Why digital reminders alone do not improve bowel preparation

Category: Colonoscopy

Author: Dan Rolls, Founder and CEO, GistMD

Published:  Aug 8 26

Written for healthcare professionals and administrators. Not medical advice.

Sending preparation instructions by text instead of on paper does not, on its own, produce cleaner colonoscopies. A randomized trial of 753 patients found no difference in attendance or preparation quality when nine automated text messages were added to usual care, while a meta-analysis of 18 randomized trials found that reinforced education nearly tripled the odds of adequate preparation. The channel is not the active ingredient. What the patient actually understands is.

That distinction matters more than it sounds, because most digital patient communication programs are budgeted and evaluated as if the move from paper to screen were the intervention. It is not. And there is a second, larger problem sitting underneath it: the definitive evidence base on why bowel preparation fails has never examined the patient's language at all.

GistMD builds clinician-approved personalized video for procedure preparation and is backed by Mayo Clinic Ventures. This article is about what the published evidence shows, and where it stops.

What did the automated text messaging trial actually find?

In a randomized clinical trial at an urban academic center, Mahmud and colleagues assigned 753 patients scheduled for outpatient colonoscopy to usual care or to usual care plus an automated series of nine educational and reminder text messages sent over the week before the procedure (Mahmud et al., JAMA Network Open, 2021).

The primary outcome, attending the appointment with good or excellent preparation, was reached by 53.1% of the intervention group and 54.4% of the control group. The p value was 0.73. The authors concluded there was no significant difference in attendance or preparation quality, and suggested that future work optimize the content and delivery of text interventions, or identify which patients benefit.

One detail in that trial is easy to skip and is the whole point. The messages were personalized: they carried the patient's name, appointment date, and clinician name. What they were not was adapted. Every patient received the same nine messages, in the same language, at the same reading level, regardless of which preparation regimen they had been assigned, what comorbidities they had, or whether they could read the message at all.

Merge fields are not personalization. They are a mail merge with a phone number attached.

What kind of communication does change preparation quality?

The comparison case is unambiguous. In a meta-analysis of 18 randomized controlled trials covering 6,536 patients, Guo and colleagues found that reinforced education, meaning instruction delivered again and in a form the patient engages with rather than once at scheduling, raised adequate preparation from 74.4% to 87.3%, an odds ratio of 2.59 (95% CI 2.09 to 3.19, p<0.001) (Guo et al., PLOS ONE, 2020).

The same analysis found improved detection downstream. Across the four included studies reporting it, adenoma detection was higher in the reinforced education arm, odds ratio 1.35 (95% CI 1.06 to 1.72, p=0.020). Polyp detection, across eight studies, showed odds ratio 1.24 (95% CI 1.02 to 1.50, p=0.030).

Put the two findings side by side and the conclusion is uncomfortable for a lot of digital programs. Generic automated reminders: no effect. Reinforced, engaged-with education: nearly triple the odds of an adequate preparation, and more precancerous lesions found. The variable is not the delivery mechanism. It is whether the instruction lands.

Why does preparation quality matter beyond the schedule?

Because it is not an operational metric. It is a cancer outcome.

In Austria's nationwide screening program, Zessner-Spitzenberg and colleagues followed 335,466 screening colonoscopies over a median of 4.6 years. Post-colonoscopy colorectal cancer deaths ran at 10.05 per 100,000 person-years after an excellent preparation, 13.60 after a good preparation, and 31.42 after a fair preparation (Zessner-Spitzenberg et al., American Journal of Gastroenterology, 2024). In adjusted analysis the study reports a 2.56-fold hazard of colorectal cancer death for fair versus excellent preparation, and 1.68-fold for good versus excellent.

Read that again with the American College of Gastroenterology (ACG) and American Society for Gastrointestinal Endoscopy (ASGE) quality benchmark in mind. Fair is not a failure by most centers' reporting standards. A patient with a fair preparation completed their colonoscopy, was told it was done, and went home with an interval that assumes the examination saw what was there.

And this is not a rare edge case. Across 67 studies, the median rate of inadequate preparation was 26%, with a range from 5% to 67% (Gandhi et al., Clinical Gastroenterology and Hepatology, 2018). One in four, at the median, before anyone starts counting fair.

What is missing from the bowel preparation evidence base?

Language.

The most comprehensive synthesis of risk factors for inadequate bowel preparation published to date pooled 154 studies and 358,257 participants and assessed 48 distinct risk factors (Beran et al., American Journal of Gastroenterology, 2024). Diabetes is in there. Opioid use, constipation, cirrhosis, inpatient status, male sex, older age, prior inadequate preparation, all in there.

Primary language is not. Limited English proficiency is not. Interpreter need is not.

Forty-eight risk factors, and the one that determines whether a patient could read the instruction sheet is not among them. This is worth stating plainly: the question has gone unanswered not because it was tested and dismissed, but because in the definitive literature on bowel preparation, it has not been asked.

Is this a motivation problem or an instruction problem?

The reflex answer in most operational conversations is that patients with limited English proficiency are harder to reach and less likely to show up. The published evidence does not support it.

In the one US study to measure gastroenterology clinic attendance directly by patient language, non-English speakers attended at higher rates than English speakers, 73.3% versus 55.4%, with adjusted odds of a missed appointment of 0.42 for Spanish speakers and 0.56 for speakers of Asian languages, in a San Francisco safety-net system of 1,833 patients (Sewell et al., Journal of Clinical Gastroenterology, 2009). That study is dated and measures clinic visits rather than procedures, but the direction is not ambiguous.

What does fail is the instruction. In a national sample of 41,453 adults, patients with limited English proficiency had received a screening colonoscopy 34% of the time versus 61% of English-proficient patients, adjusted OR 0.67, and part of that gap opens before the patient can act at all: clinicians recommended screening to 60% of them, versus 87% of English-proficient patients (Cataneo et al., The American Surgeon, 2022).

When they do act, the instruction often does not produce a usable result. At Mayo Clinic Rochester, stool-based screening tests returned a usable result 47% of the time from patients with limited English proficiency, versus 71% from English-proficient patients, with a median time from order to completion of 62.5 days versus 33 (Hill et al., Journal of Immigrant and Minority Health, 2022). A patient who mails back a kit was motivated enough to try. An unusable result is an instruction design failure.

So the framing that survives contact with the evidence is not that some patients are harder to engage. It is that some patients are handed instructions they were never in a position to follow.

What happens downstream when the instruction does not land?

It shows up as cancer, later.

In a propensity-matched analysis of 611,149 colonoscopies, patients whose primary language was not English needed a repeat colonoscopy within one year 49% more often, 1.9% versus 1.3%, OR 1.49 (95% CI 1.30 to 1.69). Cancer was found on that repeat procedure 40% more often, 0.8% versus 0.6%, OR 1.40 (95% CI 1.10 to 1.70) (Kerbage et al., Journal of Clinical Gastroenterology, 2026).

This is the strongest available evidence that a communication failure at the preparation stage is not only a throughput problem. It is measurable in cancer detection.

What does the one null result tell us?

There is a study that found no language gap in preparation quality, and it deserves attention rather than omission.

At the NIH Clinical Center, a review of 740 patients found inadequate preparation in 18% of Spanish-speaking patients and 14% of English-speaking patients, p=0.24, with identical adenoma detection rates of 35% in both groups, p=0.93 (Blaney et al., Gastro Hep Advances, 2025).

The authors explain why. Every Spanish-speaking patient in that cohort had a professional medical interpreter present, and the center provides in-person or telephone interpreters for every visit for every patient with limited English proficiency. Their own conclusion is that there was no difference between Spanish speakers with an interpreter present and fluent English speakers.

That is not evidence against a language effect. It is a demonstration of what removes it, in a language-access environment almost no community GI practice can replicate. The null result and the disparity findings are telling the same story from opposite ends: structured communication support is the variable.

What this means for your center

Four things follow directly from the evidence above, and none of them require buying anything.

Report preparation quality by patient language. Most centers cannot currently answer the question. If Beran's 154-study synthesis did not capture language, your quality dashboard almost certainly does not either. You cannot manage a gap you do not measure, and this one takes a single field in your endoscopy report joined to a single field in registration.

Separate "we sent it" from "they engaged with it." Delivery is not receipt and receipt is not comprehension. A program that reports message delivery rates is reporting on itself, not on the patient. Track whether the instruction was opened, watched, or completed, and treat non-engagement as a flag, not a statistic.

Stop counting fair as a pass. Given a 1.68-fold hazard of colorectal cancer death for good versus excellent preparation and 2.56-fold for fair, the operationally comfortable threshold and the clinically meaningful one are not the same threshold.

Ask what non-comprehension costs you specifically. The dominant reported reason for missed colonoscopy appointments is logistical, not clinical: in a survey of 160 patients who missed appointments across two large public health systems, 78.1% cited travel or scheduling issues, most often not having a ride or a companion (Bhise et al., Digestive Diseases and Sciences, 2016). That is addressable by earlier, clearer, better-timed communication, and it is invisible in a delivery-rate report.

Where GistMD fits

GistMD assembles a personalized preparation video for each patient by matching language, assigned preparation protocol, relevant comorbidity flags, and procedure type against a library of clinician-approved content objects. The architecture is deterministic: content is pre-approved and rule-assembled, so no patient receives machine-generated medical instructions. Delivery is by SMS, QR code, or email, with no app download and no portal login.

The evidence behind that approach, stated with its limits. In a prospective cohort of 2,447 ambulatory colonoscopies at Tel Aviv Sourasky Medical Center, inadequate preparation (Boston Bowel Preparation Scale, or BBPS, under 6) occurred in 27.6% of controls who received standard written instructions, 24.4% of patients sent the personalized program who did not watch it, and 19.2% of those who watched it in full, p<0.001. In adjusted analysis, both receiving the program (OR 0.70) and full use (OR 0.63) were independently associated with lower odds of inadequate preparation, both p<0.001. A cost analysis estimated a saving of roughly 25 euro per colonoscopy (Deutsch et al., presented at DDW 2024, under peer review).

Two things that study is not. It is observational, not randomized, so it establishes association and direction, not causation. And its comparator was paper, not an active digital reminder, so part of the observed benefit reflects the move from paper to digital rather than the effect of personalization specifically. The honest reading is that it demonstrates plausibility for the mechanism, not an effect size you should expect against a digital baseline.

Which is exactly why the comparison that has not been run is the one worth running: personalized, language-adapted communication against standard digital reminders, in a population where the gap is real.

Limitations

The strongest evidence here is not experimental. Guo and Mahmud are randomized; the language findings are observational, drawn from national samples and matched cohorts, and cannot establish that communication design caused the outcome difference. Sewell is 17 years old and measures clinic attendance rather than procedures. The Deutsch cohort is single-center, observational, and compares against paper. No published randomized trial has yet compared personalized multilingual communication against an active digital comparator in a low-literacy or non-English-speaking population, which is precisely the gap this article describes.

If you are building the case internally: the 90-second example of what a personalized preparation video looks like in practice is here.

Sources

• Mahmud N, Asch DA, Sung J, et al. Effect of Text Messaging on Bowel Preparation and Appointment Attendance for Outpatient Colonoscopy: A Randomized Clinical Trial. JAMA Netw Open. 2021;4(1):e2034553. PMID 33492374. DOI 10.1001/jamanetworkopen.2020.34553

• Guo X, Yang Z, Zhao L, et al. Reinforced education improves the quality of bowel preparation for colonoscopy: An updated meta-analysis of randomized controlled trials. PLOS ONE. 2020;15(4):e0231888. DOI 10.1371/journal.pone.0231888

• Zessner-Spitzenberg J, Waldmann E, Rockenbauer LM, et al. Impact of Bowel Preparation Quality on Colonoscopy Findings and Colorectal Cancer Deaths in a Nation-Wide Colorectal Cancer Screening Program. Am J Gastroenterol. 2024;119(10):2036-2044. DOI 10.14309/ajg.0000000000002880

• Beran A, Aboursheid T, Ali AH, et al. Risk Factors for Inadequate Bowel Preparation in Colonoscopy: A Comprehensive Systematic Review and Meta-Analysis. Am J Gastroenterol. 2024;119(12). PMID 39225554. DOI 10.14309/ajg.0000000000003073

• Gandhi K, Tofani C, Sokach C, et al. Patient Characteristics Associated With Quality of Colonoscopy Preparation: A Systematic Review and Meta-analysis. Clin Gastroenterol Hepatol. 2018;16(3):357-369

• Sewell JL, Kushel MB, Inadomi JM, Yee HF Jr. Non-English Speakers Attend Gastroenterology Clinic Appointments at Higher Rates Than English Speakers in a Vulnerable Patient Population. J Clin Gastroenterol. 2009;43(7):652-660. DOI 10.1097/MCG.0b013e3181855077

• Cataneo JL, Kim TD, Park JJ, Marecik S, Kochar K. Colorectal Cancer Screening Disparities in Patients With Limited English Proficiency. Am Surg. 2022;88(11):2737-2744

• Hill JC, Njeru JW, Wieland ML, et al. Colorectal Cancer Screening Test Completion Among Patients With Limited English Proficiency. J Immigr Minor Health. 2022;24(2):556-559

• Kerbage A, Haddad SF, Al Deiri D, et al. Impact of Primary Language on Early Repeat Screening Colonoscopy. J Clin Gastroenterol. 2026;60(5):418-423. PMID 40961257

• Blaney H, Rodriguez Franco G, Wright EC, Korman L, Kumar S. Evaluation of Language Disparity in Adequacy of Bowel Preparation for Colonoscopy and Adenoma Detection Rate. Gastro Hep Adv. 2025;4(6):100637. DOI 10.1016/j.gastha.2025.100637

• Bhise V, Modi V, Kalavar A, et al. Patient-Reported Attributions for Missed Colonoscopy Appointments in Two Large Healthcare Systems. Dig Dis Sci. 2016;61(7):1853-1861. DOI 10.1007/s10620-016-4096-3

• Deutsch L, Leshno M, Ritter E, Shnell M, Shibolet O, Gluck N. Personalized Digital Guidance for Pre-Colonoscopy Preparation Is Associated with Better Preparation Quality and Lower Costs Across Diverse Patient Subgroups. Tel Aviv Sourasky Medical Center. Presented at DDW 2024, under peer review