Endocrine Abstracts

Introduction: Hypo-&hyperkalaemia are common laboratory abnormalities, complicating up to 10% of all hospital admissions and contributing to mortality. Ideally, patients with mild deviations can be treated as outpatients, and only those patients with objectively severe or life-threatening levels are hospitalized. Once admitted, there are few data as to whether the degree of electrolyte disturbance consumes greater resource(s). We analysed the admission and discharge data, for acute admissions, with reference to the admission potassium value.

Methods: Clinical data for all unscheduled admissions were retrieved from the electronic health record (EHR). We used the CogStack ecosystem to access structured fields in the EHR. We analysed a 12-month cohort of all patients who had an A&E discharge summary created between 1st Jan 2017 and 1st Jan 2018 (dates of admission min 21st December 2016 – 1st Jan 2018). For each admission, the laboratory U&Es were obtained, and first potassium identified. Cox proportional hazard models evaluated the independent effect of potassium on likelihood of discharge, with eGFR, age and sex as covariates, treating potassium as categorical data (‘Hypokalaemia’ <3.5 mmol/l and ‘hyperkalaemia’ ≥5.0mmol/l) and continuously using linear spline terms (boundary knots at 3.5 and 5.0 mmol/l).

Results: In 12-months, there were 138,307 visits by 98,357 unique patients in the Emergency department. Laboratory potassium was measured in 36,631 attendances (not including haemolysed samples). Hypokalaemia was found on the initial sample in n = 2095 (5.7%), hyperkalaemia in n = 1581 (4.3%) and eukalaemia in n = 32,955 (90.0%). The eGFR was missing in n = 3310, of whom n = 2763 (83%) had potassium 3.5 to 5 mmol/l. In the categorical multivariable model, hypo- (HR : 0.644, 95% CI 0.613 - 0.676 P = <0.0001) and hyperkalaemia (HR 0.812, 95% CI: 0.765 - 0.862 P < 0.0001) remained independent predictors after adjusting for eGFR (HR 1.009 (95% CI 1.008 - 1.009; P <0.0001) and age (HR 0.990, 95% CI 0.989-0.990; P = <0.0001). The spline model demonstrated an n shaped relationship between admission potassium and hazard of discharge against time, the likelihood of discharge crossing unity between 4.0-5.0 mmol/l.

Conclusion: Both hypo-&hyperkalaemia are independently associated with prolonged length of stay for acutely hospitalised patients, even at less extreme levels of dyskalaemia.