15283-B-Blokker

20 Chapter 2 and databases provided by Statistics Netherlands (SN, a.k.a. CBS Statline) and Dutch Hospital Data (DHD) in cooperation with Kiwa Carity’s data services. The latter is a service organisation that aims to improve Dutch healthcare in various ways. We analysed the SN databases for all registered adult deaths in the Netherlands per year, and we collected tables presenting the total number of deaths and the number of deaths due to external causes (S1 Table). Kiwa Carity provided an anonymized set of aggregated data (S1 Table, S2 Table), LQFOXGLQJ DOO FDVHV RI DGXOW SDWLHQWV Ƌ \HDUV GHFHDVHG LQ 'XWFK KRVSLWDOV WKHLU age and gender, the type of hospital they died in (academic or non-academic), and whether or not an autopsy was performed. Compulsory forensic autopsies in the case of suspected unnatural death, as is the policy in the Netherlands, were excluded. Using the program Matlab®, we created files by year, consisting of one line per individual case. To ensure the privacy of individuals, the data were used according to the required protocol for data provided by DHD. Further ethical approval was not required for this part. The information on performed forensic autopsies was collected from the logbooks kept by the forensic pathologists of NFI. For each case only gender and age were extracted and registered in an anonymized file. NFI has granted us ethical approval to use this file to create an overview of these forensic cases. The emphasis of our analyses was on clinical autopsy rates. In the Netherlands, there are no extramural facilities for non-forensic autopsies. Therefore, if a person dies outside a hospital from a supposed natural cause of death, and next-of-kin ask for a post-mortem examination, the autopsy will be performed by clinical pathologists in the nearest hospital. Because this situation rarely occurs, we expect that only few cases of performed autopsies have been missed. Data analysis Excel® and SPSS®were used for data analysis. We calculatedmeans, differences, ratios and percentages. Overall numbers were plotted with the exact autopsy percentages, and, to filter randomnoisewithin the subgroups andmake trends visible, 4-year moving average plots were constructed. Linear regression was performed to show trends over time, logistic regression to analyse the effect of possible explanatory variables (year, sex, age and hospital type), and the chi-square test to analyse differences between academic and non-academic hospitals. To identify multiple trends within the collected