External validation and update of the early detection rule for severe hyperkalemia among patients with symptomatic...

Chon et al. suggested a high prevalence of severe hyperkalemia (serum potassium ≥ 6.0 mEq/L with electrocardiographic [ECG] changes) among patients with symptomatic or extreme bradycardia. Despite the urgent need to detect and treat severe hyperkalemia, serum potassium result may be available too late and is often spuriously high. Meanwhile, the traditional, descriptive ECG findings of severe hyperkalemia have shown unsatisfactory diagnostic powers. To overcome these diagnostic problems, they outlined the following quantitative rules to facilitate its early detection: Maximum precordial T wave ≥ 8.5 mV (2), atrial fibrillation/junctional bradycardia (1), heart rate (HR) ≤ 42/min (1) with (original rule)/without (ECG-only rule) diltiazem medication (2), and diabetes mellitus (1). Here we report on our external validation of these rules and the resulting updates.

This retrospective, cross-sectional study included all adults with symptomatic (HR ≤ 50/min with syncope/pre-syncope/dizziness, altered mentality, chest pain, dyspnea, general weakness, oliguria, or shock) or extreme (HR ≤ 40/min) bradycardia who visited a university emergency department from 2014 to 2019. After validating the abovementioned rules externally, we selected risk factors of severe hyperkalemia among the ECG findings and easy-to-assess clinical variables by multiple logistic regression analysis. After modelling the updated 'ECG-only' and 'ECG-plus' indices, we internally validated the better of the two by bootstrapping with 1000 iterations.

Among 455 symptomatic/extreme bradycardia cases (70.3 ± 13.1 years; 213 females [46.8%]), 70 (15.4%) had severe hyperkalemia. The previous ECG-only rule showed a c-statistic of 0.765 (95% CI: 0.706-0.825), Hosmer-Lemeshow test of p < 0.001, and a calibration slope of 0.719 (95% CI: 0.401-1.04). On updating, the ECG-plus index summing junctional bradycardia/atrial fibrillation (1), maximum precordial T wave≥8.0 mV (2), general weakness as the chief complaint (2), oxygen demand (1), and dialysis (2) outperformed the ECG-only index (c-statistic, 0.832; 95% CI, 0.785-0.880 vs. 0.764; 95% CI, 0.700-0.828; p = 0.011). On bootstrapping, the c-statistic was 0.832 (95% CI: 0.786-0.878). For scores ≥ 3 (positive likelihood ratio ≥ 5.0), the sensitivity and specificity were 0.514 and 0.901, respectively. For scores ≤ 1, negative likelihood ratio was ≤0.2.

Previous rules showed less satisfactory calibration but fair discrimination to detect severe hyperkalemia in patients with symptomatic or extreme bradycardia. We propose the ECG-plus index as the optimum tool to facilitate its early detection.