SHM Converge Scientific Abstract ePoster Gallery-Diagnostic Trajectory Analysis Improves Identification of Organizational Diagnostic Opportunities

Pdf Summary

The document describes a project called LUCID (Longitudinal Analysis of Codes to Identify Diagnostic Opportunities), which aims to improve organizational diagnostic learning by better identifying cases with potential diagnostic improvement opportunities. Traditional case-finding methods are slow and often inaccurate.<br /><br />The study used a two-stage workflow. In Stage 1, researchers developed models using patient cases and controls to predict the future development of three target diagnoses: colorectal cancer (CRC), venous thromboembolism (VTE), and spinal abscess (SA). Patients were grouped into quartiles based on predicted risk. In Stage 2, structured chart review was performed to determine which cases actually contained diagnostic opportunities, and the rate of opportunities was compared across quartiles.<br /><br />The results suggest that longitudinal diagnostic code patterns can help identify cases more likely to contain improvement opportunities. For example, the quartile with the highest predicted probability often contained the highest proportion of diagnostic opportunities. The study compared several case-selection methods, including an original rule-based method, a modified version, logistic regression, and an AI model. Across conditions, the modified and model-based approaches improved enrichment for diagnostic opportunities, with the AI and logistic regression methods performing well in some settings.<br /><br />The chart review identified several types of contributing factors, including problems with data gathering, communication, cognitive processes, context of care, affective factors, and clinician support. Example issues included delayed colonoscopy due to outside clearance delays, use of family members as interpreters, missed recognition of iron deficiency anemia, failure to discuss alternative screening options, and lack of system support to ensure referrals were completed.<br /><br />Overall, the study concludes that applying novel analytics to longitudinal diagnostic code streams shows promise for supporting organizational diagnostic learning and more efficient identification of diagnostic improvement opportunities.

Asset Subtitle

Jejo D. Koola

Meta Tag

Author List Behrooz Mamandipoor, David Laub, Jejo D. Koola, Robert El-Kareh, Shamim Nemati

Category Innovations

Concept Diagnostic Trajectory Analysis

Concept Diagnostic Opportunity

Concept Case-Finding Method

Concept Organizational Diagnostic Learning

Concept Rule-Based Approach

Distinguished Non-Finalist

Presenter Organization University of California San Diego

Presenting Author Jejo D. Koola

Track Patient Safety

Keywords

LUCID

diagnostic learning

longitudinal analysis

diagnostic opportunities

case finding

colorectal cancer

venous thromboembolism

spinal abscess

chart review

predictive analytics

Diagnostic Trajectory Analysis

Diagnostic Opportunity

Case-Finding Method

Organizational Diagnostic Learning

Rule-Based Approach