The Development and Validation of the Emergency Department Avoidability Classification

Abstract

Background: Overcrowding in emergency departments (EDs) due to avoidable visits places a significant strain on health systems. There is no known valid classification to identify avoidable ED visits in Canadian administrative data. Research Questions: Which physician interventions and patient characteristics are important to classify avoidable ED visits, and does a novel classification (the Emergency Department Avoidability Classification; EDAC), which incorporated these features, demonstrate validity? Methods: Two independent modified Delphi consensus studies determined ED physician interventions and patient characteristics that classified avoidable ED visits. These studies involved emergency and family medicine physicians across Ontario, Canada. Binary logistic regression was used to examine ED physician interventions in the National Ambulatory Care Reporting System (NACRS) database for associations with patient characteristics. These results constructed the EDAC criteria. ED physicians from an academic hospital evaluated randomly selected retrospective ED visits (n=320) which were also evaluated using the EDAC to assess their avoidability. The primary outcome of this thesis was correlation between the classification and ED physician judgements, measured using a Spearman rank correlation and ordinal logistic regression. The secondary outcome was to compare the correlations of previously published classifications with ED physician judgements. The tertiary outcome was to compare prevalence estimates of avoidable ED visits for all classifications. Results: Consensus showed strong evidence on 146 of 150 (97.3%) ED physician interventions, with 103 (68.7%) deemed suitable for non-ED care. Consensus was established on eight of nine patient characteristics, with four characteristics identified as useful in specifying avoidable ED visits: age (18-70 years), triage acuity (non-emergent), specialist consult in the ED (none) and ED visit outcome (discharged). An adjusted retrospective cohort study found the ED interventions had a strong association with patient characteristics determined in the consensus study: not aged over 65 years, having a non-emergent triage acuity and not being admitted to hospital. The classification was highly correlated with ED physician judgements (r=0.64, p<0.01), with a significant association to classify avoidable ED visits (OR=80.0, 95% CI=17.1-374.9) and strong accuracy (82.8%). The EDAC was the most accurate classifier of avoidable ED visits compared to previously published classifications. The EDAC identified a prevalence of 25.1% ED visits as avoidable and common patient conditions associated with such visits as traumatic injuries, symptoms/signs/abnormal findings, diseases of the musculoskeletal system, mental and behavioural disorders, and diseases of the respiratory system. Conclusion: My thesis developed and established the EDAC as an accurate classifier of avoidable ED visits with supporting evidence of validity and superior performance to previously published classifications. The EDAC can be easily integrated with administrative ED data and has strong potential for use in defining avoidable ED visits by health policy stakeholders.