Public health: Life Expectancy in Sheffield, UK

Use of GIS in public health

This page provides a simple example of how GIS can be used to visualise geographical relationships that may be contained within routine health information. We also show how methodology from spatial-epidemiology and spatial-statistics can be used to assess and measure spatial relationships within health data. In this example life-expectancy at birth for the population of Sheffield, in the UK, has been used as the health outcome.

Background

Health related outcomes such as disease prevalence and mortality can show some degree of spatial variation in which location, or geography, is a risk factor for the outcome.

To investigate the relationship between health and location analytical techniques that are able to incorporate spatial information are used. Geographical Information Systems (GIS), spatial-epidemiology and spatial-statistics provide the scientific framework under which these types of public health issues can be investigated.

Geographical visualisation

Life-expectancy can be calculated from routine mortality and population data. Table 1 shows male and female life-expectancy at birth for each electoral ward in Sheffield. This is the average number of years that a man, or woman, would be expected to live from birth.

The data shows that there is considerable variation across Sheffield's 28 wards. On average women living in Fulwood live 8.5 years longer than women in Burngreave. And men living in Dore and Totley are expected to live 7.6 years longer than those living in Firth Park.

In a tabular format it is difficult to assess whether life-expectancy is spatially correlated. Are the wards with longer, or shorter, life-expectancy situated close together or randomly dispersed? Using GIS the information can be mapped against the ward geography. This allows geographical patterns in life-expectancy to be seen, or visualised.
Figure 1 (males) and Figure 2 (females) show Sheffield's wards coloured to indicate whether they have low, medium or high life-expectancy. In these thematic, or choropleth, maps the wards have been ranked by life-expectancy and classified in to tertiles. Each tertile comprises an approximately equal number of wards. An Ordnance Survey map has been used as a background layer to provide geographical context.

Figure 1: Male life expectancy (years from birth)

Figure 2: Female life expectancy (years from birth)

Geographical analysis

The life-expectancy maps show an east-west gradient across Sheffield. People who live to the west of the city centre tend to live longer than those in the east. The maps do not, however, provide any explanation why this may be so. To investigate "why" requires additional data for factors that may affect life-expectancy. This data needs to be spatially referenced and of sufficient granularity to allow ward level variables to be calculated.
Socio-economic deprivation is often associated with differentials in health and can be spatially structured. In Sheffield the major industrial areas and transport networks are predominantly situated to the east of the city centre. The city's more affluent areas tend to lie to the west of the city centre, which is less industrialised and borders a National Park.
Area based indices of deprivation are calculated by the UK Government. They combine a number of indicators into a single deprivation score for small-area census units in England. The indicators are chosen to cover a range of economic, social and housing issues. Table 2 shows Sheffield's wards ranked by these deprivation indices. Ecclesall is the least deprived Sheffield ward and Manor Castle is the most deprived ward. Deprivation tertile-1 comprises the 9 Sheffield wards with the lowest levels of deprivation. Tertile-2 is the middle category (9 wards), and tertile-3 is the most deprived category comprising 10 wards with the highest levels of deprivation.
The geographical distribution of ward level deprivation across Sheffield is shown in Figure 3. The map shows a similar geographical pattern to that seen in the life-expectancy data. Based on a visual assessment we would expect that deprivation will account for some of the variation seen in male and female life expectancy at the ward level.
The scatterplot shown in Figure 4 shows an association between life-expectancy and deprivation at the ward level. The relationship is approximately linear with a negative coefficient, which indicates that life-expectancy tends to be lower in wards with higher levels of socio-economic deprivation. Fitting the data with a simple linear regression model shows that the association is statistically significant at an area level (p<0.001 for males and females). The amount of variation in life-expectancy that is explained by the model is approximately 82% for males and 51% for females. These estimates are based on a statistical model that assumes independence between deprivation and ward (geography), which as Figure 3 shows is unlikely to be true. Spatial regression models would be used in order to conduct a more robust analysis.

Figure 3: Sheffield ward deprivation tertiles

Figure 4: Life expectancy vs socio-economic deprivation

Increasing deprivation scores indicate higher levels of deprivation
R-squared from linear regression model

Summary

Use of GIS mapping shows that, for the population of Sheffield, the average male and female life expectancy varies according to the ward in which people live. The geographical pattern is similar for men and women. People who live to the east of the city centre tend to die at a younger age than those in the west of Sheffield. Socio-economic deprivation shows a similar geographical distribution to life-expectancy and appears to be a factor in explaining some of the variation in life expectancy across Sheffield. However, deprivation does not explain all the variation. Other factors will be important in determining the average length of time that a man or woman living in Sheffield can expect to live.
A linear regression model shows that deprivation appears to be more closely associated with male life-expectancy than female life-expectancy. And as the level of deprivation increases the differential between absolute male and female life expectancy tends to increase.
A more complete analysis would aim to look at other variables that were likely to affect population health and life expectancy. These may include environmental hazards such as air pollution, industrial waste, noise, and climate. Lifestyle choices such as diet can also influence health, as can access to health care. Each factor could be investigated providing data was available.
Sophisticated spatial modelling techniques can be used to provide information on the amount of variation in the outcome that is explained by each variable, as well as the proportion of spatial and non-spatial variation that each variable can explain. The level of uncertainty that is associated with each variable can also be quantified using these techniques. Examining associations using geographical areas is also complicated by the fact that neighbouring areas may be similar. This spatial dependancy causes statistical problems such as spatial autocorrelation, in which assumptions of independence among observations are not valid. Sophisticated spatial modelling techniques can be used to address these issues.
Area based investigations have their limitations. Observations that are seen at a geographical level do not necessarily hold at an individual level. This means that inference about individuals from area based, or ecological, studies is complicated. These limitations are referred to as ecological fallacy or ecological bias.

Data notes

Life expectancy calculations are based on all-cause mortality information recorded over a 9-year period from 2001 to 2010.
Wards are the key building block of UK administrative geography. The UK had 9,523 electoral wards. The district of Sheffield comprises the 28 wards. Ward population counts can vary substantially. The Sheffield average is 19,840 persons. More populous wards tend to occur in large urban areas.
Ward level deprivation scores were calculated from Lower Layer Super Output Area (LSOA) Index of Multiple Deprivation (IMD) 2010 composite scores using a population weighted method. These calculations contain National Statistics data © Crown copyright and database right [2010].

Contact

Please contact Professor Ravi Maheswaran @ r.maheswaran@sheffield.ac.uk

Page updated

Report abuse