Derivation of cohort
Magnesium sulphate use was determined from pharmacy records at Capital Health (New Halifax Infirmary and Victoria General sites). This institution is a tertiary care center for Nova Scotia (population approximately 960,000) but also provides specialized care for the other Atlantic Provinces. Since magnesium sulphate is not unit dosed, floor consumption was estimated from quarterly use of existing floor stock records. Fiscal year data was available from 2003 to the last quarter of 2013. The focus of this analysis was intravenous magnesium use on the in-hospital medical units (as opposed to operating rooms, emergency and out-patients clinics) and did not include magnesium used in parental nutrition solutions. The units of interest included all intensive care units (cardiac care unit, cardiac surgery, and two medical-surgical intensive care units), surgical floors (general (9A and 9B), urology (5B), surgical oncology (5A), thoracic (6A), orthopedics (7.2), neurosurgery (7.3), vascular (4.1 and 7.1) medical floors [general (8.2), neurology (8.1), hematology (8B), cardiology (6.2 and 6.4)] and mixed floors (6B and 8.3).
Since this institution level data does not identify which patients received intravenous magnesium and what the indications were, a detailed chart review of patients from five floors (cardiology (6.2), cardiac care unit (6.4), general surgery (9A and 9B), mixed surgery/nephrology (6B), general medicine (8.2 and 8.4)) was also undertaken. These floors were identified to have considerable increases in magnesium use over the time period.
Derivation of restricted cohort
All consecutive patients discharged (including death) in April 2006 and April 2013 were reviewed to eliminate selection bias. The year 2006 was chosen since this was the onset of the increase in use of intravenous magnesium and this was the year that hospital records were made electronic. April 2013 was selected as this allowed 6 months of follow up at the time of data collection. Patients discharged within 24 h, those who had no blood work during the admission and those who were readmitted within the month were excluded.
Outcome and potential confounders
The primary outcome of the analysis on this restricted cohort was intravenous magnesium use (yes/no). To identify factors associated with magnesium use, data was collected on discharge diagnosis, whether patients were admitted through emergency, whether their stay included time in an intensive care unit, and whether they received surgery. Information on patient co-morbidity was collected including ischemic heart disease, congestive heart failure, atrial fibrillation, chronic kidney disease, hypertension, diabetes mellitus, liver disease, alcohol abuse (as defined in the records), cancer and chronic obstructive pulmonary disease. Comorbid conditions were based on documentation through electronic patient records. All encounters including inpatient and outpatient were reviewed. Hospital admission laboratory data including serum creatinine, potassium, hemoglobin, albumin, calcium and magnesium was collected. Magnesium, albumin, calcium, and potassium levels were also collected at the time of intravenous magnesium use if available. Estimated glomerular filtration rate (GFR) was measured by the four variable Modified Diet in Renal Disease (MDRD) formula [25]. The charts were reviewed to determine if patients were on oral magnesium, calcineurin inhibitors, proton pump inhibitors or non-potassium sparing diuretics. For patients with a history of cancer the chart was reviewed to look for evidence of prior cisplatin administration. Admission and in-hospital EKG reports were examined for a diagnosis of arrhythmia (including atrial fibrillation, ventricular tachycardia or fibrillation, and long corrected QT interval-as defined by the cardiologist report). In patients receiving intravenous magnesium follow up blood work was reviewed to determine if serum magnesium levels had been done after discharge within a 6-month period. In addition to patient factors, each patient record was reviewed to determine the indication for intravenous magnesium administration. Since there are no written evidence based guidelines for intravenous magnesium administration, prior to data analysis, the authors after a review of the literature agreed indications should include severely low levels (magnesium ≤0.5 mmol/L), or a low level (0.5 to 0.74 mmol/L) in patients with long corrected QT or other arrhythmia, low potassium (≤3.0 mmol/L), low corrected calcium (≤2.0 mmol/L), neuromuscular problems or ongoing gastrointestinal loss (diarrhea, vomiting, post operative ileus or bowel obstruction) [1, 26].
Serum magnesium was measured on the hospital’s Synchron DxC (Beckman Coulter Inc., CA). The reported normal range is 0.74 to 1.03 mmol/L. The co-efficient of variation is 3.8 % in plasma. There were no changes in methodology over the 10-year period. A corrected magnesium and calcium was calculated if an albumin level was available. The formulas were as follows: corrected magnesium [mmol/L] = magnesium [mmol/L] + 0.005*(40-albumin [g/L]) and corrected calcium [mmol/L] = Calcium [mmol/L] + 0.025* (40-albumin [g/L]) respectively [27, 28].
Statistical analysis
Data for magnesium use in the institution was collected by fiscal year. Baseline characteristics of patients from 2006 and 2013 were described as means ± standard deviation (SD) for continuous normally distributed variables, median and interquartile range for non-normally distributed variables and number and percentage for categorical variables. Statistical tests of comparison were performed using independent t-tests, the Rank Sum test and Fisher’s Exact test, accordingly. Multivariable nested logistic regression models were used to determine if there was an independent association between era (2013 versus 2006) and use of intravenous magnesium among those patients with available magnesium laboratory measurements. Variables selected for inclusion in the multivariable models were performed a priori and included demographics, co-morbidities, medications associated with hypomagnesaemia, presence or absence of a low serum magnesium level at hospital admission, and whether or not there were indications to give magnesium intravenously based on rationale outlined above. For all statistical analyses a 2 sided P value <0.05 was considered statistically significant. All analyses were performed using Stata IC version 12.0 (StataCorp LP, College Station, TX). Research ethics approval to conduct this study was obtained from Capital Health Research Ethics Board.