The Problems with Risk Prediction during an Emergency Medical Admission Using Laboratory Data – Evidence from Potassium
Background: The prediction of clinical outcomes using biochemical markers is an important tool. Methods: We calculated a risk score for all emergency admissions 2002-2017. We related potassium and mortality in a multivariable fractional polynomial model. We investigated the potassium distribution and relationship of potassium to mortality over time. Results: There were 106,586 admissions in 54,928 patients. Mortality was higher for those with an admission potassium above the median – 6.1% vs 4.6% (p<0.001), OR 1.07 (95%CI: 1.06, 1.09). There was a progressive increase in mortality from the lowest – 8.9% (95%CI: 8.3%, 9.4%) to highest potassium decile – 14.2% (95%CI: 13.5%, 14.8%). The frequency of admission hypokalaemia and the mortality at any given potassium decreased over time. Conclusion: Admission potassium predicts mortality.