Study objective: To determine whether physicians blinded to the serum potassium level can predict hyperkalemia (potassium concentration of more than 5.0 mmol/L) from the ECG.
Design: ECGs of patients at high risk for hyperkalemia were interpreted retrospectively by two physicians blinded not only to the specific clinical diagnosis of the patient and to their serum potassium measurement but also to each other's interpretation. The physicians predicted the presence or absence of hyperkalemia as well as the severity of hyperkalemia on a nominal scale (mild, moderate, or severe).
Setting: The emergency department of a university-affiliated urban county hospital.
Patients: Two hundred twenty consecutive patients admitted to the hospital from the ED with a diagnosis of renal failure or hyperkalemia. Eighty-seven patients had hyperkalemia, and 133 did not.
Results: The sensitivities of the readers for predicting hyperkalemia were .43 and .34, respectively (best positive predictive value, .65). The respective specificities for detecting hyperkalemia were .85 and .86 (best negative predictive value, .69). When only patients with moderate-to-severe hyperkalemia (potassium of more than 6.5 mmol/L) were analyzed, sensitivities were .62 and .55. The readers' ability to predict the severity of hyperkalemia was equally poor.
Conclusion: The ECG is not a sensitive method of detecting hyperkalemia, even in high-risk patients. The specificity of the ECG is better for hyperkalemia, but empiric treatment of hyperkalemia based on the ECG alone will lead to mistreatment of at least 15% of patients.