentation to elaborate on the NDC intervention using a qualitative approach to categorise interventions that could then be combined with the survey data to identify the more efficacious initiatives and approaches.
Area-based initiatives represent one approach to tackling inequalities in health. At two-year follow-up, evidence does not support a positive effect of the NDC programme in England on health or the determinants of health. This study does indicate that observed changes within areas differed by socioeconomic and demographic group. Residents with no education had the least favourable health profiles at baseline and the smallest improvements throughout follow-up. Programme leaders should consider how to inform and include the least educated groups and design projects with a view to encouraging uptake among those with the lowest education. Further data collection sweeps are needed to assess progress towards this goal.
The authors would like to thank the NDC National Evaluation Team, the Centre for Regional Economic and Social Research (CRESR) at Sheffield Hallam University and in particular Paul Lawless and Christina Beatty for providing access to data and helping find their way around the complex NDC landscape when they had much more important demands on their time. The authors also wish to thank Christine McGuire, Ken Judge and Kirby Swales for their support. They are also grateful for useful comments from anonymous reviewers that were helpful in refining the final version of this paper.
Competing interests: This work was conducted as part of pilot work funded by the Department of Health on the feasibility of undertaking secondary analysis to evaluate the impact of NDC on health inequalities. MS is funded by a Department of Health personal award.
Competing interests: None declared.
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.