15283-B-Blokker

42 Chapter 3 request an autopsy. Therefore, the results may not directly apply to other hospitals with different autopsy policies. Nevertheless, the results are meaningful and in our view more generally applicable. Precisely because the study was carried out under conditions without financial and technical restraints it could trace considerations related to substance that may explain the present low autopsy rates. It is likely that where conditions and policies are less favourable towards autopsies, clinicians by similar considerations will feel even more justified to not pursue an autopsy. In this survey the clinicians reported on the consent process in the final conversation they had with next-of-kin. The risk that this self-reporting method might introduce a bias towards desired answers was accepted, because it was for practical and ethical reasons, addressed under “Materials and methods” in the paragraph “Study population and study design”, the only way to get the sought-after information. Assuming that the clinicians had always reported decisions in the consent process truthfully, we may conclude that they requested consent for autopsy in most cases (82.6%). In contrast, Burton and colleagues 82 found that consent for autopsy was requested in only 6.2% of eligible cases. In their study design, they did not investigate why clinicians did or did not request consent, because it might have introduced a bias. We believe that our results may indeed have been positively biased by our more extensive questioning and meticulous follow-up of the questionnaires, and also by the autopsy policy at our institute. Among the patient characteristics, the Chi Square test did not show significant differences between religions, probably due to the high number of unknowns (63.4%). Probably the clinicians were reluctant to ask about religion, although the questionnaire included this item. Religion was more often reported in the EPR of patients who had suffered a long illness, than of those who had died suddenly and/or unexpectedly (respectively in 59% and 41%). We were only able to evaluate univariable associations. Ideally, possible associations between variables and outcomes are evaluated with multivariable regression analyses, but to achieve a reasonable power for these analyses many more cases would have been necessary. Theoretical explanations and comparison with the literature The overall autopsy rates on surgical and neurological wards were under 10%, and those of the ICUs and the emergency room above 20%. In Sheffield, UK, autopsy rates were reported to be below 10% for many specialties, including neurology and neurosurgery, but 11.6% for general surgery. 82 In Belfast, UK, the worst decline in autopsy rates was