Improving the ethnic classification of patient registers

1
Improving the ethnic classification of patient
registers

• Centre for Advanced Spatial Analysis, UCL
• 25th May, 2005

2
Objectives of seminar

• Promote awareness of existing tools for
  targetting health communications at ethnic
  groups
• Individual patients on registers
• Surgeries and other contact points
• Local promotional activities
• 2. Discussion of ideas for collaborative use of
  individual patient registers to improve the
  quality of ethnic coding
• 3. Forum for exchange of information
• Methods
• Ethics and other data protection issues

3
Our conception of Ethnicity

• Multi-variate classification based on a
  combination of
• Cultural Origin eg religion, beliefs
• Ethnicity eg country of origin, diet
• Language

4
Background to this seminar
5
Context CASA (1)

• 1970s - Neighbourhood classifications used for
  prioritising public sector initiatives
• 1990s - Application of postcode classifications
  adopted by commercial companies
• 2002 CASA becomes involved in the application
  of Mosaic in health, policing and education 2003
• CASA work with Dr Foster on Slough Diabetes
  project

6
Context CASA (2)

• 2004 CASA sets up Knowledge Transfer
  Partnership (KTP) with Camden PCT to develop
  health applications of geodemographics
• 2004 CASA wins ESRC grant for quantitative
  analysis of names
• 2005 Camden PCT develops capability in the
  application of names as well as Mosaic to
  targetting of public health campaigns

7
Contact details

• CASA website
• E/mail addresses
• Pablo Mateos p.mateos_at_ucl.ac.uk
• Richard Webber richardwebber_at_blueyonder.co.uk
• Paul Longley p.longley_at_geog.ucl.ac.uk

www.casa.ucl.ac.uk/geonom
8
Quantitative Analysis of Names

• ESRC funded project
• Use of surname as an identifier of cultural
  origin
• Regional origins of English names
• Regional distribution of Celtic names
• Current locations of names imported from abroad
• Jewish
• Continental European and Hispanic
• Asian
• African
• Middle Eastern

9
Identification of potential applications

• Academic / Social Scientific
• Study of meaning of names
• Studies of historic migration patterns
• Social mobility of Celtic migrants to England
• Policy applications
• Measurement of social capital
• Differentiation of crude South Asian definition
• Targetting of public sector communications
  programmes
• Auditing of equal opportunities in employment

10
Key data files

• 40 million records Experian 1996 GB electoral
  roll
• First name
• Surname
• Postal area code
• Mosaic code
• 26 million records 1881 census
• Summary statistics on name frequencies by region
  from Anglophone diaspora
• US, Canada, Australia, New Zealand, North and
  Southern Ireland

11
Geography of the name Webber
12
electors with occupational names
13
electors of Welsh surnames
14
CEL assignment Phase one

• Identify 25,000 surnames with 100 occurrences
  in 1996
• Assign to hierarchy
• English general name type detailed name type
• Celtic country of origin general type
• Imported from abroad region of origin country
  of origin

15
Webber

• Level one English metonym
• Level two Metonym ending in -er
• Level three Manufacturing occupation

16
Zhang

• Level one Imported from abroad
• Level two East Asian
• Level three Chinese

17
(No Transcript)
18
Muslim and South Asian names (1462)
19
Phase one assignment method(25,000 names with
100 occurrences)

• General knowledge
• Identification of top postal area and level of
  concentration in it
• Identification of top Mosaic type and level of
  concentration in it
• Identification of concentration in 1881
• Frequencies in other Anglophone countries

20

21
(No Transcript)
22
(No Transcript)
23
Status and Asian names
24
(No Transcript)
25
(No Transcript)
26
Output

• Directory assigning a Cultural / Language /
  Ethnicity code to each name with more than 100
  occurrences on the GB electoral roll

27
Phase two assignment(all surnames 5
occurrences)

• Rank first names by frequency
• Allocate names to CEL categories where possible
• Identify for each surname the proportion of
  associated first names in known CEL categories

28

29
Selected first names
30
Output

• Database giving for 60,000 surnames imported
  from abroad
• electors by CEL of first name
• Most common cell (three level hierarchy)
• Database giving for 60,000 first names imported
  from abroad
• (3.2m occurrences)
• electors by CEL of surname
• Most common cell (three level hierarchy)

31
Evaluation of solution

• Seems to work well for all ethnic groups other
  than Caribbeans
• CEL overlaps between surname and first name
• South Asians and Muslims 80
• Africans, Turks, Cypriots, Chinese 50
• Hispanics 20
• Other Europeans 8 15
• Jew 4
• Irish, Scots, Welsh 3
• High overlap between certain CELs
• within Muslim group
• Spain, Portugal, Italy
• Netherlands, Germany and Czech Republic
• Confusion among serial migrant groups
• Hispanic migrants to India
• Chinese migrants to West Indies