Risk factors for inadequate bowel preparation: A retrospective study of 7931 paired colonoscopy cases with a predictive model

Abstract

BACKGROUND

Inadequate bowel preparation negatively impacts adenoma detection rates and overall colonoscopy quality; however, the associated risk factors have yet to be fully elucidated in large paired-cohort studies.

AIM

To identify independent risk factors associated with inadequate bowel preparation and to develop and validate a clinical prediction model for Chinese patients undergoing colonoscopy.

METHODS

A total of 7931 patients who underwent two or more colonoscopies were retrospectively enrolled in this study. Univariate and multivariate logistic regression analyses were performed to identify independent risk factors for inadequate bowel preparation. The predictive performance of the model was evaluated using 10-fold cross-validation, bootstrap resampling, temporal validation, and subgroup analyses.

RESULTS

The overall rate of inadequate bowel preparation was 13.67%. Independent risk factors included older age, male sex, afternoon colonoscopy, diabetes mellitus, constipation, and a history of inadequate bowel preparation (all P < 0.05), whereas inpatient status was identified as a protective factor (odds ratio = 0.391, P < 0.001). The prediction model demonstrated good discriminatory ability [area under the curve (AUC) = 0.699] and satisfactory calibration. In addition, the simplified scoring system yielded an AUC of 0.687 and exhibited a clear gradient of increasing risk across score categories.

CONCLUSION

A history of inadequate bowel preparation is the strongest predictor of subsequent preparation failure. Inpatient status is associated with significantly better bowel preparation quality than outpatient status. The simplified risk scoring system provides a practical tool for identifying patients at high risk of inadequate bowel preparation and facilitating personalized preparation strategies. To our knowledge, this is the first large-scale paired-colonoscopy study to quantify the recurrent nature of bowel preparation failure and to demonstrate the protective effect of inpatient status after adjustment for multiple potential confounders.

Key Words: Bowel preparation; Colonoscopy; Risk factors; Prediction model; Simplified scoring system; Inpatient status

Core Tip: Key findings of our study: In a large paired cohort of 7931 patients, prior inadequate bowel preparation was the strongest predictor of repeated inadequate preparation. Inpatient status was a significant protective factor, with substantially better bowel preparation quality than in outpatients. A simplified risk scoring system showed good discrimination and risk stratification, enabling rapid identification of high-risk patients in clinical practice.

INTRODUCTION

Colonoscopy is the gold standard for the screening and diagnosis of colorectal cancer, and its effectiveness is highly dependent on the quality of bowel preparation. Adequate bowel preparation enhances adenoma detection rates, shortens procedure time, and reduces the risk of procedure-related complications. From an endoscopic perspective, inadequate bowel preparation can prolong examination time, increase patient discomfort, impair mucosal visualization, and elevate the likelihood of missed lesions, including adenomas and polyps. Furthermore, poor preparation often necessitates repeat colonoscopy within a shorter surveillance interval, thereby increasing healthcare utilization and associated medical costs. Despite advances in bowel preparation protocols, inadequate bowel preparation remains a common clinical challenge, affecting approximately 10%-25% of patients undergoing colonoscopy[1,2]. Previous studies have identified several risk factors associated with inadequate bowel preparation, including advanced age, male sex, diabetes mellitus, constipation, and afternoon colonoscopy appointments[3-5]. However, most existing studies have employed cross-sectional designs and have not examined longitudinal outcomes across repeated colonoscopies in the same individuals. More importantly, the impact of a history of inadequate bowel preparation on subsequent colonoscopies remains poorly understood. Furthermore, the association between patient setting (inpatient vs outpatient) and bowel preparation quality remains controversial[6]. Therefore, this study utilized paired colonoscopy data to identify risk factors for inadequate bowel preparation, with particular emphasis on the predictive value of prior preparation quality for subsequent examinations. In addition, differences in bowel preparation quality between inpatients and outpatients were evaluated to provide evidence for the identification of high-risk populations and the development of individualized intervention strategies in clinical practice. This study makes three novel contributions. First, by employing a paired-colonoscopy design, it confirms the recurrent nature of bowel preparation failure and demonstrates that a history of inadequate bowel preparation is the strongest predictor of subsequent preparation failure. Second, it quantifies the protective effect of inpatient status on bowel preparation quality after adjustment for multiple potential confounders, providing evidence with important implications for targeted preventive interventions. Third, it develops and internally validates a simplified clinical risk-scoring system that facilitates rapid risk stratification and supports personalized bowel preparation management in routine clinical settings.

MATERIALS AND METHODS

Study design

This retrospective cohort study included patients who underwent at least two colonoscopies at the Affiliated Hospital of Chengdu University between January 2017 and November 2025. The study protocol was reviewed and approved by the Ethics Committee of the Affiliated Hospital of Chengdu University (approval No. PJ2025-136-02). The requirement for informed consent was waived by the Ethics Committee owing to the retrospective nature of the study.

Inclusion criteria: (1) Patients with at least two colonoscopy records; (2) Availability of complete bowel preparation score data; and (3) Availability of complete clinical data.

Exclusion criteria: (1) Emergency colonoscopy; (2) Intestinal obstruction; and (3) Missing or incomplete data.

Data source and outcome definition

Data on age, sex, examination timing, medical history, and bowel preparation assessment were extracted from the electronic medical record system. Inadequate bowel preparation was defined as a total Boston Bowel Preparation Scale (BBPS) score of < 6, or documentation in the endoscopic report containing keywords such as “poor”, “fecal water”, or “fecal residue”. The BBPS was independently assessed by two senior endoscopists, demonstrating good inter-observer consistency.

Statistical analysis

Statistical analyses were performed using R software (version 4.3). Variables with P < 0.10 in univariate analyses were entered into a multivariate stepwise logistic regression model. Model discrimination was evaluated using the area under the receiver operating characteristic curve. Internal validation was conducted using ten-fold cross-validation, bootstrap resampling with 500 iterations, and temporal stratification (2017-2021 vs 2022-2025). Sensitivity analyses, excluding patients with constipation, and subgroup analyses stratified by inpatient status were also performed. Multicollinearity among predictors was assessed using variance inflation factors (VIFs), with VIF > 5 indicating significant collinearity. All statistical tests were two-sided, and P < 0.05 was considered statistically significant.

Development of the simplified scoring system

The simplified scoring system was derived from the regression coefficients of the multivariate logistic model. Each independent predictor’s coefficient (β) was divided by the smallest absolute coefficient among all statistically significant predictors (constipation, β = 0.596). The resulting ratios were rounded to the nearest integer to assign point values. For age, which was treated as a continuous variable (β = 0.016 per 1-year increase), scores were assigned as floor (age/10) to approximate the risk gradient while maintaining clinical interpretability. Negative points were assigned to the protective factor (inpatient status, β = -0.939). The optimal cut-off value for risk stratification was determined using the Youden index. The model’s performance was evaluated using the same internal validation methods as those applied to the full logistic regression model.

RESULTS

Baseline characteristics

A total of 7931 patients who underwent at least two colonoscopies were included in the analysis. The mean age was 56.9 ± 13.0 years, and 52.0% (n = 4127) were male. Afternoon examinations accounted for 55.2% (n = 4377) of cases. The prevalence of hypertension, diabetes mellitus, and constipation was 9.5% (n = 750), 5.9% (n = 464), and 1.2% (n = 96), respectively. Prior inadequate bowel preparation was documented in 26.6% (n = 2107) of patients, and 57.3% (n = 4545) were inpatients (Table 1).

Table 1 Baseline characteristics of the study population, n (%).

Sample size	7931
Age (years)	56.9 ± 13.0
Male sex	4127 (52.0)
Afternoon examination	4377 (55.2)
Hypertension	750 (9.5)
Diabetes mellitus	464 (5.9)
Constipation	96 (1.2)
Prior inadequate preparation	2107 (26.6)
Inpatients	4545 (57.3)

Univariate and multivariate analyses

Univariate logistic regression analysis indicated that age, male sex, afternoon examination, hypertension, diabetes mellitus, and prior inadequate bowel preparation were significantly associated with inadequate bowel preparation (all P < 0.05). Constipation was not significantly associated with inadequate bowel preparation (P = 0.900). Inpatient status was identified as a protective factor [odds ratio (OR) = 0.474, 95%CI: 0.435-0.516, P < 0.001]. In multivariate analysis, age (OR = 1.016), male sex (OR = 1.447), afternoon examination (OR = 1.566), diabetes mellitus (OR = 1.736), constipation (OR = 1.814), and prior inadequate bowel preparation (OR = 3.526) remained independent risk factors (all P < 0.05), whereas inpatient status remained a protective factor (OR = 0.391). Hypertension was not statistically significant in the multivariate model (P = 0.664). Multicollinearity diagnostics showed that all VIFs were < 2, indicating no significant collinearity among predictors. A nomogram was constructed based on the multivariate model to facilitate individualized risk prediction in clinical practice (Figures 1 and 2; Table 2).

Figure 1 Multivariate analysis forest plot: Risk factors for inadequate bowel preparation. OR: Odds ratio.

Figure 2  Nomogram: Risk factor scoring system for predicting inadequate bowel preparation.

Table 2 Multivariate analysis of risk factors for inadequate bowel preparation.

Variable	OR	95%CI	P value	VIF
Age (per 1-year increase)	1.016	1.011-1.022	< 0.001	1.20
Male sex	1.447	1.262-1.660	< 0.001	1.05
Afternoon examination	1.566	1.364-1.799	< 0.001	1.04
Hypertension	1.051	0.839-1.316	0.664	1.32
Diabetes mellitus	1.736	1.358-2.219	< 0.001	1.28
Constipation	1.814	1.071-3.005	0.025	1.06
Prior inadequate preparation	3.526	3.083-4.034	< 0.001	1.10
Inpatient status	0.391	0.357-0.427	< 0.001	1.35

OR: Odds ratio; CI: Confidence interval; VIF: Variance inflation factor.

Model performance and internal validation

The prediction model demonstrated an area under the curve (AUC) of 0.699 (95%CI: 0.682-0.716) (Figure 3). Ten-fold cross-validation yielded a mean AUC of 0.700 ± 0.010, while bootstrap validation produced a mean AUC of 0.698 (95%CI: 0.697-0.699). Temporal stratification analysis further demonstrated model stability, with AUCs of 0.691 for 2017-2021 and 0.706 for 2022-2025. The Hosmer-Lemeshow test indicated good calibration (P = 0.669) (Figure 4). Decision curve analysis showed that the full model provided a higher net clinical benefit than the “intervention for all” and “intervention for none” strategies across a threshold probability range of 0.06-0.40 (Figure 5). In sensitivity analysis, exclusion of 96 patients with constipation did not materially alter the effect estimates, with ORs for all variables remaining consistent with the primary analysis. Subgroup analysis further demonstrated that ORs for individual risk factors were generally lower among inpatients than outpatients, suggesting that the inpatient setting may attenuate the impact of these risk factors on bowel preparation quality.

Figure 3 Receiver operating characteristic curves: Training vs validation. AUC: Area under the curve; ROC: Receiver operating characteristic.

Figure 4 Calibration curve. CI: Confidence interval.

Figure 5  Decision curve analysis.

Simplified clinical scoring system

Based on the multivariate analysis, a simplified scoring system was constructed: Age (per 10 years) = 1 point; male sex = 2 points; afternoon examination = 2 points; diabetes mellitus = 3 points; constipation = 3 points; prior inadequate bowel preparation = 7 points; and inpatient status = -6 points. Patients were stratified into low- (< 8 points), intermediate- (8-15 points), and high-risk (> 15 points) groups. Age-related scoring was calculated as floor (age/10), assigning 1 point for each completed 10-year age increment (e.g., 50-59 years = 5 points; 60-69 years = 6 points). The scoring system achieved an internally validated AUC of 0.687 (95%CI: 0.670-0.704), which was comparable to that of the original logistic regression model (AUC = 0.699). Using an optimal cutoff of 9.5 points determined by the Youden index, the model yielded a sensitivity of 0.642 and a specificity of 0.658. The observed rates of inadequate bowel preparation demonstrated a clear risk gradient across categories: 6.1% (95%CI: 5.3%-7.0%), 15.1% (95%CI: 14.0%-16.3%), and 24.9% (95%CI: 21.9%-28.1%) for the low-, intermediate-, and high-risk groups, respectively.

DISCUSSION

Based on a large retrospective cohort of 7931 paired colonoscopy examinations, this study provides new insights into risk factors for inadequate bowel preparation beyond the replication of previously established associations. The key novelty can be summarized in three aspects. First, by employing a paired colonoscopy design, this study demonstrates that prior inadequate bowel preparation is not merely associated with, but strongly predictive of repeat failure (OR = 3.526), thereby confirming the recurrent nature of bowel preparation failure at the individual patient level, a phenomenon that cannot be established using cross-sectional designs. Second, inpatient status was identified as a significant protective factor (OR = 0.391), suggesting that a supervised hospital environment substantially reduces the risk of inadequate bowel preparation. This finding has direct clinical implications, indicating that outpatients may represent a higher-priority group for targeted interventions. Third, a simplified scoring system was developed by translating regression coefficients into an easy-to-use clinical tool, enabling rapid risk stratification in outpatient settings without the need for complex calculations. Collectively, these findings extend beyond the validation of previously known risk factors (age, male sex, diabetes mellitus, constipation, and afternoon examination) by providing actionable insights for personalized bowel preparation strategies, particularly for patients with a history of inadequate preparation and for outpatients who cannot be hospitalized[7].

Multivariate regression analysis demonstrated that a history of inadequate bowel preparation was associated with a 3.526-fold increased risk of repeat inadequate preparation, representing the strongest predictor among all variables and indicating marked individual-level recurrence. This finding is consistent with recent large-sample studies. The recurrent nature of bowel preparation failure may be attributed to patient-specific factors that are relatively difficult to modify, including inherent differences in colonic anatomy and motility (e.g., colonic redundancy, tortuosity, and slow transit), reduced responsiveness to laxatives (e.g., alterations in mucosal secretion, osmotic regulation, or gut microbiota composition), and persistent patterns of non-adherence (e.g., insufficient dietary restriction, inadequate fluid intake, or deviations in medication timing)[8-10]. These findings support prior inadequate bowel preparation as a key high-risk indicator, highlighting the need for intensified preparation protocols, split-dose regimens, targeted patient education, and early intervention strategies to reduce the risk of repeat failure.

One of the most clinically relevant findings of this study is that inpatient status is associated with a significantly lower risk of inadequate bowel preparation (OR = 0.391). This result contrasts with the findings of Almadi et al[11] (OR = 1.59), who reported higher rates of inadequate preparation among inpatients, but is consistent with those reported by Yamaguchi et al[12]. Several mechanisms may explain this protective effect. First, direct supervision of medication administration by healthcare staff improves patient adherence. Second, hospitalization allows earlier initiation of low-residue diets and preparatory regimens, thereby extending the preparation window. Third, better control of comorbid conditions (e.g., hypertension and diabetes mellitus) during hospitalization may reduce their impact on colonic motility. Fourth, timely corrective interventions can be implemented when inadequate bowel movements are observed. Finally, inpatients may demonstrate higher levels of cooperation due to increased awareness of medical necessity. Subgroup analysis further indicated that the inpatient setting attenuates the effect sizes of most risk factors, suggesting that outpatients represent the primary high-risk population and should be prioritized for intensified interventions, including structured education, telephone follow-up, visual instructional materials, and, where feasible, pre-admission preparation strategies to reduce disparities in bowel preparation quality. Although inpatient bowel preparation is associated with superior quality, its broader implementation is limited by higher direct medical costs, including hospitalization expenses, and increased healthcare resource utilization, particularly in patients with mild disease or in resource-limited settings. To mitigate the financial burden associated with inpatient preparation, several low-cost strategies may be applied in outpatient settings, including: (1) Enhanced pre-procedure education using illustrated materials or telephone reminders; (2) Strict adherence to split-dose regimens with compliance reinforcement; (3) Mobile health applications (e.g., WeChat-based guidance) for real-time instruction and monitoring; and (4) Where resources permit, community nurse-led preparation programs may be considered as an additional low-cost intervention to improve bowel preparation quality. At our institution, the average cost of a single colonoscopy (including electrocardiography, bowel preparation agents, and intravenous anesthesia) is approximately ¥800. Repeat procedures due to inadequate bowel preparation can double direct medical costs and further increase indirect expenses, including loss of income and transportation costs.

In this study, the risk of inadequate bowel preparation was 1.74 times higher in patients with diabetes mellitus and 1.81 times higher in patients with constipation, consistent with the meta-analysis by Gandhi et al[13]. Several mechanisms may explain the association with diabetes. Chronic hyperglycemia can delay gastric emptying, thereby affecting the timing and efficacy of bowel cleansing agents. In addition, diabetic autonomic neuropathy may impair enteric nervous system function and reduce colonic motility. Diabetes is also associated with gut microbiota dysbiosis, which may further compromise bowel cleansing efficacy[14,15]. In patients with constipation, prolonged fecal retention and increased stool hardness reduce the effectiveness of standard bowel preparation regimens[16]. For patients with diabetes, optimization of glycemic control prior to colonoscopy is recommended, along with consideration of extended preparation duration or adjusted laxative dosing, and close peri-procedural monitoring of blood glucose to prevent hypoglycemic events. For patients with constipation, a low-residue diet for 3-5 days before colonoscopy, adjunctive pre-procedural laxatives to soften stool, and enhanced bowel preparation protocols are recommended.

This study confirmed that advanced age and male sex are associated with an increased risk of inadequate bowel preparation, consistent with previous reports[17]. Each one-year increase in age was associated with a 1.6% increase in risk. The underlying mechanisms related to age may include reduced colonic motility, a higher burden of comorbidities, impaired cognitive function leading to decreased adherence, and dysphagia that may limit adequate intake of bowel cleansing agents. Male sex was associated with a 44.7% higher risk compared with female sex, potentially attributable to longer colonic length, higher intra-abdominal pressure, dietary patterns characterized by higher fat and lower fiber intake, and relatively poorer adherence to preparation instructions. Afternoon colonoscopy was also associated with an increased risk of inadequate preparation (OR = 1.566), which may be explained by prolonged fasting, re-accumulation of bile and intestinal secretions leading to reduced mucosal visibility, and suboptimal alignment between laxative peak effect and examination timing. Accordingly, prioritizing morning procedures for high-risk patients and strengthening patient education and supervision are recommended[18-20].

The prediction model demonstrated an AUC of 0.699 and showed stability and reliability through ten-fold cross-validation, bootstrap resampling, temporal stratification analysis, and subgroup analyses, with satisfactory calibration. An AUC between 0.65 and 0.75 is generally considered acceptable for prediction models based on routine clinical variables. The simplified scoring system developed in this study is practical and can be completed within one minute. It demonstrated a clear gradient of inadequate bowel preparation rates across low-, intermediate-, and high-risk groups (6.1%, 15.1%, and 24.9%, respectively), facilitating rapid risk stratification in outpatient settings and supporting the implementation of intensified preparation protocols when appropriate. Decision curve analysis further confirmed that the model provides net clinical benefit across commonly used threshold probabilities, supporting its utility in clinical decision-making by reducing both unnecessary interventions and preparation failure rates. An AUC of 0.699 is acceptable for a clinical risk prediction tool based on routine variables; however, the model is intended for risk stratification rather than definitive ruling in or ruling out of inadequate bowel preparation. In clinical practice, false-positive classifications may lead to unnecessary but generally low-risk and low-cost intensified preparation measures (e.g., additional laxatives). In contrast, false-negative results may have greater clinical consequences, as missed high-risk patients may experience persistent inadequate preparation and missed lesions. Therefore, while the model’s moderate sensitivity (0.642) may be acceptable when combined with clinical judgment, its negative predictive value should be emphasized in application.

Polyethylene glycol (PEG) is widely used in North America and Europe, whereas sodium picosulfate/magnesium citrate is more commonly used in Asia and Australia. Oral sulfate solution has also gained increasing use in recent years. Recent guidelines from the European Society of Gastrointestinal Endoscopy and Asian Pacific consensus recommend split-dose regimens regardless of the specific bowel preparation agent used[21,22]. At our center, multiple bowel preparation protocols are applied based on patient characteristics and clinician preference, including 3 L split-dose PEG, 2 L PEG with ascorbate, and sodium picosulfate/magnesium citrate regimens.

This retrospective study has several limitations. First, the single-center design may introduce selection and information bias. Second, the prevalence of constipation (1.2%) was lower than that reported in the general population, likely due to reliance on medical record documentation rather than standardized diagnostic criteria (e.g., Rome IV), which may have led to underreporting. Third, several potentially relevant variables, including bowel preparation regimen, medication adherence, and laxative type, were not available. Fourth, although the model was internally validated, external validation was not performed. While the large sample size, temporal stratification (2017-2021 vs 2022-2025), and multiple internal validation strategies partially mitigate this limitation, the generalizability of the model remains uncertain. Future studies will involve collaboration with multiple tertiary hospitals in Southwest China to externally validate and further assess model performance. Fifth, data on endoscopist experience (e.g., junior, intermediate, or senior level), institutional characteristics (e.g., endoscopy unit resources), and colonoscopy indication (screening, diagnostic, or surveillance) were not collected. These unmeasured confounders may influence bowel preparation quality and should be addressed in future research.

CONCLUSION

Prior inadequate bowel preparation was the strongest predictor of repeat inadequate preparation. Inpatients demonstrated significantly better bowel preparation quality than outpatients, indicating that outpatients constitute a higher-risk population requiring targeted intervention. Diabetes mellitus and constipation were also identified as important independent risk factors. The simplified scoring system developed in this study showed good discrimination and calibration, with a clear gradient in inadequate bowel preparation rates across risk categories (6.1%, 15.1%, and 24.9% for low-, intermediate-, and high-risk groups, respectively). These findings support its practical utility for identifying high-risk patients and guiding individualized bowel preparation strategies in clinical practice.

ACKNOWLEDGEMENTS

The authors sincerely thank the medical and nursing staff of the Department of Gastrointestinal Surgery and the Department of Gastroenterology at the Affiliated Hospital of Chengdu University for their valuable support in data collection.