Introduction: With rising health care costs impairing access to care, the judicious use of diagnostic tests has become a critical issue for most jurisdictions. Among tests regularly performed in the emergency department (ED), calcium (Ca), magnesium (Mg) and phosphorus (P) laboratory testing represents an annual expenditure of more than $4 million for the Québec health care system. We then sought to determine the best indications for ordering these serum levels by identifying patient risk factors predicting abnormal results. Methods: We are conducting a retrospective cohort study in two academic hospitals of Québec City, one providing acute general care and the other providing specialized care to oncologic and nephrologic patients. We included 1000 patients who had serum Ca and/or Mg and/or P levels prescribed by an emergency physician between January 1st 2016 and May 1st 2016. We are collecting demographic (e.g. age) and clinical (e.g. comorbidities) characteristics identified from literature review as potential explanatory variables of an abnormal serum level. Predictive models of a positive test result will be derived from logistic regressions. Results: We have evaluated 143 patients. ED prevalence rates of hypo- and hyper-calcemia (10.1% and 4,3%), hypo- and hyper-magnesemia (13.0% and 7,2%), hypo- and hyper-phosphatemia (9.5% and 13,9%) were similar to those reported in literature. Preliminary bivariate analysis (p<0.05) have shown that, for patients who had serum Ca/Mg/P levels prescribed, one in four complained of weakness, one in five complained of abdominal pain and one in five presented on physical examination an abnormal mental status. Acute and chronic renal failure appears to be a strong predictor of anomalies of any of those electrolytes. Neoplasia, metastasis, hallucinations, bone pain and confusion are more specifically associated with hypercalcemia. Use of corticosteroids is associated with hypocalcemia. Conclusion: Our bivariate analyses have identified potential risk factors of abnormal Ca/Mg/P results. Multivariate logistic regressions will be conducted on the complete planned cohort to further test these preliminary results.