Published: 26/05/2005, Volume II5, No. 5957 Page 4 5 6
Geodemographic data has long been used in marketing, but now it is helping the public sector to find the causes of health inequalities and alert local authorities to the problems. Steve Mathieson reports
They are used to decide what junk mail you receive, whether a new shop will be opened near you and possibly even how much attention you got in the general election. But can geodemographic classification systems provide anything to the health service?
Professor Richard Webber, visiting professor to the centre for advanced spatial awareness at University College London, hopes so. He designed the UK's two main classifications of this kind, known as A Classification of Residential Neighbourhoods (Acorn) and Mosaic. Both divide the UK's 1.7 million postcode areas into around 60 types, based on statistical clusters of a wide range of public and private data.
Both are commercial products, but Professor Webber's first work in the field was for the state sector, and he is keen to see such techniques applied there again. 'What We are seeing now is the development of administrative databases, with health authorities holding more information about their patients, ' he says. Along with greater interest in preventing health problems, he thinks this creates the right conditions for the health service to exploit a concept that has long been used by commercial marketeers.
Geodemographic classification systems reckon they can tell a lot about you if they know where you live. Professor Webber's first work in the field was in Liverpool in the 1970s, where he categorised the city's deprived areas for the council.
The aim was to research the different needs of different areas.
He divided the city's poor districts into four areas. Inner city council estates featured lots of dockers, many of Irish descent. They were alienated from the city council, but socially cohesive.
The Liverpool 8 postcode area saw more drug addiction and single parenting, but people were better educated. In older terraces, people's problems were related to the poor quality of their houses, but social networks were strong. Recently built council estates had specific problems with youth offenders.
After the Conservative election victory of 1979, public funding dried up and Professor Webber moved to commercial classifications. He still does consultancy work for Experian, which owns Mosaic, and health information firm Dr Foster.
Although Mosaic is mainly used by the commercial sector - eg for marketing and choosing where to site new shops - Professor Webber says the state sector's needs were built in to Mosaic's 2003 revamp around 2001 census data.
'The concept of neighbourhood classification wasn't designed for marketing, but for applications of this sort, ' says Professor Webber. 'It was only by accident that marketing found this useful.
'When we rebuilt the classifications in 2003, we wanted to make them as useful to the public sector as the private. There are at least as many indicators of deprivation as of affluence.' He added that if Mosaic had been focused on the private sector, the classification would have lumped poorer people together into fewer categories.
Is targeting public health campaigns unfair if it exposes some groups to more campaign work than others? Professor Webber says not. 'I do not think with preventive campaigns [such as smoking cessation] we should complain if they are targeted, because everyone benefits, ' he says. Falling rates of lung cancer would mean more health resources for all.
CACI, which owns Acorn, Professor Webber's earlier national classification system, says several health bodies have used it.
'It is a growing field, ' says James Lennon, head of public sector solutions at the firm. 'One of the things health people always prided themselves on was having good-quality data, but they were not sure how to use it.' Mr Lennon adds that much of the interest from health has come from the private sector and insurers, but CACI is encouraging the state sector by offering lower prices than it does to the private sector.
He says that one advantage of using Acorn is that users can draw on other research, such as alcohol consumption studies for the commercial sector.
'The ability to profile a community, with all that information, means you can think about joining up information on a community and on clusters within it, ' says Mr Lennon.
'With partners, you can then think in a joined-up way as to how you can improve the situation. The health service is not going to improve housing, but being able to describe the problem to the local authority will help it address that.' Acorn is used by the Department for Education and Skills, the annual British crime survey and within local government funding calculations.
Mosaic is used by Dr Foster, which has combined the classification with its hospital episode statistics data. Its first work in this area was on the incidence of diabetes in Slough. It found that the wealthier of two Mosaic types for areas dominated by Asian ethnic groups (known as C20, Asian enterprise, 1.02 per cent of the population) saw particularly high incidence of the condition. This may be down to racial vulnerability to the condition, combined with rich diets and low levels of exercise.
The firm has since used Mosaic to analyse a variety of conditions. 'It seems to work well at most things We have looked at, ' says Dr Marc Farr, product development manager. 'I didn't think it would be particularly good at mental health, but it has produced good results.' Cross-referencing found that those of mixed race, as well as exstudents who have fallen on hard times, are particularly likely to be depressed, and that they often live in flats above shops in towns and cities.
Mosaic types are often connected to wealth, which can help to confirm disease profiles known through index of deprivation data: lung cancer is more likely to hit the poor; skin cancer affects those who are richer, because they take more foreign holidays; breast cancer affects women from all backgrounds.
'The deprivation index is a wellknown, well-used resource, but it only works at a high level of geography, ' says Dr Farr. 'Mosaic and Acorn go down to postcodes.'
Spread the message Given their use in marketing, geodemographic classification systems also provide information on how to reach people in individual groups. This can be which newspaper a type reads, or how much television it watches, for example.
Companies use this data to decide how to advertise to their target audiences most efficiently, and the NHS could follow suit, says Dr Farr.
Dr Foster used Mosaic with Brent teaching primary care trust to analyse prevalence of teenage pregnancy.
'We have taken national data on teenage pregnancy and inferred that on to Brent. Dr Foster can add what's going on with those groups, picking the groups at risk, doing focus groups, and then designing a marketing campaign, ' says Dr Farr.
The Brent Mosaic analysis found disproportionately large numbers of teenage pregnancies among the mostly black settled minorities type (D27, 1.62 per cent of the overall population), as well as those within welfare borderline (the F group, 6.43 per cent of the population).
At the same time, Dr David Newsom, then at Brent PCT but now a consultant paediatrician for Ealing Hospital trust, made use of Mosaic at no extra cost to analyse take-up of the combined measles, mumps and rubella triple vaccine.
'A big cold patch appeared in the south east of the borough, ' says Dr Newsom. 'One of the ideas was that this was an area white middle-class families were migrating to. We were looking for some sort of tool that could tell us about socio-economics in this area. That was why -6 Mosaic was useful. It doesn't just have the census data, It is much richer - the census does not, for example, tell you the wealth of households.' Brent tracked how Mosaic groups differed in their take-up, finding significant differences between the six types that make up 87 per cent of the area's population. 'The Asian enterprise group had virtually unchanged use of the vaccine, ' says Dr Newsom. He thinks they were getting health information mainly from Asian media sources, rather than the mainstream ones that campaigned against MMR.
This type makes up 30 per cent of the borough. Another mainly ethnic minority-focused Mosaic type, settled minorities, makes up 24 per cent. It too recorded little change.
The research confirmed that those living in new urban colonist postcodes, the Mosaic type Dr Newsom thought might be invading the borough, had lessened their take-up of MMR, with usage dropping from 82 per cent in 1995 to 59 per cent in 2003.
But Mosaic's data found that the group made up only 3.5 per cent of Brent's population, showing their influence was limited. Smaller falls in take-up within larger groups, such as the counter-cultural mix type (E28), which makes up 12 per cent of Brent's population, contributed more to the overall drop.
'It describes data in an interesting manner, ' says Dr Newsom, in comparison to a purely ethnic groupbased analysis. 'Because we know the socio-economic group by postcode, they are hopefully giving us a more refined picture, whereas from the PCT data we wouldn't be able to tell wealth or status. This gives us a much richer impression of what's going on.' An analysis of the data found that the groups with the greatest decline in MMR spent more time listening to the radio, indicating where the arguments over the vaccine's use would need to be won.
But Dr Newsom does have a criticism of Mosaic. 'I am not sure how well-validated it is for medical data, ' he says. 'They've described how the system is put together, but I am not sure It is been measured against other tools. Obviously they've got a vested interest in not divulging the fine details, and I understand its application is in the commercial sector.' However, Dr Newsom says the fact that the classification is private has made it difficult to get research published. 'But quite whether It is robust enough for scientific publication is separate to whether It is useful, ' he adds.
Completing the picture
Blackburn with Darwen PCT's teenage pregnancy partnership, which involves local NHS trusts, the council and local teenage pregnancy advisory services, made use of Acorn in researching teenage maternities at the local hospital, again using postcodes.
However, Blackburn with Darwen borough council research information officer Elise Carroll says the partnership used Acorn data to support other address-based classifications, such as educational attainment, school catchment areas and an index of multiple deprivation. 'Acorn backed up a lot of the issues we saw coming out of the other data sets, ' says Ms Carroll.
'It helped complete the picture.' Blackburn with Darwen used only the broadest five categories within Acorn, finding that many areas with high levels of teenage maternity came into hard-pressed, the poorest category (see box). Ms Carroll says the data may be useful in showing areas of need in supporting funding bids.
Not everyone agrees that such classifications have great uses in the health service. North Birmingham PCT director of public health Dr Jammi Rao considered using Acorn in targeting mail to encourage smoking cessation, but decided against it.
'We have direct data, ' he says, adding this is his personal view rather than that of the PCT. 'With a new brand of wine, your market is potentially anyone, so you could use the likes of Acorn. But We are sitting on a wealth of data already. As long as we use the information ethically, we do not need further categorisation based on where people live.' Dr Rao's project involved pulling out the names and addresses of those recorded as smokers from GP data, then sending them letters in the name of their GP with information on local smoking cessation services. 'We know who smokers are and where they live. We felt it was ethical to let them know what services there are, ' he says.
Dr Rao says there are clear links between some conditions and the geographical classifications, but adds that there will be individuals with needs not recognised by their area's type - and the NHS should serve individual needs. 'I am not convinced there is a great potential for this kind of marketing of health, ' he says. 'We have a wealth of information we are under-using.' .
CLASSIFICATIONS It is NO LONGER AS SIMPLE AS A, B, C1. . .
Geodemographic classifications do not rely on any one variable, unlike socio-economic groupings: the familiar classes such as A, B and C1 rely purely on the profession of the head of the family. But with many more job titles in circulation, dual-income couples and pensioners the classification has been weakened.
Experian's Mosaic was created through looking for statistical clusters in a wide range of data. The 2001 census provided 54 per cent, but the system also uses more regularly updated sources.
These include the 36 million-strong edited electoral roll, credit data based on county court judgments, consumer research surveys and house price data. Actual postal addresses can also make a difference, with house names indicating a better neighbourhood, and individuals' names are used as a clue to the ethnic mix.
Mosaic divides the UK's postcode areas into 11 groups (lettered from A to K), which are further divided into 61 types. A few of these types concern professions, such as upland hill farmers (K61, 0.41 per cent of the population), and some are primarily racial, such as South Asian industry (D26, 0.88 per cent).
Rather more are concerned with old age, such as group I, dubbed twilight subsistence (3.88 per cent), and group J, grey perspectives (7.88 per cent).
Others have names perhaps best suited to a marketing executive, such as caring professionals (E31, 1.08 per cent) and white van culture (H46, 3.17 per cent).
One public sector user, Devon and Cornwall constabulary, prefers to stick to just the type number. Dr Foster uses alternative sentence-long descriptions provided by Experian, replacing 'white van culture' with 'younger owners, mainly in good quality ex-council properties, take advantage of local economic opportunities'.
They are a 'bit drier', says Dr Foster's Dr Marc Farr, but they are also 'more appropriate'.
The names used in CACI's Acorn, such as skilled workers, semis and terraces (M41, 3.3 per cent of the population), tend to be purely descriptive. It uses three levels, compared with Mosaic's two. There are five overall categories: wealthy achievers (25.1 per cent); urban prosperity (10.7 per cent); comfortably off (26.6 per cent); moderate means (14.5 per cent) and hardpressed (22.4 per cent).
These are each divided into three or four groups, of which there are 17 overall, and these are then further subdivided into types, of which there are 57.
As with Mosaic, Acorn's types are often based around wealth or age, with a few focusing on ethnic group.
'One of the differences between Mosaic and Acorn is that we put more emphasis on age, ' says James Lennon of CACI. 'In some areas, We have split a population between a predominately older group and a predominately younger one.'
Find out more
Experian's Mosaic www. business-strategies. co. uk/ Content. asp? ArticleID=566 CACI's
Acorn www. caci. co. uk/acorn
Acorn can also be searched without registration at www. upmystreet. com
A classification architect says the NHS should make more use of geodemographic data.
Systems use data such as house prices, court judgements and consumer research.
Much of the health service's interest in the system has come from insurers and the private sector