Validated Scales for Colon Cleansing: A Systematic Review.

OBJECTIVES

Bowel cleanliness is a critical determinant of colonoscopy quality, mandating its standardized assessment, yet bowel preparation scales have been variably validated. The objective of this study was to assess validity and reliability of existing bowel preparation scales.

METHODS

A systematic review of literature from January 1980 to January 2015 was performed. Main outcomes of this study are face, content, construct, and criterion validity, and inter- and intra-observer reliability measured by associations, interclass correlations (ICC) or κ-coefficients.

RESULTS

Fourteen citations assessed seven scales. The Aronchick Scale, showed fair-to-substantial inter-observer reliability (ICC=0.31-0.76), and was the reference for subsequent validity testing. The Ottawa Bowel Preparation Quality Scale revealed superior inter-observer reliability (ICC=0.94). Five studies assessed the Boston Bowel Preparation Scale (BBPS). Increasing BBPS scores were associated with polyp detection (left colon: odds ratio (OR)=2.58 (1.34; 4.98), right colon: OR=1.6 (1.01; 2.55), less repeat colonoscopies (cutoff of 5, P<0.001), and shorter insertion/withdrawal times (P<0.001), while displaying substantial to excellent inter- and intra-observer reliability (ICC=0.74-0.91). Criterion validity of the Harefield Cleansing Scale (HCS) yielded slight-to-moderate expert-investigator agreement (ICC=0.15-0.46); HCS grades were not discriminant for adenoma detection. Inter- and intra-observer reliabilities were fair-to-moderate (ICC=0.46 (0.37; 0.54) and κ=0.28, respectively). The Chicago Bowel Preparation Scale displayed excellent inter-observer reliability (Pearson's r=0.84 (0.79; 0.88)), yet has been less studied.

CONCLUSIONS

To conclude, all the published scales displayed limitations. The BBPS is the most thoroughly validated scale and should be used in a clinical setting. Between-scale comparisons for repeat colonoscopy time interval, ease, and pertinence of use for auditing are needed.