The Scottish Government

« Previous | Contents | Next »

Listen

APPENDIX: TECHNICAL ASPECTS OF THE RESEARCH METHODS

A1. Implications of using two ICD systems for reliability of classification of deaths

During the selected time period two different ICD systems were in operation: ICD9 until 1999 and ICD10 from 2000. It is therefore important to consider the comparability (reliability) of coding between ICD9 and ICD10, in terms of both the gross distinction between suicide and undetermined deaths, and the specific cause within the suicide and undetermined categories. A bridge coding exercise, conducted in 2000, which examined the correspondence between ICD9 and ICD10 coding systems, found only minor discrepancies (Rooney and Smith, 2000) ; none is considered to be sufficiently serious as to invalidate the use of a combined dataset covering the period under review in this study.

A2. Derivation of pseudo Health Boards

There are 10,058 CATTs in Scotland. However, because CATTs do not necessarily fall completely within the boundaries of the 2001 Health Boards, the research team assigned each record to a pseudo Health Board using a point in polygon process within a geographical information system ( GIS) . The creation of pseudo health boards was necessary to ensure that data were reliable throughout the study period. Table A1 shows there was very little difference in the population distribution of the 'official' health boards and the pseudo health boards derived from CATTs (Exeter 2004) .

Table A1 Summary statistics for official health boards in 2001 and pseudo health boards derived from CATTs

Source: Exeter (2004)

A3. The Carstairs index of deprivation

The Carstairs index of deprivation was first developed following the 1981 census to identify the level of socioeconomic deprivation within postcode sectors in Scotland (Carstairs and Morris 1991) . The index comprises four dimensions (unemployment, overcrowding, lack of car ownership, and low social class) . Since 1981 there have been a number of changes to the questions asked in the Census. While these had little impact on the definition of variables used for the construction of the Carstairs index in 1991, there were significant alterations to the definition of the unemployment and low social class dimensions of the Carstairs index in 2001 (table A2) .

We downloaded the necessary census data from CASWEB²¹ for Census Output Areas ²², which were then aggregated to CATTs. In 1991, the Carstairs index ranged from -5.28 in the least deprived CATTs to 16.06 in the most deprived CATTs, while the 2001 index ranged from -5.94 to 17.47.

Table A2 The definitions of the four variables used to construct the Carstairs index of deprivation: 1991 and 2001 censuses

Appendix A4. Suicide rates by social class

Before 2001, Registrar General's Social Class ( SC) was adopted by the government as the measure of socio-economic status in official surveys and censuses. Since the classification was based on occupation, people who were retired, in full time education, in the armed forces or looking after family were not classified. Since 2001, a new classification system, the National Statistics Socio-Economic Classification ( NS- SEC) , has been adopted (see table A3) . This replaces SC in all national surveys and the 2001 census.

The NS- SEC is also an occupationally based classification but has rules to provide coverage of the whole adult population. The information required to create the NS- SEC is occupation coded to the unit groups ( OUG) of the standard occupational classification 2000 ( SOC2000) and details of employment status (whether an employer, self-employed or employee; whether a supervisor; and number of employees at the workplace) .

The version of the classification which is used for most analyses (the analytic version) has eight classes, the first of which can be subdivided. The most detailed version has 17 types with further subdivisions, called operational categories (see below) .

Table A3 Analytic classes and operational categories and sub-categories of NS- SEC

< http://www.statistics.gov.uk/methods_quality/ns_sec/analytic_operation_cat_subcat.asp#opcat >

The implementation of the new socio-economic classification results in some problems in analysis by social class in surveys/censuses relating to the year 2001. Fortunately the detailed operational categories of NS- SEC can be aggregated approximately to SC. According to the ONS, the approximation achieves a continuity level of 87 percent for SC.

For the first period (1989-1995) , counts of people by social class from the 1991 census were extracted. For the second period (1996-2002) , counts of people by NS- SEC from the 2001 census were extracted and converted into counts of people by SC using the lookup table provided by ONS.

Caution is required when comparing results using the 1991 and 2001 census data as population denominators. The table in the 1991 census was based on a 10 percent sample while the table in the 2001 census was based on a full 100 percent sample. In addition, the 1991 census table provides counts of persons aged 16 and over by social class. In contrast the 2001 census table provides counts of persons aged 16 to 74 years only by social class. For the second period analysis, persons over 74 who died by suicide were removed from the analysis (n=328) . The extent to which these differences in the calculation of denominators affect the reliability of inter-censal comparisons is unknown.

« Previous | Contents | Next »