Deriving Small Area Population Estimates
Deriving Small Area Population Estimates
1 Data Sources
The derivation of these small area population estimates relies mainly on the Registrar General's Mid-Year Estimates (MYE) and data from the Community Health Index (CHI). This section provides a brief outline of the MYE, along with background information about the CHI.
1.1 Registrar General's Mid-Year Estimate of Population
Each year the Registrar General for Scotland produces a population estimate for Scotland and its administrative areas, by age and sex. This estimate is used by a variety of users within central and local government, the health service, academia and the private sector. The estimates help to determine the allocation of resources between central government and councils and health boards; they are of fundamental importance to staff in central and local government when planning, delivering, and monitoring the provision of a wide range of services.
The starting point for the National Records of Scotland (NRS) estimates is the Census of Population. After adjusting for under-enumeration and differences in timing, the census is used to give figures for the base year (currently 1991) which are updated each year until the series is rebased after the following census. Each year the census estimate is rolled forward by adding births, subtracting deaths and adding net migration.
The total number births and deaths, along with details of age, sex and location of the birth or death, is taken from the civil registration system managed by NRS. This data is considered to be virtually complete and accurate.
The estimate of migration is much more uncertain as there is no wholly reliable source from which to estimate migration. Instead migration is estimated from the best available data sources, namely: The International Passenger Survey, the National Health Service Central Register, the Electoral Roll and the School Roll.
Using information from these data sources a 'top-down' estimate of migration for Scotland and its administrative areas is prepared. That is, a net migration total is estimated for Scotland, then net migration for each of the health boards is estimated and constrained to the Scotland total. Finally, an estimate of net migration for local authorities is derived and constrained to the health board totals.
1.2 Community Health Index
The main functions of the CHI is to: hold data which facilitates payment to GPs; enable Medical Record Envelope data to be transferred with the patient as (s)he migrates from place to place; and, to provide a common data repository for other healthcare satellite systems such as Child Health System. The CHI is also a fully integrated part of population screening and immunisation programmes.
The nearest equivalent on the CHI to the resident population definition as used in the NRS MYE is those patients that are 'currently registered' with a GP. The currently registered population can be defined as those patients currently registered with a GP for which the GP receives regular payment.
Table 1 highlights the differences in definitions of sub-populations in the currently registered and the usually resident populations.
CHI - 'Currently Registered' Population
NRS Mid-year estimates
|UK Armed Forces Personnel||Excluded||Included|
|Dependants of UK AF personnel||Included||Included|
|Overseas AF Personnel||Excluded||Included|
|Dependants of Overseas AF Personnel||Excluded||Included|
|Prisoners||Excludes prisoners sentenced to two or more years of a custodial sentence, all others included||Included|
|Psychiatric Patients||Excludes patients receiving psychiatric care for more than two years in an institution||Included|
|Students||Recorded at term-time address||Recorded at term-time address|
|Private Patients||Excluded. Persons who do not seek health care from the NHS but rather from private medical care||Included|
Of these definitional differences, only the UK Armed Forces (estimated to be about 16,500 persons in 1999) sub-group might have a substantial effect. In time, NRS will investigate the need for individual adjustments for the other sub-groups. However, because of the lack of data and the relatively small populations involved, NRS assumes that these will not have a significant effect on the interpretation of the results.
The September 1999 CHI extract provided to NRS contained nearly nine million records, which constitutes all records ever placed on the CHI. To obtain the currently registered population it is necessary to apply a number of criteria or filters to all of the patient records on the CHI (about nine million). The filters used to define the currently registered population were agreed by the working group and defined through consultation with users of the CHI and SEMA, the company managing the development and maintenance of the CHI. Details, of the five filters are given in Annex C.
Geography in the Dataset
Consultation with users indicated that small area population estimates produced for postcode sectors and 1999 wards would satisfy most user requirements. It is intended to provide population estimates for both geographies. However, this dataset only provides population estimates for postcode sectors as NRS has not yet finished digitising the 1999 ward boundaries that are needed to map postcodes to wards. When the boundaries become available, expected summer 2001, the 1999 small area population estimates will be updated to include 1999 ward estimates.
These geographies are determined by matching the patient's current postcode from the CHI against the NRS Postcode Index 99/2[Footnote 1]. A computer file derived from NRS digitised postcode boundaries, which provides an allocation of live and deleted small and large user postcodes within Scotland to higher areas such as electoral and administrative areas and other information including one metre national grid references. This match determines if a postcode is valid and provides area codes (Council areas, Health Board areas, etc.) for each valid postcode.
Postcode sectors are not contiguous with health board or council boundaries. Therefore, postcode sectors have been split to nest within health board and council boundaries. Where postcode sectors are split, a naming convention has been used to identify that it is a split postcode sector and which identifies the Health Board and Council area the split postcode sector lies within. For example taking the fictitious split postcode sector of KY15 7xxYY - 'xx' is a number between 01 and 15 representing the health board; and, 'YY' is a number between 01 and 32 representing the council area. Codelists for the health boards and council areas are provided in Annex D.
Approximately 99 per cent of the postcodes on the CHI were found to be valid. As the remaining one per cent of records were considered to be part of the currently registered population, they had to be assigned postcodes or area markers to enable them to be used in the estimation process. This assignment was done using a donor imputation method, which assumes that patients with similar characteristics will live in similar geographic locations.
For each patient record without a valid postcode, a similar record with a valid postcode is selected at random. The similarity of records was determined by comparing details of GP practice code, age, sex and health board of residence. The record with the valid postcode is the donor and its postcode (and subsequent area codes) is assigned to the record with the invalid postcode.
3 How are the small area estimates calculated?
The small area estimates presented here have been constrained to the NRS MYE as these are generally considered to be the best estimates of the population. It was also felt that small area estimates needed to be consistent with the NRS MYE to avoid any confusion over which population estimates should be used for certain geographies.
To meet the main geography requirement (postcode sectors and 1999 wards), population estimates are needed at the lowest building brick. That is, a population estimate, by postcode, sex and single year of age for each unit postcode. This enables different aggregations of the building bricks to produce different geographies. However, data at the building brick level will not be made available to the public because of confidentiality considerations and the likely inaccuracy of the data to such detail.
Finally, as shown in Table 1 the CHI does not include the AF population and therefore the CHI was used only to derive estimates of the civilian population. Data from the 10 per cent sample of 1991 Census records was used to derive estimates of the non-civilian population, i.e. Armed Forces, to a building brick level. These two components were then added together to produce the small area population estimates presented in this dataset. The resultant population estimates will be consistent by sex and single year of age, with the NRS 1999 mid-year population estimates at the council, health board and national level.
1. A computer file derived from NRS digitised postcode boundaries, which provides an allocation of live and deleted small and large user postcodes within Scotland to higher areas such as electoral and administrative areas and other information including one metre national grid references