Pathways to obesity Identifying local, modifiable determinants of physical activity and diet

1
Pathways to obesityIdentifying local,
modifiable determinants of physical activity and
diet

• Mai Stafford, Amanda Sacker Dept of Epidemiology
  Public Health, UCL
• Sally Macintyre, Anne Ellaway MRC Social Public
  Health Sciences Unit, Glasgow
• Steve Cummins Dept of Geography, Queen Mary
  College London
• R D Wiggins Dept of Sociology, City
  University

2
Background/rationale

• Neighbourhood influences on health of increasing
  interest
• Methodological developments
• Empirical evidence focuses on deprivation index
  as exposure
• Neighbourhood deprivation associated with
• mortality (all-cause, suicide, deaths from heart
  disease)
• morbidity (self-rated health, disability,
  smoking, quality of life, common mental disorder)

3
(No Transcript)
4

• Composition or context?
• Multilevel data and analysis
• Reviews of evidence conclude there is an
  association between neighbourhood deprivation and
  health
• Neighbourhood deprivation black box

5
What neighbourhood characteristics might
influence health?
Service environment e.g. health services, public
transport, retail
Neighbourhood deprivation index
Built environment
Social environment e.g. social disorder, social
capital
6
Unpacking the black box

• Identify specific, amenable neighbourhood
  determinants of health
• How do these relate to each other?
• How do they jointly influence health?
• Focus on obesity a major public health issue
• Key determinants of obesity are diet and physical
  activity

7
Service environment and diet, physical activity
and obesity
Local sport leisure facilities Greater physical activity Giles-Corti et al, 2002
Presence of supermarket Higher fruit veg consumption, lower fat intake Morland et al, 2002
Poorer access (proximity car) Lower fruit veg consumption Cheadle et al, 1991
Availability low fat/high fibre foods Greater intake low fat high fibre Rose et al, 2004
No. fast food outlets Obesity Maddock, 2004
8
Built environment
Mixed commercial-residential land use Greater levels of walking Doyle et al, 2006 Frank et al, 2006
Urban sprawl (population density) Lower levels of walking Morenoff et al, 2006
Attractive scenery open spaces Greater physical activity Brownson et al, 2001 Sugiyama et al, 2005
9
Social environment- neighbourhood disorder
Neighbourhood safety Childhood obesity Lumeng et al, 2006
Fear on streets fear of attack Less outdoor physical activity Ross, 1993
Fear of crime Less walking for pleasure Parkes Kearns, 2006
Litter and graffiti Obesity Ellaway et al, 2005
10
(No Transcript)
11
Data methodsI. Measuring obesity

• Health Survey for England (1994-1999) and
  Scottish Health Survey (1995 1998)
• Nationally representative
• Height and weight measured by trained nurse
• Obesity body mass index weight/height2
• Age, gender and occupation-based SES
• Neighbourhood unit postcode sector
• Average population 5000

12
Data methodsII. Selecting neighbourhoods for
study
Eligible neighbourhoods
35 Health Survey participants
Stratify on population density Carstairs index
of deprivation
Ensure range of environments
Select sample of neighbourhoods
Stratified random sample
Data complete for 398 neighbourhoods
Collect neighbourhood data
13
Data methodsIII. Measuring the service built
environment

• Various sources administrative data from central
  government, local government, commercial
  organisations
• e.g. violent crime per capita, number of
  supermarkets
• Difficult to obtain data
• not available
• not complete
• not comparable England and Scotland
• Data collected or converted to postcode sector

14
Data methodsIV. Measuring neighbourhood
disorder

• Local Area Social Capital Survey conducted in
  2000
• 70-item postal questionnaire
• Sent to random sample of residents 16 years
• Response rate 42
• 5 items capture neighbourhood disorder

15
(No Transcript)
16
Data methodsV. Linking data
Administrative/ commercial data Service
environment Built environment
Postal survey Neighbourhood disorder
Existing health survey Individual level obesity
Link via postcode sector identifier
17
(No Transcript)
18
Data methodsVI. Structural Equation Modeling

• Aim to investigate inter-relationships between
  various neighbourhood characteristics and their
  relationship with obesity
• Beyond estimation of direct effects
• Examine causal processes underlying observed
  relationships
• Estimate relative importance of different causal
  pathways

19
Data methods

• Some neighbourhood characteristics modeled using
  latent variables
• i.e. measured variables are indicators of
  underlying construct
• Want to know about relationships between
  underlying constructs (not between measured
  variables)

20
(No Transcript)
21
(No Transcript)
22
Data methods

• Step 1 confirmatory factor analysis to assess
  how well measured variables capture underlying
  construct
• Step 2 structural equation model linking
  neighbourhood characteristics to obesity
• Allow for clustering of participants within
  neighbourhoods
• Present standardised factor loadings and path
  coefficients
• Mplus software

23
Results

• Factor analysis confirmed that
• i) neighbourhood disorder well-measured by the 5
  items from the Social Capital Survey
• ii) high street facilities well-measured by 5
  items (3 dropped)

24
(No Transcript)
25
Summary of findings

• Previous studies show neighbourhood deprivation
  is associated with obesity
• Our findings illustrate some specific features of
  the neighbourhood environment that are associated
  with obesity
• These are neighbourhood disorder, mixed
  commercial/residential land use and urban sprawl
• In turn, policing and vacant/derelict land
  influence levels of neighbourhood disorder

26
Discussion

• Study illustrates how data from various sources
  can be combined
• Latent variables can be used when have several
  indicators of the same underlying construct
• Structural equation models can be used to explore
  theoretical causal pathways
• Limitations include
• lack of neighbourhood data capturing the
  theoretical constructs of interest
• self-selection of participants into different
  types of neighbourhood
• cross-sectional
• defining neighbourhood boundaries to fit the
  data or to fit residents perceptions and
  experience

27
Concluding remarks

• Our findings illustrate that several determinants
  of obesity are not within traditional remit of
  healthcare sector
• Private sector and non-healthcare public sector
  has an important role
• Stronger links and even representation of public
  health in police force, land use planning bodies
  and groups supporting local business are needed
• Dont forget individual characteristics also
  strongly related to obesity