Norman P. Direct effects of selective migration on health rates: an analysis based on the ONS Longitudinal Study 1971-1991. In: Estimating Small Area Populations for Use in Medical Studies: Accounting for Migration. Thesis submitted to the University of Leeds on 27 September 2002 in accordance with the requirements of the degree of Doctor of Philosophy in the Faculty of Geography. 2002.

Medical studies require numerators of mortality or morbidity events and denominator populations at risk for the geographical areas of interest. Population migration affects standardised mortality ratios (SMRs), the ratio between the ‘observed’ events in an area to the number ‘expected’, by creating some uncertainty in the denominator and because the migration process itself is directly contributing to the counts of health events and the age-sex structure of the population at risk.
Denominator uncertainties. Outside census years there is a need to estimate small area populations at risk. Whilst data on births and deaths are annually available, a lack of small area migration data may lead to incorrectly estimated age-sex structures. The expected events in an SMR are based on the population at risk but there will be uncertainty about the quality of the denominator used. Various estimation methods are used in the UK, but most do not make explicit allowances for migration. There is a need for transparency about the effect that the estimation method used has on resulting SMRs.
To address denominator uncertainties this research has provided a time-series (1990-98) of small area populations using various estimation methods so that differences can be assessed when outputs are used as denominators in SMRs. The method chosen and inclusion of specific migration elements are shown to alter estimated population age-sex structures. This affects the resulting SMRs. Various data preparation and method enhancements extend previous estimates research. These include a method to establish an annual time-series of data based on the same geography when boundary definitions change and a method of allowing for student, armed forces and communal establishment resident sub-groups. A rigorous measure of denominator uncertainty can only be made when estimates are compared with output from the 2001 Census.
Direct effects of migration. Migration is a selective process influenced by factors including an individual’s age and health status. In SMRs this selectivity potentially affects both the numerator event counts if an area receives or loses (un)healthy migrants, as well as the denominator if migration causes change in the population age-sex structure. Since the deprivation of the origin and destination locations may affect migration decisions and is used as an explanatory variable of health in epidemiological models, there is a need for improved understanding of inter-relationships between health, migration and deprivation.
Improved clarity about direct effects of migration on the reported health of an area and relationships with deprivation has been provided by this research. 1991 Census Sample of Anonymised Records (SARs) analyses show that younger healthier persons tend to migrate over longer distances to less deprived locations but younger short-distance migrants in public housing tend to be unhealthy. Older persons, both migrants and non-migrants, are less healthy than all other persons but long-distance elderly migrants tend to move into less deprived locations. The ONS Longitudinal Study (LS) reveals that between 1971 and 1991 an accumulation of healthy people migrating to less deprived locations occurs with less healthy persons either migrating into deprived locations or being immobile in locations which become more deprived. These processes will maintain health inequalities apparent between deprived and non-deprived locations but the difference is ameliorated by the migration of unhealthy, very elderly persons into less deprived areas.
In the SARs, information about origins is currently insufficient, so migration research is somewhat restricted, but additional health-related topics can be investigated when the 2001 Census SARs are released. The LS extract obtained for 1971-91 has great potential for further research. When links are made to the 2001 Census, since the health status of LS members in 1991 will be known, effects of health-selective migration on the (re-)distribution of ill health across deprivation quintiles and between regions can be determined.