Colonoscopy Indication Algorithm Performance Across Diverse Health Care Systems in the PROSPR Consortium.

Area of Science:

• Gastroenterology
• Health Informatics
• Public Health

Background:

• Characterizing colonoscopy indication is vital for quality monitoring and cancer screening evaluation.
• A standardized approach for documenting colonoscopy indication in medical records is currently lacking.
• This study addresses the need for improved data capture of colonoscopy indications.

Purpose of the Study:

• To evaluate the effectiveness of two algorithms in assigning colonoscopy indications.
• To compare the performance of the KPNC and SEARCH algorithms in classifying colonoscopy indications.
• To identify the need for standardized documentation of colonoscopy indications.

Main Methods:

• Two algorithms, KPNC and SEARCH, were applied to colonoscopy data from three healthcare systems (2010-2013).
• Algorithms utilized standard procedure, diagnostic, and laboratory codes.
• Medical record abstraction served as the gold standard for indication assessment.

Main Results:

• The KPNC algorithm demonstrated sensitivities and specificities for screening and diagnostic indications ranging from 0.74-0.91.
• The KPNC algorithm showed poor sensitivity for surveillance exams.
• The SEARCH algorithm's Area Under the Curve (AUC) for screening indication ranged from 0.76-0.84.

Conclusions:

• Standardized implementation of indication algorithms across healthcare systems resulted in imperfect data capture.
• The study recommends incorporating standard, systematic documentation of colonoscopy indication into medical records.
• Accurate and efficient data capture of colonoscopy indications is essential for quality improvement and program evaluation.