|
Public Health
Network |
 |
|
|
This page gives a basic guide to interpreting and using the most common types of public health analysis. This section also contains links to more detailed information contained in the Analytical Methods section. More comprehensive guides on interpreting data can be found on Epidemiology
for the Uninitiated - BMJ Publishing Group 1997 This section covers the following topics: The most common methods for assessing the amount of disease in a population are look at morbidity (illness) or mortality (death). There are 2 common methods used to measure morbidity; incidence and prevalence. Incidence
The incidence of a disease is the rate at which new cases occur in a population at risk during a specific period. If the population is stable, the formula used is: Incidence = Number of cases / Population at risk x time during which cases were ascertained If the population changes during the period where new cases are measured, the incidence is calculated using "total person years at risk". This is the sum of the new cases during the period which each person in the population was at risk during the measurement period. The formula is therefore Incidence = Number of new cases / Total person years at risk Example: If there were 17 cases
of flu in a school of 1000 in a 2 years, this is an incidence of 8.5/1000 per year. The proportion of a population who have a disease at a point in time is the prevalence of disease. It is often expressed as a percentage. The 'point in time' can be a single examination [point prevalence], but is often a longer time scale in order to give a better estimate of the numbers with the disease [period prevalence]. Prevalence = (cases/population) * 100 Prevalence is used for diseases that are chronic conditions, such as
asthma, diabetes, cystic fibrosis. Example: In a nursing home with 100 residents, 10 people had diabetes in a 12 month period. The prevalence of diabetes in the nursing home was therefore 10%. Rates
It is often to look at the incidence and prevalence of diseases in terms of rates. This enables comparisons between different populations. Crude
Rate A crude rate is a rate that applies to the population as a whole, and that hasn’t been adjusted to account for differences in population structures such as age and sex. It is calculated using this formula: Crude Rate = Number of Events / Total Number of People in a Population Example: If a town has a
population of 1000, and 25 people die from cancer in one year, the mortality
rate is 25/1000/year. Pros:
Con
Standardised
Rates
In order to overcome the
problems of a crude rate masking differences in particular age groups, rates can
be standardised. There are two types of standardization; direct and indirect. Direct
Standardisation
Directly standardized
rates give an indication of the number of events that would occur in a standard
population, if the population had the same age-specific rates of the local area.
The standard population that is most commonly used is the European Standard
population, however other populations such as the Avon standard population can
also be used. The rates are
calculated per 100,000 and because rates are applied to the same population,
rates across areas can be compared. Example: Between
1997-1999, Bristol Unitary Authority has a directly standardized mortality rate
for accidents of 16.7/100,000. This means that if the European Standard
population experienced the same age-specific patterns in mortality, the death
rate would be 16.7 per 100,000
Cons:
More information: See Analytical Methods section for a worked example. Indirect
Standardisation
The SMR of the reference
population is always 100, a value of lower than 100 means that fewer deaths than
expected occurred in the local population after adjusting for differences in age
and sex; more than 100 means that there have been more deaths than expected. Example: In 1998,
Bristol’s SMR from all causes, which was standardised using the England and
Wales average, was 89. This means that there were 89 deaths for
every 100 that occurred in the country as a whole. Pros:
Cons:
More information: See Analytical Methods section for a worked example. Confidence
Intervals
Confidence intervals or limits are used to give a range of values within which there is a degree of
certainly that the values are correct, and to assess if values are significantly
different from that of the reference population. This range is required as there is likely to be
some variation that occurs by chance. For 95% confidence intervals we can be 95% certain that the real value (e.g. SMR) will fall somewhere between the values of the two confidence limits 95 times out of 100. In SMRs if the confidence
intervals are both one side of 100 then the result can be termed 'statistically
significant', high if both values are above, and low if both values are below
100. A small numbers of events or a small sample population will tend to produce wide confidence intervals, i.e. a large difference between the upper and the lower confidence limit. A large number of events or a large sample population will tend to produce narrow confidence limits. Confidence intervals should be used for direct and indirectly standardised rates, and proportions. Standardised rates: Upper Confidence interval = rate + 1.96 * ((rate / sqrt. (number of events)) Lower Confidence interval = rate - 1.96 * ((rate / sqrt. (number of events)) Proportions: Upper Confidence interval = proportion + 1.96 * sqrt ((Pu * (1-Pu)) / sample population) Lower Confidence interval = proportion - 1.96 * sqrt ((Pu * (1-Pu)) / sample population) Where Pu is the proportion of cases. Example: In 1997-1999, North
Somerset had a directly standardized rate for death from accidents of
12.3/100,000. The upper confidence limit is 14.8 and the lower confidence limit
is 9.8, meaning that there is a 95% certainty that the true mortality rate for North
Somerset lies between 9.8 and 14.8/100,000. The England and Wales rate for the same period is 16.7/100,000 with an upper confidence limit of 16.9/100,000 and a lower confidence limit of 16.5/100,000. As the values within the 95% confidence limit of North Somerset UA (9.8-14.8) do not overlap with the England and Wales rate (16.5-16.9), we can say with 95% certainty that North Somerset UA has a significantly lower mortality rate for accidents than England and Wales. Caution: Confidence Intervals for age-standardised rates are unreliable where there are less than 50 cases in an area. This is because the formula commonly used is a simplified version and assumes normal distribution.
Fertility rates can be calculated in 2 ways; either taking a 'snap-shot' of current levels of fertility (period) across women of all ages at a particular point in time or by tracking a group or cohort of women as they pass through their reproductive period (usually taken as 15-44). In practice, most fertility rates are calculated using the 'snap-shot' method because they use data which are currently available. However cohort methods are useful in historical analyses examining issues such as completed family size. The following definitions are used to describe fertility. General
Fertility Rate: Total Fertility Rate: Sum of age specific fertility rates Gross Reproduction Rate: Average number of daughters produced by a woman during her reproductive life time (sum of age specific fertility rates for female births). Net Reproduction Rates: Average number of daughters, allowing for mortality, who survive to the age of the mother during her reproductive life time (sum of age specific fertility rates for female births of mothers minus age specific mortality rates of daughters). UpBirth
& Infant Death Rates
Crude Birth Rate: Infant
Mortality Rate: Still Birth
Rate: Perinatal
Mortality: Neonatal Mortality: Postneonatal Mortality: Number of infant deaths ages 28 days to 1 year / live births
There are three common methods used to estimate deprivation; the Jarman Score, the Townsend Score, and the DETR indicators. Most measures are "ecological", in other words they try to measure geographical populations, not individuals or social groups. Jarman
Score
The Jarman
score was developed as a measure of General Practice workload in the
mid-eighties that is sometimes used as a proxy for deprivation. It has been used by the Department of health to determine
additional “deprivation” payments to GPs.
The scores were re-calculated for the 1991 census, using the same census
variables as 1981. The calculation of the Jarman score consists of the three stages, data identification, weighting and aggregation. Eight census variables are used in the calculation. Each has a weight attached to it.
The mean for England and Wales is 0. An area with a high score has a greater demand for primary care based on the characteristics of the resident population, than an area with a low score. Extreme scores are those above 32 (rounded to 30 by the Department of Health). Example: South Gloucestershire has a Jarman score of -21.38 meaning that it has a lower demand for primary care than England and Wales mean, but Bristol UA has a score of 17.85, higher than the England and Wales mean. Pros:
Cons:
More Information: See Analytical Methods section for a worked example. DETR 2000
The Department of Transport, Local Government and the
Regions have produced a several ward and local authority level deprivation
scores looking at different aspects of deprivation. The 6 domains they have used
to look at deprivation are:
Each of the 6 domains uses a variety of data, mainly taken
from social services, which are then weighted according to importance and
combined to create a score. The score is then used to generate ranks the wards.
The wards are ranked so that 1 is the most deprived, and 8414 is the
least deprived in England. It is the rank, and not the score that should be
used. Example: Lawrence Hill ward in Bristol City Council is
ranked 133 using the overall score, which means it is the 133rd most
deprived ward in England out of a total of 8414 wards. However, it ranks
differently when looking at the 6 domains; 108th most deprived for
income, 294th most deprived in terms of health.
Townsend Material Deprivation ScoreThe Townsend score is made up by looking at four variables census; unemployment - % of
economically active residents aged 16-59/64 who are unemployed; The data are taken from
the 1991 census. The variables
combine to form an overall score ranking a particular area relative to others.
The higher the score, the more deprived the area. Example: Based on Avon
standardisation, the most affluent area
in Avon has a score of –9.0, the most deprived has a score of 9. Pros:
Cons:
More Information: See Analytical Methods section for a worked example. Deprivation
Quintiles
Deprivation quintiles
divide areas in fifths according to some measure of deprivation, and can be used
to analyse variations in health between deprived and affluent sections of the
population regardless of where they live. The can be of varying size, e.g.
Health Authority or enumeration districts. The national targets use
quintiles for life expectancy at Health Authority level, and the teenage
conceptions national target uses quintiles at ward level. Population Sources
There are various sources
of population data. The main sources that are used by Avon Health Authority are
the GP register, ONS estimates, local estimates and census data. GP
register Pros:
Cons
Note: Local Population Estimates are available that attempt to adjust for this. ONS
estimates Pros:
Cons:
More information: National Statistics ONS projections look at trends in fertility, mortality and population trends to produce 25 year projections. ONS produces 3 different projections that reflect differing assumptions about the underlying trends. Pros:
Cons:
More information: National Statistics
Local
Estimates Pros:
Cons:
Census
Data Pros:
Cons:
More information about Census 2001
|
|