Society: AGA

Background: The Nordic-European Initiative on Colorectal Cancer (NordICC) Study highlighted the impact of adherence on the effectiveness of colonoscopy at reducing CRC incidence and mortality. Within a large, randomized trial comparing colonoscopy to fecal immunochemical testing (FIT) for average risk screening, we previously reported that adherence with annual FIT was initially high (81.6% in year 1) and falls to 59-68% in subsequent years (Gastroenterology 2022;172.7:S-199).

Aim: To determine screening colonoscopy participation and predictors of adherence.

Methods: Eligible participants were age 50-75 years, due for CRC screening and willing to be randomized to colonoscopy vs. FIT at 46 Veterans Affairs medical centers during 2012-2017. Those with first degree relatives with CRC or signs or symptoms of CRC were excluded. Stepwise logistic regression, stratified by site, with p<0.1 for inclusion and p<0.05 criteria for remaining in the multivariable model was used to identify independent associations with colonoscopy adherence at 6- and 12-months.

Results: Among 25,065 colonoscopy-arm participants (mean age 59.1 years), 93.2% were male, 69.8% were White, 24.0% were African American or Black (AA), 1.6% were Asian, 0.8% were American Indian or Alaska Native (AIAN), and 3.7% other race. Hispanic/Latino ethnicity was reported by 11.3%. Overall, 69.3% and 70.7% completed colonoscopy within 6 and 12 months, respectively. At 6 months, the odds of adherence were 12% lower for women vs. men, 15% lower for AA and 43% lower for AIAN vs. Whites, 2% lower per year of age and 29% lower for current vs. never smokers. (Table). The odds of adherence were most positively associated with having had a prior colonoscopy (45% increase) and with having a second degree relative with CRC (17% increase), college degree (16% increase), prior screening with occult blood testing (14% increase) and former smokers (7% increase). The evaluation of 12-month adherence revealed similar findings.

Conclusions: Though colonoscopy adherence (71% at 1 year) was lower than we previously reported for FIT (82% in the first round), it was substantially higher than the 42% reported in the largest randomized colonoscopy study to date (NordICC). Predictors of adherence were different than those we previously reported for FIT. Most notably, age was inversely associated with colonoscopy adherence but positively associated with FIT adherence. Also, FIT adherence was not associated with sex or having a second degree relative with CRC. Our work extends prior studies demonstrating the importance of individual-level factors in predicting screening adherence, including identification of differences by screening strategy. Interventions, especially for those with independent risk factors for non-adherence, are needed to improve screening participation.

*The first two authors are co-first authors.

Background: Patients undergoing evaluation for heart transplant due for colorectal cancer (CRC) screening often undergo colonoscopy, but these patients are at higher risk for procedural complications. At our institution’s high-volume endoscopy suite, these endoscopies routinely involved multi-day delays to coordinate between Gastroenterology and Cardiac Anesthesiology. A multi-disciplinary program was initiated in 2016 wherein appropriate-risk patients in need of CRC screening had CT colonography as the first line screening study and then underwent reflex colonoscopy or sigmoidoscopy if concerning lesions were identified.
Aims: The primary outcome was to determine the number of colonoscopies avoided via CT colonography. A secondary outcome was to compare time-to-completion of CT colonography vs colonoscopy for hospitalized patients.
Methods: This was a retrospective observational chart review of patients who underwent heart transplant evaluation at our tertiary-care, urban hospital pre-intervention (2015) and post-intervention (2016-2022). We reviewed transplant evaluation notes for CRC screening study type. If a patient had colonoscopy or CT colonography, we collected dates and time stamps for study orders, case requests, and Gastroenterology consult orders; inpatient vs outpatient location; CT colonography outcomes; and colonoscopy indications. Time-to-completion for inpatient studies was calculated as the difference between the order or consult date/time and the study date/time. R was used to calculate quartiles and conduct Wilcoxon rank-sum test analysis.
Results: 600 patients were reviewed (56 in 2015 and 544 in 2016-2022). 162 patients had a colonoscopy or CT colonography between 0-180 days after their transplant evaluation start date. In 2015, all 13 screened patients had colonoscopies. Post-intervention, 96 patients (64%) underwent CT colonography, and 53 (36%) had colonoscopies (Fig. 1). All patients who underwent colonoscopy post-intervention were ineligible for CT colonography due to medical history requiring endoscopic evaluation. 13 of 96 (14%) CT colonography studies reflexed to colonoscopy (11 for polyps >5 mm; 2 for wall thickening). During hospitalization, CT colonography was completed in a median of 1.9 days (IQR 1.1-2.8, n=82) from study order, compared to median of 2.6 days (IQR 1.9-5.0, n=20) for colonoscopy (Fig. 2, p=0.014).
Conclusions: This CT colonography program diverted 83 patients from colonoscopy over 7 years, thereby avoiding anesthesia risk, simplifying workflow, and decompressing a high-volume endoscopy suite. CT colonography significantly decreased time-to-completion of a study compared to colonoscopy. Initial screening with CT colonography may be a practical and safe alternative for CRC screening at institutions with similarly high–risk populations. Future exploration will assess cost effectiveness of this approach.

Figure 1: Colorectal Cancer Screening Methods and Sites for Patients Undergoing Heart Transplant Evaluation. Depicts the number of patients who underwent CT colonography (n=82 inpatient at the authors' institution, n=14 outpatient) or colonoscopy (n=31 inpatient at the authors' institution, n=35 outpatient or at another hospital) within 180 days after officially starting their transplant evaluation at the authors' institution.

Figure 1: Colorectal Cancer Screening Methods and Sites for Patients Undergoing Heart Transplant Evaluation. Depicts the number of patients who underwent CT colonography (n=82 inpatient at the authors' institution, n=14 outpatient) or colonoscopy (n=31 inpatient at the authors' institution, n=35 outpatient or at another hospital) within 180 days after officially starting their transplant evaluation at the authors' institution.

Figure 2: Comparison of Time-to-Completion for Inpatient CT Colonography versus Colonoscopy in Patients Undergoing Heart Transplant Evaluation. Plots the time-to-completion (difference between the study order date and time and the study date and time) for patients who underwent CT colonography (n=82) or colonoscopy (n=20) while hospitalized at the authors' institution in 2016-2022. The difference in time-to-completion between the two groups was significant (Wilcoxon rank-sum test, p=0.014).

Figure 2: Comparison of Time-to-Completion for Inpatient CT Colonography versus Colonoscopy in Patients Undergoing Heart Transplant Evaluation. Plots the time-to-completion (difference between the study order date and time and the study date and time) for patients who underwent CT colonography (n=82) or colonoscopy (n=20) while hospitalized at the authors' institution in 2016-2022. The difference in time-to-completion between the two groups was significant (Wilcoxon rank-sum test, p=0.014).

Introduction:
Colorectal cancer (CRC) is the second leading cause of cancer deaths in the United States. CRC Screening through either stool-based testing or colonoscopies is the standard of care for all adults, age 45 to age 75 years, but national rates of procedures remain sub-optimal. Common factors contributing to low procedure adherence include but not limited to patient's no show on day of procedure, poor compliance to bowel prep regimen etc. Digital applications i.e., text-based reminders and smartphones healthcare apps, can be effective tools to overcome barriers to screening colonoscopy adherence thus improving the outcomes.

Methods:
A retrospective study on a prospectively maintained database on CRC screening adherence rates from July 2019 to March 2022 in the gastroenterology clinic of a metropolitan hospital. Screening colonoscopies (SC) scheduled during the study time period were included. Demographics and health-care resources utilization were reviewed. Primary outcome was comparison between SC performance rates, prior to and after implementation of: a digital navigation program (DNP) developed by a private company (Rx Health) that consisted of generating automated text-messages containing appointment reminders, links for driving and bowel prep intrctions and a short informative procedure video; and EPIC-EMR generated personal health app called MyChart. All participants were offered access to DNP and MyChart app set up on their smartphones.

Results:
Total of 4864 procedures scheduled during the study period. Overall age was 60±9.5 years and majority were males at 55%; Hispanics were 52% and Non-Hispanic 48%. Pre-intervention period (n=1416), 53% procedures were performed and 47% cancelled; and in post-intervention period (n=3448), 86% SC performed with 15% cancelled (p<.001) with a percent increase of 81%. In post-intervention group, DNP enrolled 920 patients of which 87% presented for procedure. Patients who did not present, received additional reminders. Bowel prep between groups showed no significant difference in quality. The majorities of cancellations were "Patient no show" without reason (54%) and this rate was substantially decreased in post-intervention period (19% vs 78%).

Discussion:
Our study highlights a significant increase in performance rates of scheduled screening colonoscopies with 81% increase after implementation of a DNP and MyChart app utilization. Improving healthcare outcomes through patient's active involvement in their own care using digital/tele medicine is an emerging concept and literature has shown supporting evidence including improving no-show rates for screening colonoscopies, however data on its implementation on a larger scale is lacking. Use of technology to increase adherence to CRC screening and surveillance seems more warranted since COVID-19 pandemic.

Background: No-shows and inadequate bowel preparations (prep) lead to significant waste of resources in endoscopy centers. While printed prep instructions and pre-procedural nursing phone calls are often used to prevent this, they are not always successful. Text message communication offers additional opportunities to improve show rate and bowel prep quality.
Methods: We created 2 protocols of short, timed text messages to send via an existing one-way messaging service, available at the Veterans Affairs (VA) medical centers throughout the United States, called “the Annie app”. Our protocols gave specific instructions that were sent at pre-specified times during the 7 days prior to appointments to those who agreed to receive text messages. The colonoscopy instructions consisted of 11 messages, and included reminders to have a driver available, to change the diet to clear liquids the day before the procedure, to take the bowel preparation in a split-dose fashion, and to call if there are additional questions. A similar, shorter, protocol was developed for upper endoscopy. These two protocols were designed to be generic and can be used by all VA endoscopy units, without needing further customization. After implementing the text messages, we surveyed patients about their experience, evaluated the no-show rate, and compared rates of adequate preparation between those who did and did not receive text messages.
Results: In the initial 60 days, text messages were offered to 440 patients. Of those, 342 (77.7%) consented to receive messages and 102 of these completed anonymous surveys. Of those surveyed, 90 (88.2%) said the messages were helpful and 92 (90.2%) said they would recommend the text message reminders to others. 85 (83.3%) found the number of messages to be appropriate, while 5 (4.9%) and 9 (8.8%) found the number of messages to be too little or excessive, respectively. Among 91 surveyed patients who received both text messages and the traditional mailed instructions, 57 (62.6%) said they would prefer receiving both forms of instructions, 16 (17.6%) would prefer Annie text messages alone, and 15 (16.5%) preferred the mailed instructions alone. In the first 60 days following the introduction of Annie messages, compared to last year, the no-show rate in endoscopy decreased from 9.2% to 5.4% (p < 0.00001). When comparing those who consented to receiving Annie messages to those who did not consent, 5.9% vs. 9% had inadequate bowel preparation (p < 0.00001).
Conclusions: A series of timed text messages reminding patients of their endoscopy appointment and prompting them to follow the major steps of bowel preparation were associated with a decreased no-show rate, an improved bowel prep quality, and were well-received by the patients. A randomized clinical trial of using text message protocols should be considered to confirm our findings.

Background
Minimal performance targets for colonoscopy quality focus on a provider adenoma detection rate of 25%: 20% for women and 30% for men. These benchmarks, largely based on single centers or local health systems, may not optimize prevention of post colonoscopy colorectal cancer, as they are not based on national assessment and are restricted to screening colonoscopies. We hypothesize that prevalence of adenomas detected at colonoscopies performed for any indication is substantially higher than current established benchmarks and can vary by patient demographics beyond gender. We aimed to conduct a national assessment of prevalence of adenomas detected in a large national cohort of individuals undergoing routine colonscopy, and to examine the association of patient demographic factors.

Methods
We identified colonoscopies for all indications using CPT codes from October 2018-September 2022 and attributed pathology to those posted within 30 days of the exam. We determined histology using a previously validated text recognition algorithm. We calculated adenoma prevalence as the percentage of colonoscopies with at least one adenoma or adenocarcinoma detected. We estimated differences in adenoma prevalence by gender, age, race, ethnicity, rurality, and geographic region by logistic regression. We considered effects associated with p<0.001 as significant, based on the large sample size.

Results
We identified 841,861 patients who underwent a colonoscopy by 1,244 unique providers (Table). The majority, 50.51% (n= 425,199), had at least one adenoma detected. Age was positively associated with adenoma, such that for each year increase in age, the likelihood of an adenoma detected increased by 4%, p<0.001. Men were significantly more likely to have an adenoma detected than women, OR 1.68 [1.63, 1.73], p<0.001. At every screening eligible age (45-75) group, the prevalence of men or women with at least one adenoma found during colonoscopy significantly exceeded current society published benchmarks; with an average prevalence difference of 13.3% higher in men than women at each age group (Figure). In addition, patients who identified as white compared to non-white, or those living in the Midwest, Continental, and Pacific compared to in the North Atlantic and Southeast regions were significantly more likely to have adenomas detected.

Conclusion
In a national colonoscopy cohort with a large number of patients and providers, the prevalence of men or women with at least one adenoma found during colonoscopy significantly exceeded current society published benchmarks. Our findings may have implications to inform patients and providers of the adenoma prevalence in different demographic groups, and to stimulate further evaluation of current quality ADR benchmarks for colonoscopies of all indications, including adjustment by gender and age.

Table. Adenoma Prevalence by Age, Gender, Race, Ethnicity, Rurality, and Geographical Region

Table. Adenoma Prevalence by Age, Gender, Race, Ethnicity, Rurality, and Geographical Region

Figure. Adenoma Prevalence by Age Group and Gender Compared to Established ADR Benchmarks. <br /> <i>Note. </i>The dashed yellow line represents the current ADR benchmark for male patients (30%), while the dotted red line represents the current benchmark for female patients (20%).

Figure. Adenoma Prevalence by Age Group and Gender Compared to Established ADR Benchmarks.
Note. The dashed yellow line represents the current ADR benchmark for male patients (30%), while the dotted red line represents the current benchmark for female patients (20%).

INTRODUCTION:
Inadequate bowel preparation, which occurs in 25% of colonoscopies, is a major barrier to the effectiveness of screening for colorectal cancer. Inadequate bowel preparation can lead to higher rates of missed colorectal polyps and cancer and need for repeat colonoscopy. There is a lack of a validated, real-time and accurate predictor of bowel preparation adequacy that can be used after a patient has completed the pre-colonoscopy bowel preparation but before the colonoscopy is performed. We aim to develop an artificial intelligence (machine learning) algorithm which assesses photos of stool output after bowel preparation to predict inadequate bowel preparation prior to colonoscopy.

METHODS:
Data were collected from June 2022 to August 2022 at Penn Presbyterian Medical Center. All patients undergoing outpatient colonoscopy were eligible. The night prior to the scheduled colonoscopy, all patients received a phone call and text message instructions to text a photo of their stool output during the morning of their colonoscopy when they believed that they neared completion of their preparation, using the phrases “when you think your preparation is clear” or when “you are nearing the end of prep related diarrhea.” After colonoscopy was completed, data were collected from the electronic medical record including Boston Bowel Preparation Scale (BBPS). For the purposes of this study, scores below 6, 6, and 7 were labeled as inadequate/fair. BBPS scores of 8 and 9 were considered good. A binary classification machine learning algorithm was designed with PyTorch, a machine learning framework, using the images of bowel preparation with the gold standard findings from colonoscopy. Approximately 20% of the collected images were set aside as a test set.

RESULTS:
A total of 792 patients underwent outpatient colonoscopy during the study period. The median age was 60.30 years with SD=12.54. Of these patients 53.5% were female. A total of 576 (72.6%) responded to phone call or text message reminders, of which 488 (84.7%) stated that they would text a photo of their stool output. A total of 352 patients submitted photos of their stool output (72.1% of patients who stated they would submit a photo and 44% of all patients scheduled for colonoscopy). In a test set of 61 images, the binary classification machine learning algorithm was able to distinguish inadequate/fair preparation (BBPS below 6, 6, and 7) from good preparation (BBPS scores 8 and 9) with a positive predictive value of 84.2% and a negative predictive value of 76.2%.

CONCLUSION:
Patients are willing to submit photos of their stool output during bowel preparation through text messages. Our machine learning algorithm demonstrates the ability to predict inadequate/fair preparation (BBPS below 6, 6, and 7) from good preparation (BBPS scores 8 and 9) based on image classification of stool output.

Figure 1. Analysis of images by machine learning algorithm

Figure 1. Analysis of images by machine learning algorithm

James Lewis

Univ of Pennsylvania

Appearances