Impact of withdrawal time on adenoma detection rate: results from a prospective multicenter trial.

BACKGROUND AND AIMS: Performing a high-quality colonoscopy is critical for optimizing the adenoma detection rate (ADR). Colonoscopy withdrawal time (a surrogate measure) of ≥6 minutes is recommended; however, a threshold of a high-quality withdrawal and its impact on ADR are not known. METHODS: We examined withdrawal time (excluding polyp resection and bowel cleaning time) of subjects undergoing screening and/or surveillance colonoscopy in a prospective, multicenter, randomized controlled trial. We examined the relationship of withdrawal time in 1-minute increments on ADR and reported odds ratio (OR) with 95% confidence intervals. Linear regression analysis was performed to assess the maximal inspection time threshold that impacts the ADR. RESULTS: A total of 1142 subjects (age, 62.3 ± 8.9 years; 80.5% men) underwent screening (45.9%) or surveillance (53.6%) colonoscopy. The screening group had a median withdrawal time of 9.0 minutes (interquartile range [IQR], 3.3) with an ADR of 49.6%, whereas the surveillance group had a median withdrawal time of 9.3 minutes (IQR, 4.3) with an ADR of 63.9%. ADR correspondingly increased for a withdrawal time of 6 minutes to 13 minutes, beyond which ADR did not increase (50.4% vs 76.6%, P < .01). For every 1-minute increase in withdrawal time, there was 6% higher odds of detecting an additional subject with an adenoma (OR, 1.06; 95% confidence interval, 1.02-1.10; P = .004). CONCLUSIONS: Results from this multicenter, randomized controlled trial underscore the importance of a high-quality examination and efforts required to achieve this with an incremental yield in ADR based on withdrawal time. (Clinical trial registration number: NCT03952611.).

High-Definition Colonoscopy Compared With Cuff- and Cap-Assisted Colonoscopy: Results From a Multicenter, Prospective, Randomized Controlled Trial.

BACKGROUND AND AIMS: Mucosal exposure devices including distal attachments such as the cuff and cap have shown variable results in improving adenoma detection rate (ADR) compared with high-definition white light colonoscopy (HDWLE). METHODS: We performed a prospective, multicenter randomized controlled trial in patients undergoing screening or surveillance colonoscopy comparing HDWLE to 2 different types of distal attachments: cuff (CF) (Endocuff Vision) or cap (CP) (Reveal). The primary outcome was ADR. Secondary outcomes included adenomas per colonoscopy, advanced adenoma and sessile serrated lesion detection rate, right-sided ADR, withdrawal time, and adverse events. Continuous variables were compared using Student's t test and categorical variables were compared using chi-square or Fisher's exact test using statistical software Stata version16. A P value <.05 was considered significant. RESULTS: A total of 1203 subjects were randomized to either HDWLE (n = 384; mean 62 years of age; 81.3% males), CF (n = 379; mean 62.7 years of age; 79.9% males) or CP (n = 379; mean age 62.1 years of age; 80.5% males). No significant differences were found among 3 groups for ADR (57.3%, 59.1%, and 55.7%; P = .6), adenomas per colonoscopy (1.4 ± 1.9, 1.6 ± 2.4, and 1.4 ± 2; P = .3), advanced adenoma (7.6%, 9.2%, and 8.2%; P = .7), sessile serrated lesion (6.8%, 6.3%, and 5.5%; P = .8), or right ADR (48.2%, 49.3%, and 46.2%; P = .7). The number of polyps per colonoscopy were significantly higher in the CF group compared with HDWLE and CP group (2.7 ± 3.4, 2.3 ± 2.5, and 2.2 ± 2.3; P = .013). In a multivariable model, after adjusting for age, sex, body mass index, withdrawal time, and Boston Bowel Preparation Scale score, there was no impact of device type on the primary outcome of ADR (P = .77). In screening patients, CF resulted in more neoplasms per colonoscopy (CF: 1.7 ± 2.6, HDWLE: 1.3 ± 1.7, and CP: 1.2 ± 1.8; P = .047) with a shorter withdrawal time. CONCLUSIONS: Results from this multicenter randomized controlled trial do not show any significant benefit of using either distal attachment devices (CF or CP) over HDWLE, at least in high-detector endoscopists. The Endocuff may have an advantage in the screening population. (ClinicalTrials.gov, Number: NCT03952611).

Tailoring Surveillance Colonoscopy in Patients With Advanced Adenomas.

BACKGROUND & AIMS: Patients with advanced colorectal adenomas (AAs) are directed to undergo intensive surveillance. However, the benefit derived from surveillance may be outweighed by the risk of death from non-colorectal cancer (CRC) causes, leading to uncertainty on how best to individualize follow-up. The aim of this study was to derive a risk prediction model and risk index that estimate and stratify the risk for non-CRC cancer mortality (NCM) subsequent to diagnosis and removal of AA. METHODS: We conducted a retrospective cohort study of veterans ≥40 years old who had colonoscopy for diagnostic or screening indications at 13 Veterans Affairs Medical Centers between 2002 and 2009 and had 1 or more AAs. The primary outcome was NCM using a fixed follow-up time period of 5 years. Logistic regression using the lasso technique was used to identify factors independently associated with NCM, and an index based on points from regression coefficients was constructed to estimate risk of 5-year NCM. RESULTS: We identified 2943 veterans with AA (mean age [standard deviation] 63 [8.6] years, 98% male, 74% white), with an overall 5-year mortality of 16.7%, which was nearly all due to NCM (16.6%). Age, comorbidity burden, specific comorbid conditions, and hospitalization within the preceding year were independently associated with NCM. The risk prediction model had a goodness of fit (calibration) P value of .41 and c-statistic (discrimination) of 0.74 (95% confidence interval, 0.71-0.76). On the basis of comparable 5-year risks of NCM, the scores comprised 3 risk categories: low (score of 0-1), intermediate (score of 2-4), and high (score of ≥5), in which NCM occurred in 6.5%, 14.1%, and 33.2%, respectively. CONCLUSIONS: We derived a risk prediction model that identifies veterans with advanced adenomas who are at high risk of NCM within 5 years, and who are thus unlikely to benefit from further surveillance.

Prevalence of Advanced Colorectal Neoplasia in Veterans: Effects of Age, Sex, and Race/Ethnicity.

GOAL: We sought to quantify the independent effects of age, sex, and race/ethnicity on risk of colorectal cancer (CRC) and advanced neoplasia (AN) in Veterans. STUDY: We conducted a retrospective, cross-sectional study of Veterans aged 40 to 80 years who had diagnostic or screening colonoscopy between 2002 and 2009 from 1 of 14 Veterans Affairs Medical Centers. Natural language processing identified the most advanced finding and location (proximal, distal). Logistic regression was used to examine the adjusted, independent effects of age, sex, and race, both overall and in screening and diagnostic subgroups. RESULTS: Among 90,598 Veterans [mean (SD) age 61.7 (9.4) y, 5.2% (n=4673) were women], CRC and AN prevalence was 1.3% (n=1171) and 8.9% (n=8081), respectively. Adjusted CRC risk was higher for diagnostic versus screening colonoscopy [odds ratio (OR)=3.79; 95% confidence interval (CI), 3.19-4.50], increased with age, was numerically (but not statistically) higher for men overall (OR=1.53; 95% CI, 0.97-2.39) and in the screening subgroup (OR=2.24; 95% CI, 0.71-7.05), and was higher overall for Blacks and Hispanics, but not in screening. AN prevalence increased with age, and was present in 9.2% of men and 3.9% of women [adjusted OR=1.90; 95% CI, 1.60-2.25]. AN risk was 11% higher in Blacks than in Whites overall (OR=1.11; 95% CI, 1.04-1.20), was no different in screening, and was lower in Hispanics (OR=0.74; 95% CI, 0.55-0.98). Women had more proximal CRC (63% vs. 39% for men; P=0.03), but there was no difference in proximal AN (38.3% for both genders). CONCLUSIONS: Age and race were associated with AN and CRC prevalence. Blacks had a higher overall prevalence of both CRC and AN, but not among screenings. Men had increased risk for AN, while women had a higher proportion of proximal CRC. These findings may be used to tailor when and how Veterans are screened for CRC.

Development, validation, and proof-of-concept implementation of a two-year risk prediction model for undiagnosed atrial fibrillation using common electronic health data (UNAFIED).

BACKGROUND: Many patients with atrial fibrillation (AF) remain undiagnosed despite availability of interventions to reduce stroke risk. Predictive models to date are limited by data requirements and theoretical usage. We aimed to develop a model for predicting the 2-year probability of AF diagnosis and implement it as proof-of-concept (POC) in a production electronic health record (EHR). METHODS: We used a nested case-control design using data from the Indiana Network for Patient Care. The development cohort came from 2016 to 2017 (outcome period) and 2014 to 2015 (baseline). A separate validation cohort used outcome and baseline periods shifted 2 years before respective development cohort times. Machine learning approaches were used to build predictive model. Patients ≥ 18 years, later restricted to age ≥ 40 years, with at least two encounters and no AF during baseline, were included. In the 6-week EHR prospective pilot, the model was silently implemented in the production system at a large safety-net urban hospital. Three new and two previous logistic regression models were evaluated using receiver-operating characteristics. Number, characteristics, and CHA2DS2-VASc scores of patients identified by the model in the pilot are presented. RESULTS: After restricting age to ≥ 40 years, 31,474 AF cases (mean age, 71.5 years; female 49%) and 22,078 controls (mean age, 59.5 years; female 61%) comprised the development cohort. A 10-variable model using age, acute heart disease, albumin, body mass index, chronic obstructive pulmonary disease, gender, heart failure, insurance, kidney disease, and shock yielded the best performance (C-statistic, 0.80 [95% CI 0.79-0.80]). The model performed well in the validation cohort (C-statistic, 0.81 [95% CI 0.8-0.81]). In the EHR pilot, 7916/22,272 (35.5%; mean age, 66 years; female 50%) were identified as higher risk for AF; 5582 (70%) had CHA2DS2-VASc score ≥ 2. CONCLUSIONS: Using variables commonly available in the EHR, we created a predictive model to identify 2-year risk of developing AF in those previously without diagnosed AF. Successful POC implementation of the model in an EHR provided a practical strategy to identify patients who may benefit from interventions to reduce their stroke risk.

Provider-specific quality measurement for ERCP using natural language processing.

BACKGROUND AND AIMS: Natural language processing (NLP) is an information retrieval technique that has been shown to accurately identify quality measures for colonoscopy. There are no systematic methods by which to track adherence to quality measures for ERCP, the highest risk endoscopic procedure widely used in practice. Our aim was to demonstrate the feasibility of using NLP to measure adherence to ERCP quality indicators across individual providers. METHODS: ERCPs performed by 6 providers at a single institution from 2006 to 2014 were identified. Quality measures were defined using society guidelines and from expert opinion, and then extracted using a combination of NLP and data mining (eg, ICD9-CM codes). Validation for each quality measure was performed by manual record review. Quality measures were grouped into preprocedure (5), intraprocedure (6), and postprocedure (2). NLP was evaluated using measures of precision and accuracy. RESULTS: A total of 23,674 ERCPs were analyzed (average patient age, 52.9 ± 17.8 years, 14,113 were women [59.6%]). Among 13 quality measures, precision of NLP ranged from 84% to 100% with intraprocedure measures having lower precision (84% for precut sphincterotomy). Accuracy of NLP ranged from 90% to 100% with intraprocedure measures having lower accuracy (90% for pancreatic stent placement). CONCLUSIONS: NLP in conjunction with data mining facilitates individualized tracking of ERCP providers for quality metrics without the need for manual medical record review. Incorporation of these tools across multiple centers may permit tracking of ERCP quality measures through national registries.

Informative Patterns of Health-Care Utilization Prior to the Diagnosis of Pancreatic Ductal Adenocarcinoma.

Early-detection tests for pancreatic ductal adenocarcinoma (PDAC) are needed. Since a hypothetical screening test would be applied during antecedent clinical encounters, we sought to define the variability in health-care utilization leading up to PDAC diagnosis. This was a retrospective cohort study that included patients diagnosed with PDAC in the Indianapolis, Indiana, area between 1999 and 2013 with at least 1 health-care encounter during the antecedent 36-month period (n = 1,023). Patients were classified by unique patterns of health-care utilization using a group-based trajectory model. The prevalences of PDAC signals, such as diabetes mellitus (DM) and chronic pancreatitis, were compared. Four distinct trajectories were identified, the most common (42.0%) being having few clinical encounters more than 6 months prior to PDAC diagnosis (late acceleration). In all cases, a minority of persons had DM (30.6%, with 9.5% <1.5 years before PDAC) or any pancreatic disorder (39.9%); these were least common in the late-acceleration group (DM, 14.7%; any pancreatic disorder, 32.1% (P < 0.001)). The most common pattern of antecedent care was having few clinical encounters until shortly before PDAC diagnosis. Since the majority of patients diagnosed with PDAC do not have an antecedent PDAC signal, early-detection strategies limited to these groups may not apply to the majority of cases.

Multi-center colonoscopy quality measurement utilizing natural language processing.

BACKGROUND: An accurate system for tracking of colonoscopy quality and surveillance intervals could improve the effectiveness and cost-effectiveness of colorectal cancer (CRC) screening and surveillance. The purpose of this study was to create and test such a system across multiple institutions utilizing natural language processing (NLP). METHODS: From 42,569 colonoscopies with pathology records from 13 centers, we randomly sampled 750 paired reports. We trained (n=250) and tested (n=500) an NLP-based program with 19 measurements that encompass colonoscopy quality measures and surveillance interval determination, using blinded, paired, annotated expert manual review as the reference standard. The remaining 41,819 nonannotated documents were processed through the NLP system without manual review to assess performance consistency. The primary outcome was system accuracy across the 19 measures. RESULTS: A total of 176 (23.5%) documents with 252 (1.8%) discrepant content points resulted from paired annotation. Error rate within the 500 test documents was 31.2% for NLP and 25.4% for the paired annotators (P=0.001). At the content point level within the test set, the error rate was 3.5% for NLP and 1.9% for the paired annotators (P=0.04). When eight vaguely worded documents were removed, 125 of 492 (25.4%) were incorrect by NLP and 104 of 492 (21.1%) by the initial annotator (P=0.07). Rates of pathologic findings calculated from NLP were similar to those calculated by annotation for the majority of measurements. Test set accuracy was 99.6% for CRC, 95% for advanced adenoma, 94.6% for nonadvanced adenoma, 99.8% for advanced sessile serrated polyps, 99.2% for nonadvanced sessile serrated polyps, 96.8% for large hyperplastic polyps, and 96.0% for small hyperplastic polyps. Lesion location showed high accuracy (87.0-99.8%). Accuracy for number of adenomas was 92%. CONCLUSIONS: NLP can accurately report adenoma detection rate and the components for determining guideline-adherent colonoscopy surveillance intervals across multiple sites that utilize different methods for reporting colonoscopy findings.

Current and future applications of natural language processing in the field of digestive diseases.

Natural language processing (NLP) is a technology that uses computer-based linguistics and artificial intelligence to identify and extract information from free-text data sources such as progress notes, procedure and pathology reports, and laboratory and radiologic test results. With the creation of large databases and the trajectory of health care reform, NLP holds the promise of enhancing the availability, quality, and utility of clinical information with the goal of improving documentation, quality, and efficiency of health care in the United States. To date, NLP has shown promise in automatically determining appropriate colonoscopy intervals and identifying cases of inflammatory bowel disease from electronic health records. The objectives of this review are to provide background on NLP and its associated terminology, to describe how NLP has been used thus far in the field of digestive diseases, and to identify its potential future uses.

Clinical decision support with natural language processing facilitates determination of colonoscopy surveillance intervals.

BACKGROUND & AIMS: With an increased emphasis on improving quality and decreasing costs, new tools are needed to improve adherence to evidence-based practices and guidelines in endoscopy. We investigated the ability of an automated system that uses natural language processing (NLP) and clinical decision support (CDS) to facilitate determination of colonoscopy surveillance intervals. METHODS: We performed a retrospective study at a single Veterans Administration medical center of patients age 40 years and older who had an index outpatient colonoscopy from 2002 through 2009 for any indication except surveillance of a previous colorectal neoplasia. We analyzed data from 10,798 reports, with 6379 linked to pathology results and 300 randomly selected reports. NLP-based CDS surveillance intervals were compared with those determined by paired, blinded, manual review. The primary outcome was adjusted agreement between manual review and the fully automated system. RESULTS: κ statistical analysis produced a value of 0.74 (P < .001) for agreement between the full text annotation and the NLP-based CDS system. Fifty-five reports (18.3%; 95% confidence interval, 14.1%-23.2%) differed between manual review and CDS recommendations. Of these, NLP error accounted for 30 (54.5%), incomplete resection of adenomatous tissue accounted for 14 (25.5%), and masses observed without biopsy findings of cancer accounted for 4 (7.2%). NLP-based CDS surveillance intervals had higher levels of agreement with the standard (81.7%) than the level agreement between experts (72% agreement between paired reviewers). CONCLUSIONS: A fully automated system that uses NLP and a guideline-based CSD system can accurately facilitate guideline-recommended adherence surveillance for colonoscopy.

Lower provider volume is associated with higher failure rates for endoscopic retrograde cholangiopancreatography.

BACKGROUND: Among physicians who perform endoscopic retrograde cholangiopancreatography (ERCP), the relationship between procedure volume and outcome is unknown. OBJECTIVE: Quantify the ERCP volume-outcome relationship by measuring provider-specific failure rates, hospitalization rates, and other quality measures. RESEARCH DESIGN: Retrospective cohort. SUBJECTS: A total of 16,968 ERCPs performed by 130 physicians between 2001 and 2011, identified in the Indiana Network for Patient Care. MEASURES: Physicians were classified by their average annual Indiana Network for Patient Care volume and stratified into low (<25/y) and high (≥25/y). Outcomes included failed procedures, defined as repeat ERCP, percutaneous transhepatic cholangiography or surgical exploration of the bile duct≤7 days after the index procedure, hospitalization rates, and 30-day mortality. RESULTS: Among 15,514 index ERCPs, there were 1163 (7.5%) failures; the failure rate was higher among low (9.5%) compared with high volume (5.7%) providers (P<0.001). A second ERCP within 7 days (a subgroup of failure rate) occurred more frequently when the original ERCP was performed by a low-volume (4.1%) versus a high-volume physician (2.3%, P=0.013). Patients were more frequently hospitalized within 24 hours when the ERCP was performed by a low-volume (28.3%) versus high-volume physician (14.8%, P=0.002). Mortality within 30 days was similar (low=1.9%, high=1.9%). Among low-volume physicians and after adjusting, the odds of having a failed procedure decreased 3.3% (95% confidence interval, 1.6%-5.0%, P<0.001) with each additional ERCP performed per year. CONCLUSIONS: Lower provider volume is associated with higher failure rate for ERCP, and greater need for postprocedure hospitalization.

Natural language processing accurately categorizes findings from colonoscopy and pathology reports.

BACKGROUND & AIMS: Little is known about the ability of natural language processing (NLP) to extract meaningful information from free-text gastroenterology reports for secondary use. METHODS: We randomly selected 500 linked colonoscopy and pathology reports from 10,798 nonsurveillance colonoscopies to train and test the NLP system. By using annotation by gastroenterologists as the reference standard, we assessed the accuracy of an open-source NLP engine that processed and extracted clinically relevant concepts. The primary outcome was the highest level of pathology. Secondary outcomes were location of the most advanced lesion, largest size of an adenoma removed, and number of adenomas removed. RESULTS: The NLP system identified the highest level of pathology with 98% accuracy, compared with triplicate annotation by gastroenterologists (the standard). Accuracy values for location, size, and number were 97%, 96%, and 84%, respectively. CONCLUSIONS: The NLP can extract specific meaningful concepts with 98% accuracy. It might be developed as a method to further quantify specific quality metrics.

Low yield of significant findings on endoscopic retrograde cholangiopancreatography in patients with pancreatobiliary pain and no objective findings.

BACKGROUND: Due to the challenging nature of the type III sphincter of the Oddi dysfunction (SOD) patient, the suspected low diagnostic yield from endoscopic retrograde cholangiopancreatography (ERCP), the high complication rate, and the potential for litigation it is surprising that diagnostic ERCP continues to be performed in this patient population. AIMS: The purpose of this study was to determine the incidence of significant findings on ERCP alone in patients with disabling abdominal pain of suspected pancreatobiliary origin and no objective findings. METHODS: Entry criteria of this study included: (1) ERCP with attempt at visualization of both the biliary tree and pancreatic duct, (2) suspected of having abdominal pain of pancreatobiliary origin, (3) biliary or pancreatic type III by the modified Geenen-Hogan classification, (4) never undergone sphincterotomy, (5) attempted manometry of both sphincters. A total of 265 patients met entry criteria. RESULTS: Significant findings were found in seven patients (2.6 %): choledococoele (1), anomalous pancreatobiliary ductal union (2), mild-moderate chronic pancreatitis (2), and pancreatic duct filling defect suspicious for IPMN (2). Potentially significant in 25 patients (9.4 %) were: equivocal chronic pancreatitis (1), incomplete (4) and complete pancreas divisum (20). SOD was diagnosed in 77.7 %. 11.3 % had undergone a previous diagnostic ERCP. CONCLUSION: ERCP in this high-risk population requires detailed informed consent, availability of SOM to increase the diagnostic yield, and skills in placing prophylactic pancreatic stents. It is our belief that patients without objective findings of pancreatobiliary disease that would explain their subjective complaints should not undergo diagnostic ERCP.

The effect of colonoscopy preparation quality on adenoma detection rates.

BACKGROUND: Colonoscopy reduces the risk of colorectal cancer mortality by removing precancerous adenomas. The detection rate of subcentimeter (<10 mm) polyps is lower for procedures with inadequate preparation quality. OBJECTIVE: To compare the adenoma detection rates of small (6-9 mm) and diminutive (≤ 5 mm) adenomas in patients with poor and fair quality preparations with those with adequate quality preparations. DESIGN: Cross-sectional study and multivariable, hierarchical model. SETTING: Roudebush Veterans Affairs Medical Center. PATIENTS: This study involved 8800 colonoscopies performed from 2001 to 2010. MAIN OUTCOME MEASUREMENTS: Preparation quality rating, polyp size, and polyp histology. RESULTS: Preparation quality was rated as fair in 2809 (31.9%) and poor in 829 (9.4%) colonoscopies. In patients with poor compared with adequate quality, the detection rate was lower for diminutive adenomas (odds ratio [OR] 0.57; 95% CI, 0.47-0.70) but not for small adenomas (OR 0.84; 95% CI, 0.65-1.07). There were no differences in the detection rate of diminutive (OR 1.08; 95% CI, 0.94-1.24]) or small (OR 1.09; 95% CI, 0.94-1.27) adenomas in patients with fair compared with adequate quality preparation. Detection of advanced histology in patients with poor preparation quality was lower than in those with adequate quality (P = .027; 3.3% vs 5.0%), but there was no difference in those with fair compared with adequate quality (P = .893; 4.9% vs 5.0%). LIMITATIONS: Single-center study; no standardization of preparation quality or size measurements. CONCLUSIONS: A fair preparation quality rating does not decrease the detection rate for adenomas of any size or for advanced histology, suggesting that fair quality may be considered adequate and that follow-up intervals may not need to be shortened. Poor preparation quality decreases the detection rate of diminutive adenomas and advanced histology, suggesting substandard colonoscopy performance.