ABSTRACT: Spatial microsimulation models can be used to estimate previously unknown data at the micro-level, although validation of
these models can be challenging. This paper seeks to describe an approach to validation of these models. Obesity data in adults
were estimated at the small area level using a static, deterministic, spatial microsimulation model called SimObesity. This
model utilised both Census 2001 data and the Health Survey for England for 2004–2006. Regression analysis was used to identify
the covariates that were the strongest predictors of obesity and these were used as the model input variables. The model was
calibrated using regression and equal variance t-tests. Two methods of external validation were undertaken; aggregating obesity
data to a coarser geographical level at which obesity data was available, and secondly using small area level cancer data
for tumour sites known to be correlated to obesity. The output obesity data were mapped and statistically significant hot
(cold) spots of high (low) prevalence of obesity identified. Both internal and external validation showed low errors, suggesting
this was a satisfactory simulation. Statistically significant hot and cold spots of (simulated) obesity prevalence existed,
even after adjusting for age. This paper emphasises three steps to validation of spatial microsimulation models: 1. Accurate
simulations require strong correlations between the input and output variables; 2. It is essential to internally validate
the models; 3. Use all means possible to externally validate the model.
KeywordsObesity–Small area estimation–Spatial microsimulation modelling
Applied Spatial Analysis and Policy 04/2012; 4(4):281-300.
ABSTRACT: The School Fruit and Vegetable Scheme (SFVS) provides children in government-run schools in England with a free piece of fruit or a vegetable each school day for the first 3 years of school. The present study examines the impact of the SFVS, in terms of its contribution towards the total daily intake of fruit and vegetables by children across England. Quantitative dietary data were collected from 2306 children in their third year of school, from 128 schools, using a 24 h food diary. The data were examined at different spatial scales, and variations in the impact of the scheme across areas with different socio-economic characteristics were analysed using a deprivation index and a geodemographic classification. The uptake of the SFVS and the total intake of fruit and vegetables by children varied across different parts of England. Participation in the SFVS was positively associated with fruit and vegetable consumption. That is, in any one area, those children who participated in the SFVS consumed more fruit and vegetables. However, children living in deprived areas still consumed less fruit and vegetables than children living in more advantaged areas: the mean daily frequency of fruit and vegetables consumed, and rates of consumption of fruit or vegetables five times or more per d, decreased as deprivation increased (r -0.860; P = 0.001; r -0.842; P = 0.002). So the SFVS does not eliminate the socio-economic gradient in fruit and vegetable consumption, but it does help to increase fruit and vegetable consumption in deprived (and affluent) areas.
The British journal of nutrition 02/2012; 108(4):733-42. · 3.45 Impact Factor
ABSTRACT: To monitor growth trends in young children in order to ascertain success (or otherwise) in halting the rapid rise in childhood obesity prevalence, and to assess the suitability of using routinely measured data for this purpose.
Retrospective serial cross-sectional analyses of the proportion of obese children (logistic regression) and BMI standard deviation score (linear regression/maps) were undertaken. BMI coverage was calculated as percentage of sample with data ('usual'), percentage of total births and percentage of census values. BMI was standardised for age and sex (British reference data set).
Metropolitan Leeds, UK.
Children aged 3 to 6 years. Weight, height, sex, age and postcode data were collected from Primary Care Trust records.
Data were collected on 42 396 children, of whom 13 020 (31 %) were excluded due to missing data/data problems. Seventy-two per cent of 3-year-olds and 92 % of 5-year-olds had data recorded ('usual' coverage). From 1998 to 2003 there was a significant increase in the proportion of obese children (4.5 % to 6.6 %; P < 0.001); children were 1.5 times more likely to be obese in 2003 than in 1998.
Childhood obesity rose significantly between 1998 and 2003. Routinely measured data are an important means of monitoring population-level obesity trends, although more effort is required to reduce the quantity of data-entry errors, for relatively low marginal cost.
Public Health Nutrition 01/2011; 14(1):56-61. · 2.17 Impact Factor
ABSTRACT: To analyse the association between childhood overweight and obesity and the density and proximity of fast food outlets in relation to the child's residential postcode.
This was an observational study using individual level height/weight data and geographic information systems methodology.
Leeds in West Yorkshire, UK. This area consists of 476 lower super-output areas.
Children aged 3-14 years who lived within the Leeds metropolitan boundaries (n=33,594).
The number of fast food outlets per area and the distance to the nearest fast food outlet from the child's home address. The weight status of the child: overweight, obese or neither.
27.1% of the children were overweight or obese with 12.6% classified as obese. There is a significant positive correlation (p<0.001) between density of fast food outlets and higher deprivation. A higher density of fast food outlets was significantly associated (p=0.02) with the child being obese (or overweight/obese) in the generalised estimating equation model which also included sex, age and deprivation. No significant association between distance to the nearest fast food outlet and overweight or obese status was found.
There is a positive relationship between the density of fast food outlets per area and the obesity status of children in Leeds. There is also a significant association between fast food outlet density and areas of higher deprivation.
Health & Place 11/2010; 16(6):1124-8. · 2.67 Impact Factor
ABSTRACT: The aim of this paper was to investigate variations in childhood obesity globally and spatially at the micro-level across Leeds.
Body mass index data from three sources were used. Children were aged 3-13 years. Obesity was defined as above the 98th centile (British reference dataset). The data were analysed by age group and gender, then tested for significant micro-level hot spots of childhood obesity using a spatial scan statistic and a two-level multilevel model.
Older children (13 years) were 2.5 times (95% CI 2.1 to 3.1) more likely to be obese than younger children (3 years). Childhood obesity was significantly associated with deprived and affluent areas. 'Blue collar communities,' 'Constrained by circumstances' and 'Multicultural' had significantly higher (relative risk (RR): 1.1, 1.2, 1.2; 95% CI 1.0 to 1.2, 1.1 to 1.2, 1.1 to 1.3, respectively) obesity levels, and 'Typical traits' and 'Prospering suburbs' had significantly lower (RR: 0.9, 0.8; 95% CI 0.8 to 1.0, 0.7 to 0.9, respectively) obesity levels. In the unadjusted model, obesity 'hot spots' were found in deprived (RR 1.5) and affluent (RR 6.1) areas. After adjusting for demographic covariates, hot spots were found only in affluent areas (RR 1.6 to 1.9), and cold spots in affluent (RR 1.3 to 4.4) and deprived (RR up to 1.1) areas.
These results suggest there is either a spread of obesity across socio-economic groups and/or something special about the high-/low-prevalence areas that affects the likelihood of obesity. The microlevel spatial analyses displayed the variations in obesity across Leeds thoroughly, identifying high-risk populations.
Archives of Disease in Childhood 11/2009; 95(2):94-9. · 2.88 Impact Factor
ABSTRACT: Obesogenic environments are a major explanation for the rapidly increasing prevalence in obesity. Investigating the relationship between obesity and obesogenic variables at the micro-level will increase our understanding about local differences in risk factors for obesity. SimObesity is a spatial microsimulation model designed to create micro-level estimates of obesogenic environment variables in the city of Leeds in the UK: consisting of a plethora of health, environment, and socio-economic variables. It combines individual micro-data from two national surveys with a coarse geography, with geographically finer scaled data from the 2001 UK Census, using a reweighting deterministic algorithm. This creates a synthetic population of individuals/households in Leeds with attributes from both the survey and census datasets. Logistic regression analyses identify suitable constraint variables to use. The model is validated using linear regression and equal variance t-tests. Height, weight, age, gender, and residential postcode data were collected on children aged 3-13 years in the Leeds metropolitan area, and obesity described as above the 98th centile for the British reference dataset. Geographically weighted regression is used to investigate the relationship between different obesogenic environments and childhood obesity. Validation shows that the small-area estimates were robust. The different obesogenic environments, as well as the parameter estimates from the corresponding local regression analyses, are mapped, all of which demonstrate non-stationary relationships. These results show that social capital and poverty are strongly associated with childhood obesity. This paper demonstrates a methodology to estimate health variables at the small-area level. The key to this technique is the choice of the model's input variables, which must be predictors for the output variables; this factor has not been stressed in other spatial microsimulation work. It also provides further evidence for the existence of obesogenic environments for children.
Social Science [?] Medicine 09/2009; 69(7):1127-34. · 2.70 Impact Factor
ABSTRACT: The aims of this study were to model jointly the incidence rates of six smoking related cancers in the Yorkshire region of England, to explore the patterns of spatial correlation amongst them, and to estimate the relative weight of smoking and other shared risk factors for the relevant disease sites, both before and after adjustment for socioeconomic background (SEB).
Data on the incidence of oesophagus, stomach, pancreas, lung, kidney, and bladder cancers between 1983 and 2003 were extracted from the Northern & Yorkshire Cancer Registry database for the 532 electoral wards in the Yorkshire region. Using postcode of residence, each case was assigned an area-based measure of SEB using the Townsend index. Standardised incidence ratios (SIRs) were calculated for each cancer site and their correlations investigated. The joint analysis of the spatial variation in incidence used a Bayesian shared-component model. Three components were included to represent differences in smoking (for all six sites), bodyweight/obesity (for oesophagus, pancreas and kidney cancers) and diet/alcohol consumption (for oesophagus and stomach cancers).
The incidence of cancers of the oesophagus, pancreas, kidney, and bladder was relatively evenly distributed across the region. The incidence of stomach and lung cancers was more clustered around the urban areas in the south of the region, and these two cancers were significantly associated with higher levels of area deprivation. The incidence of lung cancer was most impacted by adjustment for SEB, with the rural/urban split becoming less apparent. The component representing smoking had a larger effect on cancer incidence in the eastern part of the region. The effects of the other two components were small and disappeared after adjustment for SEB.
This study demonstrates the feasibility of joint disease modelling using data from six cancer sites. Incidence estimates are more precise than those obtained without smoothing. This methodology may be an important tool to help authorities evaluate healthcare system performance and the impact of policies.
International Journal of Health Geographics 02/2008; 7:41. · 2.62 Impact Factor
Social Science & Medicine. 70(12):2096-2096.