Intravenous Magnesium and Hospital Outcomes in Children Hospitalized With Asthma

Abstract

BACKGROUND:

Use of intravenous magnesium (IVMg) for childhood asthma exacerbations has increased significantly in the last decade. Emergency department administration of IVMg has been shown to reduce asthma hospitalization, yet most children receiving IVMg in the emergency department are subsequently hospitalized. Our objective with the study was to examine hospital outcomes of children given IVMg for asthma exacerbations.

METHODS:

We conducted a retrospective cohort study using data from the Pediatric Health Information System. We used propensity score matching to compare children who received IVMg on the first day of hospitalization with those who did not. Primary outcomes were initiation and duration of noninvasive positive pressure ventilation. Secondary outcomes included mechanical ventilation (MV) initiation, duration of MV, length of stay, and subsequent tertiary medication use. Primary analysis was restricted to children admitted to nonintensive care inpatient units.

RESULTS:

Overall, 91309 hospitalizations met inclusion criteria. IVMg was administered in 25882 (28.4%) children. After propensity score matching, IVMg was not significantly associated with lower initiation (adjusted odds ratio 0.88; 95% confidence interval [CI] 0.74–1.05) or shorter duration of noninvasive positive pressure ventilation (rate ratio 0.94; 95% CI 0.87–1.02). Similarly, no significant associations were seen for MV initiation, MV duration, or length of stay. IVMg was associated with lower subsequent tertiary medication use (adjusted odds ratio 0.66; 95% CI 0.60–0.72). However, the association was lost when ipratropium was removed from the tertiary medication definition.

CONCLUSIONS:

IVMg administration was not significantly associated with improved hospital outcomes. Further study is needed to inform the optimal indications and timing of magnesium use during hospitalization.

National Heart, Lung, and Blood Institute (NHLBI) guidelines recommend intravenous magnesium (IVMg) sulfate for the management of severe asthma exacerbations that are unresponsive to short-acting β-agonists, anticholinergics, oxygen, and systemic steroids with a goal of avoiding endotracheal intubation.¹ Pediatric studies demonstrate that IVMg improves pulmonary function and may reduce hospital admission when administered early in the emergency department (ED) management course.^2–4 Although these studies were relatively small in size, the use of IVMg for the treatment of pediatric asthma has been rising, increasing from 17% to 36% over the past decade but with no appreciable reduction in inpatient or ICU admission rate.⁵

Recent studies suggest providers are using IVMg to avoid respiratory failure despite the current evidence supporting use of IVMg to decrease hospital admission rate. A cross-sectional survey of pediatric emergency medicine physicians reported that they discharged <5% of their patients from the hospital after receiving IVMg.⁶ The physicians reported the purpose of using IVMg was to reduce ICU admission in children with impending respiratory failure and not to prevent admission in general. Furthermore, in a recent multicenter study of US pediatric EDs, researchers found the medication was typically administered late in the ED course and a large majority of children were admitted to the hospital after IVMg exposure.⁷ Despite IVMg reducing the risk of hospital admission, most children given IVMg are still admitted. For example, <2% of children who receive IVMg for asthma exacerbation are discharged from the ED.⁵

Although most patients who receive IVMg in the ED are admitted to the hospital, little is known about the effect of IVMg on hospital outcomes. Our objective with this study was to examine whether IVMg administration in children hospitalized for asthma exacerbation was associated with improved hospital outcomes.

METHODS

Study Design

We conducted a multicenter retrospective cohort study of children admitted to children’s hospitals that contribute data to the Pediatric Health Information System (PHIS) database. The PHIS database includes administrative data from 49 tertiary care children’s hospitals across the United States that are affiliated with the Children’s Hospital Association (Lenexa, KS). Data quality and reliability are ensured through a joint effort between the Children’s Hospital Association and participating hospitals. Hospitals submit discharge and encounter data, including demographics, diagnoses, procedures, and International Classification of Diseases, 10th Revision, Clinical Modification codes, as well as detailed pharmacy information.

Study Population

We included all hospitalized patients aged 2 to 18 years of age who were discharged from a PHIS hospital between October 15, 2015, and October 14, 2018, with a primary diagnosis of asthma (diagnosis related group 141). We included data from the 49 hospitals that consistently participated in PHIS during the study time frame. Patients transferred from outside facilities including EDs, were excluded from the study because outside hospital medication administration was not available. Children were allowed to contribute >1 admission because previous IVMg exposure likely does not affect the clinical decision to administer IVMg. Children admitted to the ICU from the ED were analyzed as a secondary stratified analysis.

Primary Exposure

The primary exposure was IVMg administered on the calendar day of ED presentation. The PHIS database does not capture location of administration. Nebulized magnesium was excluded from the study because this route of administration is less common in the United States and has limited efficacy.³

Outcomes

The primary outcome measures were initiation and duration of noninvasive positive pressure ventilation (NPPV), the latter defined as the number of calendar days a patient received bilevel positive airway pressure or continuous positive airway pressure. The primary outcome was chosen because NPPV use is a more objective surrogate of respiratory decompensation than other hospital outcomes like ICU use, which varies substantially by hospital.⁸ NPPV is also more common than other outcomes like mechanical ventilation (MV) or death and can be readily identified within the PHIS database. Lastly, NPPV has plausibly fewer determinants than more process-oriented outcomes like duration of hospitalization, which can be affected by considerations as varied as exacerbation severity to extenuating social circumstances.⁹ Secondary outcome measures included MV initiation (defined as invasive MV), duration of MV, subsequent tertiary medication use (defined as the patient receiving ipratropium, terbutaline, aminophylline, theophylline, or epinephrine) after day 1, and hospital length of stay (LOS), defined as the duration of hospitalization in calendar days.

Propensity Score Matching

Covariates included age, number of complex chronic conditions (CCCs), number of ED visits or hospitalizations in the previous year, non-Hispanic Black race, Hispanic ethnicity, and intravenous (IV) steroid, ipratropium, IV fluids, blood gas (defined as capillary blood gas, arterial blood gas, or venous blood gas), and performance of a chest radiograph, all administered or performed on day 1. These day-1 covariates were chosen because they are interventions that occur in severe asthma, usually in the ED, typically before admission and NPPV administration. These test and treatment covariates were chosen by the study team as potential markers for severe asthma exacerbation that would predict receiving IVMg. The balance of covariates across magnesium-treated and non–magnesium-treated patients after matching 1:1 were assessed by using standardized differences. The standard cutoff of an absolute difference of >0.25 was considered significant.^10,11 In brief, the propensity score is the probability that the patient would receive the treatment (in this case IVMg) on the basis of the characteristics of the patient, clinician, and clinical setting (study covariates listed above). Matching patients on the basis of propensity score is used to minimize bias and confounding when analyzing nonrandomized data. The number of cases included at each hospital can be found in Supplemental Table 4.

A propensity score was calculated by first constructing a multivariable logistic regression model with IVMg exposure as the dependent variable (model c statistic = 0.802). Those who received IVMg and those who did not receive IVMg were matched on propensity score with use of a greedy nearest-neighbor matching with a caliper set at one-quarter of the SD of the logit of the propensity scores. We forced matches within hospitals to avoid biases introduced as a result of across-hospital variability in outcomes and on the receipt of IV steroids.

Statistical Analysis

Characteristics of the exposure groups were summarized by using frequencies and percentages for categorical variables and χ² tests for bivariate comparisons. For the main propensity score matched model, hierarchical models with a random effect for each hospital were used. We used binomial distributions for binary outcomes and normal distributions with log-transformed for continuous outcomes. A sensitivity analysis was performed, restricting the study population to those who received IV steroids on the day of admission. A sensitivity analysis was also performed excluding ipratroprium from the definition of tertiary medications. All analyses were performed by using SAS version 9.4 (SAS Institute, Inc, Cary, NC), and P < .05 was considered statistically significant. The institutional review board at our medical center determined that this study of deidentified data did not constitute human subjects research.

RESULTS

Study Population

Of the 91 309 hospitalizations that met the inclusion criteria, IVMg was administered in 25 882 (28.4%) either in the ED or inpatient floor on the day of admission (Table 1). Those who received IVMg were more likely to be older and more likely to be transferred to the ICU. (Table 1).

TABLE 1.

Demographic Characteristics of Children Admitted to the Inpatient Floor

	Overall	No IVMg	IVMg
n (%)	91 309	65 427 (71.7)	25 882 (28.4)
Age, median (IQR), y	5 (2–8)	4 (2–8)	6 (3–10)
Age group, n (%), y
<5	43 245 (47.4)	33 979 (51.9)	9266 (35.8)
5–11	37 900 (41.5)	25 241 (38.6)	12 659 (48.9)
12–17	10 164 (11.1)	6207 (9.5)	3957 (15.3)
Sex, n (%)
Male	56 458 (61.8)	40 732 (62.3)	15 726 (60.8)
Female	34 849 (38.2)	24 694 (37.7)	10 155 (39.2)
Race, n (%)
Non-Hispanic white	26 045 (28.5)	20 325 (31.1)	5720 (22.1)
Non-Hispanic Black	38 159 (41.8)	25 376 (38.8)	12 783 (49.4)
Hispanic	17 955 (19.7)	13 059 (20)	4896 (18.9)
Asian American	2109 (2.3)	1445 (2.2)	664 (2.6)
Other	7041 (7.7)	5222 (8)	1819 (7)
Payer, n (%)
Government	60 380 (66.1)	42 877 (65.5)	17 503 (67.6)
Private	27 363 (30)	19 892 (30.4)	7471 (28.9)
Other	3566 (3.9)	2658 (4.1)	908 (3.5)
ICU flag, n (%)	909 (1)	356 (0.5)	553 (2.1)
No. asthma encounters in previous y (%)
0	62 901 (68.9)	46 539 (71.1)	16 362 (63.2)
1–2	22 478 (24.6)	14 988 (22.9)	7490 (28.9)
3+	5930 (6.5)	3900 (6)	2030 (7.8)
LOS, median (IQR), d	1 (1–2)	1 (1–2)	1 (1–2)

All comparisons significant at P < .001. IQR, interquartile range.

Use of IVMg Across Hospital and Locations

There was wide variation of IVMg use across hospitals, ranging from 0.0% to 64.1% of hospitalized children (Fig 1).

FIGURE 1.

Hospital variation in IVMg exposure in children with asthma. The percentage of those receiving IVMg by hospital is shown.

Propensity Match Characteristics

Matching on propensity scores reduced differences between IVMg recipients and nonrecipients (Table 2). The absolute standard difference was <0.25 across all postmatch covariates.

TABLE 2.

Pre— and Post—Propensity Match Cohort Characteristics of Patients Admitted to the Inpatient Floor

	Prematch		Postmatch
	No IVMg	IVMg	No IVMg	IVMg	Std Diff
N or n (%)	65 427 (71.7)	25 882 (28.4)	15 030	15 030	—
Age group, n (%), y
<5	33 979 (51.9)	9266 (35.8)	6364 (42.3)	6276 (41.8)	0.006
5–11	25 241 (38.6)	12 659 (48.9)	6713 (44.7)	6825 (45.4)	−0.007
12–17	6207 (9.5)	3957 (15.3)	1953 (13.0)	1929 (12.8)	0.002
CCC count, n (%)
0	60 340 (92.2)	24 417 (94.3)	14 042 (93.4)	13 996 (93.1)	0.003
1	3831 (5.9)	1277 (4.9)	833 (5.5)	884 (5.9)	−0.003
2+	1256 (1.9)	188 (0.7)	155 (1.0)	150 (1.0)	0.000
No. asthma encounters in previous y (%)
0	46 539 (71.1)	16 362 (63.2)	10 096 (67.2)	9782 (65.1)	0.021
1–2	14 988 (22.9)	7490 (28.9)	3906 (26)	4140 (27.5)	−0.016
3+	3900 (6.0)	2030 (7.8)	1028 (6.8)	1108 (7.4)	−0.005
Non-Hispanic Black race, n (%)	25 914 (39.6)	12 979 (50.1)	6957 (46.3)	6964 (46.3)	0.000
Hispanic, n (%)	13 059 (20.0)	4896 (18.9)	2779 (18.5)	3202 (21.3)	−0.028
IV steroids use on d 1, n (%)	15 991 (24.4)	13 430 (51.9)	6750 (44.9)	6750 (44.9)	0.000
Ipratropium use on d 1, n (%)	40 888 (62.5)	21 276 (82.2)	11 324 (75.3)	1 309 (75.2)	0.001
IV fluids on d 1, n (%)	17 733 (27.1)	18 691 (72.2)	9112 (60.6)	9131 (60.8)	−0.001
Chest radiograph, n (%)	23 013 (35.2)	10 202 (39.4)	6115 (40.7)	5991 (39.9)	0.008
Blood gas on d 1, n (%)	2399 (3.7)	2816 (10.9)	1206 (8)	1212 (8.1)	0.000

All prematch comparisons are significant at P < .001. Std Diff, standardized difference; —, not applicable.

Magnesium and Hospital Outcomes in Patients Admitted to the Inpatient Floor

In our multivariable propensity matched inpatient unit model, 293 (1.9%) of those receiving IVMg received NPPV compared with 321 (2.1%) of those not receiving IVMg. IVMg was not associated with NPPV initiation or duration of NPPV use (Table 3). In the matched cohort, a total of 2026 (13.4%) patients received tertiary medication after admission. Exposure to IVMg was associated with decreased tertiary medication use. Using sensitivity analysis excluding ipratropium from tertiary medications, we found no association between IVMg and inpatient floor tertiary medication use (adjusted odds ratio 0.88; 95% confidence interval [CI] 0.66–1.17). There was no association between IVMg and use of MV, duration of MV, or LOS (Table 3, Fig 2).

TABLE 3.

Outcomes Post–Propensity Score of Matched Cohort of Patients Admitted to the Inpatient Floor

	Prematch			Postmatch
	No IVMg	IVMg	P	No IVMg	IVMg	Adjusted OR or RR (95% CI)	P
NPPV	694 (1.1)	583 (2.3)	<.001	321 (2.1)	293 (1.9)	0.88 (0.74–1.05)	.151
NPPV in d	1.8 (1.7–1.9)	1.5 (1.5–1.6)	<.001	1.7 (1.7–1.9)	1.6 (1.5–1.7)	0.94 (0.87–1.02)	.152
MV	433 (0.7)	348 (1.3)	<.001	176 (1.2)	182 (1.2)	1.09 (0.86–1.38)	.483
MV in d	1.2 (1.1–1.2)	1.1 (1.1–1.1)	.154	1.2 (1.1–1.2)	1.1 (1.1–1.2)	0.99 (0.93–1.05)	.676
Tertiary medication use	3861 (5.9)	1752 (6.8)	<.001	1211 (8.1)	895 (6)	0.66 (0.6–0.72)	<.001
Hospital LOS in d	1.3 (1.3–1.3)	1.4 (1.4–1.4)	<.001	1.4 (1.4–1.4)	1.4 (1.4–1.4)	1 (0.99–1.01)	.804

Listed values are geometric means (95% CI). OR, odds ratio; RR, rate ratio.

FIGURE 2.

Forest plot of IVMg and hospital outcomes.

Magnesium and Hospital Outcomes in Patients Admitted to the ICU

In the secondary model including only patients admitted to the ICU, 22.8% of those who received IVMg received NPPV compared with 21.9% of those not receiving IVMg (Supplemental Table 5). Matching on propensity scores reduced differences between IVMg recipients and nonrecipients (Supplemental Table 6). The absolute standard difference was <0.25 across all postmatch covariates. IVMg was not associated with NPPV use overall or duration of NPPV use (Supplemental Table 7) among patients admitted to the ICU. Lastly, the sensitivity analysis restricting the study population to those receiving IV steroids did not alter the outcome findings (Supplemental Table 8).

DISCUSSION

In this multicenter cohort study leveraging propensity matching methodology, there were 3 main findings. First, we found wide variation in IVMg use across US children’s hospital in those admitted to both the inpatient unit and ICU. Second, IVMg was not associated with changes in initiation or duration of NPPV or MV. Third, among children receiving IVMg, approximately half of the patients received IV steroids and 20% did not receive ipratropium, suggesting many patients are treated per NHLBI guidelines.

Previous studies have revealed that confounding by illness severity is a prominent challenge in studying the effectiveness of IVMg using nonexperimental designs. In a recent study, researchers examined the relationship between IVMg use in the non-ICU inpatient setting using retrospective data, matching 33 children who received IVMg with those who did not receive IVMg using a respiratory assessment score measured at a similar duration of continuously aerosolized albuterol therapy. They found that the children receiving IVMg had nearly twice the length of albuterol therapy (34 vs 18 hours), along with longer LOS (72 vs 49 hours).¹² In another study, researchers pooled data across 6 children’s hospitals in Ohio to assess outcomes associated with adjunctive medications (ipratropium alone versus ipratropium plus IVMg versus IVMg alone) for 242 children receiving continuous albuterol for asthma. They also found that children receiving IVMg had longer durations of continuous albuterol and LOS. Although both of these previous studies attempted to adjust for factors that may be associated with exacerbation severity, they were limited by small sample sizes across a single or few institutions and limited severity adjustment.

In the current study, we used a large, multicenter hospital database and employed several strategies to minimize the impact of confounding by indication. First, we created cohorts with likely different severities of asthma exacerbation (those admitted to the inpatient unit and those admitted to the ICU) and excluded children admitted to the ICU from our primary analysis to reduce cohort heterogeneity. Second, we employed a propensity score matching analysis to compare patients who were otherwise treated similarly within each cohort within each hospital. After propensity score matching, the exposed and unexposed cohorts remained balanced (ie, standard difference <0.25) on components of the propensity score (ipratropium and IV steroid use, blood gas, IV fluids, and chest radiograph, all on the day of presentation). Third, to evaluate a more homogeneous severity population, we performed a sensitivity analysis restricting the population to only those who received IV steroids.

We did not find statistically significant associations between IVMg and NPPV, MV (either initiation or duration), or LOS. The proportion of patients receiving tertiary medication was 13.4% and we did find less tertiary medication use among those receiving IVMg admitted to the inpatient floor. When ipratropium is removed from the tertiary medication analysis, the association disappears, suggesting ipratropium use may be driving the association between IVMg and tertiary medication use. It is possible that IVMg may have more prominent effects in patients presenting with less severe exacerbations. Only a well-conducted randomized controlled trial would help to address this hypothesis.

Limitations of our findings include the retrospective nature of the study and the lack of clinical patient-level data in the PHIS data set. For this reason, we could not account for patient-level physiologic and laboratory assessments to determine asthma severity in the ED or specify timing of IVMg in hours rather than day of administration. The PHIS database does not reliably measure the duration of albuterol (ie, continuous), high-flow nasal cannula, or oxygen supplementation and these could not be used as outcomes in this study. The PHIS database only includes duration in calendar days. Furthermore, previous ED visits or hospitalization were only captured if they occurred at a PHIS hospital. The PHIS database does not contain the location of medication administration; therefore, we were unable to determine if IVMg delivered on day 1 was given in the ED or inpatient unit. The PHIS data set only includes children’s hospitals, which may limit the generalizability of our results to community hospitals or hospitals without training programs. We were also unable to determine the dose of IVMg delivered, which may have an impact on efficacy.^13,14 The PHIS database contained clinical and pharmaceutical billing information, and it is possible that a medication or respiratory therapy was billed but not administered. Lastly, the overall proportion of patients who experienced the primary outcome of NPPV was relatively low, particularly in the non-ICU inpatient setting, suggesting that children hospitalized for asthma exacerbation tend to have generally good outcomes.

Despite these limitations, our findings represent an important step in understanding the impact of IVMg on hospital outcomes in children hospitalized for asthma exacerbation. This is the first large study used to assess the association between IVMg and NPPV by using more advanced adjustments for exacerbation severity. Further studies using prospectively derived data are needed to determine the effect of IVMg on respiratory support and tertiary medication use provided to children hospitalized for asthma exacerbations.

CONCLUSIONS

Despite increasing use of IVMg in children’s hospitals, it was not associated with changes in NPPV or MV among hospitalized patients. Among patients receiving IVMg, approximately half received IV steroids and 20% did not receive ipratropium, suggesting nonadherence to NHLBI guidelines in many patients. Further study is needed to inform the optimal indications and timing of magnesium use to improve hospital outcomes.