Did medical doctors who order abdominal CT scans during on-call hours truly become worse at clinical reasoning? Yes, they did

Objective To investigate temporal changes in clinical reasoning quality of physicians who requested abdominal CT scans at a tertiary care center during on-call hours within a 15-year period. Methods This retrospective study included 531 patients who underwent abdominal CT at a tertiary care center during on-call hours on 36 randomly sampled unique calendar days in each of the years between 2005 and 2019. Clinical reasoning quality was expressed as a percentage (0–100%), taking into account the degree by which the differential diagnoses on the CT request form matched the CT diagnosis. Temporal changes in the quality of clinical reasoning and number of CT scans were assessed using Mann-Kendall tests. Associations between the quality of clinical reasoning with patient age and gender, requesting department, and time of CT scanning were determined with linear regression analyses. Results The median annual clinical reasoning score was 14.7% (interquartile range: 12.2 to 16.0%; range: 7.7 to 34.6%). The quality of clinical reasoning significantly decreased between 2005 and 2019 (Mann-Kendall Tau of −0.390, p = 0.048), while the number of abdominal CT scans significantly increased (Mann-Kendall tau of 0.790, p < 0.001). Conclusion The clinical reasoning quality of physicians who request abdominal CT scans during on-call hours has deteriorated over time. Clinical reasoning appears to be worse in younger patients. Key Points • In patients with suspected acute abdominal pathology who are scheduled to undergo CT scanning, referring physicians generally have difficulties in making an accurate pretest (differential) diagnosis. • Clinical reasoning quality of physicians who request acute abdominal CT scans has deteriorated over the years, while the number of CT scans has shown a significant increase. • Clinical reasoning quality appears to be worse in younger patients in this setting.


Introduction
Medical imaging utilization keeps on increasing in the Western world, and this particularly applies to CT examinations [1,2]. This growth in CT examinations is due to the relatively widespread availability of CT systems, continuous improvements in technology (which increase image quality, reduce scan time, and decrease radiation dose), decreasing costs of CT scans, and new clinical CT applications.
Although medical imaging is generally considered beneficial to patients, a part of its increase has been attributed to overutilization [3]. Anecdotal data indicate that clinicians request more and more imaging simply because it is available, and to rule out disease rather than to confirm disease that is most likely based on probabilistic hypothetico-deductive reasoning [4][5][6]. Based on these anecdotal data and our own experience, it is hypothesized that the clinical reasoning quality of physicians who request medical imaging has deteriorated over the years. However, establishing a good differential diagnosis based on clinical history, physical examination, and laboratory findings, is considered important to optimize the interpretation and diagnostic yield of an imaging examination [4][5][6][7]. In addition, a good differential diagnosis is necessary to avoid unnecessary imaging studies that may be costly, that may have potential side effects due to the use of ionizing radiation and contrast agents, and that may yield incidental findings.
During on-call hours, CT scans of the abdomen are frequently ordered, because the list of acute abdominal diseases is long and they may be challenging to diagnose based on clinical grounds [8]. As such, they provide an excellent setting to test our hypothesis. The purpose of our study was therefore to investigate temporal changes in clinical reasoning quality of physicians who requested abdominal CT scans at a tertiary care center during on-call hours within a 15-year period.

Study design
This retrospective study was approved by the institutional review board and the requirement for informed consent was waived. A random sample of 36 unique calendar days was selected using the Random Calendar Date Generator [9]. All patients who underwent abdominal CT at a tertiary care center in The Netherlands during on-call evening and night hours on each of these 36 unique calendar days in each of the years between 2005 and 2019, were potentially eligible for inclusion. The University Medical Center Groningen is one of the seven university medical centers in the Netherlands, with more than 2 million people in its direct catchment area in the north-east of the country. It has more than 10,000 employees and 1,300 beds, and all medical specialties are represented. The hospital is also a recognized Level 1 trauma center, and performs all organ transplants. Patients were included if CT scanning was done to evaluate for acute abdominal pathology, and if the scan range included the region from the diaphragm to the pubic symphysis. Patients who underwent CT scanning to guide an intervention (e.g., drainage), patients whose CT scan was performed for non-acute reasons, and patients whose CT report or clinical records were missing, were excluded. Note that the cases that were selected for the present study were derived from a previously published study that had a different purpose, namely to investigate temporal changes in the utilization and patient impact of abdominal CT during duty shifts in terms of negative scan proportion, same-day hospital discharge, and incidental findings [10]. When ordering abdominal CT during on-call hours, the referring physician both calls the attending radiology resident or radiologist, and submits a digital request form. Other details regarding on-call logistics, CT acquisition, and CT interpretation can be read elsewhere [10].

Data extraction
Patient age, patient gender, indication for CT scanning, requesting department, use of intravenous CT contrast agents, use of gastrointestinal CT contrast agents, time of CT scanning (before or after midnight), and year in which the CT scan was performed, were recorded. The digital request form of each CT scan was evaluated to extract the differential diagnosis that was compiled by the referring physician, and the CT report was evaluated to extract the diagnosis made.

Data analysis
The quality of referring physicians' clinical reasoning before ordering a CT scan was quantified using a scoring system. To that end, the differential diagnosis made by the referring physician on the CT request form was compared to the CT diagnosis. Each of the differential diagnoses on the CT request form that matched the CT diagnosis received 1 point. If no differential diagnosis was provided, or if the CT scan diagnosis did not match any of the differential diagnoses on the CT request form, 0 points were assigned. The clinical reasoning score was then calculated by dividing the total number of points by the total number of differential diagnoses that was listed on the CT request form, and expressed as a percentage. For example, if the clinical differential diagnosis included ruptured abdominal aortic aneurysm, urolithiasis, and cholecystitis, and CT demonstrated urolithiasis, the clinical reasoning score would be 33%. Mann-Kendall tests were performed to test for any temporal changes in both clinical reasoning scores in each year between 2005 and 2019, and number of abdominal CT scans during on-call hours on the 36 randomly sampled unique calendar days in each year between 2005 and 2019. Subsequently, linear regression analyses were performed to determine the association between the quality of clinical reasoning (all years combined) and the following variables: patient age, patient gender, requesting department (surgery, internal medicine, emergency medicine, intensive care, or other), and time of CT scanning (before or after midnight). pvalues < 0.05 were considered statistically significant. Statistical analyses were executed using R version 3.6.3 software (R Foundation for Statistical Computing) and IBM Statistical Package for the Social Sciences (SPSS) version 23 (SPSS).  (Fig. 2).

Clinical reasoning skills: determinants
There was a significant association between the quality of clinical reasoning and patient age (β coefficient of 0.210, p = 0.002). The quality of clinical reasoning was not significantly associated with patient gender, requesting department, or time of CT scanning (Table 3).

Discussion
Our study has four main findings. First, in patients with suspected acute abdominal pathology who are scheduled to undergo CT scanning during on-call hours, referring physicians generally have difficulties in making an accurate pretest (differential) diagnosis. Second, clinical reasoning quality of physicians who request abdominal CT scans in these patients has significantly deteriorated over the years. Third, the number of CT scans has shown a significant increase during the same time frame. Fourth, clinical reasoning quality was significantly lower in younger patients in this setting. Our results underline the importance of CT scanning complementary to clinical history, physical examination, and laboratory findings, in increasing or decreasing the likelihood of acute diseases in the abdomen in the appropriate clinical setting. Most importantly, they confirm our hypothesis that the clinical reasoning quality of physicians who request acute abdominal CT scans has deteriorated over the years, while the number of CT scans has shown an opposite temporal trend. We speculate that this is mainly due to the CT scan capacity that increased over the years. In 2005, our hospital had two CT systems, which increased to three, four, and five CT systems in the course  (Fig. 2). Because of the virtually unlimited CT scan capacity, the need to carefully select those patients who would benefit most from CT scanning has been eliminated. We believe this has contributed to a culture change in which medical doctors spend less efforts to and have become less skilled at clinical reasoning, in essence exchanging clinical reasoning quality for CT scan quantity. Perhaps an increase in clinical productivity and other tasks (e.g., administrative) of referring physicians has also resulted in worse clinical reasoning, although there are no data available to support this notion. It remains unclear why clinical reasoning was worse in younger patients. Whether or not this is related to younger generations demanding more imaging, defensive medicine practices in younger patients, or other reasons, requires further investigation. Our study may have some implications. On one hand, attempts may be made to improve clinical reasoning of physicians who request CT scans. Education has been suggested as the most effective approach to reach this goal, and radiologists have been urged to accept this educational responsibility [3]. This may be attempted by participating in undergraduate medical education and by providing feedback to referring physicians (e.g., on an individual basis or during multidisciplinary meetingshighlighting the importance of communication with referring physicians), all of which take place at our institution. On the other hand, if poor clinical reasoning of referring physicians will be accepted as the status quo, healthcare policy makers may consider assigning more value to the role of medical imaging and interpreting radiologists. This may be realized by allocating more funds to train and hire radiology staff, because they save healthcare costs elsewhere, namely referring physician's efforts and time spent on comprehensive clinical history taking, physical examination, requesting and interpreting laboratory findings, and establishing an accurate differential diagnosis. In light of ongoing technical developments in cross-sectional imaging that further decrease acquisition time (CT, MRI and PET) or decrease radiation dose (CT and PET) [11][12][13][14], it is not inconceivable that CT and other cross-sectional imaging modalities will take over more and more upfront diagnostic tasks from referring physicians. Experts in the field of abdominal imaging have previously reported that many patients with acute abdominal pain are referred for imaging without a clear pretest diagnosis, and that in only a small proportion of patients a confident and accurate diagnosis can be made solely on the basis of medical history, physical examination, and laboratory test findings [8]. These expert statements are completely in line with the findings of the present study. Previous studies have also shown that the use of cross-sectional imaging (particularly CT and MRI in adults) keeps on rising, including abdominal CT utilization during on-call hours [1,2]. These findings are again no different from that of the present study. Most importantly, however, studies investigating (temporal changes in) clinical reasoning quality of physicians who request cross-sectional imaging have been lacking so far.
The present study had some limitations. First, this study was performed in a tertiary care center in a Western European country without any CT scan capacity issues. The results may be  However, providing adequate clinical information with a differential diagnosis on the CT request form can be considered a requisite to avoid miscommunication and errors, and in our experience the telephonic and written communications match in this setting. Fourth, annual data on the number of patient consultations related to acute abdominal pathology, patient expectations, clinical status and gravity of their clinical presentation, staffing of the different departments in our hospital, training level of the persons who filled in the CT request forms (e.g., medical specialist vs. non-medical specialist, and years of experience), allegations against medical doctors, and defense medicine practices, were not available, as a result of which it was not possible to correlate these metrics to clinical reasoning quality. These topics should be investigated by further studies. It also remains unclear if new guidelines have an effect on clinical reasoning quality. Nevertheless, after a national guideline on diagnostic tests for acute abdominal pain in adults (excluding abdominal pain after trauma, known intra-uterine pregnancy, gastrointestinal hemorrhage, hemorrhagic shock, chronic conditions, recurrent abdominal pain, and abdominal pain within 30 days after abdominal surgery) was published in 2013 (which recommends ultrasonography as first-line imaging, followed by CT if ultrasonography is negative or inconclusive, unless the patient is critically ill) [15], there was no clear trend change with regard to clinical reasoning quality. Fifth, CT scans were not evaluated in terms of appropriateness, influence on patient management, or effect on patient outcome. This was beyond the scope of the present study. Previous work has already shown that CT results in emergency department settings frequently change physicians' diagnoses and admission decisions [16].
In conclusion, the clinical reasoning quality of physicians who request abdominal CT scans during on-call hours has deteriorated over time. Clinical reasoning appears to be worse in younger patients.