Adding geographical detail to social surveys: Estimating local disability prevalence Alan Marshall ESDS Government 15 th April 2010.

Adding geographical detail to social surveys:

Estimating local disability prevalence

Alan MarshallESDS Government

15th April 2010

The problem

• Researchers require detailed local information, for example, to provide appropriate services. BUT:

• The data that is available for small areas (districts, wards) often lacks detail– Census question on limiting long term illness

and disability

• Detailed data sources lack geography (confidentiality) – HSE has information on specific disabilities but

does not distinguish district of residence

• Combining locally available data with survey data offers a solution to this problem

Disability age pattern

0.1

.2.3

.4.5

0 20 40 60 80age

Observed rates Model rates

Disability rates (Higher severity) by age in England (females)

Source: Health Survey for England 2001

Many disability types are strongly linked to age

Prevalence ratio method

• Multiply the national disability rate (HSE) and the local population count at each age.

• This approach is used by the POPPI and PANSI websites to estimate mobility and personal care disability.

• Developed by the Institute for Public Care.

• Designed to help explore the possible impact that demography and certain conditions may have on populations.

• http://www.pansi.org.uk/• http://www.poppi.org.uk/

3000 2000 1000 0 1000 2000 3000 4000 5000

0

5

10

15

20

25

30

35

40

45

50

55

60

65

70

75

80

88A

ge

Population

Bolton population pyramid – 2001 and 2021

Males Females

Grey bars indicate the population in 2001Clear bars indicate the population in 2021 Source: ONS MYE and pop projections

10000 8000 6000 4000 2000 0 2000 4000 6000 8000

0

5

10

15

20

25

30

35

40

45

50

55

60

65

70

75

80

88A

ge

Population

Manchester population pyramid – 2001 and 2021

Males Females

Grey bars indicate the population in 2001Clear bars indicate the population in 2021 Source: ONS - MYE and pop projections

Including further local information

• Disability is linked to characteristics other than age that are available for local areas in the census

• E.g. LLTI increases the risk of having a personal care disability

If two districts with same age structure But one has higher level of LLTIThen we would expect higher levels of personal care disability

Relational models

0.2

.4.6

.8P

reva

len

ce

20 40 60 80 100Age

LLTI - Census Personal care - HSEPersonal care - Modelled

England - MalesLLTI and Personal care disability prevalence

Two adjustments relate the LLTI curve to the PC curve.

Comparing PANSI and relational estimates of personal care disability

Personal care prevalence (2008) (18-64)

SIR PANSI

Relational estimate

South Bucks 0.66 5.0 2.4 Brighton 0.98 4.1 4.3 Bury 1.05 4.7 5.2 Pendle 1.16 4.8 6.1 Wakefield 1.22 4.8 6.3 Easington 1.63 4.8 9.9

SIR = Standardised illness ratio

Other approaches – Area classifications

• People with similar socio-demographic characteristics cluster together within certain areas

• Classify areas into groups according to the socio-demographic characteristics of their populations

• ACORN MOSAIC – commercial classifications

• ONS classifications (since 1970s)

• HSE (2001) includes ONS area classification and urban/rural classifications

ONS Area classifications

ONS district classification (2001) SIRs – Census 2001SIR (2001)

Evaluating local estimates

• Catch 22 – if there were local estimates then we wouldn’t need to create them!

• Compare with proxy data - administrative records

• Seek opinions of local experts

• Compare with locally conducted surveys

• Compare estimates produced using different methods

Disability estimates from mythesis will soon be at:

http://www.ccsr.ac.uk/staff/am.htm

POPPI and PANSI websites:

http://www.pansi.org.uk/http://www.poppi.org.uk/

Output area classification group

http://areaclassification.org.uk/

Bajekal, M., Scholes, S.,Pickering, K. and Purdon, S.(2004). Synthetic estimation ofhealthy lifestyle indicators: Stage1 report. NatCen, London

Skinner, C. (1993). The Use ofSynthetic Estimation Techniquesto Produce Small AreaEstimates. New MethodologySeries NM18. OPCS. London.

Resources References