Introduction
Year on year, medical imaging increases inexorably [
1,
2], outstripping growth in radiology staffing [
3]. Radiologists are therefore under increasing pressure to report more studies more rapidly, while maintaining accuracy. Yet these two demands conflict fundamentally; faster reporting may compromise image scrutiny, leading to increased error. For simpler examinations consisting of a single or very few images (e.g. conventional radiographs), rapid interpretation may be possible with relatively little deterioration in diagnostic performance [
4]. Conversely, for complex studies with multiple images (e.g. CT scanning), greater interpretation speed is commonly deleterious [
5].
Computed tomographic colonography (CTC) is both complex and time-consuming to interpret. Moreover, it is fatiguing, as the interpretive task (of flying or scrolling through the colon) is repetitive, and the majority of examinations are negative for the primary target condition (colorectal cancer or large polyps) [
6‐
9], a phenomenon that is known to reduce vigilance [
10]. Anecdotally, radiologists often admit that they find it tiresome to report more than a handful of CTC examinations in a given reporting session, and that their concentration often wavers if they attempt to do so. It is therefore tempting to interpret CTC rapidly, particularly for the final few examinations in a given reporting session. However, this may well reduce detection rates; in a laboratory environment, more rapid fly-through at endoluminal CTC reduces both the proportion of colonic mucosa viewed by the radiologist and the polyp detection rate [
11]. For colonoscopy, endoscopists with shorter withdrawal times (i.e. providing less time to inspect the colon) have lower adenoma detection rates (ADR) [
12‐
14] and higher interval cancer rates [
14]. Moreover, ADR tends to drop towards the end of the day and even towards the end of an individual colonoscopy list [
15,
16], implying a “fatigue effect” when performing multiple examinations consecutively. Whether the same is true for CTC is unknown.
We therefore aimed to determine if polyp detection rates (PDR) and positive predictive value (PPV) at CTC are associated with (a) the number of CTC examinations interpreted by a radiologist on any given day (i.e. a fatigue effect) and (b) the length of time radiologists spend on interpretation (as a proxy for completeness of image scrutiny).
Discussion
CTC interpretation is time-consuming and fatiguing. We found that as radiologists interpreted more CTC examinations on a given day, their detection rate dropped by roughly 40% after five or more studies had been reported. Moreover, radiologists who spent longer interpreting cases that they ultimately called negative had higher detection rates than their colleagues who interpreted more quickly, with no corresponding detriment to their positive predictive value. These data strongly suggest that radiologists reporting CTC must be protected from pressures to report too quickly, or for too long—or missed pathology will be the consequence.
Although, in most cases, the primary goal of CTC is to confirm or refute colorectal cancer (or an alternative cause for patient symptoms, such as diverticulosis), it also represents an opportunity to reduce future colorectal cancer incidence by detection and subsequent removal of precursor polyps (i.e. adenomas and certain serrated lesions). Accordingly, radiologists interpreting CTC must be vigilant not only for large masses that may underpin symptoms, but also for smaller polyps; otherwise, patients may return in the future (usually many years later) with a post-investigation colorectal cancer (PICRC) [
19]. Indeed, the majority of PICRCs occurring after CTC are visible in retrospect, either as a mass lesion that was overlooked or as a polyp that subsequently became malignant [
20]. Such errors will be impossible to prevent entirely, but systems and methods that diminish this clinical and medicolegal risk would improve patient care substantially. Our findings suggest that relatively simple changes to radiologist workflow might be valuable; avoiding fatigue by reducing the number of CTC studies reported consecutively and introducing a minimum “negative interpretation time” before a scan are deemed normal. This is highly plausible, because eye-tracking experiments show that over-rapid endoluminal fly-through reduces the amount of colonic surface that a radiologist can bring into their central vision [
21]—slowing down would mitigate this risk. A minimum interpretation time of 20 min per case would seem reasonable, since this was the average time taken for the first scan interpreted each day at the quicker of the two centres, with 30 min per negative case being a desirable (and achievable) standard.
Of note, for the most directly comparable test to CTC, namely colonoscopy, the importance of prolonging inspection of the colonic mucosa to maximise detection has been recognised for many years [
12]. Gastroenterologists who spent less than 6 min withdrawing the colonoscope had detection rates that were less than half that of their colleagues spending longer [
12]. More recently, data from the English Bowel Cancer Screening Programme (in which endoscopists are already highly trained and accredited) show that extending the examination towards 10 min yields further benefits in detection rate [
13]. Moreover, this translates to clinical outcomes; colonoscopists with longer withdrawal times have lower post-colonoscopy colorectal cancer (PCCRC; similar to “interval cancers” in a screening programme) rates than those who withdraw the scope (too) rapidly [
14]. Negative withdrawal time (i.e. calculated only for cases where no polyps are found) is now recognised as a key performance indicator (KPI) for the quality of many colonoscopy services, including in the UK [
22], Europe [
23] and the USA [
24].
The concept of slowing down to improve accuracy is not new, nor is it specific to CTC. Requirements to report large numbers of examinations rapidly (to reduce wait times and reporting backlogs) must be balanced against the risks of making errors. If scans are acquired but languish on the PACS, remaining unreported, this is a worse situation than them being reported, even suboptimally. This clinical risk has been highlighted in England by the Care Quality Commission (CQC) [
25]. On the other hand, patients will rightly not accept that their cancer or polyp was missed due to time pressures and underfunding. It is highly iniquitous and counter-intuitive that a patient may have colonoscopy, where they receive the undivided attention of an accredited endoscopist who will examine their colon for a minimum length of time (i.e. the negative withdrawal time), or—based on local pathways or the whim of a referring doctor—instead undergo CTC where the radiologist may be interrupted repeatedly and without any minimum standard for the duration of interpretation. Such infrastructural and process shortcomings highlight the need for robust minimum standards that protect both patients and radiologists in the face of increasing demand.
We also found that reporting multiple CTC examinations in sequence was associated with progressive deterioration in detection, suggesting a “fatigue effect”. This phenomenon has been described in many other areas, including colonoscopy. The adenoma detection rate (ADR) falls as colonoscopy lists progress [
15,
16], and is typically higher in the morning than evening. However, this finding is not universal, and some studies have found the effect either weak [
26] or absent entirely [
27]. Nonetheless, anecdotally, many radiologists become fatigued after reporting several CTC examinations consecutively, and avoid doing so where possible. Given our findings, it may be prudent to avoid reporting large numbers (four or more) of CTC without a break. A 4-h session of approximately eight CTC studies reported in two blocks with a half-hour break would seem an appropriate guideline, since it permits both the minimum negative interpretation time and no more than four cases criteria to be met.
This study has several limitations. Firstly, we investigated just two tertiary care centres, and only seven radiologists, which may not represent wider practice. Secondly, the data are retrospective and observational, and therefore it is not possible to exclude bias. For example, scans interpreted earlier in a sequence may have been highlighted to the radiologist for prioritisation (for example, marked as “urgent” on the RIS), although it was the practice at both institutions to report in date order. Even so, if we ignore fatigue, it is difficult to explain why the effect of scan sequence was consistent across two centres with different workflows and for as many as five successive scans. Thirdly, we were forced to make some assumptions when estimating radiologist negative interpretation time, specifically calculating the time spent interpreting a CTC study by using the time at which the report was verified and relating this to the immediate prior report, and excluding some reports with implausibly long or short interpretation times. The reporting time also includes time spent scrutinising the image for extracolonic findings, which may partly explain the relatively large difference in average interpretation time between the two centres. We mitigated against this by using each radiologist’s negative interpretation time normalised to the centre average, but this may have altered the size of the effect that we observed.
In summary, in a retrospective observational study from two NHS hospitals, we found that the proportion of positive CTC examinations and polyp detection rates reduced as radiologists reported multiple examinations, suggesting a “fatigue effect”, and radiologists with longer interpretation times had higher polyp detection rates with no corresponding reduction in positive predictive value. CTC services should protect their radiologists and patients by removing the need to report too fast or for too long.
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.