Table 3.

Potential solutions to improve informatic classification and CRC screening

Reasons for true negative classification	Potential solutions
Examinations ordered but not completed
Colonoscopy ordered but not completed	More transparent documentation of referral status and outcome Clinic-based patient outreach
Faecal immunochemical test ordered but not completed	Clinic-based patient outreach
CT colonography ordered but not completed	More transparent documentation of referral status and outcome Clinic-based patient outreach
Lack of documentation or incorrect documentation	Improved primary care education Improved gastroenterologist-primary care communication
Declined screening	Improved patient education
Insufficient time to discuss	Clinic-based strategies to encourage follow-up
Reasons for false-negative misclassification
Screening outside of UCSF	Patient-approved data sharing, harmonisation and interoperability Natural language processing Optical character recognition Deep learning
Screening prior to Epic EHR implementation	Institutional investment in clinical data integration and harmonisation
Database and query errors	Recruitment, training and funding for more clinical informaticians, especially clinician-investigators Institutional investment in clinical data integration and harmonisation
Misclassified as eligible for screening
Poor life expectancy, or risks outweighing benefits	Deep learning with natural language processing
Above risk (personal or family history of polyps)	Natural language processing Improved family history taking practices Patient consent for EHR data-sharing, chart-linkage by familial relationship
Not primary care empanelled	Deep learning with natural language processing
Reasons for false-positive misclassification
Ordered but incomplete faecal immunochemical test	EHR flag/reminders to repeat screening
Performed colonoscopy revealed inadequate bowel preparation	Natural language processing EHR flag/reminders to repeat screening

EHR, electronic health records; UCSF, University of California, San Francisco.