15283-B-Blokker

169 General discussion 9 Along the same lines, pathologists have to be trained in the interpretation of CT and MRI scans, and in the collaboration with radiologists: they have to learn their language. They also have to get confidence in reading the MIA biopsies, avoiding both overinterpretation, and being too cautious. In fact, like the specialized pathologist proposed for CA, 43 MIA will require dedicated, specialized teams of radiologists and pathologists. Diagnostic accuracy Deaths are usually not the result of a single isolated disease or event. 37 Especially among elderly with multiple medical problems, death often occurs due to a coexistence and possible interaction of several diseases, and (side) effects from given therapies. 257 Therefore, in most cases, CA is a complex investigation, and so is the MIA. As a consequence, different observers may interpret the results of both techniques differently, a situation that is not uncommon in clinical practice. In fact, this is the main reason why the cause of death determined by CA and MIA had to be compared to a consensus cause of death. As several studies 56,90,258 and previous experience 57 demonstrated, we could not assume that the diagnostic accuracy of CA in our study would be infallible. This was another reason for using consensus cause of death, and consensus diagnoses, rather than CA-results as gold standard in the comparison of MIA and CA. To determine the consensus cause of death, we set up a ‘reference standard process’ containing three review steps with different professionals independently deciding per case whether the presumed cause of death was concordant, discordant or undecided. If cases were discordant, the cause of death reported by either MIA or CA was defined as the consensus. The first review was based on the 99 MIA and CA reports only. Two parties had to agree that the two methods reported the same cause of death, in order to accept it as the consensus cause of death. The remaining 40 cases were subjected to the second review, which included both the reports and the radiological and macroscopic images and histological slides, and, if necessary, a correction of erroneous interpretation of the CA histology (to compensate for bias due to the research setting of MIA versus the daily routine of CA). The third review required a “reference standard committee” (RSC) of several specialists discussing the ten remaining cases and jointly defining their causes of death. In the end, there was concordance for cause of death between MIA and CA in 91 cases and discordance in eight cases. For the latter, the consensus cause of death was based on the MIA report in five cases, and on the CA report in three cases. Thus, CA reported the correct COD in 95% of cases and MIA in 97%. chapter 5