Key Points
Question
Do adenoma and proximal sessile serrated polyp detection rates differ based on endoscopist characteristics?
Findings
This cohort study demonstrated differences in the adenoma and proximal sessile serrated polyp detection rates through replication of analyses of previous studies that have examined this question. However, after adjusting for additional factors, no difference in detection based on endoscopist characteristics was found.
Meaning
Differences in the adenoma and proximal sessile serrated polyp detection rates by endoscopist characteristics found by previous studies may be owing to residual confounding; per this analysis, patients or health systems need not select endoscopists based on the examined characteristics for the purposes of colorectal cancer screening.
This cohort study uses data from colonoscopies to attempt to replicate and expand on findings from previous studies that showed associations between adenoma and proximal sessile serrated polyp detection rates and endoscopist characteristics.
Abstract
Importance
Research demonstrates adenoma detection rate (ADR) and proximal sessile serrated polyp detection rate (pSSPDR) are associated with endoscopist characteristics including sex, specialty, and years in practice. However, many studies have not adjusted for other risk factors associated with colonic neoplasia.
Objective
To assess the association between endoscopist characteristics and polyp detection after adjusting the factors included in previous studies as well as other factors.
Design, Setting, and Participants
This cohort study was conducted in the Cleveland Clinic health system with data from individuals undergoing screening colonoscopies between January 2015 and June 2017. The study analyzed data using methods from previous studies that have demonstrated significant associations between endoscopist characteristics and ADR or pSSPDR. Multilevel mixed-effects logistic regression was performed to examine 7 endoscopist characteristics associated with ADRs and pSSPDRs after controlling for patient demographic, clinical, and colonoscopy-associated factors.
Exposures
Seven characteristics of endoscopists performing colonoscopy.
Main Outcomes and Measures
The ADR and pSSPDR, with a hypothesis created after data collection began.
Results
A total of 16 089 colonoscopies were performed in 16 089 patients by 56 clinicians. Of these, 8339 patients were male (51.8%), and the median (range) age of the cohort was 59 (52-66) years. Analyzing the data by the methods used in 4 previous studies yielded an association between endoscopist and polyp detection; surgeons (OR, 0.49 [95% CI, 0.28-0.83]) and nongastroenterologists (OR, 0.50 [95% CI 0.29-0.85]) had reduced odds of pSSPDR, which was similar to results in previous studies. In a multilevel mixed-effects logistic regression analysis, ADR was not significantly associated with any endoscopist characteristic, and pSSPDR was only associated with years in practice (odds ratio, 0.86 [95% CI, 0.83-0.89] per increment of 10 years; P < .001) and number of annual colonoscopies performed (odds ratio, 1.05 [95% CI, 1.01-1.09] per 50 colonoscopies/year; P = .02).
Conclusions and Relevance
The differences in ADRs that were associated with 7 of 7 endoscopist characteristics and differences in pSSPDRs that were associated with 5 of 7 endoscopist characteristics in previous studies may have been associated with residual confounding, because they were not replicated in this analysis. Therefore, these characteristics should not influence the choice of endoscopist for colorectal cancer screening. However, clinicians further from their training and those with lower colonoscopy volumes have lower adjusted pSSPDRs and may need additional training to help increase pSSPDRs.
Introduction
Health care expenditures in the United States have steadily increased despite substantial deficits in health care quality relative to other countries that spend less money.1,2 To justify expenses, third-party payers, large health systems, and medical societies have increasingly emphasized quality and outcome measures.3 Given wide variation in adenoma detection rates (ADRs) between clinicians, and given the inverse correlation of ADRs with interval cancer, it has been widely adopted as a quality benchmark for colonoscopy.4,5,6,7 Although performance targets for proximal sessile serrated polyp (pSSP) detection are not established, there is considerable evidence suggesting that these lesions are a major cause of interval colorectal cancer and a greater variability between endoscopists exists in proximal sessile serrate polyp detection rates (pSSPDRs) than ADRs.7,8,9,10,11
Research into endoscopist characteristics as factors associated with ADRs, to date, demonstrated mixed results. For example, some studies have shown greater ADRs among gastroenterologists,12,13 female endoscopists,12 and more recently trained physicians.12 Others11 have found no difference based on endoscopist specialty. Importantly, these studies11,12,13 did not adjust for many factors associated with the prevalence of polyps, such as comorbid conditions, medication use, and colonoscopy procedure–associated characteristics. Further, there are limited data on the association of endoscopist characteristics with pSSPDRs.14,15
We sought to assess the association between endoscopist characteristics and ADRs and pSSPDRs. To assess if previous study results were owing to residual confounding and not endoscopist characteristics, we first analyzed our data by methods used in previous studies.12,13,14,15 We hypothesized that adjustment for patient and procedural factors (eAppendix 1 in the Supplement) would negate any observed differences in ADRs and pSSPDRs by specialty, clinician sex, or other endoscopist characteristics.
Methods
Study Setting and Patient Selection
The Cleveland Clinic institutional review board approved this study. All normal-risk screening colonoscopies performed in the Cleveland Clinic system between January 2015 and June 2017 were identified from electronic health records. Informed consent was not needed owing to the deidentified nature of the data.
Data Collection and Definitions
Patient demographic characteristics, substance use (tobacco use [former or current], alcohol abuse per diagnostic codes or description in electronic health records), comorbidities, medication use (at the time of colonoscopy per data on patient medical record), colonoscopy timing (time of the day and month), and location of endoscopy centers were also obtained from the electronic health records. Elements of colonoscopy and pathology reports were ascertained by validated natural language processing. The number of polyps, quality of bowel preparation (per the Aronchick scale, descriptions of good and excellent preparation were considered adequate, and descriptions of fair, poor, and inadequate preparation were considered inadequate), withdrawal time, cecal intubation rate, location of polyps (proximal defined as the cecum to and including the splenic flexure), endoscopist’s name, and presence of a trainee during colonoscopy were obtained. The adenoma detection rate (ADR) and pSSPDR were defined as the percentage of screening colonoscopies with the associated pathological changes detected, over all screening colonoscopies. The annual number of colonoscopies performed by each endoscopist was calculated. Only patients undergoing normal-risk screening colonoscopies and colonoscopies performed by clinicians who performed more than 100 normal-risk screening colonoscopies during the study period were included. Information regarding endoscopist specialty, sex, location of the medical school attended (United States vs non-United States), years since completion of final subspecialty training, and primary practice setting (with academic settings defined as main campus or tertiary referral centers and private settings defined as family health and regional endoscopy centers) were obtained using institutional physician biographic sketches.
Statistical Analysis
Baseline characteristics of all patients who underwent a colonoscopy and endoscopists performing colonoscopy were described. Differences in ADRs (overall and by patient sex) and pSSPDRs (overall and by patient sex) were computed based on endoscopist characteristics: specialty, sex of the endoscopist, location of medical school attended by the endoscopist, years since completion of training, number of colonoscopies performed per year, practice setting, and presence of trainee during colonoscopy. Advanced endoscopists were defined as endoscopists with additional training in endoscopy. Coefficients of variation (SD divided by mean) were calculated, and bar graphs were created to demonstrate the distribution of pSSPDRs and ADRs. Categorical variables were reported as frequencies with percentages. Significance of differences in categorical variables was measured using χ2 analysis. Next, bivariate logistic regression was performed to identify the odds of detecting adenomas and proximal SSPs (which does not adjust for random effects of individual endoscopist). We also used the data set to replicate the analysis performed in prior studies (eAppendix 2 in the Supplement) to determine if endoscopist characteristics were associated with polyp detection rates.12,13,14,15
A multilevel mixed-effects logistic regression analysis was performed to identify the odds of detecting adenomas and proximal SSPs based on various endoscopist characteristics after adjusting for random effects of individual clinicians and fixed effects (listed in eAppendix 1 in the Supplement). We also performed a multivariable logistic regression analysis for the aforementioned outcomes that adjusts for the fixed effects alone to assess the influence of random effects of the individual endoscopist. All statistical analyses were conducted using Stata SE, version 15.1 (StataCorp). Two-sided P values less than .05 were considered significant. Data analysis occurred from January 2015 to December 2017.
Results
Baseline Characteristics of Patients and Endoscopists
The baseline demographic and clinical characteristics of the 16 089 patients are presented in Table 1. Briefly, 8339 patients were male (51.8%), and the median (range) age of the cohort was 59 (52-66) years.
Table 1. Baseline Demographic, Clinical, and Colonoscopy Features.
| Variable | No. (%)a |
|---|---|
| Colonoscopies | 16 089 (100) |
| Age, median (IQR), y | 59 (52-66) |
| Sex | |
| Male | 7750 (48.2) |
| Female | 8339 (51.8) |
| Race/ethnicity | |
| White | 12 115 (75.3) |
| Black | 2911 (18.1) |
| Other | 1063 (6.6) |
| Insurance | |
| Private | 10 440 (64.9) |
| Medicare | 4416 (27.5) |
| Medicaid | 1112 (6.9) |
| Other | 121 (0.8) |
| BMI | |
| Underweight (<18.5) | 90 (0.6) |
| Normal (18.5-24.9) | 6129 (38.1) |
| Overweight (25-29.9) | 4302 (26.7) |
| Obese (30-39.9) | 4666 (29) |
| Morbid obesity (≥40) | 900 (5.6) |
| Tobacco use | 5532 (34.4) |
| Alcohol abuse | 807 (5.0) |
| Comorbidities | |
| Diabetes | 2941 (18.3) |
| Cirrhosis | 787 (4.9) |
| Dementia | 173 (1.1) |
| Stroke | 71 (0.4) |
| Constipation | 1408 (8.8) |
| Coronary artery disease | 1492 (9.3) |
| Congestive heart failure | 545 (3.4) |
| Medication use | |
| Aspirin | 5094 (31.7) |
| Statin | 5247 (32.6) |
| Calcium supplement | 5810 (36.1) |
| Vitamin D | 8511 (52.9) |
| Estrogen | 1206 (7.5) |
| Angiotensin-converting enzyme inhibitors | 5027 (31.2) |
| Angiotensin receptor blocker | 2004 (12.5) |
| Calcium channel blocker | 3165 (19.7) |
| Bisphosphonates | 1184 (7.4) |
| History of cholecystectomy | 245 (1.5) |
| Month colonoscopy performed | |
| January-March | 4862 (30.2) |
| April-June | 4591 (28.5) |
| July-September | 3353 (20.8) |
| October-December | 3283 (20.4) |
| Time of day of colonoscopy | |
| Morning | 12 432 (77.3) |
| Afternoon | 3657 (22.7) |
| Sedation type | |
| Conscious sedation | 15 740 (97.8) |
| Monitored or general anesthesia | 349 (2.2) |
| Bowel preparation | |
| Inadequate | 2050 (12.7) |
| Adequate | 14 039 (87.3) |
| Location where colonoscopy performed | |
| Main campus | 3227 (20.1) |
| Community hospital | 1770 (11) |
| Ambulatory endoscopy center | 11 092 (68.9) |
| Withdrawal time, median (IQR), min | 10 (7-10) |
| Cecal intubation rate | 15 652 (97.3) |
Abbreviations: BMI, body mass index (calculated as weight in kilograms divided by height in meters squared); IQR, interquartile range.
Categorical variables are listed as frequency with percentages; continuous variables are listed as median with interquartile ranges.
Fifty-six endoscopists, including gastroenterologists (34 [61%]), surgeons (15 [27%]), and advanced endoscopists (7 [13%]), performed the colonoscopies. Twenty-five percent of the endoscopists were female (n = 14), and 25% were international medical graduates (n = 14). Endoscopists had a median time from training of 16.3 years (interquartile range [IQR], 6.4-26.0 years) and performed a median (IQR) of 267.0 (158.5-436.2) colonoscopies per year. One in every 8 clinicians (n = 7 [13%]) were in private practice, and 8.2% of colonoscopies (n = 1319) involved a trainee.
Analyzing the Data According to Methods in Previous Studies
When we used the data to perform a similar analysis as was done by Mehrotra et al,12 we found that a lower ADR was noted in surgeons (odds ratio [OR], 0.67 [95% CI, 0.62-0.72]) and clinicians with longer time spans since training (OR, 0.86 [95% CI, 0.83-0.90] per increment of 10 years). These findings were similar to those of Mehrotra et al.12 Unlike Mehrotra et al, significant differences in adenoma detection were not found based on endoscopist sex (OR, 1.06 [95% CI, 0.97-1.15]; Table 2).
Table 2. Previous Study Results and Replicated Analyses With Current Dataa.
| Analysis | OR (95% CI) | ||||
|---|---|---|---|---|---|
| Specialtyb | Sex | Years in Practice | Annual Colonoscopy Volume | Trainee Participation | |
| Adenoma detection rate | |||||
| Mehrotra et al12 | |||||
| Original | 0.58 (0.47-0.72) | 1.26 (1.00-1.59) | 0.88 (1.16-1.82)c | 0.90 (0.74-1.09); 0.88 (0.72-1.08); 1.08 (0.88-1.31)d | NA |
| Replicated | 0.67 (0.62-0.72) | 1.06 (0.97-1.15) | 0.86 (0.83-0.90)c | 1.00 (0.99-1.01)d | NA |
| Zorzi et al13 | |||||
| Original | 0.87 (0.76-0.96) | NA | NA | NA | NA |
| Replicated | 0.68 (0.54-0.85) | NA | NA | NA | NA |
| Proximal sessile serrate polyp detection rate | |||||
| Parikh et al14 | |||||
| Original, mean (SD) | 3.9 (3.5) vs 2.2 (3.0)e | NA | NA | NA | 2.4 (13.4) vs 3.5 (3.8)e |
| Replicated | 0.49 (0.28-0.83) | NA | NA | NA | 0.88 (0.66-1.18) |
| Crockett et al15 | |||||
| Original | 0.53 (0.37-0.75) | 1.10 (0.84-1.44) | 1.52 (1.14 - 2.04)f | 1.77 (1.27-2.46)g | NA |
| Replicated | 0.50 (0.29-0.85) | 1.07 (0.91-1.25) | 0.70 (0.56-0.88)f | 1.05 (1.01-1.11)g | NA |
Abbreviations: NA, not available; OR, odds ratio.
Both the original and replicated analysis account for random effects of or clustering by individual endoscopist. All values are reported as ORs, except where stated otherwise.
Nongastroenterologists were compared with gastroenterologists (reference variable). For the sake of uniformity, in studies where nongastroenterologists were used as a reference, inverse values of ORs were used to ensure gastroenterologists were the reference variable.
Mehrotra et al12 compared endoscopists who had 9 or fewer years of practice since residency (OR, 0.88 [95% CI, 1.16-1.82]) with physicians who had 27 to 51 years of practice. The replicated analysis reports the OR per increment of 10 years in practice.
Mehrotra et al12 compared a 2-year colonoscopy volume of 116 to 278 colonoscopies (OR, 0.90 [95% CI, 0.74-1.09]), 279 to 771 colonoscopies (OR, 0.88 [95% CI, 0.72-1.08]), and 772 to 2654 colonoscopies (OR, 1.08 [95% CI, 0.88-1.31]) with a reference-year colonoscopy volume of 30 to 115 procedures. The replicated analysis reports the OR per increment of 50 colonoscopies per year.
Parikh et al14 did not provide ORs for their adjusted analysis, but rather mean (SD) values for gastroenterologists (3.9 [3.5]) and nongastroenterologists (2.2 [3.0]; P = .03; P = .001 on multivariable analysis) and colonoscopies performed with fellows (2.4 [13.4]) and without fellows (3.5 [3.8]; P = .049).
Being in practice for 9 years or less was associated with greater odds of proximal sessile serrated polyp detection (OR, 1.52 [95 % CI, 1.14 – 2.04]) compared with being in practice for 27 years or more. The replicated analysis reports ORs per increment of 10 years in practice.
Being in the highest quartile of annual colonoscopy volume was associated with greater odds of proximal sessile serrated polyp detection (OR, 1.77 [95 % CI, 1.27-2.46]) compared with being in the lowest quartile. The replicated analysis reports the OR per increment of 50 colonoscopies per year.
Next, the data were subjected to a multivariable analysis according to Zorzi et al.13 We found a lower ADR among nongastroenterologists (OR, 0.68 [95% CI, 0.54-0.85]), which was similar to the finding of Zorzi et al (Table 2).
Performing a multivariable analysis as in Parikh et al,14 we found that surgeons had reduced odds of pSSPDR (OR, 0.49 [95% CI, 0.28-0.83]). This was similar to their results.
Performing multivariable analysis using the same parameters as Crockett et al,15 we found that being a nongastroenterologist (OR, 0.50 [95% CI 0.29-0.85]) and completing a greater annual volume of colonoscopies (OR, 1.05 [95% CI, 1.01-1.11] per increment of 50 colonoscopies per year) both had significant associations with pSSPDR outcomes. However, years since completion of training was negatively associated with pSSPDRs (OR, 0.70 [95% CI, 0.56-0.88] per increment of 10 years).
Bivariate Analysis
Overall, the mixed-sex ADR was 31.3% (36.9% in male patients [n = 2860 of 7749] and 26.0% in female patients [n = 2167 of 8339]). In the bivariate analysis, ADR differed by endoscopist specialty (male patients: gastroenterologist, 1791 of 4462 [40.1%]; surgeon, 753 of 2586 [29.1%]; advanced endoscopist, 316 of 701 [45.1%]; P < .001; female patients: gastroenterologist, 1427 of 5070 [28.1%]; surgeon, 534 of 2511 [21.3%]; advanced endoscopist, 206 of 758 [27.2%]; P < .001), endoscopist sex (male patients: male clinician, 2215 of 6116 [36.2%]; female clinician, 645 of 1633 [39.5%]; P < .001; female patients: male clinician, 1516 of 6080 [24.9%]; female clinician, 651 of 2259 [28.8%]; P < .001), and time since training (male patients: <10 years, 1042 of 2407 [43.3%]; 10-20 years, 692 of 1861 [.37.2%]; >20 years, 1126 of 3481 [32.3%]; P < .001; female patients: <10 years, 865 of 2894 [29.9%]; 10-20 years, 542 of 2137 [25.4%]; >20 years, 760 of 3308 [23.0%]; P < .001; Table 3; eFigure 1 in the Supplement).
Table 3. Bivariate Analysis of Adenoma Detection Rate (ADR) by Endoscopist Characteristics.
| Variable | Overall Adenoma Detection Rate | Adenoma Detection Rate in Men | Adenoma Detection Rate in Women | |||
|---|---|---|---|---|---|---|
| No./Total No. (%) | P Value | No./Total No. (%) | P Value | No./Total No. (%) | P Value | |
| All colonoscopies | 5028/16 089 (31.3) | NA | 2860/7749 (36.9) | NA | 2167/8339 (26) | NA |
| Specialty | ||||||
| Gastroenterologist | 3219/9533 (33.8) | <.001 | 1791/4462 (40.1) | <.001 | 1427/5070 (28.1) | <.001 |
| Surgeons | 1287/5097 (25.3) | 753/2586 (29.1) | 534/2511 (21.3) | |||
| Advanced endoscopist | 522/1459 (35.8) | 316/701 (45.1) | 206/758 (27.2) | |||
| Endoscopist sex | ||||||
| Male | 3732/12 197 (30.6) | .002 | 2215/6116 (36.2) | .02 | 1516/6080 (24.9) | <.001 |
| Female | 1296/3892 (33.3) | 645/1633 (39.5) | 651/2259 (28.8) | |||
| Location of medical school | ||||||
| United States | 3786/12 222 (31.0) | .18 | 2166/5860 (37.0) | .86 | 1619/6361 (25.5) | .05 |
| Non–United States | 1242/3867 (32.1) | 694/1889 (36.7) | 548/1978 (27.7) | |||
| Years since completion of training | ||||||
| <10 | 1908/5302 (36) | <.001 | 1042/2407 (43.3) | <.001 | 865/2894 (29.9) | <.001 |
| 10-20 | 1234/3998 (30.9) | 692/1861 (37.2) | 542/2137 (25.4) | |||
| >20 | 1886/6789 (27.8) | 1126/3481 (32.3) | 760/3308 (23.0) | |||
| Colonoscopies per year | ||||||
| <250 | 1324/4266 (31) | .72 | 804/2104 (38.2) | .15 | 519/2161 (24.0) | .02 |
| ≥250 | 3704/11 823 (31.3) | 2056/5645 (36.4) | 1648/6178 (26.7) | |||
| Practice setting | ||||||
| Academic | 4676/14 868 (31.5) | .06 | 2646/7127 (37.1) | .18 | 2029/7740 (26.2) | .09 |
| Private practice | 352/1221 (28.8) | 214/622 (34.4) | 138/599 (23.0) | |||
| Presence of trainee | ||||||
| No | 4613/14 828 (31.1) | .19 | 2632/7142 (36.9) | .73 | 1980/7685 (25.8) | .11 |
| Yes | 415/1261 (32.9) | 228/607 (37.6) | 187/654 (28.6) | |||
Abbreviation: NA, not applicable.
The pSSPDR was 4.6% (n = 741) in both male and female patients (eFigure 2 in the Supplement) and differed by endoscopist specialty (gastroenterologist, 3219 of 9533 [33.8%]; surgeons, 1287 of 5097 [25.3%]; and advanced endoscopist, 522 of 1459 [35.8%]; P < .001) and time since training (<10 years, 1908 of 5302 [36.0%]; 10-20 years, 1234 of 3998 [30.9%]; >20 years, 1886 of 6789 [27.8%]; P < .001) (Table 4).
Table 4. Bivariate Analysis of Proximal Sessile Serrate Polyp Detection Rate (pSSPDR) by Endoscopist Characteristics.
| Variable | Overall pSSPDR | pSSPDR in Men | pSSPDR in Women | ||||
|---|---|---|---|---|---|---|---|
| No./Total No. (%) | P Value | No./Total No. (%) | P Value | No./Total No. (%) | P Value | ||
| All colonoscopies | 741/16 089 (4.6) | NA | 355/7749 (4.6) | NA | 386/8339 (4.6) | NA | |
| Specialty | |||||||
| Gastroenterologist | 536/9533 (5.6) | <.001 | 247/4462 (5.5) | <.001 | 289/5070 (5.7) | <.001 | |
| Surgeons | 117/5097 (2.3) | 60/2586 (2.3) | 57/2511 (2.3) | ||||
| Advanced endoscopist | 88/1459 (6.0) | 48/701 (6.8) | 40/758 (5.3%) | ||||
| Endoscopist sex | |||||||
| Male | 509/12 197 (4.2) | <.001 | 258/6116 (4.2) | .003 | 251/6080 (4.1) | <.001 | |
| Female | 232/3892 (5.9) | 97/1633 (5.9) | 135/2259 (6.0) | ||||
| Location of medical school | |||||||
| United States | 548/12 222 (4.5) | .19 | 246/5860 (4.2) | .004 | 302/6361 (4.7) | .35 | |
| Non–United States | 193/3867 (5.0) | 109/1889 (5.8) | 84/1978 (4.2) | ||||
| Years since completion of training | |||||||
| <10 | 352/5302 (6.6) | <.001 | 164/2407 (6.8) | <.001 | 188/2894 (6.5) | <.001 | |
| 10-20 | 205/3998 (5.1) | 105/1861 (5.6) | 100/2137 (4.7) | ||||
| >20 | 184/6789 (2.7) | 86/3481 (2.5) | 98/3308 (3.0) | ||||
| Colonoscopies per year | |||||||
| <250 | 154/4266 (3.6) | <.001 | 91/2104 (4.3) | .51 | 63/2161 (2.9) | <.001 | |
| ≥250 | 587/11 823 (5.0) | 264/5645 (4.7) | 323/6178 (5.2) | ||||
| Practice setting | |||||||
| Academic | 703/14 868 (4.7) | .01 | 330/7127 (4.6) | .49 | 373/7740 (4.8) | .003 | |
| Private practice | 38/1221 (3.1) | 25/622 (4.0) | 13/599 (2.2) | ||||
| Presence of trainee | |||||||
| No | 686/14 828 (4.6) | .67 | 327/7142 (4.6) | .97 | 359/7685 (4.7) | .53 | |
| Yes | 55/1261 (4.4) | 28/607 (4.6) | 27/654 (4.1) | ||||
Abbreviation: NA, not applicable.
The coefficient of variation between clinicians for ADR was 26 and for pSSPDR was 77. Bivariate logistic regression analysis (not adjusting for random effects of individual endoscopist or other fixed effects) was used to demonstrate these differences in the form of odds ratios along with 95% CI (Table 5). On bivariate analysis, surgeons had lower ADRs than gastroenterologists (OR, 0.66 [95% CI, 0.61-0.71]; P < .001) and advanced endoscopists (OR, 0.61 [95% CI, 0.54-0.69]; P < .001). Female endoscopists had higher ADRs relative to male endoscopists (OR, 1.13 [95% CI, 1.05-1.22]; P = .002), and ADRs were lower for clinicians with more years since training was completed (OR, 0.86 [95% CI, 0.83-0.89] per 10-year increment; P < .001).
Table 5. Endoscopist Characteristics and Polyp Detectiona.
| Variable | Adenoma Detection Rate | Proximal Sessile Serrate Polyp Detection Rate | ||
|---|---|---|---|---|
| Odds Ratio (95% CI) | P Value | Odds Ratio (95% CI) | P Value | |
| Bivariate Analysis | ||||
| Specialty | ||||
| Surgeon vs gastroenterologist | 0.66 (0.61-0.71) | <.001 | 0.39 (0.32-0.48) | <.001 |
| Advanced endoscopist vs gastroenterologist | 1.09 (0.97-1.23) | .13 | 1.08 (0.85-1.36) | .53 |
| Surgeon vs advanced endoscopist | 0.61 (0.54-0.69) | <.001 | 0.37 (0.28-0.49) | <.001 |
| Female endoscopist | 1.13 (1.05-1.22) | .002 | 1.46 (1.24-1.71) | <.001 |
| International medical graduate | 1.05 (0.98-1.14) | .18 | 1.12 (0.95-1.32) | .19 |
| Years since training, per 10-y increment | 0.86 (0.83-0.89) | <.001 | 0.67 (0.61-0.72) | <.001 |
| Colonoscopies per year, per increment of 50 procedures/year | 1.00 (0.99-1.01) | .47 | 1.07 (1.05-1.09) | <.001 |
| Privately employed physician | 0.88 (0.78-1.00) | .06 | 0.65 (0.46-0.90) | .01 |
| Presence of trainee | 1.09 (0.96-1.23) | .19 | 0.94 (0.71-1.25) | .67 |
| Multivariable Analysis | ||||
| Specialty | ||||
| Surgeon vs gastroenterologist | 1.22 (0.84-1.76) | .29 | 0.75 (0.54-1.03) | .08 |
| Advanced endoscopist vs gastroenterologist | 0.99 (0.6-1.65) | .98 | 1.16 (0.78-1.73) | .47 |
| Surgeon vs advanced endoscopist | 1.23 (0.70-2.14) | .47 | 0.64 (0.41-1.01) | .06 |
| Female endoscopist | 1.05 (0.71-1.53) | .82 | 1.14 (0.84-1.55) | .40 |
| International medical graduate | 0.92 (0.64-1.34) | .68 | 1.15 (0.85-1.55) | .38 |
| Years since training, per 10-y increment | 1.06 (0.91-1.25) | .45 | 0.85 (0.74-0.98) | .02 |
| Colonoscopies per year, per increment of 50 procedures/year | 1.00 (0.95-1.05) | .90 | 1.05 (1.01-1.09) | .02 |
| Privately employed physician | 0.84 (0.51-1.38) | .48 | 0.84 (0.53-1.34) | .47 |
| Presence of trainee | 1.14 (0.96-1.34) | .14 | 0.90 (0.66-1.22) | .49 |
This Table does not adjust for random effects of individual endoscopist or other fixed effects. A multivariable analysis adjusting for random effects of the endoscopist and fixed effects is in eAppendix 1 in the Supplement.
On bivariate analysis, surgeons had lower pSSPDRs than gastroenterologists (OR, 0.39 [95% CI, 0.32-0.48]; P < .001) and advanced endoscopists (OR, 0.37 [95% CI, 0.28-0.49]; P < .001). Female endoscopists had higher pSSPDRs relative to male endoscopists (OR, 1.46 [95% CI, 1.24-1.71]; P < .001), and pSSPDRs were lower for clinicians with more years since training was completed (OR, 0.67 [95% CI, 0.61-0.72] per 10-year increment; P < .001) and higher for those who completed more colonoscopies per year (OR, 1.07 [95% CI, 1.05-1.09]; P < .001). Privately employed physicians also had significantly different ORs than public employee physicians (OR, 0.65 [95% CI, 0.46-0.90]; P = .01).
Multivariable Analysis—Endoscopist Characteristics
After adjusting for patient and endoscopist characteristics in the multilevel mixed-effects logistic regression analysis (eAppendix 1 in the Supplement), no significant differences in ADR were noted based on colonoscopist specialty, sex, location of medical school (international vs within the United States), practice setting, or the presence of a trainee during the examination. For pSSPDR, years in practice (per 10-year increment) was associated with reduced odds of detection (OR, 0.85 [95% CI, 0.74-0.98]; P = .02), and number of colonoscopies performed per year (in increments of 50) were associated with increased detection (OR, 1.05 [95% CI, 1.01-1.09]; P = .02). The multivariable analysis is presented in Table 5. Variables with greater than 5% of data missing were not included in the multivariable model. As such, none of the examined variables were missing more than 5% of data.
Discussion
In this large, retrospective analysis, we found that after adjustment for patient-level and procedure-associated factors, ADR was not associated with any of the examined endoscopist characteristics. The time elapsed since the completion of training and the annual volume of colonoscopies performed were the only significant factors associated with pSSPDR. Although previous studies noted differences in ADR and pSSPDR based on endoscopist characteristics,11,12,13,14,15 it is likely that residual confounding from patient demographic, clinical, and colonoscopic factors known to be associated with the presence of polyps rather than endoscopist characteristics explain these differences. In fact, using these data to replicate previous analyses, which did not account for potential confounders, we demonstrated an association between endoscopist characteristics and polyp detection.
The average ADR in this study was 31.3%, which is above the minimum standards suggested by American Society for Gastrointestinal Endoscopy–American College of Gastroenterology quality task force guidelines.4 The mean pSSPDR was 4.6%, which was approximately similar to the results from other contemporary studies.8,16 Consistent with a previous study,11 we noted a greater variation in pSSPDR (coefficient of variation, 77) than in ADR (26). However, the coefficient of variation in this study was significantly lower than in the previous study (ADR, 53; pSSPDR, 163). This is may be because of the attempt to reduce sampling bias by excluding endoscopists who had performed less than 100 colonoscopies during the study period.
Multiple studies have examined differences in the quality metrics of colonoscopy performance between gastroenterologists and surgeons. After negative colonoscopies performed by surgeons, patients had a greater incidence of colorectal cancer and associated deaths compared with those who received negative results from gastroenterologists.17,18 Ko et al19 demonstrated a 20% reduction in odds of polyp detection and a 30% to 40% reduction in odds of polypectomy in colonoscopies performed by surgeons compared with gastroenterologists. Further, multiple studies12,13,20 have noted 14% to 89% increased odds of adjusted ADR in gastroenterologists when compared with nongastroenterologists. In this study, surgeons had an 8.8% lower ADR than gastroenterologists, consistent with the magnitude of difference noted in previous studies. Indeed, after adjusting for similar factors as previous studies using these data, we were able to replicate the findings of previous studies12,13 that had shown higher odds of adenoma detection in gastroenterologists compared with other endoscopists. However, after adjusting for additional patient demographic and clinical characteristics (eg, comorbidities, medication use, body mass index, tobacco use), as well as colonoscopic factors (eg, location where the procedure was performed, time of day and month of procedure, withdrawal time, preparation adequacy) that play a role in the prevalence of polyps (eAppendix 1 in the Supplement), we demonstrated that the differences in ADR between specialties became nonsignificant.21,22,23 Therefore, it is likely that residual confounding effects rather than true differences between specialists’ abilities were responsible for the discordance noted in ADRs between gastroenterologists and surgeons. Further well-adjusted studies are likely needed to verify if the differences between colorectal cancer incidence and mortality based on endoscopist specialty17,18 are also associated with residual confounding effects.
A limited number of studies have evaluated differences in pSSPDR between specialists. A recent study demonstrated that gastroenterologists have a greater pSSPDR than surgeons.14 We found that gastroenterologists’ pSSPDR was 3.3% greater than that of surgeons. After replicating the analysis by Parikh et al14 with these data, surgeons were noted to have decreased odds of proximal SSP detection. In this study, after adjusting for fixed effects alone, the differences in pSSPDR by specialty remain significant. However, after adjusting for random effects of the endoscopists, the difference in pSSPDR by specialty is no longer significant. This suggests that random effects of individual endoscopists may be at least partially responsible for the difference in pSSPDR noted by endoscopist specialty on bivariate analysis.
Whereas Butterly et al24 showed a decreased ADR among female endoscopists, it is likely that this is owing to confounding; there is a tendency for female patients to seek female endoscopists.25,26 A more recent study has in fact noted a greater ADR among female clinicians.12 In this study, although unadjusted ADR and pSSPDR were greater in female clinicians, after adjusting for multiple characteristics, these differences were no longer significant. Also, we noted that the differences in pSSPDR reached significance when we did not adjust for random effects of individual endoscopists. Therefore, the random effects of individual endoscopists may be at least partially responsible for the difference in pSSPDRs noted by specialty on bivariate analysis.
Bias against the selection of international medical graduates for residency training, especially among surgical residencies has been demonstrated.27 However, multiple studies have confirmed that the care provided by international medical graduates is noninferior to the care provided by US medical school graduates.28,29 Adding to the aforementioned studies demonstrating noninferiority of care provided by international medical graduate colleagues, this study demonstrated that ADRs and pSSPDRs are not significantly different between international and US medical graduates.
Mehrotra et al12 noted greater ADRs among more recently graduated trainees. Indeed, replicating the analysis performed by Mehrotra et al12 using these data, we noted a similar trend, and the initial unadjusted analysis corroborated these findings. However, after adjusting for multiple characteristics, the association of years since training with ADR was mitigated. Yet, as noted by Crockett et al15 and confirmed by this study, the number of years since completion of training was negatively associated with pSSPDRs. A few factors may have contributed to this association. First, because the serrated polyp neoplasia pathway was more recently identified, it is possible that recent graduates had greater emphasis in their training on SSP detection compared with clinicians who trained in previous decades, when the importance of serrated lesions of the colon were either not discovered or well known by either gastroenterologist or pathologists.30,31,32
Consistent with previous analyses, we found the annual volume of colonoscopy was not associated with ADR.33 Because the differences in ADR by annual colonoscopy volume and years since training become significant when we do not adjust for random effects of individual endoscopists, it is likely that the random effects of individual endoscopists may be at least partially responsible for the differences in ADRs seen on bivariate analysis. Also, this study demonstrated colonoscopy volume is positively associated with pSSPDR. Although no prior research examined the association between pSSPDRs and colonoscopy volume, this finding is analogous to surgical literature linking high procedural volume with improved outcomes.34 These data suggest educational interventions to detect SSPs in endoscopists who are more distant from their training or performing fewer colonoscopies may be warranted, because SSPs are less common and more difficult to detect than conventional adenomas.
To our knowledge, this is one of the largest single-center studies (both in number of colonoscopies evaluated and number of endoscopists included) assessing endoscopist characteristics as they associate with ADRs and pSSPDRs. We evaluated of a broad variety of endoscopist characteristics after adjusting for numerous patient, clinical, and colonoscopic variables known to be associated with ADRs and pSSPDRs. Since we used these data and were able to replicate some of the results on endoscopist factors and polyp detection noted from previous studies that were mitigated or attenuated in the adjusted analysis, we feel we have established that it is likely the lack of controlling for residual confounding in other studies that was responsible for endoscopist-based differences seen in prior studies. Further, exclusion of clinicians who performed less than 100 colonoscopies—a factor that was not always assessed in previous studies—reduced sampling error when calculating ADRs and pSSPDRs for individual endoscopists.
Limitations
Certain limitations need to be considered in interpreting these findings, owing to the retrospective nature of this study. As with any such analysis, these data are prone to misreporting and omissions. Although the natural language processing algorithm was assessed to be more than 99% accurate on included measures, we were not able through this method to measure factors such as level of water immersion, air insufflation, retroflexion, patient positioning, use of antispasmodic medications, and other advanced technologies that may be associated with ADRs or pSSPDRs.35 Another potential limitation of this study stems from the data derivation from a single health system, which may not adequately reflect the performance characteristics of all endoscopists in the United States. However, the endoscopists included in this study are by no means homogenous. We included endoscopists of various backgrounds who are practicing divergent practice settings and different stages of their career.
Conclusions
Multiple studies have demonstrated differences in polyp detection based on endoscopist characteristics. In this large, single-center study, we examined differences in ADRs and pSSPDRs for a wide range of endoscopist characteristics. We found differences in ADRs and pSSPDRs based on endoscopist characteristics were no longer significant on adjusting for multiple confounding factors. Further, using these data, we were able to replicate the results from analyses performed in previous studies with similar conclusions, showing differences in polyp detection based on endoscopist specialty, sex, and years elapsed since completion of training. However, after adjusting for additional patient and colonoscopy characteristics, most of these differences in ADRs and pSSPDRs between clinicians disappeared (except differences in pSSPDRs by years elapsed since training and annual volume of colonoscopies). This suggests that residual confounding rather than endoscopist characteristics themselves may be responsible for differences in ADRs or pSSPDRs found in other studies.
eAppendix 1. Adjustments for patients and procedural factors.
eAppendix 2. Analyses performed in prior studies.
eFigure 1. Adenoma detection rate (ADR) in all included endoscopists.
eFigure 2. Proximal sessile serrated polyp detection rate (pSSPDR) in all included endoscopists.
References
- 1.Lorenzoni L, Belloni A, Sassi F. Health-care expenditure and health policy in the USA versus other high-spending OECD countries. Lancet. 2014;384(9937):83-92. doi: 10.1016/S0140-6736(14)60571-7 [DOI] [PubMed] [Google Scholar]
- 2.Chen W, Okunade A, Lubiani GG. Quality-quantity decomposition of income elasticity of U.S. hospital care expenditure using state-level panel data. Health Econ. 2014;23(11):1340-1352. doi: 10.1002/hec.2986 [DOI] [PubMed] [Google Scholar]
- 3.Centers for Medicare and Medicaid Services What are the value-based programs? https://www.cms.gov/Medicare/Quality-Initiatives-Patient-Assessment-Instruments/Value-Based-Programs/Value-Based-Programs.html. Published July 25, 2018. Accessed March 18, 2019.
- 4.Cohen J, Pike IM. Defining and measuring quality in endoscopy. Gastrointest Endosc. 2015;81(1):1-2. doi: 10.1016/j.gie.2014.07.052 [DOI] [PubMed] [Google Scholar]
- 5.van Rijn JC, Reitsma JB, Stoker J, Bossuyt PM, van Deventer SJ, Dekker E. Polyp miss rate determined by tandem colonoscopy: a systematic review. Am J Gastroenterol. 2006;101(2):343-350. doi: 10.1111/j.1572-0241.2006.00390.x [DOI] [PubMed] [Google Scholar]
- 6.Pohl H, Robertson DJ. Colorectal cancers detected after colonoscopy frequently result from missed lesions. Clin Gastroenterol Hepatol. 2010;8(10):858-864. doi: 10.1016/j.cgh.2010.06.028 [DOI] [PubMed] [Google Scholar]
- 7.Imperiale TF, Glowinski EA, Juliar BE, Azzouz F, Ransohoff DF. Variation in polyp detection rates at screening colonoscopy. Gastrointest Endosc. 2009;69(7):1288-1295. doi: 10.1016/j.gie.2007.11.043 [DOI] [PubMed] [Google Scholar]
- 8.Hetzel JT, Huang CS, Coukos JA, et al. Variation in the detection of serrated polyps in an average risk colorectal cancer screening cohort. Am J Gastroenterol. 2010;105(12):2656-2664. doi: 10.1038/ajg.2010.315 [DOI] [PubMed] [Google Scholar]
- 9.Kahi CJ, Hewett DG, Norton DL, Eckert GJ, Rex DK. Prevalence and variable detection of proximal colon serrated polyps during screening colonoscopy. Clin Gastroenterol Hepatol. 2011;9(1):42-46. doi: 10.1016/j.cgh.2010.09.013 [DOI] [PubMed] [Google Scholar]
- 10.Arain MA, Sawhney M, Sheikh S, et al. CIMP status of interval colon cancers: another piece to the puzzle. Am J Gastroenterol. 2010;105(5):1189-1195. doi: 10.1038/ajg.2009.699 [DOI] [PubMed] [Google Scholar]
- 11.Sanaka MR, Gohel T, Podugu A, et al. Adenoma and sessile serrated polyp detection rates: variation by patient sex and colonic segment but not specialty of the endoscopist. Dis Colon Rectum. 2014;57(9):1113-1119. doi: 10.1097/DCR.0000000000000183 [DOI] [PubMed] [Google Scholar]
- 12.Mehrotra A, Morris M, Gourevitch RA, et al. Physician characteristics associated with higher adenoma detection rate. Gastrointest Endosc. 2018;87(3):778-786.e5. doi: 10.1016/j.gie.2017.08.023 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Zorzi M, Senore C, Da Re F, et al. ; Equipe Working Group . Quality of colonoscopy in an organised colorectal cancer screening programme with immunochemical faecal occult blood test: the EQuIPE study (Evaluating Quality Indicators of the Performance of Endoscopy). Gut. 2015;64(9):1389-1396. doi: 10.1136/gutjnl-2014-307954 [DOI] [PubMed] [Google Scholar]
- 14.Parikh MP, Muthukuru S, Jobanputra Y, et al. Proximal sessile serrated adenomas are more prevalent in Caucasians, and gastroenterologists are better than nongastroenterologists at their detection. Gastroenterol Res Pract. 2017;2017:6710931. doi: 10.1155/2017/6710931 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Crockett SD, Gourevitch RA, Morris M, et al. Endoscopist factors that influence serrated polyp detection: a multicenter study. Endoscopy. 2018;50(10):984-992. doi: 10.1055/a-0597-1740 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Payne SR, Church TR, Wandell M, et al. Endoscopic detection of proximal serrated lesions and pathologic identification of sessile serrated adenomas/polyps vary on the basis of center. Clin Gastroenterol Hepatol. 2014;12(7):1119-1126. doi: 10.1016/j.cgh.2013.11.034 [DOI] [PubMed] [Google Scholar]
- 17.Rabeneck L, Paszat LF, Saskin R. Endoscopist specialty is associated with incident colorectal cancer after a negative colonoscopy. Clin Gastroenterol Hepatol. 2010;8(3):275-279. doi: 10.1016/j.cgh.2009.10.022 [DOI] [PubMed] [Google Scholar]
- 18.Baxter NN, Warren JL, Barrett MJ, Stukel TA, Doria-Rose VP. Association between colonoscopy and colorectal cancer mortality in a US cohort according to site of cancer and colonoscopist specialty. J Clin Oncol. 2012;30(21):2664-2669. doi: 10.1200/JCO.2011.40.4772 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Ko CW, Dominitz JA, Green P, Kreuter W, Baldwin LM. Specialty differences in polyp detection, removal, and biopsy during colonoscopy. Am J Med. 2010;123(6):528-535. doi: 10.1016/j.amjmed.2010.01.016 [DOI] [PubMed] [Google Scholar]
- 20.Leyden JE, Doherty GA, Hanley A, et al. Quality of colonoscopy performance among gastroenterology and surgical trainees: a need for common training standards for all trainees? Endoscopy. 2011;43(11):935-940. doi: 10.1055/s-0030-1256633 [DOI] [PubMed] [Google Scholar]
- 21.Lee TJ, Rees CJ, Blanks RG, et al. Colonoscopic factors associated with adenoma detection in a national colorectal cancer screening program. Endoscopy. 2014;46(3):203-211. doi: 10.1055/s-0033-1358831 [DOI] [PubMed] [Google Scholar]
- 22.Adams MA, Saleh A, Rubenstein JH. A systematic review of factors associated with utilization of monitored anesthesia care for gastrointestinal endoscopy. Gastroenterol Hepatol (N Y). 2016;12(6):361-370. [PMC free article] [PubMed] [Google Scholar]
- 23.Haque TR, Bradshaw PT, Crockett SD. Risk factors for serrated polyps of the colorectum. Dig Dis Sci. 2014;59(12):2874-2889. doi: 10.1007/s10620-014-3277-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Butterly L, Robinson CM, Anderson JC, et al. Serrated and adenomatous polyp detection increases with longer withdrawal time: results from the New Hampshire Colonoscopy Registry. Am J Gastroenterol. 2014;109(3):417-426. doi: 10.1038/ajg.2013.442 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Butterly L, Robinson CM, Anderson JC, et al. Serrated and adenomatous polyp detection increases with longer withdrawal time: results from the New Hampshire Colonoscopy Registry. Am J Gastroenterol. 2014;109(3):417-426. doi: 10.1038/ajg.2013.442 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Gerson LB, Twomey K, Hecht G, et al. Does gender affect career satisfaction and advancement in gastroenterology? results of an AGA institute-sponsored survey. Gastroenterology. 2007;132(4):1598-1606. doi: 10.1053/j.gastro.2007.02.045 [DOI] [PubMed] [Google Scholar]
- 27.Desbiens NA, Vidaillet HJ Jr. Discrimination against international medical graduates in the United States residency program selection process. BMC Med Educ. 2010;10:5. doi: 10.1186/1472-6920-10-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Tsugawa Y, Jena AB, Orav EJ, Jha AK. Quality of care delivered by general internists in US hospitals who graduated from foreign versus US medical schools: observational study. BMJ. 2017;356:j273. doi: 10.1136/bmj.j273 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Norcini JJ, Boulet JR, Dauphinee WD, Opalek A, Krantz ID, Anderson ST. Evaluating the quality of care provided by graduates of international medical schools. Health Aff (Millwood). 2010;29(8):1461-1468. doi: 10.1377/hlthaff.2009.0222 [DOI] [PubMed] [Google Scholar]
- 30.Torlakovic E, Snover DC. Serrated adenomatous polyposis in humans. Gastroenterology. 1996;110(3):748-755. doi: 10.1053/gast.1996.v110.pm8608884 [DOI] [PubMed] [Google Scholar]
- 31.Hawkins NJ, Ward RL. Sporadic colorectal cancers with microsatellite instability and their possible origin in hyperplastic polyps and serrated adenomas. J Natl Cancer Inst. 2001;93(17):1307-1313. doi: 10.1093/jnci/93.17.1307 [DOI] [PubMed] [Google Scholar]
- 32.Huang CS, O’Brien MJ, Yang S, Farraye FA. Hyperplastic polyps, serrated adenomas, and the serrated polyp neoplasia pathway. Am J Gastroenterol. 2004;99(11):2242-2255. doi: 10.1111/j.1572-0241.2004.40131.x [DOI] [PubMed] [Google Scholar]
- 33.Jover R, Zapater P, Bujanda L, et al. ; COLONPREV Study Investigators . Endoscopist characteristics that influence the quality of colonoscopy. Endoscopy. 2016;48(3):241-247. doi: 10.1055/s-0042-100185 [DOI] [PubMed] [Google Scholar]
- 34.Birkmeyer JD, Stukel TA, Siewers AE, Goodney PP, Wennberg DE, Lucas FL. Surgeon volume and operative mortality in the United States. N Engl J Med. 2003;349(22):2117-2127. doi: 10.1056/NEJMsa035205 [DOI] [PubMed] [Google Scholar]
- 35.Cadoni S, Falt P, Sanna S, et al. Impact of colonoscopy insertion techniques on adenoma detection. Dig Dis Sci. 2016;61(7):2068-2075. doi: 10.1007/s10620-016-4053-1 [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
eAppendix 1. Adjustments for patients and procedural factors.
eAppendix 2. Analyses performed in prior studies.
eFigure 1. Adenoma detection rate (ADR) in all included endoscopists.
eFigure 2. Proximal sessile serrated polyp detection rate (pSSPDR) in all included endoscopists.