T Blakely1, A Woodward, C Salmond. 1. Department of Public Health, Wellington School of Medicine, University of Otago, New Zealand. tblakely@wnmeds.ac.nz
Abstract
BACKGROUND: The New Zealand Census-Mortality Study (NZCMS) aims to investigate socio-economic mortality gradients in New Zealand, by anonymously linking Census and mortality records. OBJECTIVES: To describe the record linkage method, and to estimate the magnitude of bias in that linkage by demographic and socio-economic factors. METHODS: Anonymous 1991 Census records, and mortality records for decedents aged 0-74 years on Census night and dying in the three-year period 1991-94, were probabilistically linked using Automatch. Bias in the record linkage was determined by comparing the demographic and socio-economic profile of linked mortality records to unlinked mortality records. RESULTS: 31,635 of 41,310 (76.6%) mortality records were linked to one of 3,373,896 Census records. The percentage of mortality records linked to a Census record was lowest for 20-24 year old decedents (49.0%) and highest for 65-69 year old decedents (81.0%). By ethnic group, 63.4%, 57.7%, and 78.6% of Maori, Pacific, and decedents of other ethnic groups, respectively, were linked. Controlling for demographic factors, decedents from the most deprived decile of small areas were 8% less likely to be linked than decedents from the least deprived decile, and male decedents from the lowest occupational class were 6% less likely to be linked than decedents from the highest occupational class. CONCLUSION: The proportion and accuracy of mortality records linked was satisfactorily high. Future estimates of the relative risk of mortality by socio-economic status will be modestly under-estimated by 5-10%.
BACKGROUND: The New Zealand Census-Mortality Study (NZCMS) aims to investigate socio-economic mortality gradients in New Zealand, by anonymously linking Census and mortality records. OBJECTIVES: To describe the record linkage method, and to estimate the magnitude of bias in that linkage by demographic and socio-economic factors. METHODS: Anonymous 1991 Census records, and mortality records for decedents aged 0-74 years on Census night and dying in the three-year period 1991-94, were probabilistically linked using Automatch. Bias in the record linkage was determined by comparing the demographic and socio-economic profile of linked mortality records to unlinked mortality records. RESULTS: 31,635 of 41,310 (76.6%) mortality records were linked to one of 3,373,896 Census records. The percentage of mortality records linked to a Census record was lowest for 20-24 year old decedents (49.0%) and highest for 65-69 year old decedents (81.0%). By ethnic group, 63.4%, 57.7%, and 78.6% of Maori, Pacific, and decedents of other ethnic groups, respectively, were linked. Controlling for demographic factors, decedents from the most deprived decile of small areas were 8% less likely to be linked than decedents from the least deprived decile, and male decedents from the lowest occupational class were 6% less likely to be linked than decedents from the highest occupational class. CONCLUSION: The proportion and accuracy of mortality records linked was satisfactorily high. Future estimates of the relative risk of mortality by socio-economic status will be modestly under-estimated by 5-10%.
Authors: Hye-Chung Kum; Ashok Krishnamurthy; Ashwin Machanavajjhala; Michael K Reiter; Stanley Ahalt Journal: J Am Med Inform Assoc Date: 2013-11-07 Impact factor: 4.497
Authors: Megan A Bohensky; Damien Jolley; Vijaya Sundararajan; Sue Evans; David V Pilcher; Ian Scott; Caroline A Brand Journal: BMC Health Serv Res Date: 2010-12-22 Impact factor: 2.655