Standardisation Anthea Springbett
Jan 12, 2016
Standardisation
Anthea Springbett
Topics covered in this session
• Population rates
• Why do we standardise?
• How do we standardise?
• Comparing standardised rates
• Which method is best?
Diet: In 2008, 25% of men and 29% of women in England & Wales
reported meeting the government ‘5 a day’ guidelines of consuming
five or more portions of fruit and vegetables a day.
Exercise cuts risk of developing bowel cancer polyps
People who lead an active lifestyle are up to three times less
likely to develop polyps which can develop into bowel
cancer, according to a study.
Passive smoking ‘raises breast cancer risk’
Exposure to second-hand tobacco smoke as a child or adult appears
to increase a woman's risk of breast cancer, experts say. Their study of
nearly 80,000 women found breast cancer risk was a third higher among
those who had clocked up decades of passive smoking.
Population RatesCannabis use 'raises psychosis risk'
Using cannabis as a teenager or young adult increases the risk of
psychosis, a report suggests.
Coronary Heart DiseaseThe estimated prevalence of coronary heart disease based on admission to hospital is 3.3% of the Scottish population. Prevalence is higher in males (4.2%) than in females (2.5%) and is strongly related to age.An estimated 16% of the Scottish population aged 75+ is living with coronary heart disease (CHD). In some, more deprived, community health partnerships around 25% of men aged 75+ have CHD.
Registrations with the NHS General Dental Service in Scotland
The 6-12 age group had the highest level of population registered
(94.7%) with an NHS GDS dentist.
NHS Ayrshire & Arran had the highest level of population (all ages)
registered with an NHS GDS dentist (77.6%).
Why do we standardise
• Comparison of rates over time or between geographical areas etc.
• Populations differing in structure (age, sex, deprivation …).
• Comparisons of crude rates may not be sensible.
0
6,000,000
12,000,000
18,000,000
24,000,000
30,000,000
2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
Num
ber o
f Pre
scrib
ed It
ems
0
50,000,000
100,000,000
150,000,000
200,000,000
250,000,000
Gross Ingredient Cost (£)
Number of Prescribed Items Gross Ingredient Cost (£)
Cardiovascular prescribing 2001-10
Jargon
• Target population(s) – the population(s) that we are interested in.
• Standard population – the population that we use to construct comparisons with and between target populations.
• Directly & indirectly standardised rates – two different forms of standardisation.
Health warning
• This talk concentrates on age standardisation.
• In real life it is likely that you will need to standardise by age, sex and possibly additional variables (eg SIMD).
How do we standardise?Miami and Alaska
Example (deaths in USA)
Age (yrs)Population
sizeDeaths
Population size
Deaths
<45 328,049 401 89,893 114
45-64 142,670 1,016 14,947 90
65+ 92,168 3,605 2,077 81
Total 562,887 5,022 106,917 285
Miami Alaska
Example (deaths in USA)
Crude death rate for Miami or Alaska =
(deaths for all age groups)
(popns for all age groups)
Age-specific death rate is the crude death rate for a specific age group.
Example (deaths in USA) Miami and Alaska
Miami crude death rate = 5022/562,887
= 8.9 per thousand
Alaska crude death rate = 285/106,917
= 2.7 per thousand
Is Miami really that much worse?
Example (deaths in USA) State population distributions and age-specific rates
Miami Alaska
Age (yrs) Population size
Population (%)
Age-specific
rate*Population
sizePopulation
(%)
Age-specific
rate*
<45 328,049 58 1.2 89,893 84 1.3
45-64 142,670 25 7.1 14,947 14 6.0
65+ 92,168 16 39.1 2,077 2 39.0
Total 562,887 100 8.9 106,917 100 2.7
*Age-specific death rate per thousand population
Example (deaths in USA) State population distributions and age-specific rates
Miami Alaska
Age (yrs) Population size
Population (%)
Age-specific
rate*Population
sizePopulation
(%)
Age-specific
rate*
<45 328,049 58 1.2 89,893 84 1.3
45-64 142,670 25 7.1 14,947 14 6.0
65+ 92,168 16 39.1 2,077 2 39.0
Total 562,887 100 8.9 106,917 100 2.7
*Age-specific death rate per thousand population
Example (deaths in USA)State population distributions and age-specific rates
Miami Alaska
Age (yrs) Population size
Population (%)
Age-specific
rate*Population
sizePopulation
(%)
Age-specific
rate*
<45 328,049 58 1.2 89,893 84 1.3
45-64 142,670 25 7.1 14,947 14 6.0
65+ 92,168 16 39.1 2,077 2 39.0
Total 562,887 100 8.9 106,917 100 2.7
*Age-specific death rate per thousand population
Example (deaths in USA)State population distributions and age-specific rates
Miami Alaska
Age (yrs) Population size
Population (%)
Age-specific
rate*Population
sizePopulation
(%)
Age-specific
rate*
<45 328,049 58 1.2 89,893 84 1.3
45-64 142,670 25 7.1 14,947 14 6.0
65+ 92,168 16 39.1 2,077 2 39.0
Total 562,887 100 8.9 106,917 100 2.7
*Age-specific death rate per thousand population
Directly standardised rate
• Relative sizes of age groups affect crude rate comparison.
• Weighting of age specific rates differs between target populations.
• Use standard age group sizes and apply age specific rates to these.
• Result is a directly standardised rate.
Directly standardised rateStandard population (USA)
USA Miami Alaska
Age (yrs) Population (%)
Population (%)
Population (%)
<45 67 58 84
45-64 22 25 14
65+ 12 16 2
Total 100 100 100
Directly standardised rate
Directly age standardised rate for Miami =
Σ(standard weight x age specific rate)
Σ(standard weights)
where sum (Σ) is over all age groups, and
weight = size of standard population for each age group
Directly standardised rate
Directly age standardised rate for Miami =
Σ(standard weight x age specific rate)
Σ(standard weights)
where sum (Σ) is over all age groups, and
weight = size of standard population for each age group
USA
Miami
Directly standardised rate
Age (yrs) US Population (%)
Age-specific rate (Miami)
Age-specific rate (Alaska)
<45 67 1.2 1.3
45-64 22 7.1 6.0
65+ 12 39.1 39.0
Total 100 6.9 6.7
Directly standardised rates for Miami and Alaska
Miami: (67 x 1.2 + 22 x 7.1 + 12 x 39.1)/100 = 6.9
Directly standardised rate
Age (yrs) US Population (%)
Age-specific rate (Miami)
Age-specific rate (Alaska)
<45 67 1.2 1.3
45-64 22 7.1 6.0
65+ 12 39.1 39.0
Total 100 6.9 6.7
Directly standardised rates for Miami and Alaska
Miami: (67 x 1.2 + 22 x 7.1 + 12 x 39.1)/100 = 6.9
Directly standardised rate
Age (yrs) US Population (%)
Age-specific rate (Miami)
Age-specific rate (Alaska)
<45 67 1.2 1.3
45-64 22 7.1 6.0
65+ 12 39.1 39.0
Total 100 6.9 6.7
Directly standardised rates for Miami and Alaska
Crude rates 8.9
2.7
Direct standardisation
• Direct standardisation applies age specific rates from the target population(s) to the age group sizes in a standard population.
• Answers the question: What would the rate in the standard population be if it had the same age specific rates as the target population?
• Allows comparison between target populations.
Direct standardisationStandard populations
• How do you choose the right standard population?– Relevant to target population(s)
• eg Scottish population for HB comparisons
– Appropriate for comparison being made• eg hospital population for surgery outcome data
Direct standardisation
• Direct standardisation applies age-specific rates from the target population to the age group structure of a standard population.
• What do you do if you cannot get age-specific rates for the target population or if these rates are unstable (eg because of low numbers in some age groups)?
Indirect standardisation
• Indirect standardisation applies age-specific rates from the standard population to the age group structure of the target population.
• Then constructs ratio of observed to expected population rates.
• Answers question: How does the observed rate compare with the expected rate?
Methods of standardisation
Direct Indirect
Target population
Group specific rates
Group population
sizes
Standard population
Group population
sizes
Group specific rates
Indirect standardisation
Indirectly standardised rates are usually presented as ratios (eg Standardised Mortality Ratio):
Σ(target age specific rates x weights)
Σ(standard age specific rates x weights)
where sum (Σ) is over all age groups, and
weight = target population for each age group
Indirect standardisationMiami
Indirectly standardised rates for Miami:
Σ(target age specific rates x weights)
Σ(standard age specific rates x weights)
where sum (Σ) is over all age groups, and
weight = target population for each age group
Miami
USA Miami
Indirect standardisationMiami
Miami Miami US MiamiAge (yrs) Population Observed
DeathsAge-specific
rate*Expected
deaths
<45 328,049 401 1.2 383
45-64 142,670 1,016 7.1 1,019
65+ 92,168 3,605 49.5 4,563Total 562,887 5,022 8.1 5,965
*Age-specific death rate per thousand population
Indirect standardisationMiami
Miami Miami US MiamiAge (yrs) Population Observed
DeathsAge-specific
rate*Expected
deaths
<45 328,049 401 1.2 383
45-64 142,670 1,016 7.1 1,019
65+ 92,168 3,605 49.5 4,563Total 562,887 5,022 8.1 5,965
*Age-specific death rate per thousand population
Expected deaths <45 yrs for Miami = 328,049 x 1.2 / 1,000 = 383
Indirect standardisationMiami
Miami Miami US MiamiAge (yrs) Population Observed
DeathsAge-specific
rate*Expected
deaths
<45 328,049 401 1.2 383
45-64 142,670 1,016 7.1 1,019
65+ 92,168 3,605 49.5 4,563Total 562,887 5,022 8.1 5,965
*Age-specific death rate per thousand population
Expected deaths <45 yrs for Miami = 328,049 x 1.2 / 1,000 = 383
Indirect standardisationMiami
Miami Miami Miami US MiamiAge (yrs) Population Observed
DeathsAge-specific
rate*Age-specific
rate*Expected
deaths
<45 328,049 401 1.2 1.2 383
45-64 142,670 1,016 7.1 7.1 1,019
65+ 92,168 3,605 39.1 49.5 4,563
Total 562,887 5,022 8.9 8.1 5,965
Miami SMR = observed deaths/expected deaths
= 5022/5965
= 0.84
*Age-specific death rate per thousand population
Indirect standardisationMiami
Miami Miami Miami US MiamiAge (yrs) Population Observed
DeathsAge-specific
rate*Age-specific
rate*Expected
deaths
<45 328,049 401 1.2 1.2 383
45-64 142,670 1,016 7.1 7.1 1,019
65+ 92,168 3,605 39.1 49.5 4,563
Total 562,887 5,022 8.9 8.1 5,965
Miami SMR = observed deaths/expected deaths
= 5022/5965
= 0.84
*Age-specific death rate per thousand population
Indirect standardisationMiami
Miami Miami Miami US MiamiAge (yrs) Population Observed
DeathsAge-specific
rate*Age-specific
rate*Expected
deaths
<45 328,049 401 1.2 1.2 383
45-64 142,670 1,016 7.1 7.1 1,019
65+ 92,168 3,605 39.1 49.5 4,563
Total 562,887 5,022 8.9 8.1 5,965
Miami SMR = observed deaths/expected deaths
= 5022/5965
= 0.84
*Age-specific death rate per thousand population
Indirect standardisationMiami
Miami Miami Miami US MiamiAge (yrs) Population Observed
DeathsAge-specific
rate*Age-specific
rate*Expected
deaths
<45 328,049 401 1.2 1.2 383
45-64 142,670 1,016 7.1 7.1 1,019
65+ 92,168 3,605 39.1 49.5 4,563
Total 562,887 5,022 8.9 8.1 5,965
Miami SMR = observed deaths/expected deaths
= 5022/5965
= 0.84
*Age-specific death rate per thousand population
Indirect standardisationMiami
Miami Miami Miami US MiamiAge (yrs) Population Observed
DeathsAge-specific
rate*Age-specific
rate*Expected
deaths
<45 328,049 401 1.2 1.2 383
45-64 142,670 1,016 7.1 7.1 1,019
65+ 92,168 3,605 39.1 49.5 4,563
Total 562,887 5,022 8.9 8.1 5,965
Miami SMR = observed deaths/expected deaths
= 5022/5965
= 0.84
*Age-specific death rate per thousand population
Indirect standardisationAlaska
Alaska Alaska Alaska US AlaskaAge (yrs) Population Observed
DeathsAge-specific
rateAge-specific
rateExpected
deaths
<45 89,893 114 1.3 1.2 105
45-64 14,947 90 6.0 7.1 107
65+ 2,077 81 39.0 49.5 103
Total 106,917 285 2.7 8.1 315
Alaska SMR = observed deaths/expected deaths
= 285/315
= 0.91
*Age-specific death rate per thousand population
Indirect standardisationAlaska
Alaska Alaska Alaska US AlaskaAge (yrs) Population Observed
DeathsAge-specific
rateAge-specific
rateExpected
deaths
<45 89,893 114 1.3 1.2 105
45-64 14,947 90 6.0 7.1 107
65+ 2,077 81 39.0 49.5 103
Total 106,917 285 2.7 8.1 315
Alaska SMR = observed deaths/expected deaths
= 285/315
= 0.91
*Age-specific death rate per thousand population
Comparison of standardised rates
• Direct standardisation: Weighted average of target population age specific
rates. Can compare standardised rates for two target
populations that were calculated using same standard population weights.
• Indirect standardisation: Comparisons can be made only if certain conditions are
met (not usually the case).
Comparison of standardised ratesConfidence limits
• There are several methods for calculating confidence limits for comparison of directly and indirectly standardised rates.
• References supplied on last slide and in folder.
Additional points
• Standardisation for multiple categories Age, sex, SIMD, ….
• Is standardisation the right solution? Group specific rates may be more appropriate
• Indirect vs direct What sort of comparison do you want? What sort of data have you got?
References
1. Eastern Region PHO technical briefing on standardisation (INphoRM 6), which calculates and comments on both directly and indirectly standardised rates and includes the calculation of confidence intervals.
http://www.erpho.org.uk/
2. APHO website has spreadsheets for calculating direct and indirectly standardised rates (Technical Briefing 3).
http://www.apho.org.uk/
3. Standardisation of rates and ratios (hard copy in course materials)
4. ISD guide to standardisation (available today)