Related Experiment Videos
[The quality of mortality statistics. Galicia 1987]
Insights
This study assessed the accuracy of mortality statistics in Galicia by comparing autopsy reports to death certificates. Results indicate an acceptable level of reliability, though improvements in cause-of-death selection and definition are recommended.
Area of Science:
- Public Health
- Epidemiology
- Medical Statistics
Context:
- Mortality statistics are crucial for public health surveillance and resource allocation.
- Accurate cause-of-death data is essential for understanding disease burden and trends.
- Previous studies have highlighted potential discrepancies in death certification accuracy.
Purpose:
- To evaluate the reliability of mortality statistics in Galicia using autopsy reports as a gold standard.
- To compare the basic cause of death from death certificates with causes determined from autopsy reports.
- To identify specific error types and their frequencies in mortality data.
Summary:
- Ninety autopsy reports were independently reviewed by physicians to determine the basic cause of death.
- Agreement levels were assessed using ICD-9 coding: 41.1% at 4-digit, 45.6% at 3-digit, and 63.3% at the ICD-9 group level.
- Identified errors included incomplete definitions (13.3%), logic sequence errors (10.0%), incorrect cause selection (15.6%), and unrelated causes (20.0%).
- Predictive Positive Value, Sensitivity, and Specificity were calculated for cardiovascular diseases and tumors, showing acceptable performance.
Impact:
- Provides valuable insights into the quality of mortality statistics in Galicia.
- Highlights areas for improvement in death certification practices and medical training.
- Contributes to the broader understanding of challenges in international mortality data accuracy.
Abstract:
We intend to study the reliability of the mortality statistics in Galicia (1987), comparing the basic cause of death, with that we obtain by the Autopsy Report (AR). After independently studying 90 AR, three physicians selected the basic cause of death, subsequently coded by the same team. All this persons were blind in relation with the previous certification. 41.1% of the causes of death agreed in 4 digit, 45.6% agreed in 3 digit and the 63.3% remained in the same ICD-9 group. Taking into account all the mistakes found, the 13.3% took place due to an incomplete definition, but the classification remained correct. The 10.0% due to an error in the logic sequence, the 15.6% by an error in the selection of the cause of death and, in the 20.0% there was no relationship between causes. For cardiovascular diseases, the Predictive Positive Value (PPV) was 76.9%, the Sensibility (S) was 83.0% and the Specificity (Sp) 90.9%. For the tumors, the value of the PPV and the S was 73.7% and the Sp was 92.9%. In spite of selection bias because of the method used, we think that the information produced is useful for evaluating the quality of mortality statistics in Galicia. When comparing our results with similar studies, we find an acceptable level.