The total incidence of QM events in the radiological workup process of intensive care patients decreased over time with statistical significance after the implementation of a new quality management initiative consisting of regular multidisciplinary conferences with a direct feedback mechanism between radiologists and ICU physicians. The majority of imaging-related QM events in our analysis affected the radiological report. The most frequent consequence was a revision or retrospective adjustment of the respective report. A standardised protocol for providing written feedback was frequently made use of.

Structured bilateral feedback is associated with a decline of imaging-related adverse events for patients. The implementation of an effective mechanism for timely identification of errors may contain their detrimental effects.

Dedicated investigation of QM events confirms the hypothesis that unsafe acts appear in recurrent patterns rather than being individual mistakes [22]. Hence, a systematic approach to combat the risk of error on an organisational level as presented here is crucial to ensure patient safety. The introduction of reporting systems was highly recommended by the Committee on Quality of Health Care in America [1]. In a more recent report, the committee further encouraged a greater focus on patient-centredness and effectiveness, both of which were addressed by this QM initiative [23]. The integration of both, clinical, and radiological aspects within a multidisciplinary team allows for a more holistic patient view and therefore a more personalised patient management as special attention is paid to each patient’s current condition rather than simply their diagnoses. The error rate we found is lower than reported by other investigators [5, 6, 24, 25]. One explanation might be that imaging examinations are routinely interpreted by two readers in our department. This standardised double reading was not mentioned in the studies above. Most published diagnostic error studies retrospectively identified adverse events over a certain time period. Conversely, our approach was to integrate both the detection of errors and the implementation of a constant feedback loop into clinical routine. An effect in terms of reducing QM event rates appears to have occurred early after implementation. Our results suggest that, regarding imaging-related errors, the radiological report is especially prone to error. This is in accordance with published studies, which focus mainly on the radiological report as a source of adverse events [10,11,12,13,14,15]. Tracking radiologists’ eye movements during the reading of verified chest imaging studies showed that they spent longer time dwelling on overlooked findings [26]. If a finding is initially detected, but wrongly interpreted as insignificant, retrospective multidisciplinary discussion possibly allows for interpretation improvements as agreement within the teams is reached. Our study population included more male than female patients (60% versus 40%), which reflects the usual sex distribution of ICU patients in many countries [27,28,29,30]. The high number of CTs in comparison to X-rays presented in MDCs may be explained by their higher complexity both in clinical questions to be answered and their interpretation. Hence, there may be more need for multidisciplinary consultation. Additionally, ICU physicians may request a higher-resolution imaging modality in case of inconclusive X-ray findings before registering the patient for an MDC. We have previously investigated the use of CT for the identification of septic foci and in doing so, have found that unclear infectious sources are associated with low reader confidence [31]. This may explain why most registered examinations focussed on infectious source identification, often concerning septic patients. Systematic error identification may therefore be especially important in septic patients.

While error eradication seems impossible, error prevention and reduction are still worth pursuing. Another aim should be to identify imaging-related QM events before they cause harm. Correcting errors by early reporting may impede adverse events in patients. Establishing a regular structured feedback mechanism should benefit all participants and may thus reduce errors as well as improve the overall quality of the diagnostic workup process. Furthermore, an established feedback culture creates an environment encouraging everyone, especially junior physicians, to speak up more. With declining QM event rates, the structured collection of written feedback during MDCs might still prove important for maintaining awareness of quality control in clinical routine.

The retrospective design of this study results in several limitations. For example, comparison between pre- and post-interventional periods was not possible. The documentation of QM events, the dependent variable, was part of the intervention itself and had thus not been performed previously. To some extent, the observed decline in QM events may be due to learning effects. Importantly, both variable detail of the description in the MDC protocols and handwritten notes rendered the retrieval of some information challenging. Moreover, analysis of errors and related consequences based on medical records alone is limited, especially in ICU patients with complex conditions and is further limited by German data protection law. Our analysis therefore primarily focussed on measurable imaging-related errors. Besides, the time period analysed does not account for the emergence of the COVID-19 pandemic. Its possible impact is being investigated in an ongoing analysis and not within the scope of this manuscript.