Optimizing Ambulance Transport of Hemodialysis Patients to the Emergency Department: A Cohort Study

BACKGROUND: Dialysis patients who require ambulance transport to the emergency department (“ambulance-ED”) may subsequently require timely dialysis in a monitored setting (“urgent dialysis”). OBJECTIVE: The purpose of this study was to develop and internally validate a risk prediction model for urgent dialysis based on patient characteristics at the time of paramedic assessment before ambulance-ED. DESIGN: Cohort Study SETTING: Region of Nova Scotia, Canada, covered by a single emergency medical services provider PATIENTS: Thrice-weekly hemodialysis patients who initiated dialysis between 2009 and 2013 (follow-up to 2015) and experienced one or more ambulance-ED events. MEASUREMENTS: The primary outcome (“urgent dialysis”) was defined as dialysis within 24 hours of an ambulance-ED in a monitored setting or dialysis within 24 hours of an ambulance-ED with an initial ED potassium of >6.5 mmol/L. Predictors of urgent dialysis based on paramedic assessment before ambulance-ED included presenting complaint, vital signs and time from last dialysis to ambulance dispatch. METHODS: Associations with urgent dialysis were analyzed using logistic regression from which a risk prediction model was created. The model was internally validated using bootstrapping and model performance was assessed by discrimination and calibration. RESULTS: Among 197 patients, there were 624 ambulance-ED events and 87 episodes of urgent dialysis. Weakness as a presenting complaint (odds ratio [OR]: 4.62, 95% confidence interval [CI]: 1.23-17.29), >24 hours since last dialysis (OR: 2.09, 95% CI: 1.15-3.81), and vital signs, including heart rate <60 beats/minute (OR: 3.06, 95% CI: 1.09-8.61), oxygen saturation <90% (OR: 3.04, 95% CI: 1.55-5.94), elevated respiratory rate (≥20 breaths/min), and systolic blood pressure>160 mmHg, were associated with urgent dialysis after ambulance-ED. A risk prediction model incorporating these variables had very good discrimination (C-statistic: 0.81, 95% CI: 0.76-0.86). The negative predictive value was 93.6% using the optimal cut point. Of patients who were predicted to need urgent dialysis but were transported to a facility incapable of providing it, 31% were re-transported for urgent dialysis. LIMITATIONS: Findings of our study may not be generalizable to other centers where the practice of ambulance transfer and availability of monitored dialysis may differ, and data were lacking for potential missed dialysis sessions or changes in routine dialysis scheduling. CONCLUSIONS: Patient characteristics at the time of paramedic assessment are associated with urgent dialysis after ambulance-ED. This risk prediction model has the potential to guide dialysis patient transport to dialysis-capable facilities when needed.