Aims and Objectives Increased mortality has been observed among older people whose systolic pressure was at least 7mmHg below their baseline primary care value when they attended the emergency department (ED). This study aimed to (1) assess feasibility of identifying this ‘relative hypotension’ using readily available ED data, (2) externally validate the 7mmHg threshold, and (3) refine a threshold for clinically important relative hypotension.

Method and Design This single-centre retrospective cohort study of people aged over 64 linked year 2019 ED attendance data to vital signs at hospital discharges within the previous eighteen months. Hospital frailty risk (HFRS) and Charlson comorbidity scores were calculated. Previous discharge (‘baseline’) vital signs were subtracted from initial ED values to give individuals’ relative change. Cox regression analysis compared relative hypotension exceeding 7mmHg with mean time to mortality censored at 30 days. The relative hypotension threshold was refined using a fully adjusted risk tool formed of logistic regression models. Receiver operating characteristics were compared to NEWS2 models with and without incorporation of relative systolic pressure.

Results and Conclusion 5136 (16%) of 32548 ED attendances were linkable with recent discharge vital signs. Relative hypotension exceeding 7mmHg was associated with increased 30-day mortality (HR: 1.98; 95%CI: 1.66-2.35). The adjusted risk tool (AUC: 0.69; sensitivity: 0.61; specificity: 0.68) estimated each 1mmHg relative hypotension to increase 30-day mortality by 2% (OR: 1.02; 95%CI: 1.02-1.02). 30-day mortality prediction was marginally better with NEWS2 alone (AUC: 0.73; sensitivity: 0.59; specificity: 0.78) and NEWS2 + relative systolic (AUC: 0.74; sensitivity: 0.62; specificity: 0.75).

Comparing ED vital signs with recent discharge observations was feasible for 16% individuals. The association of relative hypotension exceeding 7mmHg with 30-day mortality was externally validated. Indeed, any relative hypotension appeared to increase risk, but model characteristics were poor. These findings are limited to the context of older people with recent hospital admissions.