Evaluating colonoscopy quality by performing provider type

Abstract

Background

Colorectal cancer is the third most diagnosed cancer and the second leading cause of cancer death in the United States. Colonoscopy is an essential tool for screening, used as a primary approach and follow-up to an abnormal stool-based colorectal cancer screening result. Colonoscopy quality is often measured with 4 key indicators: bowel preparation, cecal intubation, mean withdrawal time, and adenoma detection. Colonoscopies are most often performed by gastroenterologists (GI), however, in rural and medically underserved areas, non-GI providers often perform colonoscopies. This study aims to evaluate the quality and safety of screening colonoscopies performed by non-GI practitioner, comparing their outcomes with those of GI providers.

Methods

Descriptive statistics were used to characterize the study population. Results for quality indicators were stratified by provider type and compared. Statistical significance was determined using a P value of less than .05 as the threshold for all comparisons; all P values were 2-sided.

Results

No statistical difference was found when comparing performance by provider type. Median performance for gastroenterologists, general surgeons, and family medicine providers ranged from 98% to 100% for cecal intubation; 97.4% to 100% for bowel preparation; 57.4% to 88.9% for male adenoma detection rate; 47.7% to 62.13% for female adenoma detection rate; and 0:12:10 to 0:20:16 for mean withdrawal time. All provider types met and exceeded the goal metric for each of the quality indicators (P < .001).

Conclusions

As a result of this analysis, we can expect non-GI practitioner to perform colonoscopies with similar quality to GI practitioner given the performance outcomes for the key quality metrics.

Colorectal cancer (CRC) is the third most diagnosed cancer and the second leading cause of cancer deaths in the United States (1). Screening for colorectal cancer can reduce incidence by detecting precancerous polyps and mortality by identifying cancer at earlier stages where it is more easily treated (1). Colonoscopy is an essential tool for screening, used both as a primary approach and follow-up to an abnormal stool-based CRC screening result. The quality of a colonoscopy can impact patient outcomes, which led the Quality Assurance Task Group of the National Colorectal Cancer Roundtable to develop a standardized colonoscopy reporting and data system in 2007, based on the key indicators for continuous quality improvement identified by the Multi-Society Task Force on Colorectal Cancer in 2002 (2,3). These indicators were selected to serve as a quality improvement instrument, providing the referring physician with a colonoscopy report illustrating the quality of the examination along with specific recommendations for follow-up.

Access to high-quality colonoscopies is associated with reduced CRC incidence and mortality. Historically, studies found gastroenterologists (GI) provided higher-quality colonoscopies than non-gastroenterologist specialties (such as general surgery, family medicine, or internal medicine), but these studies were primarily focused in urban areas or academic medical institutions (4). A recent systematic review and meta-analysis on the association of specialty and colonoscopy quality reviewed 36 studies, and authors found lower rates of adenoma detection associated with surgeons in screening colonoscopies when comparing gastroenterologists with surgeons (odds ratio [OR] = 0.85, 95% confidence interval [CI] = 0.78 to 0.94) and nonsurgeon endoscopists (OR = 0.91, 95% CI = 0.87 to 0.96). There was no difference in cecal intubation rates found between gastroenterologists and surgeons in screening colonoscopies (OR = 1.04, 95% CI = 0.89 to 1.22) or among other endoscopists and GIs (OR = 0.88, 95% CI = 0.64 to 1.26). This study reported considerable heterogeneity in most studies, resulting in low overall certainty for the evidence (I² = 85%-97.5%) (5). Additionally, only one-quarter of studies included were completed in the last 5 years. Many rural and underserved areas do not have access to gastrointestinal specialists. Instead, colonoscopies in these regions are often performed by family physicians, general surgeons, or advanced practice professionals including nurse practitioners and physician assistants, and quality data published for these specialties are limited (6,7).

This study combines colonoscopy quality data from 2 high-volume programs serving rural populations, with GI and non-GI specialists performing colonoscopies. Key quality measures include adenoma detection rates for male and female patients, cecal intubation rate, rate of adequate bowel preparation, and mean withdrawal time. This study also aims to add data noted as missing in the recent systematic review and meta-analysis, including bowel preparation quality at the per-patient level stratified by endoscopist specialty.

Methods

Study population and setting

The study combines data from colonoscopies completed through programs offered by the University of Texas (UT) Southwestern Moncrief Cancer Institute (MCI) and University of North Dakota (UND) Center for Family Medicine. Specifically, MCI’s Colorectal Cancer Screening and Navigation (CSPAN) program is a multiphase project offering a mailed fecal immunochemical test to facilitate colorectal cancer screening across north and central Texas. The program was originally implemented through the Tarrant County safety-net system and John Peter Smith Health Network and has been described in detail previously (8-12). It has since transitioned to an open system, serving rural and medically underserved Texans across a 67-county service region. Participants were either identified by referrals from local health-care networks, community clinics, federally qualified health centers, and primary care clinics or can self-enroll through community outreach events, health fairs, or a dedicated 800 number. Patients with an abnormal fecal immunochemical test result were contacted by phone and mail to arrange a diagnostic colonoscopy with a partnering provider within their community. Performing providers include gastroenterologists, general surgeons, and family medicine physicians, as well as midlevel providers who completed colonoscopies at local hospitals or surgical centers within the service area. All clinical services were provided at no cost to the patient. Of the near 20 000 patients who completed at least 1 fecal immunochemical test from November 1, 2013, to December 31, 2021, as part of 1 or more phases of the CSPAN Program, 12.4% (n = 2456) had a positive test with 68.3% (n = 1678) completing the necessary follow-up colonoscopy. All colonoscopies completed through the CSPAN program were follow-up to an abnormal fecal immunochemical test result.

The UND Center for Family Medicine offers a rural training track centralized in Bismarck with a rural training site in Hettinger. This program provides colonoscopy training to 1 resident per year through the residency clinic. All procedures are completed in the endoscopy suite located within the clinic, and a nurse anesthetist provides monitored anesthesia care sedation. As a result, no facility fee is charged as would be done for a hospital-based colonoscopy. In addition to the residency clinic, outreach clinics are provided at the Fort Yates Indian Health Service hospital, focusing on preoperative visits for colonoscopy. The residency clinic volume varies from 8 to 15 colonoscopies per month, resulting in 572 colonoscopies completed from November 5, 2009, to August 30, 2022. Colonoscopies completed through the UND program were a mixture of initial colon cancer screening and abnormal fecal immunochemical test follow-up; no diagnostic procedures investigating other symptoms of GI disease were included in the sample.

Definitions

For this study, we stratify colonoscopy quality measures by performing provider specialty, comparing GI performance with general surgeons, as well as family medicine physicians and midlevel providers. These include adenoma detection rate, cecal intubation rate, rate of adequate bowel preparation, and mean withdrawal time. Adenoma detection rate is the proportion of colonoscopies in which 1 or more adenomas are detected; the established performance target is at least 45% for males and at least 35% for females when colonoscopy is performed as follow-up to a positive fecal immunochemical test (13). The cecal intubation rate is the proportion of procedures where visualization of the cecum is documented by the notation of landmarks; the expected performance target is at least 90% (14). Rate of adequate bowel prep is the proportion of procedures performed where bowel preparation is deemed adequate to allow the use of recommended surveillance or screening intervals; the expected performance target is at least 85% for outpatient examinations (14). Mean withdrawal time is the average of withdrawal times documented for the performing provider with the performance target being at least 6 minutes in negative-result screening colonoscopies (14).

Analytic approach

Patient characteristics were described using race and ethnicity, age (derived from date of birth), and sex abstracted from the electronic medical record at UT Southwestern Moncrief Cancer Institute and UND Center for Family Medicine. Values are self-reported by patients on intake; “other” is a designated value in the electronic medical record for race.

Colonoscopy performance was evaluated for each provider. Metric data was then aggregated using median and interquartile range based on the discipline of the performing provider, as either family medicine, general surgeon, or GI. Family medicine was further subdivided by exam location to either Texas or North Dakota to eliminate any confounding effect. Performance within each specialty group was also stratified by number of colonoscopies completed using 25 as the cut point. We used Mood median test, a special case of Pearson χ² test, to evaluate whether the population medians could be defined as equal. Each provider group’s performance was also compared with the target performance specified for each metric. Statistical significance was determined using a P value of less than .05 as the threshold in all comparisons; all P values were 2-sided. These analyses were performed using SAS software, Version 9.4 (SAS Institute Inc, Cary, NC, USA).

Like many other aspects of colonoscopy performance, documentation has also been shown to vary, resulting in missing data within the performance metrics. Again, to minimize confounding, we performed a sensitivity analysis using complete case and imputed datasets. These datasets were created in R (version 4.3.1; R Foundation for Statistical Computing, Vienna, Austria), specifically using the mice package to create the imputed dataset through multiple imputation (15). All authors had access to the study data and reviewed and approved the final manuscript.

Ethical approval

Study approval was provided by the UT Southwestern Medical Center (STU-2022-0650) institutional review board. A copy of the approved protocol is available on request. The Cancer Prevention and Research Institute of Texas (PP120229, PP150061, PP200009) was the primary funding source for the CSPAN program, with additional project support provided by the National Institutes of Health (P30CA142543; PI: Arteaga, Study Lead: Argenbright).

Results

Colonoscopies were analyzed for 2170 individuals. Patient demographics are summarized in Table 1. The population was largely female (61.94%) but racially and ethnically diverse. Family medicine specialists performed 1169 (53%) colonoscopies, GIs performed 822 (37.9%) colonoscopies, and general surgeons performed 179 (8.25%) colonoscopies. All colonoscopies by general surgeons and GIs were performed in Texas, and colonoscopies performed by family medicine providers were evenly split between Texas (n = 590) and North Dakota (n = 572).

Table 1.

Patient demographics for colonoscopy completers (n = 2170)

Demographics	No. (%)
Median age (IQR)a	58.68 (8.66)
Provider type
Family medicine	1169 (53.87%)
Gastroenterology	822 (37.88%)
General surgery	179 (8.25%)
Patient sex
Female	1344 (61.94%)
Male	824 (37.97%)
Unknown	2 (0.09%)
Patient race and ethnicity
American Indian and Alaska Native	169 (7.79%)
Black	316 (14.56%)
Hispanic and Latino	491 (22.63%)
Other	64 (2.95%)
White	995 (45.85%)
Unknown	135 (6.22%)

a

IQR = interquartile range.

Missingness varied somewhat by metric and provider type, yet GIs and family medicine providers had similar proportions of missing data, ranging from nearly 1% (adenoma detection and cecal intubation; P = .9207) to approximately 10% (withdrawal time; P = .835), while general surgeons were consistently almost 3 times that, ranging from nearly 4% (adenoma detection and cecal intubation) to 27.9% (withdrawal time). Regardless of which dataset was considered, performance remained consistent for each provider specialty for each quality metric (Table 2). Also, certain provider groups were more heavily influenced by the number of colonoscopies performed by a participating provider but only for certain metrics. Adequate bowel preparation rates were significantly different for the general surgeon and GI groups (P = .04 and P = .01, respectively), while cecal intubation rate was significantly different only for the GI group (P < .001). No statistical difference was found in the comparison for mean withdrawal time or either male or female adenoma detection rates, and no statistical difference was found among the family medicine group, regardless of the performance metric. As a result, the findings presented are based on the original dataset, including those providers who had completed 25 or more colonoscopies.

Table 2.

Sensitivity analysis for colonoscopy metrics by provider type^a

	Adenoma detection rate, female		Adenoma detection rate, male		Cecal intubation		Adequate bowel preparation		Withdrawal time, h:min:sec
	Median (IQR)	P	Median (IQR)	P	Median (IQR)	P	Median (IQR)	P	Median (IQR)	P
Moncrief FM
Original data	47.7% (5.09%)	.3114	65.8% (9.45%)	.3114	100% (1.44%)	.9252	100% (1.24%)	.3638	0:17:20 (0:01:39)	.747
Complete cases	51.8% (1.84%)		72.4% (4.59%)		98.9% (1.12%)		100% (0%)		0:15:40 (0:01:59)
Imputed data	47.7% (5.2%)		65.8% (8.82%)		100% (1.42%)		98.7% (1.35%)		0:14:48 (0:01:58)
North Dakota FMb
Original data	62.13% (—)	.3679	57.38% (—)	.3679	98.95% (—)	.3679	97.37% (—)	.3679	0:20:16 (—)	.3679
Complete cases	62.55% (—)		58.13% (—)		98.92% (—)		97.31% (—)		0:20:16 (—)
Imputed data	62.13% (—)		57.38% (—)		98.82% (—)		97.20% (—)		0:19:94 (—)
Moncrief general surgeon
Original data	50.0% (8.33%)	.589	88.9% (25.0%)	.589	98.2% (7.69%)	.5134	100% (0.86%)	.5134	0:12:10 (0:00:56)	.5134
Complete cases	32.0% (0%)		50% (0%)		100% (0%)		100% (0%)		0:11:40 (0:00:00)
Imputed data	50.0% (8.33%)		88.9% (25.0%)		98.3% (7.69%)		100% (0%)		0:11:42 (0:00:53)
Moncrief gastroenterologist
Original data	50.0% (9.64%)	.5594	61.1% (15.1%)	.899	98.0% (1.96%)	.1559	100% (1.32%)	.2418	0:12:36 (0:03:25)	.8703
Complete cases	56.6% (12.3%)		61.5% (17.6%)		100% (1.36%)		100% (0%)		0:11:54 (0:02:53)
Imputed data	52.9% (11.3%)		63.2% (16.1%)		98.0% (1.92%)		100% (0%)		0:13:10 (0:04:01)

a

Median estimates and IQR of each performance metric by provider type and data source, along with comparative P value. IQR = interquartile range.

b

Note colonoscopies in North Dakota were performed by a single provider, as such the values included reflect actual performance rather than a median estimate, and IQRs are not reported (signified with “—”).

Adenoma detection rate exceeded the target performance rates for both male and female patients. Values ranged from 47.7% to 62.13% for female patients and 61.1% to 88.9% for male patients. No statistical difference was found when comparing performance by provider type (family medicine vs general surgeon vs GI) for female adenoma detection rate (P = .77) or male adenoma detection rate (P = .80). Cecal intubation rates were also similar by provider type, with no statistical difference when comparing performance by provider type (P = .57). Adequate bowel preparation rates ranged from 97.37% to 100%, again with no statistical difference in performance by provider type (P = .92). Mean withdrawal time showed the greatest variation by provider type with the lowest times for GI and general surgeon providers at 0:12:10 and 0:12:36, respectively, and the highest for family medicine providers at 0:17:20  for MCI and 0:20:16 for North Dakota. Despite this range, no statistical difference was found in comparison (P = .06).

Figure 1 illustrates metric performance by provider specialty, along with the established threshold. For each quality metric, we found median performance to be above the target threshold for all providers (P < .001).

Figure 1.

Colonoscopy metric performance by program and provider type for (A) cecal intubation rate, (B) adequate bowel preparation rate, (C) adenoma detection rate for male patients, (D) adenoma detection rate for female patients, and (E) mean withdrawal time. The dashed line denotes the performance target associated with the given metric, and the asterisk indicates an outlying measurement. FM = family medicine; GI = gastroenterologist; GS = general surgeon.

Discussion

The National Colorectal Cancer Roundtable launched an initiative to increase the CRC screening rate to 80% in the eligible US population by 2018 (16). It has since been extended and expanded to encourage 80% screening in all populations to reduce disparities, yet many health-care systems, particularly in rural and medically underserved communities across the United States, do not have adequate capacity or resources to provide screening colonoscopies in pursuit of this goal (17,18). Recent changes to the national screening guidelines have further exacerbated the capacity issue by adding individuals aged 45-49 years to the eligible screening cohort (19). Although alternative screening approaches are often implemented to supplement CRC screening capacity, 1) colonoscopy still constitutes more than 60% of all CRC screenings and 2) alternatives like fecal immunochemical test or a combined fecal immunochemical test–multitarget DNA test only provide CRC prevention when positive tests are followed by colonoscopy (18,20). We instead look to primary care physicians to fill the gap, particularly in these resource-strained communities. However, data detailing the safety and efficacy of non-GI providers performing screening colonoscopies are limited. In a 2009 meta-analysis, 12 papers published between 1992 and 2006 were summarized to define colonoscopy quality based on 18 292 colonoscopies performed by 73 colonoscopists (17). In this paper, colonoscopies performed by primary care providers were reported to be safe and effective based on certain quality indicators published for specialists performing the procedure (17), yet a key limitation of their study was the underrepresentation of colonoscopies performed by other primary care specialties (17). We address this limitation in our study as we evaluate colonoscopy quality data from 2 large CRC screening programs for 2170 colonoscopies performed from 2009 to 2021 by nearly 90 unique providers, including GIs, general surgeons, and family medicine providers. Some variation in performance is expected, however, effective colonoscopy requires high-quality bowel preparation, safe and complete colonoscope insertion to the cecum, and sufficient withdrawal time, as it is highly correlated with greater adenoma detection. As a result, we not only expand on the works considered previously in the meta-analysis by evaluating performance using these 4 quality measures but also in our direct comparison to the defined thresholds and gastroenterologist performance.

Adenoma detection rate is a patient-based measure defined as the proportion of colonoscopies where at least 1 conventional adenoma is detected and verified by pathology (18). When colonoscopy is performed in follow-up to an abnormal fecal immunochemical test, the target for adenoma detection rates is at least 45% in the male population and 35% in the female population (13). This does not reflect the maximum rates of detection, but just below the mean prevalence observed in available combined studies (13,14,18). We found performance to be similar when stratifying by provider specialty, with rates of 45% in all provider groups regardless of patient sex. This is worth noting as postcolonoscopy CRC reduction has been shown to progressively increase with adenoma detection rates more than 30% (18).

Cecal intubation is characterized by visualization of the ileocecal valve, and terminal ileum if intubated, so that the appendiceal orifice and entire medial wall of the cecum are evident for inspection (14,18). This should be achieved in at least 90% of colonoscopies, as poor cecal intubation rates are associated with a greater risk for postcolonoscopy CRC (14,18). We found no difference in our comparison (P = .57), with all provider specialties exceeding the prespecified threshold ranging from 95.93% to 98.95%. Cecal intubation rate is considered a priority quality indicator, yet there is some debate regarding its clinical relevance, particularly when considered over time (18). Specifically, high levels of performance are expected to remain stable or increase over time but not decrease, suggesting intermittent checks could be sufficient to document adequate performance for this metric (18).

Bowel prep efficacy is associated with improved lesion detection and cost effectiveness, reducing the need for repeat procedures (18). Although the performance target is for at least 85% of outpatient examinations to have adequate bowel preparation, multiple studies suggest individual centers may have rates of inadequate bowel preparation during colonoscopy as high as 25% (14,18). Median bowel prep rates ranged from 96.0% to 98.3% for included providers, well exceeding the quality threshold. A difference in performance was not detected when comparing by performing provider.

Short withdrawal times are considered an indicator of poor technique for colonoscopists with an adenoma detection rate below the threshold (18). Specifically, 6 minutes of withdrawal time in a normal colon produces a reasonable separation when comparing high and low adenoma detection rate performers (14,18). Although all providers exceeded the standard, results varied more by provider type yet did not reach statistical significance (P = .06). GIs and general surgeons had the lowest times at 0:12:36 and 0:12:10 , respectively, while the MCI-family medicine providers had mean withdrawal time of 0:17:20 and North Dakota family medicine providers had an average of 0:20:16. Colonoscopies by North Dakota family physicians often involved a resident, which can increase withdrawal time, as a result there was greater variability by provider type.

Our results indicate family medicine physicians and general surgeons are performing safe and effective colonoscopies of comparable quality to gastroenterologists based on these key quality metrics. We anticipate the difference in our study as compared with the 2009 meta-analysis may be associated with our collection of quality data. Specifically, providers contracted to perform colonoscopies for the CSPAN program are required to submit quality metrics for each procedure as part of the criteria for reimbursement, while the UND program is often providing training to residents. In both programs, it is well understood that quality data will be reviewed and as such may improve provider performance. Moreover, if we consider only the CSPAN providers not meeting an established metric (adenoma detection rate female: n = 23; adenoma detection rate male: n = 13; cecal intubation rate: n = 4; bowel prep: n = 2; mean withdrawal time: n = 0), we find, apart from 1 GI provider, this sample is limited to those having performed at least 5 colonoscopies for the program. More specifically, nearly half (47.8%) had completed only 1 colonoscopy, suggesting these outcomes are not reflective of the provider’s true performance.

A key limitation to this study is the regional data. Though the sample is large and diverse, there may be some regional limitations for generalizing nationally. Also, general surgeons were likely underrepresented relative to gastroenterologists and family medicine providers. Similarly, our dataset does not include advanced practice providers, yet it is known they and other provider types may also be performing colonoscopies in rural and underresourced areas. Our findings highlight the need for additional research in the following areas: 1) defining training standards for non-GI endoscopists to ensure high-quality colonoscopy performance and 2) determining how increasing the number of endoscopists, regardless of specialty, might influence colonoscopy uptake in rural and resource-constrained communities.

Family medicine physicians and general surgeons can perform colonoscopies with similar performance outcomes to gastroenterologists for key quality metrics, having met and exceeded the target metrics. With the rapidly increasing need for colonoscopy, the time is ripe to consider expanding endoscopy capacity through the training and inclusion of other specialties.

Data availability

The data that support the findings of this study are available from UT Southwestern, but restrictions apply to the availability of these data, which were used under license for the current study, and so are not publicly available. Data are however available from the authors upon reasonable request and with permission of UT Southwestern.

Author contributions

Emily Berry, MSPH, PhD (Conceptualization; Data curation; Formal analysis; Funding acquisition; Project administration; Visualization; Writing—original draft; Writing—review & editing), Jeff Hostetter, MD, MS, FAAFP (Conceptualization; Data curation; Validation; Writing—review & editing), Joseph Bachtold, DO, MPH, FAAFP (Data curation; Writing—review & editing), Sarah Zamarripa, MPH (Data curation; Formal analysis; Visualization; Writing—original draft; Writing—review & editing), and Keith E Argenbright, MD, MMM (Supervision; Writing—review & editing).

Funding

This work was supported by The Cancer Prevention and Research Institute of Texas (PP120229, PP150061, PP200009, Argenbright—PI) and the National Institutes of Health (P30CA142543; PI: Arteaga, Study Lead: Argenbright).

Conflicts of interest

The authors declare no potential conflicts of interest.

Acknowledgements

The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; or decision to submit manuscript for publication.