High-Powered Magnet Exposures in Children: A Multi-Center Cohort Study

This is a multicenter, retrospective cohort study. Subjects were identified through standardized institutional chart review, and data were entered into case report forms on a centralized REDCap database maintained by Nationwide Children’s Hospital.^13,14  REDCap (Research Electronic Data Capture) is a secure, web-based software platform designed to support data capture for research studies, providing (1) an intuitive interface for validated data capture; (2) audit trails for tracking data manipulation and export procedures; (3) automated export procedures for seamless data downloads to common statistical packages; and (4) procedures for data integration and interoperability with external sources. The study was conducted under a waiver of informed consent obtained from the institutional review boards of participating hospitals.

A list of >200 children’s hospitals in the United States was reviewed, and researchers within the pediatric emergency medicine, gastroenterology, and surgical divisions at sites were contacted, irrespective of logistics or demographics. Principal investigators were selected on the basis of the researchers’ experience, clinical familiarity with high-powered magnet products, and ability to participate. A total of 25 sites participated in the study representing all 5 main regions of the United States (ie, Northeast, Southeast, West, Southwest, Midwest). States with a participating site included Alabama, California, Colorado, Connecticut, Florida, Georgia, Illinois, Kentucky, Massachusetts, Maryland, Michigan, New York, Ohio, Oklahoma, Pennsylvania, South Carolina, Texas, Washington, and Washington, D.C.

Cases were identified through a query of hospital medical records for patients with any foreign body ICD-10 diagnosis code and any radiographic study within 7 days of diagnosis. The search was limited to patients ages 0 to 21 years seeking care between January 1, 2017, and December 31, 2019 (ie, after magnets re-entered the market). Natural language processing (NLP), often used when analyzing large data sets, was used to limit the number of charts eligible for manual review.^15–17  Researchers at institutions without NLP capabilities manually reviewed each chart to determine eligibility criteria, whereas those with NLP reviewed only those charts containing the keyword “magnet.”

Cases were then reviewed for inclusion utilizing pertinent history, imaging, and operative reports if available. Guidelines for determining a high-powered magnet product were disseminated to researchers. Only cases involving high-powered magnets, as determined by the site principal investigator, were included. In cases in which the magnet type was unclear, history and imaging were reviewed with study principal investigators to determine inclusion eligibility. If the case could not be confirmed, it was not included. Cases with a clear ferrite or ceramic magnet by imaging or history were excluded (Supplemental Information). In addition, all aggregate data were reviewed by the study PIs and any outlying data were clarified with the site PI (ie, quality control).

Upon case identification, clinical, demographic, and outcome data were extracted utilizing a standardized, electronic case report form. Self-reported demographic data including gender, race, and ethnicity were extracted from the medical record when available. This information was collected to assess injury disparities between groups. The patient’s home zip code at the time of injury was also recorded for geographic and socioeconomic data analysis.

All researchers underwent training with one of the study principal investigators (L.K.M.) before site initiation. Training included a deidentified practice chart for which each researcher reviewed and completed data entry. The principal investigator (L.K.M.) reviewed each practice case report form and provided individual feedback to each participating site investigator; no site had REDCap access until these steps were completed. In addition, frequently asked questions with a response were published to the group regularly. Research documents, such as data source hierarchy, detailed explanations for each question within the questionnaire, and a manual of operations were available to site investigators throughout the study. Data quality was reviewed by the same principal investigator (L.K.M.) for each site and clarifying or corrective information was sought as needed.

Clinically significant magnet exposures were defined a priori as injuries that required endoscopic or surgical intervention or resulted in morbidity. Endoscopic procedures included rigid or flexible bronchoscopy, esophagogastroduodenoscopy, antegrade small bowel push or balloon enteroscopy, colonoscopy sigmoidoscopy, and retrograde small bowel balloon enteroscopy. Surgical procedures included diagnostic or exploratory laparoscopy, laparoscopic small bowel repair or bowel resection, open small bowel repair or resection, colonic repair or resection, ileostomy, colostomy, appendectomy, abdominal wash out, and cystotomy. Morbidity was defined as injuries known to occur directly from magnets, which includes perforation, fistula formation, obstruction, bleeding, infection, volvulus, and/or bowel herniation.

All deidentified case data were entered directly into a REDCap database maintained at Nationwide Children's Hospital.

Descriptive statistics were calculated for demographics, injury characteristics, and clinical and outcome variables of interest within the entire cohort and separately by age. Associations between predictor variables (demographic data, symptoms, magnet location, whether magnets passed spontaneously, and need for transfer or imaging) and outcome (injury) were assessed via univariate logistic regression, for which unadjusted odds ratios (ORs) and 95% confidence intervals (CIs) were reported. Injury proportions for the two predictor variables believed to be most likely associated with morbidity a priori (magnet size and number) were also calculated, and differences between groups were compared to the χ² test. Results were considered significant if the P value was <.05. SAS Enterprise Guide version 8.1 (SAS Institute Inc., Cary, NC, USA) was used for all analyses.

A total of 596 patients were identified. The mean age at the time of exposure was 7.6 years, ranging from 2 months to 18 years. The majority of patients were male (60.7%), White (71.6%), and non-Hispanic (73.3%). Of all patients, 103 children (17.3%) had developmental delays, and 30 (5%) had a comorbid health condition including potentially noncontributory conditions such as asthma, congenital heart disease, epilepsy, and diabetes, or more relevant conditions such as gastroschisis (n = 1), eosinophilic esophagitis (n = 2), and eosinophilic colitis (n = 1) (Table 1).

The median exposure was with 2 magnets (interquartile range, IQR, 1–4), with 189 patients (32.5%) having a single magnet exposure. The maximum exposure was 93 magnets. Magnet size could be identified by chart review in 348 patients (58.4%). The median size of these magnets was 5 mm (IQR 5–11); 4 (1.1%) were 2 mm; 21 (6%) were 3 mm; 25 (7.2%) were 4 mm; 184 (52.9%) were 5 mm; and 114 (32.8%) were >5 mm in diameter.

Magnet exposures frequently occurred in the child or family member’s home (34.9%), although many occurred in school or at a child care facility (12.9%). However, a specific exposure location could not be identified on the basis of available medical records in >50% of the cases. Ingestion was the mechanism of exposure in 562 children (96.2%), whereas 17 (2.9%) children presented for management of nasal or aural magnet foreign bodies, 4 (0.7%) for magnets in their genitourinary tract, and 1 patient (0.2%) had magnets in their respiratory tract. Most children (55.7%) required hospitalization, with 4 patients (0.7%) requiring admission to an ICU. The median length of stay was 3 days (IQR 2–4). During diagnosis and management, 81.4% of children received more than 1 radiograph, with a median of 5 radiographs (IQR 3–8) per patient. One child received 43 radiographs, and 36 children (6%) required a computed tomography (CT) scan (Table 2).

In more than half of patients, the magnet(s) passed spontaneously (53.7%). Of the 320 cases of spontaneous passage, 161 patients (50.3%) had a single magnet exposure, 95 patients (29.7%) had 2 magnets, 61 patients (19%) had ≥3 magnets, and 3 patients (0.9%) did not have the number of magnets stated in their record.

A total of 276 (46.3%) children required a procedure for magnet removal or treatment of complications, with 191 (32%) requiring endoscopy alone, 58 (9.7%) requiring surgery alone, and 27 (4.5%) requiring both endoscopy and surgery (Fig 1). Magnet exposure led to morbidity in 57 (9.6%) patients. These included perforation (6%), fistula formation (3.7%), obstruction (2.7%), bleeding (0.7%), infection (0.5%), volvulus (0.2%), and/or bowel herniation (0.2%). This includes 19 children (3.2%) that suffered >1 of these morbidities.

Several clinical and demographic factors were associated with morbidity. Children aged <2 years had the highest likelihood of morbidity. Children between 5 and 9 years old were 33% less likely to suffer morbidity compared to children aged <2 years (95% CI, 0.13–0.89), although there was no difference in injury likelihood among children of other ages (data not shown). Those with developmental delays were >3 times more likely to have an injury (OR 3.25; 95% CI, 1.81–5.85). The greatest risk for morbidity occurred among children seeking care for high-powered magnet exposure symptoms more than once before a definitive diagnosis (ie, a delayed diagnosis) (OR 21.28; 95% CI, 6.84–66.19); those who received a CT scan compared to those who did not (OR 17.67; 95% CI, 8.45–36.92); and those in whom the magnets did not pass spontaneously compared to magnets that did (OR 39.57; 95% CI, 9.55–163.92). Most clinical signs and symptoms were also associated with morbidity, with abdominal distension (OR 51.73; 95% CI, 5.93–451.18) and systemic complaints like fever or malaise (OR 15.74; 95% CI, 4.3–57.59) having the strongest associations. Morbidity was not associated with sex, race, or ethnicity (Table 3).

The number and size of magnets affected both injury patterns and outcomes. Specifically, magnets ≤5 mm or less were seven-fold more likely to be associated with injury compared to magnets >5 mm in diameter (12% vs 1.8%, P = .001). Further, there were no injuries in children with only 1 magnet exposure, but children with 2 or ≥3 high-powered magnets had a 7.5% and 18.7% chance of morbidity, respectively (P <.001 for both) (Fig 2).

The average number of children seeking ED care for high-powered magnet injuries increased 6.1% a year since 2009, although incidence has changed over time after actions by the CPSC and judiciary.⁸  Incidence decreased from an aggregate mean of 3.58 to 2.83 ED visits per 100 000 persons after CPSC enforcement in 2012, but increased to 5.15 visits per 100 000 persons after the CPSC ruleset was overturned in late 2016.⁹  These data, generated from NEISS by 2 independent research groups, have since been supported by injury trends in the National Poison Data System,¹⁸  the database for American Association of Poison Control Centers.

Outcome data in the United States, however, are limited to case reports,^19–32  single center retrospective studies,^33–39  a survey of providers,⁴⁰  and a single retrospective study of 2 institutions.⁴¹  There are no previous large or multicenter studies with clinical care or outcome data. Additional studies are, therefore, needed to quantify morbidity, identify risk mitigation strategies or treatment, and better inform public policy decisions. The latter point is of particular significance given possible reintroduction of federal legislation and a new CPSC ruleset effectively banning high-powered magnet sets, although passage of either is highly uncertain and one that, if successful, will likely take years.^{10–12,42,43} 

Here we present data on 596 cases from 25 children’s hospitals in the 3 years after high-powered magnets reentered the US market (2017 to 2019). These cases were compiled from geographically diverse regions within the United States and included patients evaluated for a confirmed high-powered magnet exposure in any clinical setting (ie, in-patient, outpatient, emergency, or urgent care). Data were obtained by retrospective chart review with trained researchers. The dataset is, therefore, a robust tool from which to quantify risk of high-powered magnets.

The current study reports several important results. Of the 596 children seeking care vis-à-vis high-powered magnets, nearly one in ten had life-threatening morbidity such as bowel obstruction, perforation, infection, bleeding, fistulae, or volvulus. Virtually all care was for ingestion (96.3%) and the majority (55.7%) of children required hospitalization. Although researchers could not identify the location where the injury occurred from the medical chart in most cases, they did identify schools or day cares as a location in a large percent (12.9%). This is consistent with previously published data showing school environments as a frequent location of injury of outside children’s homes, and it indicates parental education is not enough to prevent injuries.⁹  These data also indicate that advocacy and education campaigns within school districts may prevent some injuries.

Importantly, morbidity did not occur in children in which only 1 magnet was ingested and increased with the number of magnets. Many patients (46.3%) required an endoscopy, surgery, or both, although procedures were more common in older as compared to younger children. The children who required a surgical procedure, such as a laparotomy, now carry a 4.6% lifetime risk of adhesive bowel obstruction.⁴⁴  In aggregate, these data suggest that high-powered magnets are among the most dangerous consumer products available today.⁴⁵ 

Under present law, high-powered magnet sets may only be marketed to “adults” ≥14 years of age.¹²  However, 95% of care in our study was for children <14 years of age; the mean patient age within our cohort (7.6 years) is also higher than that of most other foreign bodies.⁴⁶  These data highlight the ease with which children access high-powered magnets. It is also important to note that, whereas most cases and morbidity occurred with “desk toy” size magnets (ie, ≤5 mm), which is the focus of previous CPSC action, a sizeable number of cases (n = 114, 32.8%) and a few (n = 2, 0.34%) injuries occurred with high-powered magnets >5 mm (Fig 2). This suggests that future legislative action may warrant targeting these products as well.

The primary strength of this study is a large sample size gained through a multicenter design with standardized data collection processes among trained researchers. Alternatively, the retrospective design limited the availability of some data and may have led to misclassification bias. For example, identification of developmental delays required a definitive diagnosis in the chart but not multiple confirmatory factors, as used elsewhere.^47,48  Similarly, children with “delayed diagnoses” would be misclassified if symptoms at a previous visit were unrelated to the magnet exposure, as attributed by researchers.

The multicenter design may also have captured differences in management, although variable treatment approaches in a variety of clinical scenarios are widely accepted.⁴⁵  In addition, only using data from children's hospitals may have skewed results toward more significant morbidity because of referral patterns. Future studies exploring optimal treatment algorithms, the role of socioeconomic status as a risk factor for injury, and the utility of current preventative efforts (eg, warning labels) are warranted.

High-powered magnet foreign bodies lead to morbidity and frequent need for recurrent radiation exposure, hospitalization, and endoscopic or surgical intervention in children.

We thank Erica L. Riedesel, MD; Diane Studzinski, BS; Kyra Shreeve, BS; Amir Kimia, MD; Aaron M. Lipskar, MD; Katherine Battisti, MD; Naureen Islam, MPH; Megan W. Wong, BA; Carley Bright, MD; Julia Lieberman, MD; Shelby L. Hall, MD; Caroline Morris, MD; Roberto Gugig, MD; Edaire Cheng, MD; and Christian J Streck, MD.

CI

confidence interval

CPSC

Consumer Product Safety Commission

CT

computed tomography

ED

emergency department

IQR

interquartile range

NEISS

National Electronic Injury Surveillance System

NLP

natural language processing

OR

odds ratio

REDCap

research electronic data capture