Mortality prediction and equity in emergency care

Introduction: Emergency departments (EDs) face crowding and long patient waiting times due to resource-demand imbalances. Waiting exposes patients to sufferings and risks, such as pain, anxiety, permanent organ damage, or death. Elderly patients, in particular elderly frail, are at greater risk and have longer ED lengths of stay than others. Triage is applied at the ED for timely allocation of medical resources, but makes little compensation for risks associated with patient age. The extent to which age is associated with ED risks and how it can be used for better triage is unknown.

Aims and Hypotheses: The overarching aim of this thesis was to improve risk assessment and care for emergency department patients, particularly elderly and frail patients with low triage priorities. The hypotheses were that age-related risks aren’t fully considered in the Rapid Emergency Triage and Treatment System (RETTS), better algorithms could improve early high-risk patient identification, socioeconomic factors influence physician waiting times, and longer stays at the ED for elderly frail patients are primarily due to process differences.

Methods and Results: This thesis includes four observational cohort studies using three different data sets derived from electronic health records from seven hospitals in the Stockholm Region in 2012-2016 and 2019 combined with government registry data on mortality and socioeconomic factors. The studied data included ca 1.7 million visits by ca 800 thousand individual patients. Five different analytical approaches were applied.

Study I applied logistic regression models to 639.387 triage priority stratified ED visits. It found older age to be associated with increased 7- and 30-day mortality in all triage priorities. The association increased with higher age and with lower priority. In the sex-adjusted model, the 7day mortality odds ratio of priority 4 patients was 15.4 for patients aged 60-79, and 52.4 for patients aged ≥80 when compared to patients aged 18-59.

Study II developed and evaluated three 3-day mortality prediction models based on triage priority and/or age on 1.7 million ED visits. Adding age information to a RETTS triage priority model improved predictive performance. A model based on patient age alone was typically superior and always non-inferior to a triage priority-based model in predicting 3day mortality. The largest difference was seen in patients with abdominal or flank pain where the area under the curve increased from 0.729 to 0.925 by adding age information to the triage priority model, while the age only model had an area under the curve of 0.908.

Study III applied stepwise linear regression to over 1.7 million ED visits to seven EDs while adjusting for morbidity and process variables, finding residual differences in waiting time to see a physician associated with age, sex, birth region and education level, with a dose-response relationship. While each studied exposure alone had a small impact on waiting times, the most disadvantaged patient category had a 17% longer waiting time.

Study IV applied frailty criteria to nurse-assessed care needs and combined it with patient age to identify elderly frail patients. The identified group’s mortality, admission rate, and ED length of stay was similar to previous studies’ elderly frail patient characteristics. Process mining of 45,114 ED visits showed that elderly frail patients spent similar or less time in most ED process steps, but spent longer time in the discharge process and more frequently followed longer arrival-todisposition process paths.

Conclusions: Age-related mortality risks are not fully reflected in the RETTS algorithms and age can improve mortality prediction early in the ED visit. Age, sex and socioeconomic background are associated with waiting times to be seen by a physician although the clinical relevance is unclear. In order to reduce ED lengths of stay for elderly frail patients an in-depth understanding of ED process differences and earlier patient disposition decision-making is needed.

List of scientific papers

I. Ruge T*, Malmer G*, Wachtler C, Ekelund U, Westerlund E, Svensson P, Carlsson AC. Age is associated with increased mortality in the RETTS-A triage scale. BMC Geriatr. 2019 May 23;19(1):139. * = Shared first authorship.
https://doi.org/10.1186/s12877-019-1157-4

II. Malmer G, Åhlberg R, Svensson P, Af Ugglas B, Westerlund E. Age in addition to RETTS triage priority substantially improves 3-day mortality prediction in emergency department patients: a multi-center cohort study. Scand J Trauma Resusc Emerg Med. 2023 Oct 18;31(1):55.
https://doi.org/10.1186/s13049-023-01123-8

III. Malmer G, Fällman A, Åhlberg R, Svensson P, Westerlund E, af Ugglas B. Equity in the Emergency Department. Sociodemographic characteristics are associated with waiting time to physician. [Manuscript]

IV. Malmer G, Fernández Llatas C, Seoane Martinez F, Åhlberg R, Svensson P, af Ugglas B, Westerlund E. Elderly frail patients are subject to longer and more time-consuming pathways in the Emergency Department - a descriptive study leveraging process mining. [Manuscript]