Reconstructing long term trends in causes of death in the Netherlands
Statistics Netherlands: Wiet Koren (project leader), Ronald van der Stegen (principal researcher), Peter Harteloh and Jan Kardaun
This project deals with the reconstruction of long term trends in causes of death for the Netherlands (1950-2008). These cause-specific mortality trends are hindered by sudden shifts in the trends due to the regular revisions of the International Classification of Diseases (ICD) and other changes in coding causes of death. A new method is developed to control for the effects of these coding changes, which involves a time series approach.
We shall examine and validate this time series approach by calculating comparability ratios (CRs) for 17 ICD10 chapters with respect to the ICD9 in the Netherlands and comparing these to the CRs in Italy and Canada where bridge coding was applied.
In addition, the Dutch cause of death time series shall be reconstructed, using the time series approach. This involved the following steps:
1)Constructing a concordance table for ICD6 to ICD10 using nosology, medical knowledge and literature.
2)The automatic detection and quantification of statistically significant shifts in cause of death numbers at ICD transition years and in-between transitions by a state space model in STAMP software.
3)Selection of the corrections to be implemented. Considerations were a maximum variance in time series, therefore excluding series below 40 cases per year. Large shifts were only corrected for, if nosologically sound and in line with medical knowledge. Causes of death with historically poor diagnosis and changes in the propensity to report are more likely not corrected.
|Last modified:||19 September 2014 10.04 a.m.|