Background
Health in Britain varies from place to place, neighbourhood to neighbourhood
and region to region. Does this mean that policies to tackle poor
health should be directed at the places with particularly poor health?
Are indicators of area deprivation important signposts in the allocation
of health resources or imperfect substitutes for targeting poor
people? To direct policy effectively we need to understand how places
and people come to be unhealthy and unpack the social and environmental
influences on poor health. Identifying the operation of these influences
at different spatial levels helps to establish a balance of policies
aimed at people and at places.
This project
investigates health in both social and spatial settings, combining
information on people with information on the places where they
live. It takes innovative statistical evidence from three large-scale
British data sets. Each follows individuals through time, which
enables the 'cause and effect' of ill-health to be disentangled.
Ill-health appears in various guises: as premature death, physical
complaints and psychological malaise. Its predictors are measured
within narrower or wider boundaries, at the level of the person,
the family, the neighbourhood, the district, or the region. We explore
the interplay of these different levels of influence as precursors
of poor health.
Aims and
Objectives
We aim to quantify the relationship between social and economic
status and health, allowing for the geographical setting by answering
the following questions:
- How far does
locality have an effect on its own? Do the effects of 'place'
apply to all inhabitants indiscriminately or depend on personal
circumstances? Do area effects reflect the social composition
of the area or other more strictly ecological features of its
environment?
- In which
localities is health particularly good or bad after allowing for
measured socio-economic characteristics of the inhabitants?
- Do the different
dimensions of health have common socio-economic predictors?
- Does the
experience of change, for people or in places, affect the socio-economic
patterns of health?
It is also planned
to provide the research community with important enhancements to
two of the datasets, bringing new statistical procedures (multi-level
modelling) to the Office for National Statistics Longitudinal Study
(ONS-LS) and bringing the 1991 Census Small Area Statistics into
the National Child Development Study (NCDS).
Study Design
This is an interdisciplinary quantitative study, doing secondary
analyses of three existing datasets, each with a longitudinal element.
They are the ONS-LS, linking one percent of the population of England
and Wales across the 1971, 1981 and 1991 censuses with vital registration;
the NCDS, the cohort followed from birth in 1958 to age 33 in 1991,
and the Linked Health and Lifestyle Survey (HALS), where two thirds
of a nationally representative (but clustered) sample from 1984
were re-interviewed about symptoms and behaviour in 1991. Each includes
indicators of socio-economic advantage and disadvantage and various
dimensions of health. The individuals in each dataset can be related
to the social profile - and some other characteristics - of the
places where they live (or have died).
The ONS-LS Study
is used to explore how the relationship between deprivation and
mortality or long-standing illness varies by social setting. It
also throws light on the experience of change, by individuals or
in places. The NCDS is used to highlight the life course of young
adults in different types of locality. All three datasets (including
HALS) are used to explore the extent to which different aspects
of ill-health (mortality, chronic and acute illness) have a common
underlying pattern of relation to social disadvantage, attitudes
or behaviour.
Policy Implications
Our central goal is to inform policies about resource allocation
for health care and prevention, balancing interventions targeted
on people and on places. It is also to improve understanding of
the pros and cons of identifying 'disadvantaged areas' by crude
nationally available indicators of deprived areas, in contrast to
more specific classification of local conditions. The former are
nationally available and apparently objective, but are they target
efficient?
If the social disadvantages associated with premature death are
also more closely linked to psychological rather than physical illness,
this would help inform resource allocation between psychiatric and
physical medicine, and between the preventative and curative sectors.
Project Summary
Rates of morbidity and mortality are higher in disadvantaged areas
and these geographical inequalities in health have been increasing
(see projects by Dorling and by Southall). Do poor areas have a
poor health profile because of the poor people who live there or
do areas have an independent influence on health? The project addressed
this question using a range of health outcomes, including limiting
long-term illness (LLTI) which was included in the census for the
first time in 1991.
The project
examined a range of mechanisms by which areas may have an effect,
including de-industrialization, social capital and climate. It linked
ecological and individual data in three longitudinal datasets, the
Office for National Statistics-Longitudinal Study (ONS-LS), National
Child Development Study (NCDS) and the Health and Lifestyle Survey
(HALS). For these analyses, it developed the facility to use the
appropriate software (MLn) on the ONS-LS and enhanced the capacity
of the NCDS to be used with geo-referenced data.
Key findings
- There are
marked area differences in LLTI, with higher rates in areas described
as 'coalfields' and 'ports and industry' and lower rates in areas
of growth and prosperity.
- Locality
is a less important predictor of LLTI than individual socio-economic
characteristics. These individual characteristics include education,
social class and ethnicity, as well as social and geographical
trajectories (time spent out of employment, downward social mobility,
time spent in the most prosperous areas of the country). Individual
characteristics over the 20-year period of observation provided
by the ONS-LS accounted for about half of the area variation in
LLTI. Change or continuity in individual circumstances affected
the chances of reporting a limiting long-term illness and each
observed occurrence of a disadvantageous state (e.g. being unemployed)
added to the chances of an adverse health outcome.
- Individual
characteristics do not however fully explain the regional variation
in LLTI. Over and above these characteristics, those areas described
as 'coalfields' and 'ports and industry' had a higher than expected
risk of LLTI and those in the most prosperous areas had a lower
than expected risk. Area type (as classified by ONS) accounts
for around a quarter of the geographical variation in rates of
LLTI. This leaves around one quarter of the variation (age standardized)
between districts unexplained.
- The areas
of higher risk were ones of deindustrialisation and its specific
effects were examined. Residents in areas which had once had a
high level of heavy industry and then lost a large number of such
jobs did indeed report poorer heath, whether or not they were
unemployed or employed in manual work. Feeling part of the community
did not appear to protect people against the adverse health-effects
of deindustrialisation: the effects of industrial decline were
no milder in those who felt part of their community. However,
deindustrialisation did not explain the whole of the area differences
in health. It should be noted that the reporting of LLTI may be
particularly high in areas where jobs are scarce, with workers
displaced from manual work more likely to move into economic inactivity
than in areas where jobs remain plentiful.
- Further analyses
enabled area and individual influences to be tested on outcomes
other than LLTI. There was a clear and approximately linear relationship
between area deprivation (measured at ward level in the ONS-LS)
and a range of poor health outcomes and adverse life events among
women, including underweight birth and, particularly, teenage
birth and sole registered birth. However, when adjustment was
made for personal disadvantage, these ecological associations
were largely, if not entirely, accounted for by individual measures
of disadvantage. Whatever contextual factors may be influencing
these fertility outcomes, they are not well identified by the
census indicators of social composition.
- The potential
influence of social capital on a further measure of health, symptoms
in the past two weeks, was examined in HALS. This suggested an
association with current health, but poor areas had health disadvantages
over and above those associated with the disadvantaged circumstances
of individuals and with low social capital.
- Climate and
housing may combine in ways which contribute to regional inequalities
in health. For people living in good housing, bad weather made
a smaller difference to their respiratory health, while for those
in poor housing, damp and cold had more of an effect on their
lung function. There was also evidence of an 'inverse housing
law' in which the worst quality housing is found in areas with
severe climate.
- The project
also had the methodological objective of developing the capacity
to do multi-level modelling of the ONS-LS and to enhance the NCDS
by attaching geographical information. Both these objectives were
achieved. The project made a pioneering application of multi-level
modelling to both datasets.
|