1475 Derivation of a tool to predict mortality in a population of frequent attenders to an inner city emergency department

Aims, Objectives and Background The Royal College of Emergency Medicine defines Frequent Attenders (FA) as anyone who attends the Department(ED) five or more times per year. This group has a high mortality is significant burden on services. Bristol infirmary (BRI) city-centre adult-only ED, where 1.8% our patients are FAs, with 5 year rate 20% in this group. Our aim was to further develop triage tool used by BRI High Impact User team, determining which factors increase population. Method Design Data collected retrospectively from 250 electronic patient records, randomly selected 1780 FAs attending 2016. Six variables were chosen for analysis: current mental health problems(MHP), homelessness, injecting drug use, alcohol misuse, chronic medical problems(CMP) number attendances that age, gender 5-year also recorded. Logistic regression modeling performed determine best predicted mortality. Results Conclusion Univariate analysis, found age CMPs only two independently associated 10–20 MHP(n=22) had higher (31.3%) than those >20 (n=6) 0%. Multivariate analysis suggested different predictors depending presence absence MHP. Therefore algorithms derived; both AUROC over 0.7. small-scale service development using specific cohort FAs. findings not generalisable. However they highlight complex relationship between risk factors, attendance frequency We must question whether efforts reduce ED risks inadvertently discouraging accessing potentially protective interactions. hope encourage other EDs similar tools their FA populations.