Abstract
Background and Objective: Hypomagnesemia, defined as a serum magnesium (Mg) level <1.5 mg/dL (0.62 mmol/L), is often asymptomatic. The goals of this study were to determine the incidence of clinically significant abnormal Mg levels in the inpatient setting and to identify diagnoses for which testing would be diagnostically helpful.
Methods: We obtained data from 2010 through 2011 on charges for serum Mg levels and Mg supplementation for all non-ICU inpatients from the 43 tertiary care children’s hospitals in the Pediatric Health Information System database. A manual chart review was performed for all patients at our institution with charges for both Mg levels and Mg supplementation.
Results: A median of 13.5% (interquartile range: 7.7–22.1) of non-ICU inpatients from Pediatric Health Information System centers had charges for Mg levels, at a total charge of $41 million in the 2010–2011 period. At our institution, 19.1% of non-ICU inpatients had charges for Mg levels, at a charge of $67.32/patient-day. Of the 4608 patients with Mg laboratory charges at our institution, 171 (3.7%) had an intervention, defined as addition or modification of an Mg supplement dose in response to a serum Mg level. The 4 most common groups of diagnoses (oncologic, abdominal surgery requiring total parenteral nutrition, solid organ transplant, and short bowel syndrome) accounted for 143 (83.6%) of these interventions.
Conclusions: Serum Mg levels were frequently ordered in non-ICU inpatients, but levels were seldom abnormal and rarely resulted in changes in clinical management. These findings raise concerns about resource overutilization and provide a target for more judicious laboratory ordering practices.
Increases in health care spending continue to be a major area of concern among health policy experts, and the current trajectory of spending has been described as unsustainable,1 with estimates placing health care at 20% of the gross domestic product by 2020. The overtreatment of patients, which includes unnecessary laboratory evaluations, has been estimated to be responsible for up to $226 billion of wasted health care money in 2011 alone. The overuse of laboratory resources is one such area in which quality improvement interventions can help reduce costs.2
Magnesium (Mg) is an intracellular cation with roles in multiple physiologic processes, such as parathyroid metabolism, cardiovascular tone, nerve conduction, and proper function of adenosine triphosphate complexes. However, serum hypomagnesemia (defined as <1.5 mg/dL [0.62 mmol/L]) is often asymptomatic and “not often clinically significant.”3 Studies in PICUs have reported a 20% to 60% incidence of hypomagnesemia on arrival but are divided on its role in predicting mortality and length of ICU stay.4–6 Outside of the ICU, hypomagnesemia has been reported in children with malignancy7 and in up to 15% of those being treated for malnutrition,8 but its role as a routine laboratory assessment has not been studied.
The goal of the present study was to investigate the ordering of serum Mg levels to determine the necessity of such testing and the potential financial impact of its overuse. Based on our observations, we hypothesized that the vast majority of Mg levels ordered on non-ICU patients are not clinically meaningful and result in unnecessary laboratory utilization. We also sought to identify diagnoses most commonly associated with abnormal Mg levels at an individual institution to help target cost-effective ordering practices.
METHODS
This study was a cross-sectional mixed methods study using Pediatric Health Information Systems (PHIS) database (Children’s Hospital Association, Overland Park, KS) data as well as a chart review of local data. The PHIS database contains de-identified administrative data, detailing demographic characteristics, diagnostics, procedures, and pharmacy billing, from 43 freestanding tertiary care children’s hospitals. This database accounts for ∼20% of all annual pediatric hospitalizations in the United States. Data quality is ensured through a joint effort between the Children’s Hospital Association and participating hospitals.9 All 43 PHIS hospitals were represented in the present study.
Inclusion criteria were all children aged 0 to 18 years, admitted in the years 2010 through 2011, with laboratory charges for serum Mg levels. Patients with ICU-level charges were excluded. PHIS data were used to compare Mg ordering practices among similar academic centers. Chart review of local data obtained through PHIS was used to identify both the frequency of clinically meaningful Mg levels and associated diagnoses.
A clinically meaningful Mg level was defined as any result that was followed by an intervention. We defined these interventions as orders for either a new Mg supplement or a modification of an existing Mg supplement, including changes to Mg concentration in total parenteral nutrition (TPN). Patients receiving TPN were included to assess the need for routine Mg-level monitoring for those with consistent electrolyte concentrations. We excluded orders for Mg sulfate for asthma, orders for Mg salts for constipation, and orders for home medications as determined by nursing intake medication reconciliation, unless any of these were preceded by a low Mg level as defined according to reference values. Timing of these interventions varied from within 60 minutes to within 24 hours (in the case of modifying TPN electrolyte concentrations).
National Data Extraction
Aggregate data were extracted from each PHIS institution, including case mix index (CMI) for all patients and for patients with Mg laboratory charges, total patient-days for patients with Mg laboratory charges, total Mg laboratory and supplement charges, and percentage of all cases with Mg laboratory charges. We excluded any patients with ICU charges.
Medians and interquartile ranges were calculated for percentage of patients with Mg laboratory charges and for Mg laboratory charges per patient-day at each PHIS institution. To determine the relationship between patient complexity and Mg utilization, we performed a goodness-of-fit analysis with mean institutional CMI as our independent variable. Using Microsoft Excel (Microsoft Corporation, Redmond, WA), the coefficient of determination value was calculated between CMI and the following: (1) the percentage of patients with Mg laboratory charges at each institution; and (2) Mg supplement charges per patient, as a proxy for frequency of intervention for clinically meaningful Mg levels.
Local Data Extraction
De-encrypted medical record numbers were obtained from PHIS for patients with Mg laboratory charges at our institution, and we performed manual chart reviews on the subset of patients with both Mg laboratory charges and Mg supplement charges. Data were collected on discharge diagnoses, number of Mg levels ordered, timing of Mg level orders, actual serum Mg levels, types of Mg supplements ordered (including TPN), timing of Mg supplement orders, and concentrations of Mg in TPN orders. Appendix 1 presents the reference laboratory values at our institution.
RESULTS
Over the 2-year study period, 823 696 non-ICU inpatient cases were identified; 16.4% (135 085) had Mg laboratory charges. Total laboratory charges over the study period equaled $41 million. Patients with Mg laboratory charges had higher mean CMI (2.2 vs 0.9) and longer mean length of stay (7.4 vs 3.1 days) than other patients. Comparing individual hospitals, there was a wide variation in the percentage of patients with Mg laboratory charges (median: 13.5% [interquartile range: 7.7–22.1]) (Fig 1) and in charges per patient-day for those patients (median: 13.5% [interquartile range: 20.77–55.29]) (Fig 2). Mean CMI did not correlate with percentage of patients with Mg laboratory charges or Mg supplement charges per patient-day (Figs 3 and 4).
Percentage of patients with Mg laboratory charges at all PHIS hospitals. IQR, interquartile range.
Mg laboratory charges per patient-day at all PHIS hospitals (for patients with Mg levels drawn).
CMI versus percentage of patients with Mg levels drawn.
CMI versus Mg supplement charges per patient.
At our institution, there were 24 177 non-ICU inpatient cases during the study period; 4608 patients (19%) had Mg laboratory charges. Demographic characteristics are summarized in Table 1. A total of 764 (17%) of the 4608 patients with Mg laboratory charges also had Mg supplement charges, including TPN. A total of 171 patients had a clinically meaningful Mg level; this group comprised ∼4% of total patients with Mg laboratory charges (Fig 5). They accounted for $316 000 in Mg laboratory charges (16% of all Mg laboratory charges at our institution).
Demographic Characteristics of Pediatric Inpatients With Serum Mg Levels Obtained
Local study population breakdown.
The majority of interventions were in response to low Mg levels as defined according to reference ranges (Fig 5). Four major diagnosis categories (oncologic [including bone marrow transplant], abdominal surgery requiring TPN, intestinal failure/short bowel syndrome, and solid organ transplant) comprised 83% of interventions (Table 2). Appendix 2 contains a comprehensive list of primary and secondary discharge diagnoses in patients with a clinically meaningful Mg level. Some patients account for >1 of the diagnoses listed.
Most Common Diagnosis Groups With Clinically Meaningful Serum Mg Levels (n = 171)
Our study identified 367 patients receiving TPN at our institution who had Mg laboratory charges; 80% of these patients (n = 295) had all Mg levels within the normal reference range. Seventy-nine patients (22%) had interventions for their levels, including 32 with levels within normal reference range; 78 of these patients were in 1 of the 4 major diagnosis categories listed earlier.
There were 13 cases of hypermagnesemia (range: 2.3–2.8 mg/dL) that led to interventions; 11 were associated with TPN, and 10 were in 1 of the 4 major diagnosis categories described earlier. The remaining 3 cases were as follows: (1) a 9-year-old with chronic pancreatitis who had Mg levels in TPN adjusted after a level of 2.6 mg/dL was recorded; (2) a 16-year-old with inflammatory bowel disease who received Mg citrate for bowel preparation before a colonoscopy and had Mg levels in TPN adjusted after a level of 2.4 mg/dL was recorded; and (3) a 14-year-old with status asthmaticus and aortic stenosis who received intravenous furosemide after a dose of intravenous Mg sulfate (peak level: 2.7 mg/dL).
We also identified 68 patients at our institution with a discharge diagnosis of asthma who had both an Mg sulfate supplement charge and an Mg laboratory charge. Twenty-three (34%) patients had a high Mg level (range: 2.3–2.8 mg/dL), but only 1 patient (discussed earlier) received treatment. We also reviewed charts for patients admitted with 4 common pediatric diagnoses (asthma, pneumonia, bronchiolitis, and cellulitis) and found a very low overall incidence of clinically meaningful Mg levels (Table 3).
Common Pediatric Diagnoses and Clinically Meaningful Serum Mg Levels
DISCUSSION
Our study found that the majority of Mg levels ordered in non-ICU pediatric inpatients at our institution did not require intervention. Individual PHIS hospitals exhibited wide variations in frequency of Mg laboratory ordering, raising the question of whether the level of care provided can be just as effective with decreased laboratory usage. Across all PHIS hospitals, there was an association between patients’ CMI and Mg laboratory charges, but data at our institution suggest that the majority of Mg laboratory charges are still ordered for conditions with low incidence of hypomagnesemia or hypermagnesemia. Clinically meaningful Mg levels were clustered in 4 major disease categories, all of which carry high medical complexity. We also found that those patients with clinically meaningful Mg levels only accounted for 16% of the total Mg laboratory charges at our institution. The remaining levels represented $1.6 million in total charges over the study period, many of which may not have been necessary.
Kaplinsky and Alon7 previously described the link between chemotherapy and hypomagnesemia, and we found a high incidence in children undergoing bone marrow transplant, stem cell transplant, and treatment of osteosarcoma. Many of these regimens included cisplatin, which has been linked to renal Mg wasting in children and adults.10,11 We also found several cases of abnormal Mg levels in children with a history of solid organ transplant, which had previously been reported by Staatz et al12 and Al-Rasheed et al13 in association with tacrolimus and cyclosporine, respectively. Many solid organ transplant patients in our study had hypomagnesemia on admission and required intravenous Mg sulfate boluses. Similarly, intestinal failure and short bowel syndrome seemed to predispose children to abnormal Mg levels, which has previously only been described in case reports of adults.14,15 Our data, in conjunction with previous research, suggest routine assessment of Mg levels for children hospitalized for chemotherapy or with a history of a bone marrow or solid transplant, conditions which likely do not represent overuse.
In contrast to previous research,8 we found no cases of hypomagnesemia in patients with eating disorders or protein–calorie malnutrition. We identified 26 patients aged ≥10 years with Diagnosis-Related Group codes for malnutrition or eating disorders. These patients had 190 Mg laboratory charges totaling $24 000; all of the tested levels were in the normal reference range for age. Three patients were started on TPN during their hospitalization, and none received other oral or intravenous Mg supplements.
The role of Mg-level monitoring in children receiving TPN has not been thoroughly studied. We found abnormal Mg levels in patients receiving TPN specifically after abdominal surgery, as opposed to consistently normal levels in patients receiving TPN for nonsurgical illnesses such as acute pancreatitis. This lack of association of low Mg levels with pancreatitis in children is in contrast to the 2012 report by Huang et al,16 who found a 28% incidence of adults with the disease. Baker et al17 reported that most TPN-associated hypomagnesemia was present before initiation of TPN, and our subanalysis found a low incidence of Mg abnormalities with TPN, in addition to the 4 diagnosis groups previously listed. Ali et al18 described 2 cases of iatrogenic hypermagnesemia after pharmacy errors, but no other such associations have been reported.
Hypermagnesemia has also been reported in patients taking oral Mg oxide, but these levels do not require intervention.19,20 In our study population, some of whom took daily Mg supplements at home, the highest Mg level found was 2.8 mg/dL; symptoms usually do not appear until the serum Mg level is >4.5 mg/dL. The only pediatric case reports of life-threatening hypermagnesemia have been in patients receiving excessive doses of Mg supplements21–24 or with late-stage chronic kidney disease.25 Based on our data and literature review, the yield of Mg levels in patients taking oral Mg supplements seems low outside of the 4 major diagnosis categories (oncologic, short bowel, solid organ transplant, and abdominal surgery requiring TPN).
To our knowledge, the present study is the only trial specific to Mg-level ordering practices in pediatrics; however, the issue of unnecessary diagnostic and laboratory evaluations has been well documented in medical literature, especially at academic centers similar to ours.26,27 Our study identified an example of laboratory overuse that lends itself to efforts for quality improvement, such as using audits to change existing behaviors.28–30 A systematic review by Jamtvedt et al28 demonstrated varying degrees of effectiveness of such techniques in achieving change. Interventions tended to be more effective with more aggressive feedback regimens and if compliance with the desired behavior is low at the start of an intervention. Our finding of a limited set of diagnoses associated with hypomagnesemia could be used as a feedback intervention to help target Mg laboratory ordering practices.
Administrative solutions have also been used successfully in decreasing laboratory overutilization. Providing physicians with the price of laboratory tests at time of ordering has been shown to reduce costs in a pediatric ED setting,31 but results have been inconsistent overall.32 In addition, computerized provider order entry can be used to restrict a provider’s ability to order laboratory tests.32,33 Since the completion and analysis of these data, our institution has removed serum Mg levels from most general pediatric admission order sets in an attempt to eliminate indiscriminate ordering.
There are several limitations to the present study. Although we were able to obtain charge data, the true reduction in cost from eliminating all unnecessary Mg studies is unclear. We recognize that the marginal cost of Mg testing may be insignificant compared with the total hospital stay, but bringing continued awareness to the issue of resource overuse could lead to future, more substantial reductions in fixed cost, as suggested by Auerbach and Wachter.34
We were only able to analyze patient-level data at 1 institution; the PHIS database as used in the present study does not contain individual laboratory results for each center. This restriction also prevented us from calculating the significance of the association between patients’ CMI and Mg laboratory charges across all institutions. Clinicians at different institutions may have varying diagnosis-specific protocols for checking Mg levels, and their thresholds for intervention could change depending on the clinical scenario. Our measure of “clinically meaningful” was deliberately subjective and dependent on physician intervention, not on an assessment of symptoms. Even at our institution, interventions for Mg levels did not always correlate to the actual serum level, and we found many interventions for Mg levels that were normal according to reference values.
Other children’s hospitals may also have different proportions of patients with malignancies, solid organ transplants, and short bowel syndrome, which could account for the interhospital variations in percentages of patients with Mg laboratory charges. Laboratory charges themselves could have been undercounted based on individual hospitals’ metabolic panels. In addition, Mg supplement charge data may not include similar elements at different institutions; TPN, for example, may carry several different standard formulations, and their Mg-specific charges could have been underestimated.
Future directions for this project should involve the aforementioned quality improvement methods as well as the development of an educational module by using dissemination and implementation (D&I) methods. D&I is an emerging multidisciplinary science that investigates integration of evidence-based interventions into practice settings, as well as the process of spreading knowledge and information to these settings. A D&I follow-up project would pinpoint behavioral drivers of laboratory overuse and build an effective intervention for a pre-/post-analysis.35,36
CONCLUSIONS
At our institution, serum Mg levels were frequently ordered in non-ICU inpatients who never required intervention for the resulting levels. This finding represented charges of $1.6 million from 2010 through 2011. Most clinically meaningful Mg levels were found in select subsets of patients. At individual children’s hospitals, there was no correlation between patient complexity and the frequency of Mg laboratory ordering. Future directions of study to discourage laboratory overuse include education of medical staff by using a D&I approach and the use of diagnosis-specific admission order sets to facilitate focused ordering of Mg levels.
APPENDIX 1 Reference Ranges for Mg Levels at Children’s of Alabama
| Age | Range (mg/dL) |
|---|---|
| <7 d | 1.2–2.6 |
| 7 d to 2 y | 1.6–2.6 |
| 3 to 6 y | 1.5–2.4 |
| 7 to 10 y | 1.6–2.3 |
| 11 to 14 y | 1.6–2.2 |
| ≥15 y | 1.5–2.3 |
APPENDIX 2 Diagnoses Associated With Clinically Meaningful Serum Mg Levels
| Diagnosis | No. |
|---|---|
| Oncologic | 88 |
| Bone marrow transplant | 32 |
| Osteosarcoma | 22 |
| Hepatoblastoma | 8 |
| Stem cell transplant | 6 |
| Hepatocellular carcinoma | 4 |
| Acute lymphoblastic leukemia | 3 |
| Medulloblastoma | 3 |
| Rhabdomyosarcoma | 3 |
| Acute myelogenous leukemia | 2 |
| Lymphoma | 2 |
| Neuroblastoma | 2 |
| Ovarian carcinoma | 1 |
| Abdominal surgery requiring TPN | 29 |
| Bowel obstruction | 8 |
| Perforated appendicitis | 6 |
| Enterocutaneous fistula | 3 |
| Inflammatory bowel disease | 3 |
| Colectomy | 2 |
| Duplication cyst | 2 |
| Common bile duct obstruction | 1 |
| Malrotation | 1 |
| Splenectomy | 1 |
| Splenic laceration | 1 |
| Umbilical mass | 1 |
| Intestinal failure | 14 |
| Short bowel syndrome | 12 |
| Small bowel transplant | 2 |
| Solid organ transplant | 11 |
| Renal transplant | 9 |
| Liver transplant | 2 |
| Endocrine | 9 |
| Hypocalcemia | 6 |
| Diabetic ketoacidosis | 3 |
| Pulmonary | 4 |
| Status asthmaticus | 3 |
| Cystic fibrosis | 1 |
| Cardiovascular | 3 |
| Congestive heart failure | 1 |
| Hypoplastic left heart syndrome | 1 |
| Tetralogy of Fallot | 1 |
| Hematologic | 3 |
| Sickle cell disease | 3 |
| Renal | 3 |
| Acute renal failure | 2 |
| Fanconi syndrome | 1 |
| Gastrointestinal | 2 |
| Bowel obstruction | 1 |
| Pancreatitis | 1 |
| Infectious | 2 |
| Bacteremia | 2 |
| Neurologic | 2 |
| Muscular dystrophy | 2 |
| Surgical | 1 |
| Burn injury | 1 |
Footnotes
FINANCIAL DISCLOSURE: Drs Narayanan and Scalici have indicated they have no financial relationships relevant to this article to disclose.
FUNDING: No external funding.
POTENTIAL CONFLICT OF INTEREST: Drs Narayanan and Scalici have indicated they have no potential conflicts of interest to disclose.
COMPANION PAPER: A companion to this article can be found on page 42, and online at www.hospitalpediatrics.org/cgi/doi/10.1542/hpeds.2014-0188.
- CMI
- case mix index
- D&I
- dissemination and implementation
- Mg
- magnesium
- PHIS
- Pediatric Health Information Systems
- TPN
- total parenteral nutrition
REFERENCES
- Copyright © 2015 by the American Academy of Pediatrics
In this issue
