Using the HSfE to estimate small area indicators of health need Graham Moon, Steve Barnard, Gemma Holt and Liz Twigg Institute for the Geography of Health,

Using the HSfE to estimate small area indicators of

health need Graham Moon, Steve Barnard, Gemma Holt

and Liz TwiggInstitute for the Geography of Health,

University of Portsmouth Bill Blyth

TNS (Taylor Nelson Sofres)

Aim

• To provide an oversight of the outcomes of a programme of integrated projects focussed on the derivation of small area indicators of health need using synthetic estimation.

• To outline key issues arising in this research

Objectives

• Citizen-based public heath – NSF drivers– Identifying the local geography of risk– Facilitating resource allocation and

monitoring at a local scale – Flexible, consistent, bottom-up, robust

(Wanless agenda)

• Academic– Substantive advance in modelling

methodology– Determinants of health geographies

Data and Methods

• Consolidated 1998-2003 – HSfE,

– census

– TNS data

– + rest of UK)

• Multilevel modelling

The Indicators to Date I

• Known Heart Disease• Known Heart Disease Risk• Undiagnosed Heart Disease Risk• Hypertension• Obesity +• Known Type II Diabetes• Undiagnosed Type II Diabetes

NB population health measures not service users

The Indicators to Date II

• Flexible Organisational Geographies:– PCTs,

– Wards,

– General Practices

– Other

The Indicators to Date III

Robust high quality indicators

Indicator Var Exp % Match %

IHD 89 91

Undiag IHD 54 88

Ttwo diag 72 95

Obesity 62 72

Linkage Issues

• Question comparability • Changing geographies over time

– The disappearance of DHAs • Geographical comparability of

spatial levels– PSU = ward

• Spatial identifiers– Not disclosed for general use

Scale issues

• Variation is scale dependent

L1 L2 L3 L4

CHD

U-CHD

TTWO

OBE

PCT ScaleDiagnosed CHD

Undiagnosed Risk of CHD

Ward Scale

Diagnosed CHD

Undiagnosed Risk of CHD

Scale

GP Surgeries

Alternatives

• HES

• QoF

• IMD

• Area typologies

Alternatives - HESCHD

R2 = 0.1121

0

5

10

15

20

25

30

35

30 50 70 90

GeoMedics chd

SB

PC

T c

hd

Alternatives QoFCHD

R2 = 0.2104

0

10

20

30

40

50

60

70

80

30 50 70 90

GeoMedics

SB

PC

T

Alternatives –IMD

Overall index of multiple deprivation score

100806040200

Pe

rce

nta

ge

est

ima

ted

to

ha

ve t

ype

2 d

iab

ete

s

8

6

4

2

0 Rsq = 0.2018

Health deprivation and disability score

43210

Pe

rce

nta

ge

est

ima

ted

to

ha

ve t

ype

2 d

iab

ete

s

8

6

4

2

0 Rsq = 0.1729

Alternatives - Mosiac

ON

S W

ard

Cla

ssific

atio

n -

Clu

ste

rs1

5

9

13

17

21

25

29

33

37

41

Smoking Prevalence Prediction %

605040302010

What next

• Extend into other disease/health domains

• Extend into non-health areas

• Extend internationally

• Link to health service inputs

What next

Conclusions

• Healthy England– University environs– Army camps– New developments– Bristol

• Unhealthy England– Tyne-Tees– Lincolnshire resorts– Black Country

Conclusions

• Major utility in health profiling• Clear advantages in

methodological terms and through linkage with other sources