How will the NHS work in 2005?
Download
Report
Transcript How will the NHS work in 2005?
Epidemiologic Prevalence
Modelling Project
Michael Soljak
Informing Healthy Choices ImplementationTeam
Julian Flowers
Eastern Region PHO
Why Model Prevalences?
Epidemiologic disease and risk factor prevalence models can be
used for:
assessing the completeness of disease registers in primary
care or assessing the completeness of case finding
comparing outcomes such as complication rates or admission
rates after adjustment for variation in expected prevalence
comparing service provision with population need
Planning and commissioning health & social care services,
including projecting future levels of demand
undertaking health equity audits
Existing examples: PBS diabetes model
http://www.yhpho.org.uk/PBS_diabetes.aspx#1
•a spreadsheet model that generates expected total numbers of persons
with Type 1 and Type 2 diabetes mellitus (diagnosed plus undiagnosed
combined) in 2001 for England, GO Regions, SHAs, LAs, PCTs, electoral
wards and user-defined populations including GP practices
•applies age/sex/ethnic group-specific estimates of diabetes prevalence
rates, derived from epidemiological population studies, to 2001 Census
resident populations
•forecasts of 2010 diabetes prevalence are also presented for subnational areas based on projected population change and trends in
obesity:
•Scenario 1 population change only, holding 2001 BMI pattern
constant
•Scenario 2 population change and predicted BMI in 2010 if trends
in obesity prevalence continue
•Scenario 3 Population change and a return to 1995 BMI patterns
Existing examples: PBS diabetes model
http://www.yhpho.org.uk/PBS_diabetes.aspx#1
See also: forecasting the burden of diabetes on secondary care
http://www.nwph.net/nwpho/Publications/AintreeDiabetes.pdf
Existing examples: PBS diabetes model
http://www.yhpho.org.uk/PBS_diabetes.aspx#1
HPA HIV Prevalence Model
http://www.hpa.org.uk/infections/topics_az/hiv_and_sti/publicatio
ns/publications.htm
Direct method= categorising the population into a set of
mutually exclusive risk groups of known size, and applying
estimates of risk group-specific HIV prevalence to each
group
Model was based on Bayesian multi-parameter evidence
synthesis (MPES) of surveillance data
Uses WinBUGS package for Bayesian Markov chain Monte
Carlo to incoporate multiple evidence sources
estimates the risk group size, group-specific HIV prevalence
and proportion diagnosed in 13 risk groups, in each of 3
regions (Inner London, Outer London, Rest of England and
Wales)
UK estimates of prevalent HIV infections – adults aged 15-59, 2005
Multi-parameter Evidence Synthesis method - Goubar A et al. 2005, SOPHID, Health
Protection Scotland, Natsal 2000, Unlinked Anonymous Programme, National Study
of HIV in Pregnancy & Childhood, ICH.
COPD Prevalence Model: PCT example
http://tools.erpho.org.uk/copdsel.aspx
•Based on data from 2001 Health Survey for England
• Logistic regression analysis used to choose risk factors for
inclusion based on the strength of association between selected
risk factors and COPD
•Risk factors are age, sex, smoking and ethnicity, degrees of
urbanisation and deprivation
•Validated against a direct model obtained from epidemiologic
studies
•7-fold variation in the prevalence across subgroups of the
population, with lowest values in Asian women from wealthy rural
areas (1.7 %), and highest in black men from deprived urban areas
(12.5 %)
COPD Prevalence Model: PCT example
http://tools.erpho.org.uk/copdsel.aspx
Smoking prevalence - All males
Smoking prevalence - All females
Gender
Prevalence
15-24
F
1.4606
1.5503
25-34
F
1.4352
M
2.7933
35-44
F
1.9148
45-54
M
3.5525
45-54
F
3.5671
55-64
M
10.0185
55-64
F
7.1250
65-74
M
14.3512
65-74
F
8.2524
75+
M
16.5440
75+
F
9.4431
Age range
Gender
Prevalence
15-24
M
1.3870
25-34
M
35-44
Smoking prevalence - White
males
Age range
Age range
Smoking prevalence - White
females
Gender
Prevalence
Age range
Gender
Prevalence
15-24
M
1.3671
15-24
F
1.3789
25-34
M
1.5280
25-34
F
1.3548
35-44
M
2.7539
35-44
F
1.8091
45-54
M
3.5030
45-54
F
3.3789
55-64
M
9.8915
55-64
F
6.7825
65-74
M
14.1790
65-74
F
7.8642
75+
M
16.3510
75+
F
9.0048
APHO CHD & hypertension models
http://www.apho.org.uk/apho/models.aspx
•Hypertension: April 2006 PCT registered populations multiplied by
hypertension prevalence rates identified in the 2003 and 2004 HSfE,
modified by ethnic-group age-standardised risk ratios from the 2004
HSfE
•CHD: stage 1- prevalence of doctor diagnosed CHD in each age/sex
stratum based on national data from the HSfE; stage 2- assumes that
areas with higher CHD mortality rates have comparably higher
prevalence with a linear relationship CHD SMR = (2.604 × UV67) +
25.97. Using UV67 scores calculated for each PCT, get a multiplying
factor for each PCT
•Both models need validation and further development e.g. effect of
diabetes prevalence on CHD prevalence
•NB NatCen obesity prevalence modelling to 2010, by GOR
O:E CHD Prevalence, English PCTs
Patient-reported doctor-diagnosed CHD: England
Observed relative to
expected (%)
20
10
0
-10
-20
-30
-40
-50
0
10000
30000
20000
Expected No. of Patients
40000
Benchmarking tool developed by Paul Fryers, Doncaster PCT
50000
Healthcare for London: impact of prevalence on utilisation
Changes in prevalence rates for selected long term conditions have a small
impact on related hospital activity*
Units: Spells and regular attendances
Diabetes
137,918
2005/06
activity
Hypertension
13,470
6,821
158,209
Impact of
Impact of
2016/17
demographics prevalence
activity
rate changes
COPD
2005/06
activity
1,110
-27
14,348
Impact of
Impact of
2016/17
demographics prevalence
activity
rate changes
CHD
31,092
2005/06
activity
Source:
13,265
2,655
-5
33,742
Impact of
Impact of
2016/17
demographics prevalence
activity
rate changes
65,740
2005/06
activity
* Includes the impact on directly related and indirectly related i npatient HRG (see Appendix)
NHS London
6,337
-195
71,882
Impact of
Impact of
2016/17
demographics prevalence
activity
rate changes
Projecting Older People Population Information System
(POPPI) http://www.poppi.org.uk/
•will support Joint Strategic Needs Assessment
•includes LA population projections to 2025
•includes data on prevalence of depression, dementia, heart attack,
stroke, bronchitis\emphysema, falls, continence, visual impairment,
mobility, obesity
Forecasts are provided for numbers:
Helped to live at home
Intensive home care
Community based services
Supported residents in care homes
Admissions to permanent residential and nursing care
Carers receiving services
2007-8 Work Programme
Update/improve/validate existing models and methodology
Consider development of case-finding strategies
Develop new models for a cancer site (with ACR), chronic
kidney disease, and (serious?) mental illness (with
NIMHE/CSIP)
Further develop intervention modelling- ASSET stroke
model, health inequalities intervention model, overall
mortality?
With stakeholders e.g. academia, encourage a consensus
view about methodology and future requirements