Who is the best colonoscopist?

Evolving Role and Future Directions of Natural Language Processing in Gastroenterology

In line with the current trajectory of healthcare reform, significant emphasis has been placed on improving the utilization of data collected during a clinical encounter. Although the structured fields of electronic health records have provided a convenient foundation on which to begin such efforts, it was well understood that a substantial portion of relevant information is confined in the free-text narratives documenting care. Unfortunately, extracting meaningful information from such narratives is a non-trivial task, traditionally requiring significant manual effort. Today, computational approaches from a field known as Natural Language Processing (NLP) are poised to make a transformational impact in the analysis and utilization of these documents across healthcare practice and research, particularly in procedure-heavy sub-disciplines such as gastroenterology (GI). As such, this manuscript provides a clinically focused review of NLP systems in GI practice. It begins with a detailed synopsis around the state of NLP techniques, presenting state-of-the-art methods and typical use cases in both clinical settings and across other domains. Next, it will present a robust literature review around current applications of NLP within four prominent areas of gastroenterology including endoscopy, inflammatory bowel disease, pancreaticobiliary, and liver diseases. Finally, it concludes with a discussion of open problems and future opportunities of this technology in the field of gastroenterology and health care as a whole.

Detecting Deficient Coverage in Colonoscopies

Colonoscopy is tool of choice for preventing Colorectal Cancer, by detecting and removing polyps before they become cancerous. However, colonoscopy is hampered by the fact that endoscopists routinely miss 22-28% of polyps. While some of these missed polyps appear in the endoscopist's field of view, others are missed simply because of substandard coverage of the procedure, i.e. not all of the colon is seen. This paper attempts to rectify the problem of substandard coverage in colonoscopy through the introduction of the C2D2 (Colonoscopy Coverage Deficiency via Depth) algorithm which detects deficient coverage, and can thereby alert the endoscopist to revisit a given area. More specifically, C2D2 consists of two separate algorithms: the first performs depth estimation of the colon given an ordinary RGB video stream; while the second computes coverage given these depth estimates. Rather than compute coverage for the entire colon, our algorithm computes coverage locally, on a segment-by-segment basis; C2D2 can then indicate in real-time whether a particular area of the colon has suffered from deficient coverage, and if so the endoscopist can return to that area. Our coverage algorithm is the first such algorithm to be evaluated in a large-scale way; while our depth estimation technique is the first calibration-free unsupervised method applied to colonoscopies. The C2D2 algorithm achieves state of the art results in the detection of deficient coverage. On synthetic sequences with ground truth, it is 2.4 times more accurate than human experts; while on real sequences, C2D2 achieves a 93.0% agreement with experts.

Association Between Endoscopist Personality and Rate of Adenoma Detection

BACKGROUND & AIMS

There is significant variation among endoscopists in their adenoma detection rates (ADRs). We explored associations between ADR and characteristics of endoscopists, including personality traits and financial incentives.

METHODS

We collected electronic health record data from October 2013 through September 2015 and calculated ADRs for physicians from 4 health systems. ADRs were risk-adjusted for differences in patient populations. Physicians were surveyed to assess financial motivations, knowledge and perceptions about colonoscopy quality, and personality traits. Of 140 physicians sent the survey, 117 responded.

RESULTS

The median risk-adjusted ADR for all surveyed physicians was 29.3% (interquartile range, 24.1%-35.5%). We found no significant association between ADR and financial incentives, malpractice concerns, or physicians' perceptions of ADR as a quality metric. ADR was associated with the degree of self-reported compulsiveness relative to peers: among endoscopists who described themselves as much more compulsive, the ADR was 33.1%; among those who described themselves as somewhat more compulsive, the ADR was 32.9%; among those who described themselves as about the same as others, the ADR was 26.4%; and among those who described themselves as somewhat less compulsive, the ADR was 27.3%) (P = .0019). ADR was also associated with perceived thoroughness (much more thorough than peers, ADR = 31.5%; somewhat more, 31.9%; same/somewhat less, 27.1%; P = .0173). Physicians who reported feeling rushed, having difficulty pacing themselves, or having difficulty in accomplishing goals had higher ADRs. A secondary analysis found the same associations between personality and adenomas per colonoscopy.

CONCLUSIONS

We found no significant association between ADR and financial incentives, malpractice concerns, or perceptions of ADR as a quality metric. However, ADRs were higher among physicians who described themselves as more compulsive or thorough, and among those who reported feeling rushed or having difficulty accomplishing goals.

Physician characteristics associated with higher adenoma detection rate

BACKGROUND AND AIMS

Patients who receive a colonoscopy from a physician with a low adenoma detection rate (ADR) are at higher risk of subsequent colorectal cancer. It is unclear what drives the variation across physicians in ADR. We describe physician characteristics associated with higher ADR.

METHODS

In this retrospective cohort study a natural language processing system was used to analyze all outpatient colonoscopy examinations and their associated pathology reports from October 2013 to September 2015 for adults age 40 years and older across physicians from 4 diverse health systems. Physician performance on ADR was risk adjusted for differences in patient population and procedure indication. Our sample included 201 physicians performing at least 30 colonoscopy examinations during the study period, totaling 104,618 colonoscopy examinations.

RESULTS

The mean ADR was 33.2% (range, 6.3%-58.7%). Higher ADR was seen among female physicians (4.2 percentage points higher than men, P = .020), gastroenterologists (9.4 percentage points higher than nongastroenterologists, P < .001), and physicians with ≤9 years since their residency completion (6.0 percentage points higher than physicians who have had 27-51 years of practice, P = .004).

CONCLUSIONS

Gastroenterologists, female physicians, and more recently trained physicians had higher performance in adenoma detection.

Assessment of colorectal polyp recognition skill: development and validation of an objective test

BACKGROUND

The quality of colonoscopy is known to vary. The extent to which colonoscopists can recognize the presence of subtle colorectal lesions by visually distinguishing them from the surrounding mucosa (i.e., polyp recognition skill) may be one of several attributes that influence polyp detection rates. The aim of the present study was to develop and validate the first objective test of polyp recognition skill.

METHODS

Validation study. Twenty-eight experienced colonoscopists and eighty novices took a preliminary 280-item computer-based polyp recognition test. Items were genuine endoscopic images which participants assessed for the presence of "likely polyps." Half included clinically identified polyps. Participants clicked on a suspected lesion or a button marked "no likely polyp", and the main outcome measures were accuracy and response latency. The best items were selected for the final 50-item test.

RESULTS

In the preliminary test, experienced colonoscopists correctly identified more polyps than novices (P < .0001) and better discriminated between clinically identified polyps and non-polyp features (as measured by d', P < .0001). For polyp items, the experienced group also responded faster (P < .01). Effect sizes were large for accuracy (Cohen's d = 3.22) and d' (Cohen's d = 3.22). The 50 final test items produced comparable results for accuracy, d', and response latency. For both versions of the test, score scale reliability was high for both polyp and non-polyp items (α = .82 to .97).

CONCLUSIONS

The observed experienced-novice differences support the construct validity of the performance measures derived from the tests, indicating that polyp recognition skill can be quantified objectively. The final test may potentially be used to assess trainees, but test sensitivity may be insufficient to make fine-grained distinctions between different skill levels among experienced colonoscopists. More sensitive future tests may provide a valuable supplement to clinical detection rates, allowing objective comparisons between skilled colonoscopists.

Applying a natural language processing tool to electronic health records to assess performance on colonoscopy quality measures

BACKGROUND

Gastroenterology specialty societies have advocated that providers routinely assess their performance on colonoscopy quality measures. Such routine measurement has been hampered by the costs and time required to manually review colonoscopy and pathology reports. Natural language processing (NLP) is a field of computer science in which programs are trained to extract relevant information from text reports in an automated fashion.

OBJECTIVE

To demonstrate the efficiency and potential of NLP-based colonoscopy quality measurement.

DESIGN

In a cross-sectional study design, we used a previously validated NLP program to analyze colonoscopy reports and associated pathology notes. The resulting data were used to generate provider performance on colonoscopy quality measures.

SETTING

Nine hospitals in the University of Pittsburgh Medical Center health care system.

PATIENTS

Study sample consisted of the 24,157 colonoscopy reports and associated pathology reports from 2008 to 2009.

MAIN OUTCOME MEASUREMENTS

Provider performance on 7 quality measures.

RESULTS

Performance on the colonoscopy quality measures was generally poor, and there was a wide range of performance. For example, across hospitals, the adequacy of preparation was noted overall in only 45.7% of procedures (range 14.6%-86.1% across 9 hospitals), cecal landmarks were documented in 62.7% of procedures (range 11.6%-90.0%), and the adenoma detection rate was 25.2% (range 14.9%-33.9%).

LIMITATIONS

Our quality assessment was limited to a single health care system in western Pennsylvania.

CONCLUSIONS

Our study illustrates how NLP can mine free-text data in electronic records to measure and report on the quality of care. Even within a single academic hospital system, there is considerable variation in the performance on colonoscopy quality measures, demonstrating the need for better methods to regularly and efficiently assess quality.

Gastronet survey on the use of one- or two-person technique for colonoscopy insertion

BACKGROUND

Usually, colonoscopy insertion is performed by the colonoscopist (one-person technique). Quite common in the early days of endoscopy, the assisting nurse is now only rarely doing the insertion (two-person technique). Using the Norwegian national endoscopy quality assurance (QA) programme, Gastronet, we wanted to explore the extent of two-person technique practice and look into possible differences in performance and QA output measures.

METHODS

100 colonoscopists in 18 colonoscopy centres having reported their colonoscopies to Gastronet between January and December 2009 were asked if they practiced one- or two-person technique during insertion of the colonoscope. They were categorized accordingly for comparative analyses of QA indicators.

RESULTS

75 endoscopists responded to the survey (representing 9368 colonoscopies) - 62 of them (83%) applied one-person technique and 13 (17%) two-person technique. Patients age and sex distributions and indications for colonoscopy were also similar in the two groups. Caecal intubation was 96% in the two-person group compared to 92% in the one-person group (p < 0.001). Pain reports were similar in the groups, but time to the caecum was shorter and the use of sedation less in the two-person group.

CONCLUSION

Two-person technique for colonoscope insertion was practiced by a considerable minority of endoscopists (17%). QA indicators were either similar to or better than one-person technique. This suggests that there may be some beneficial elements to this technique worth exploring and try to import into the much preferred one-person insertion technique.

Deep sedation compared with moderate sedation in polyp detection during colonoscopy: a randomized controlled trial

AIM

The detection rate of adenomas is one of the current quality indicators in high-quality colonoscopy. We compared the performance of colonoscopy for the detection of polyps in patients sedated with deep and moderate sedation. Secondary objectives included the patient's and the endoscopist's satisfaction, recovery time and the adverse events related to sedation between the two groups.

METHOD

Five hundred and twenty patients submitted for colonoscopy were prospectively randomized into a deep sedation group (DS group, n = 258) and a moderate sedation (MS group, n = 262) group. In both, sedation and analgesia were performed using midazolam with pethidine.

RESULTS

There were no differences between the two groups in the following three areas: (1) The overall detection of polyps (DS, 1 [0-20]; MS, 1 [0-15]; P = 0.67), (2): polyp size ≥ 10 mm (DS, 0.00 [0-7]; MS, 0.00 [0-6]; P = 0.30), and (3) polyp size < 10 mm (DS, 0.00 [0-20]; MS, 0.00 [0-13]; P = 0.83). There was no significant difference in the rate of adenoma detection (DS, 1 [0-10]; MS, 1 [0-9]; P = 0.99).

CONCLUSION

The study demonstrated no difference in the detection of polyps by colonoscopy using deep or moderate sedation.

Association between visual gaze patterns and adenoma detection rate during colonoscopy: a preliminary investigation

OBJECTIVES

Eye gaze tracking (EGT) technology follows a person's gaze and records the resulting visual gaze pattern (VGP). Adenoma detection rate (ADR) is a validated measure of colonoscopy quality. Higher ADRs are associated with prolonged withdrawal times and other endoscopic maneuvers that allow a better visualization of the mucosa; however, the influence of VGP has yet to be explored. We aim to quantify the VGP for endoscopists observing colonoscopy videos and describe the association between VGP and ADR. Furthermore, we will evaluate the relationship between VGP and the endoscopists' years of experience.

METHODS

Eleven endoscopists watched three videos while their VGP was recorded. The videos corresponded to 3 min of three different colonoscopy withdrawals. We divided the screen into a 3 × 3 grid of nine segments: eight peripheral and one central. We compared percent of gaze time (GT) in the central vs. peripheral segments using a paired t-test. VGP with ADR and years of practice were evaluated using Pearson's test.

RESULTS

Subjects spent more GT in the screen's central segment (65 vs. 33%, P<0.001). ADR was significantly associated with increased percentage of central GT (r = 0.67, P = 0.024) and increased mean GT in the central segment (r = 0.70, P = 0.017). There was negative correlation between endoscopists' years of practice and the percentage of central GT (r = -0.67, P = 0.025), but no correlation between years of practice and percentage of peripheral GT (r = 0.24, P = 0.47).

CONCLUSIONS

This reveals an association between a centrally focused VGP and ADR. Future steps include confirming in a larger sample and exploring if VGP can retrain low ADR endoscopists to perform higher quality colonoscopies.

Efficacy and effectiveness of colonoscopy: how do we bridge the gap?

Colonoscopy is sometimes considered the preferred colorectal cancer screening modality, yet this modality may be subject to variation in operator performance more than any other screening test. Failures of colonoscopy to consistently detect precancerous lesions threaten the effectiveness of this technique for the prevention of colorectal cancer. Studies on high-level adenoma detectors under optimal conditions have begun to establish the true efficacy of colonoscopy and further widen the gap between efficacy and effectiveness. Research is required to establish the component skills, attitudes, and behaviors for high-level mucosal inspection competence necessary for training and assessment. Interventions to bridge the gap between efficacy and effectiveness are lacking, yet they should emphasize quality measurement and operate at various levels within the health system to motivate change in endoscopist behavior.

Cap-fitted colonoscopy: a randomized, tandem colonoscopy study of adenoma miss rates

BACKGROUND

Failures of adenoma detection diminish the effectiveness of colonoscopy.

OBJECTIVE

This study investigated the impact of cap-fitted colonoscopy (CFC) on the adenoma miss rate at colonoscopy.

DESIGN

Randomized, tandem colonoscopy study.

SETTING

University hospital.

PATIENTS

This study involved patients undergoing elective screening or surveillance colonoscopy.

INTERVENTION

Patients were randomized to undergo cap-fitted (n = 52) or regular, high-definition (n = 48) colonoscopy before undergoing a second colonoscopy by the alternate method. During CFC, a plastic cap or hood was attached to the tip of the colonoscope, which was used to flatten haustral folds and improve mucosal exposure.

MAIN OUTCOME MEASUREMENTS

The primary outcome measure was the miss rate for adenomas between patients who underwent CFC first and patients who underwent regular colonoscopy first.

RESULTS

A total of 238 adenomas were detected in 67 patients (67%), with a combined overall miss rate of 27.7%, comprising 66 missed adenomas in 38 patients. Patients undergoing initial CFC had a significantly lower miss rate for all adenomas compared with that of patients undergoing regular colonoscopy (21% vs 33%, P = .039). Miss rates with CFC were significantly lower for adenomas of ≤5 mm (22% vs 35%; P = .037). There was no significant difference in per-patient miss rates between the initial CFC group (51%, n = 18) and the initial regular colonoscopy group (63%, n = 20, P = .36).

LIMITATIONS

Single-center study with two endoscopists.

CONCLUSION

CFC reduces miss rates for all adenomas and specifically for small adenomas. (

CLINICAL TRIAL REGISTRATION NUMBER

NCT00577083).

Hood-assisted colonoscopy is more effective in detection of colorectal adenomas than narrow-band imaging

BACKGROUND & AIMS

Colonoscopy, using either a transparent retractable extension device or narrow band imaging, is thought to improve colon adenoma detection. We compared the abilities of a transparent retractable extension device and narrow band imaging to detect colorectal adenomas.

METHODS

One hundred and seven patients with colonic adenomas that were detected by traditional colonoscopy were randomly assigned to groups that underwent a second colonoscopy that used either a transparent retractable extension or narrow band imaging; adenomas were removed. The principal outcome parameters were the number, size, shape, and location of adenomas detected. The patients' demographic characteristics, indications for colonoscopy, and cecal intubation times were similar between groups.

RESULTS

Use of the transparent retractable extension resulted in detection of 31% more adenomas than the initial procedure (P < .0001). The majority of newly discovered adenomas were sessile (79%; 26/33) and less than 5 mm in size (73%; 24/33). There was no significant increase in adenoma detection (5%) between first colonoscopy and second colonoscopy using narrow band imaging. Additional adenomas were found in 40.7% of patients that were examined using the transparent hood (22/54) versus 13.2% of those examined using narrow band imaging (7/53) (P = .0028).

CONCLUSIONS

Colonoscopy with a transparent retractable extension significantly improved the adenoma detection rate compared with repeat colonoscopy using narrow band imaging.

Characterization of forces applied by endoscopists during colonoscopy by using a wireless colonoscopy force monitor

BACKGROUND

To perform a colonoscopy, the endoscopist maneuvers the colonoscope through a series of loops by applying force to the insertion tube. Colonoscopy insertion techniques are operator dependent but have never been comprehensively quantified.

OBJECTIVE

To determine whether the Colonoscopy Force Monitor (CFM), a device that continually measures force applied to the insertion tube, can identify different force application patterns among experienced endoscopists.

DESIGN

Observational study of 6 experienced endoscopists performing routine diagnostic and therapeutic colonoscopy in 30 patients.

SETTING

Outpatient ambulatory endoscopy center.

PATIENTS

Adult male and female patients between 30 and 75 years of age undergoing routine colonoscopy.

INTERVENTIONS

CFM monitoring of force applied to the colonoscope insertion tube during colonoscopy.

MAIN OUTCOME MEASUREMENTS

Maximum and mean linear and torque force, time derivative of force, combined linear and torque vector force, and total manipulation time.

RESULTS

The CFM demonstrates differences among endoscopists for maximum and average push/pull and mean torque forces, time derivatives of force, combined push/torque force vector, and total manipulation time. Endoscopists could be grouped by force application patterns.

LIMITATIONS

Only experienced endoscopists using conscious sedation in the patients were studied. Sample size was 30 patients.

CONCLUSIONS

This study demonstrates that CFM allows continuous force monitoring, characterization, and display of similarities and differences in endoscopic technique. CFM has the potential to facilitate training by enabling trainees to assess, compare, and quantify their techniques and progress.

Flat adenomas of the large bowel: a single endoscopist study

PURPOSE

Flat colorectal adenomas may not be recognized or they may be seen and dismissed as normal mucosa. As a result, the "true" incidence of flat lesions is unknown. The aim of this study was to report the incidence and character of flat adenomas in a large series of patients in the United States.

METHODS

Details of colorectal polyps seen during colonoscopy are entered into a prospective database that was queried for flat adenomas. No magnification or dye spray was used. The colonoscopy completion rate was 96.8 percent. The flat adenoma detection rate of the senior author is in line with that of Japanese colonoscopists.

RESULTS

During 2,659 colonoscopies, 5,749 colorectal lesions were seen in 2,003 patients. Of these patients, 1,125 were men and 878 were women, with a mean age of 65 years (+/- 12 years). There were 3,115 adenomas (54 percent of lesions); 315 were flat (10.1 percent), 23 (0.8 percent) were depressed, 2,575 were sessile (82.7 percent), and 200 were pedunculated (6.4 percent). Two hundred twelve patients had flat adenomas (10.6 percent of patients); 172 patients had one and 40 had multiple (2 to 8) flat adenomas. Eight patients with flat adenomas had more flat adenomas on follow-up.

CONCLUSIONS

Colonoscopists in the United States must be aware of the existence of flat adenomas and sensitive to their appearance.

Time trends in colon cancer incidence and distribution and lower gastrointestinal endoscopy utilization in Manitoba

OBJECTIVES

There are limited data on recent trends in subsite-specific colon cancer incidence and utilization of lower gastrointestinal endoscopy from Canada. The aim of our study was to determine the concomitant trends in right-sided colon cancer incidence and utilization of colonoscopy and flexible sigmoidoscopy (FS) in Manitoba.

METHODS

Cases of colon cancer diagnosed from 1964 to 2004 were identified from the Manitoba Cancer Registry. Lower gastrointestinal endoscopies performed between 1984 and 2003 were identified from Manitoba Health's Physician Claims database. Trends of age-standardized incidence rates were determined using Joinpoint analyses.

RESULTS

Rates of right-sided colon cancer showed a monotonic increase in both sexes (annual percent change [APC] in both sexes 1.04%, P < 0.001). The most rapid increase (200%) occurred in individuals of 70 yr of age and older. While rates of colonoscopies without polypectomies quadrupled between 1985 (257 per 100,000) and 2003 (1,083 per 100,000, APC 8.89%, P < 0.001), rates of colonoscopies with polypectomies quadrupled from 35 per 100,000 in 1985 to 140 per 100,000 in 2000, and then increased more rapidly in the subsequent 4 yr (233 per 100,000 in 2003, APC 20%, P < 0.001). Rates of FS declined between 1999 (342 per 100,000) and 2003 (257 per 100,000, APC -6.68%, P= 0.01).

CONCLUSIONS

The rates of right-sided colon cancer are continuing to increase in Manitoba, with the most rapid increase occurring in older individuals. Reasons for the increasing incidence of right-sided colon cancer despite increasing rates of colonoscopy need to be further explored, and may reflect increased detection of asymptomatic cancers or a real increase in right-sided colon cancer incidence.

Adenoma detection rate and the quality of colonoscopy: the sword has two edges

With the increasing demand for colonoscopy for colorectal cancer screening and surveillance of colorectal adenomas and carcinomas, there has been an increasing emphasis on the quality of colonoscopy. Few publications have considered the full implications of pushing the adenoma detection rate as a prime quality indicator. In this article, some of the potential problems with this course are discussed and an alternative approach is suggested. Although defining and measuring quality in colonoscopy is important, further work needs to be done to arrive at a practical, clinically meaningful way of quality assessment.

A program to enhance completion of screening colonoscopy among urban minorities

BACKGROUND & AIMS

Although colonoscopy is becoming the preferred screening test for colorectal cancer, screening rates, particularly among minorities, are low. Little is known about the uptake of screening colonoscopy or the factors that predict colonoscopy completion among minorities. This study investigated the use of patient navigation within an open-access referral system and its effects on colonoscopy completion rates among urban minorities.

METHODS

This was a cohort study that took place at a teaching hospital in New York. Participants were mostly African Americans and Hispanics directly referred for screening colonoscopy by primary care clinics from November 2003 to May 2006. Once referred, a bilingual Hispanic female patient navigator facilitated the colonoscopy completion. Completion rates, demographic factors associated with completing colonoscopy, endoscopic findings, and patient satisfaction were analyzed.

RESULTS

Of 1169 referrals, 688 patients qualified for and 532 underwent navigation. Two thirds (66%) of navigated patients completed screening colonoscopies, 16% had adenomas, and only 5% had inadequate bowel preps. Women were 1.31 times more likely to complete the colonoscopy than men (P = .014). Hispanics were 1.67 times more likely to complete the colonoscopy than African Americans (P = .013). Hispanic women were 1.50 times more likely to complete the colonoscopy than Hispanic men (P = .009). Patient satisfaction was 98% overall, with 66% reporting that they definitely or probably would not have completed their colonoscopy without navigation.

CONCLUSIONS

By using a patient navigator, the majority of urban minorities successfully completed their colonoscopies, clinically significant pathology was detected, and patient satisfaction was enhanced. This approach may help increase adherence with screening colonoscopy efforts in other clinical settings.

Screening for colorectal cancer

Although there are several methods available for colon cancer screening, none is optimal. This article reviews methods for screening, including fecal occult blood tests, flexible sigmoidoscopy, colonoscopy, CT colonography, capsule endoscopy, and double contrast barium enema. A simple, inexpensive, noninvasive, and relatively sensitive screening test is needed to identify people at risk for developing advanced adenomas or colorectal cancer who would benefit from colonoscopy. It is hoped that new markers will be identified that perform better. Until then we fortunately have a variety of screening strategies that do work.