Directly to content
  1. Publishing |
  2. Search |
  3. Browse |
  4. Recent items rss |
  5. Open Access |
  6. Jur. Issues |
  7. DeutschClear Cookie - decide language by browser settings

The existence of standard-biased mortality ratios due to death certificate misclassification - a simulation study based on a true story

Deckert, Andreas

In: BMC Medical Research Methodology, 16 (2016), Nr. 8. pp. 1-9. ISSN 1471-2288

PDF, English
Download (2MB) | Lizenz: Creative Commons LizenzvertragThe existence of standard-biased mortality ratios due to death certificate misclassification - a simulation study based on a true story by Deckert, Andreas underlies the terms of Creative Commons Attribution 3.0 Germany

Citation of documents: Please do not cite the URL that is displayed in your browser location input, instead use the DOI, URN or the persistent URL below, as we can guarantee their long-time accessibility.


Background: Mortality statistics are used to compare health status of populations; optimally, they base on individual death certificates. However, determining cause of death is error-prone. E.g. cardiovascular disease (CVD) death determination is characterized by sensitivity (SE) and specificity (SP) lower than 85%. Furthermore, differential misclassification may be present in case of homogenous target populations. We investigate the bias of standardized mortality ratios (SMR), based on real-world data. Methods: CVD mortality of 6378 ethnic German repatriates was assessed and the SMR calculated. Non-differential age-dependent misclassification was introduced into data by scenarios of equal SE and SP in a range of 0.7 to 0.85. The bias between originally reported and actual SMR was calculated for each pair of values. Additionally, four differential misclassification scenarios were simulated, reflecting two extreme scenarios of both quality criteria varied in the cohort but fixed to either higher or lower in the reference, and two scenarios of crossed criteria values. Results: In case of non-differential misclassification the bias is always towards the null-hypothesis. The lowest bias was 13.5% (SE, SP = 0.85 constantly), the maximum bias was 40% (SP = 0.7). However, in case of differential misclassification the observed SMR can be on the wrong track. If SP is high but SE low in the cohort, negative bias up to −10% can occur. In case SE is low but SP is high in the reference, the bias remains always positive. In the opposite case plus SP is high in the cohort, the bias can reach −30%. Conclusion: SMR values are always biased due to the diagnostic test character of death determination. In majority of epidemiological studies the bias should be towards the null-hypothesis (non-differential misclassification). However, caution is needed in case of differential misclassification, possibly experienced in studies on homogenous subgroups, and in large prospective cohorts with specifically trained personnel.

Item Type: Article
Journal or Publication Title: BMC Medical Research Methodology
Volume: 16
Number: 8
Publisher: BioMed Central; Springer
Place of Publication: London; Berlin; Heidelberg
Date Deposited: 22 Feb 2016 13:59
Date: 2016
ISSN: 1471-2288
Page Range: pp. 1-9
Faculties / Institutes: Medizinische Fakultät Heidelberg > Institut für Public Health (IPH)
Subjects: 610 Medical sciences Medicine
About | FAQ | Contact | Imprint |
OA-LogoDINI certificate 2013Logo der Open-Archives-Initiative