Endoscopist specialty is associated with colonoscopy quality

Colorectal cancer (CRC) screening programs are either in preparation or ongoing in all Canadian provinces and many developed countries [1]. Colonoscopy is central to CRC screening, and colonoscopy quality assurance initiatives are underway in several countries [2–4]. Measures of quality in colonoscopy, such as cecal intubation rate, colonoscope withdrawal time, and adenoma detection rate (ADR) [5, 6], vary substantially among endoscopists [7–10]. The establishment of benchmarks for these quality indicators aims to standardize colonoscopy practice quality.

Of the many quality indicators identified, only ADR has been shown to independently predict CRC diagnosis after colonoscopy, an important public health outcome that suggests failure of screening [11]. However, ADR is difficult to assess using population-based data due to the lack of pathology report information in health administrative databases. Polypectomy rate, the proportion of colonoscopies that result in the removal of one or more polyps, has been proposed as an alternative quality indicator because it is known at the time of colonoscopy, strongly correlated with ADR, and available from health administrative data [12–14]. Quality benchmarks for polypectomy rates have been proposed [12, 14].

Some studies have shown endoscopist specialty to be associated with cancers missed by colonoscopy [15–18], raising concerns about training and quality assurance. Polypectomy rate is an upstream marker of quality because failure to remove and diagnose pre-cancerous and cancerous lesions leads to missed cancers. It is conceivable, therefore, that differences in missed cancers between specialties may originate with specialty differences in polypectomy rates. The relationship between endoscopist specialty and polypectomy rate has not been examined. If this relationship echoes that between specialty and missed cancers, this finding would lend support to the usefulness of polypectomy rate as a quality indicator.

The objectives of the present study were to determine whether there is a difference in polypectomy rates between surgeons and gastroenterologists beyond what is attributable to differences in patient risk profiles, and to assess the variability in polypectomy rates within each specialty.

Using a combination of primary and health administrative data from Montreal and Calgary, we estimated the association between endoscopist specialty and polypectomy rate. We found a clinically important difference between endoscopist specialities in Montreal, with surgeons being approximately 50% less likely than gastroenterologists to remove polyps. A similar trend was observed in Calgary but wide credible intervals preclude a definitive conclusion. Patient risk factors yielded odds ratios consistent with the literature, including increase in the probability of polypectomy with increasing age. However, despite statistical adjustment, residual confounding due to patient-level risk factors may contribute the observed difference between specialties.

Several prior studies have examined the association between endoscopist speciality and missed CRCs. In a retrospective cohort study using population level data from Ontario, Bressler et al. identified endoscopist specialty as a risk factor for incident CRCs post-colonoscopy among both male and female patients [16]. In contrast to our study, the difference was found mainly between gastroenterologists and primary care physicians, rather than between gastroenterologists and surgeons. This study also found that office colonoscopies are more likely to be associated with missed lesions than hospital-based colonoscopies. A Manitoba study also showed no difference in missed cancers between surgeons and gastroenterologists. This model included colonoscopy volume and practice location (rural vs. urban) as covariates, neither of which was significantly associated with missed cancers [18]. Rabeneck et al. reported an odds ratio of 1.39 (95% CI: 1.16–1.67) for the risk of CRC diagnosis after negative colonoscopy in the patients seen by surgeons vs. those seen by gastroenterologists in the hospital setting [17]. Colonoscopy volume was not a predictor of missed cancers in this study. Baxter et al. used administrative data from Ontario and found no difference between surgeons and gastroenterologists in post-colonoscopy CRCs, but indicated that non-surgeon and non-gastroenterologist specialties were at increased risk for missed cancers [15]. Similar to previous findings, practice setting was associated with post-colonoscopy CRCs, while colonoscopy volume was not. Only one of the aforementioned studies that used post-colonoscopy CRCs as an outcome detected a difference between surgeons and gastroenterologists. Although many of these studies indicated a non-significant trend towards higher rates of missed cancers among surgeons, the failure to detect a statistically significant difference may have been due to a lack of power because the outcome is rare. In contrast to province-wide studies from Ontario and Manitoba on endoscopy specialty and quality, our study was restricted to urban hospitals and to ambulatory care patients; thus, our findings are less generalizable. Nevertheless, in province-wide studies, the power to detect an interaction between location and specialty may be of concern, since the majority of colonoscopies are performed by gastroenterologists in urban areas and by surgeons in rural areas [22].

We determined the variability in polypectomy rates within each specialty, and found considerable variation in both specialties. Several studies have reported important variation in ADRs by specialty. Barclay et al. examined 2,053 screening colonoscopies by 12 endoscopists and found that ADRs varied from 9.4 to 32.7% [23]. Chen et al. studied the variation among 9 endoscopists who performed 10,034 colonoscopies. After adjusting for patient age and sex, detection rates for at least one adenoma ranged between 15.5 and 41.1% [24]. Imperiale et al. found that ADRs ranged from 7% to 44% among 46 endoscopists who performed 2,664 screening colonoscopies [7]. Our results mirror those of others showing variation in endoscopist performance, and further current knowledge in that considerable variation exists between as well as within specialties.

Our results suggest a clinically important difference in polypectomy rates between surgeons and gastroenterologists after adjusting for patient-level factors, as well as substantial variation among all endoscopists. Some of this variation may be explained by endoscopist-level factors such as training, practice factors, and technical factors. Recently, performance quality was compared in trainees with similar endoscopy experience; gastroenterology trainees outperformed surgery trainees on several quality indicators including colonoscopy completion rate, polypectomy rate, ADR, and withdrawal time [10]. One Calgary study found that gastroenterology trainees performed considerably more colonoscopies than surgery trainees during their training [25]. Further, all of the gastroenterology fellows fulfilled the minimum number of colonoscopies recommended by the American Society for Gastrointestinal Endoscopy for assessment of competency, while none of the surgery residents did. These findings suggest that differences in colonoscopy quality between the two specialties may arise from differences in training.

Practice factors such as setting and annual case volume may also contribute to variation among endoscopists. For example, Bressler et al. and Baxter et al. both found non-hospital based colonoscopies to be associated with increased risk of missed cancers [15, 16]. A recent study from the U.K. showed that volume and accreditation were significantly associated with colonoscopy quality indicators [26]. Interestingly, in this study surgeons had a higher polyp detection rate compared to physicians. However, this difference was minimal after adjustment for patient-level factors. Studies in Manitoba and Ontario have not found colonoscopy volume to predict missed cancers [15, 17, 18].

In terms of technical factors, withdrawal time has been identified as a significant predictor of ADR in some studies. Results from the Bowel Cancer Screening Programme in England show a significant increase in ADR among endoscopist with mean withdrawal time of 11 minutes or longer compared to those with mean withdrawal time of less than 7 minutes [27]. Investigators in Spain examined bowel cleansing, sedation, cecal intubation, and withdrawal time as potential predictors of ADR. Only withdrawal time longer than 8 minutes was independently associated with ADR [28]. However, a German study found neither annual case volume nor withdrawal time to be correlated with ADR [29]. The mixed findings regarding practice and technical factors that affect colonoscopy quality may be due to differences in training, accreditation, practice settings, and CRC screening delivery models between countries.

A major strength of our study is primary data collection on patient level CRC risk factors that enabled adjustment for colonoscopy indication, family history of CRC and previous colonoscopy. This is important because polypectomy rate is a function of both patient risk and endoscopist performance.

One study limitation is residual confounding. Although we adjusted for many important patient risk factors, it is possible that differences in the characteristics of patients referred to surgeons and gastroenterologists were not adequately captured by the patient-level covariates. A second limitation is potential misclassification, as data on polypectomy status were derived from provincial health administrative databases. To address this issue, we employed methods to adjust for the imperfect accuracy of health administrative data in both provinces. Using health administrative data whilst acknowledging and accounting for its limitations is good practice in clinical and health services research. However, our approach to misclassification adjustment for the Montreal data may have introduced bias if misclassification was differential between surgeons and gastroenterologists. Differences in administrative data quality between specialties may arise from differences in billing practices. Nevertheless, it is reassuring that the Calgary sample, where misclassification was reduced by combining two data sources, showed results in the same direction as the Montreal sample. Thirdly, hierarchical modeling was used to estimate the rates as it has the advantage of conservatively bringing unstable estimates closer to the overall mean so that they are less likely to affect the range of variation in polypectomy rates [30]. While use of this statistical technique allowed us to estimate the polypectomy rates of endoscopists with few study patients, there were wide credible interval limits around some individual estimates.

Although our findings highlight the important issue of the discrepancy in quality between surgeons and gastroenterologists, the lack of endoscopist-level variables in our study preclude us from isolating the modifiable predictors of endoscopist performance. Future studies aimed at teasing out such factors would help inform changes to training, accreditation, and quality assurance programs. We hope that our findings will serve as impetus for such investigations.

In conclusion, we found that surgeons were less likely to remove polyps compared to gastroenterologists, and that considerable variation in polypectomy rates exists within each specialty. Using polypectomy rate as an indicator of colonoscopy quality, our findings showed a difference in practice quality between the two specialties despite controlling for variations in patient risk profiles. Potential reasons for this discrepancy, such as training, practice volume, and technical factors, need to be investigated in future studies. Our findings suggest that the difference in ADRs may begin at the level of polyp removal, and lend further support for the use of polypectomy rate as a colonoscopy quality indicator.