Risk Factors and Outcomes After a Brief Resolved Unexplained Event: A Multicenter Study

For this multicenter retrospective cohort study, we used hospital administrative data to identify children <1 year of age presenting over a 3-year period to 15 EDs participating in the BRUE Research and Quality Improvement Network. The institutional review board of each participating hospital approved this study.

Administrative data and medical record review were used for this study. The Pediatric Health Information System (PHIS) was used to identify children from the 11 children’s hospitals that provide data to PHIS. The PHIS database contains demographic, clinical, administrative, and billing data from 50 children’s hospitals affiliated with the Children’s Hospital Association (Lenexa, KS).⁶  The 4 hospitals that do not participate in PHIS performed an identical query of their institutional electronic medical record. For all sites, administrative data were validated and supplemented for each subject by having local trained investigators perform medical record review to (1) confirm eligibility, (2) determine risk factors and event characteristics, (3) determine final diagnoses, and (4) review all visits within their hospital system’s electronic medical record to subspecialty outpatient clinics, ED, and hospital in the first year of life. A single investigator from each site reviewed the record and entered the data in a deidentified fashion into a Research Electronic Data Capture database.⁷  Each chart reviewer trained until an interrater reliability of 0.80 was achieved.

Children aged <1 year were included if they were discharged between October 1, 2015, and September 30, 2018, and they had a discharge code from the International Classification of Diseases, 10th Revision (ICD-10) known to be associated with a BRUE diagnosis (Fig 1). We excluded patients with codes indicating transfer from another hospital or extreme prematurity (P07.01, P07.02, P07.21-P07.25).

To identify potential study subjects, we systematically applied 4 ICD-10 coding strategies designed to minimize sampling bias. This approach involves the review of many patients who by medical record review will not qualify for study participation, but it also reduces underclassification of rare causes and overclassification from BRUE ICD-10 code use. This approach has been used to validate ICD-10 code use to identify urinary tract infection, pneumonia, and febrile infant.^8–10  First, to identify all possible codes used for patients with a BRUE, we rank ordered the PHIS discharge diagnoses used in conjunction with patients assigned the ALTE/BRUE ICD-10 code (R6813). Then, we classified these codes and the ALTE/BRUE code into 4 discrete coding groups or “cohorts.” Cohort 1 included patients with a primary or secondary ALTE/BRUE ICD-10 discharge code (R6813). Cohort 2 identified patients meeting BRUE criteria but discharged from the ED or hospital with a nonspecific diagnosis. The cohort included patients with discharge codes indicating common BRUE symptoms, such as “altered consciousness” or “change in muscle tone.” We included cohorts 3 and 4 to identify patients who presented with a BRUE but did not receive the ALTE/BRUE code because the event was eventually explained by a specific diagnosis. Cohort 3 included serious conditions, such as child abuse and airway abnormalities. Cohort 4 included less serious diagnoses, such as GER or viral infections. Finally, we estimated sample size on the basis of the least specific cohorts 3 and 4 because they were least likely to have a qualifying case. We then performed a weighted convenience sample of each cohort on the basis of the perceived probability of a BRUE and to maximize sensitivity and specificity (40%, 36%, 12%, and 12% for cohorts 1–4, respectively).^2,3  Cohorts 1 and 2 were oversampled, given their specificity, prevalence in previous ALTE/BRUE publications, and higher likelihood of qualifying patients.^3,4 

We used medical record review to validate BRUE and study qualification on the basis of published criteria by the AAP. We excluded patients if they (1) sought ED care for an unrelated reason (eg, ear infection), (2) were transferred from another hospital (because reliability of data could not be guaranteed), (3) had a known history of a comorbid condition recognized as a plausible cause of a BRUE (eg, airway abnormality), (4) presented with a history of objective symptoms precluding a diagnosis of BRUE (eg, fever), (5) did not have at least 1 BRUE characteristic (eg, color change), or (6) had an abnormal physical examination finding precluding a diagnosis of BRUE (eg, fever or sustained change in breathing or mental status). Patients developing viral respiratory symptoms after the initial presentation were not excluded.

Patients were stratified by ED disposition (discharged from the ED versus hospitalized) because risk factors and management may differ. Discharge documentation was primarily used to determine final discharge diagnoses. Patients with a probable or definite diagnosis that explained the event were classified as having a brief resolved explained event (BREE). Patients without a probable or definite diagnosis that explained the event were classified as having a BRUE. A probable or definite diagnosis was defined as at least 1 single diagnosis and, when indicated, a dedicated treatment plan (eg, pneumonia treated with antibiotics). A committee of investigators independently reviewed each case with a probable or definite final diagnosis to determine if the diagnosis could plausibly explain the BRUE and to identify if it qualified as a serious underlying diagnosis. Site investigators provided additional chart detail when indicated to adjudicate any disagreements.

Medical record review was conducted to identify patients with any potential factors placing them at an increased risk for the study outcomes. The factors used were based on the AAP guideline and other BRUE publications.⁴  Risk factors included were age <2 months; prematurity (35–37, 30–34, <30 weeks) and corrected gestational age <45 weeks; history of a similar event, multiple events, or event clusters; event duration >1 minute; indicated use of cardiopulmonary resuscitation (CPR); abnormal medical history such as hospitalization or underlying medical condition; family history of genetic problems associated with BRUEs; and social history concerning for abuse. Although not included in the AAP guideline as a risk factor, each qualifying BRUE characteristic (eg, color change) was also examined as a potential risk factor. Covariates included sex, race or ethnicity, language, and insurance.

Age was categorized as a dichotomous variable in the models at 2 months because the AAP classifies infants <2 months of age as higher risk. In a separate analysis, we evaluated age as a continuous variable using a receiver operating characteristic curve model to further evaluate the predictive value of age with the outcomes of interest and to try to identify an optimal age cut point for maximizing sensitivity and specificity.

The primary outcome was documentation of a serious underlying diagnosis that explained the index event. A serious underlying diagnosis was defined as a condition for which a delay in diagnosis or treatment could potentially increase morbidity or mortality (eg, child abuse). The diagnosis could have been identified before discharge from the index ED or hospital stay or at any time during the first year of life at subsequent visits to the ED, hospital, or subspecialty outpatient clinic. We evaluated 4 secondary outcomes. First, we determined any ED or hospital revisits that occurred before 1 year of age that were potentially related to the index visit or BRUE(s). For example, an encounter was included if the reason for the revisit was another event or, in retrospect, a potential cause of the index event (eg, seizure). Second, to assess the potential benefit of an ED or hospital visit, we evaluated whether a recurrent event occurred during the ED or hospital visit. Third, we evaluated whether observation of a recurrent event contributed to a probable or definitive diagnosis. For example, if there was documentation that an event was observed by a provider, that allowed the event to be characterized as a GER or a seizure. Fourth, we assessed a composite measure for presence of the primary or any secondary outcomes.

Patient factors and event characteristics were summarized for the entire population and stratified by ED disposition. Characteristics of children classified as having a BREE on discharge were compared with characteristics of those without a clear diagnosis (eg, BRUE) within each disposition stratum by using χ² tests for categorical variables and Wilcoxon rank tests for continuous variables. Bivariate analyses of each potential risk factor were conducted to evaluate the independent associations with each of the 4 outcomes and a composite measure of all outcomes. Factors included all the following: age < 60 days, sex, language, insurance, gestational age, corrected gestational age ≤ 45 weeks, history of similar event, event duration > 1 minute, color change, respiratory changes, tone change, altered responsiveness, and abnormal medical, social, or family history. A multivariable regression model adjusting for the same risk factors and covariates was performed for each of the 4 outcomes and a composite of all outcomes. The model accounted for clustering of patients within the hospital with a random intercept for each hospital. The reduced model displays only the significant factors for each of the outcomes stratified by ED disposition. In a separate analysis, researchers calculated the sensitivity, specificity, and negative predictive value (NPV) or positive predictive value (PPV) of the AAP higher-risk criteria on the outcomes of interest.

All statistical analyses were performed by using SAS version 9.4 (SAS Institute, Inc, Cary, NC) and P < .05 was considered statistically significant.

We initially identified over 267 368 encounters and reviewed 4971 medical records (Fig 1). The final study population included 2036 encounters, and 87% percent (1771) met AAP higher-risk criteria because they had at least 1 AAP risk factor. Of the patients who presented to the ED with a BRUE, 1286 (63%) were hospitalized. In Table 1, we compare patient characteristics, risk factors, BRUE characteristics, and outcomes by patient disposition and an explanatory diagnosis.

There were statistically significant differences in patients with explanations compared with patients with no explanation (Table 1). The most common explanations were less serious and self-limited (43.1% [878]). These included GER (18.5% [376]), choking or gagging (8.2% [166]), viral respiratory infections (4.4% [89]), and breath-holding spells (4.1% [84]) (Fig 2). A serious diagnosis was identified in 4.0% (82) of patients, and 45% (37) of these diagnoses were made after discharge from the index visit without an explanation. There was substantial heterogeneity in the serious explanations. The most common diagnoses included seizures and infantile spasms that required treatment with antiepileptic drugs (1.1% [23]), airway abnormalities that required surgery (0.64% [13]), and abusive head traumas (0.34% [7]). In 6 of the 7 patients with abusive head trauma, the diagnosis was made during readmission and was based on imaging findings. No patients died during the initial medical encounter. One patient died after readmission for cardiorespiratory failure at home and subsequent anoxic brain injury not attributed to child abuse.

Of patients discharged from the ED, 2.7% had another BRUE while in the ED and 6.9% (52) had a revisit to the ED or hospital before 1 year of age for reasons related to the index BRUE visit. Of hospitalized patients, 18.2% (234) had another event during the index hospitalization and 10.7% (137) revisited the ED or hospital (Table 1).

In an unadjusted univariate analysis, a history of a similar event was the only AAP risk factor or event characteristic associated with an increased risk of all outcomes (Table 2). Having at least 1 AAP risk factor (ie, higher-risk criteria) was associated with a recurrent event in the ED or hospital (odds ratio [OR] 5.9; 95% confidence interval [CI] 2.7–12.6) and a recurrent event that led to an explanation (OR 15.1; 95% CI 2.1–108.6). Adjusting for covariates, significant risk factors and event characteristics varied between patients discharged from the ED and hospital (Table 3). A history of a similar event was the only factor associated with an increased risk for all outcomes across ED and hospital discharges (P < .05). For patients discharged from the ED, a history of a similar event, event duration >1 minute, abnormal medical history, and altered responsiveness were associated with an increased risk of a serious underlying diagnosis. When age was modeled in a separate analysis as a continuous variable, it did not correlate with worse outcomes, so an age cutoff point could not be determined (Supplemental Fig 3). For a serious underlying diagnosis, having at least 1 of the AAP risk factors had a 97% NPV and 4% PPV. For all outcomes, they had a 90% NPV and 23% PPV (Supplemental Table 4).

Findings from this large multicenter study of patients presenting with a BRUE to the ED provide a better understanding of epidemiological data and risk factors to identify patients at increased risk for important outcomes. First, only 13% of patients did not have any of the AAP higher-risk factors and therefore would have qualified for AAP management recommendations. Second, presence of any of the AAP higher-risk factors had an excellent NPV but a poor PPV for all outcomes. Only a subset of the AAP risk factors helped identify the 4% of patients at risk for a serious underlying condition, event recurrence during the index visit (2.7% in the ED and 18.2% in the hospital), and a revisit (6.9% after ED discharge and 10.7% after hospital discharge). Third, in this study, we quantify the risk of certain outcomes after BRUE and provide guidance that can be used in shared decision-making with families. The low risk of mortality and serious underlying diagnoses provide some reassurance to families and providers. Finally, our study presents comprehensive data on the spectrum of common and less serious explanations for a BRUE. Most of these diagnoses relate to infant immaturity, such as GER, choking and gagging, and breath-holding spells.

Importantly, we found that the majority (87%) of BRUE presentations qualify as higher risk under AAP criteria because they have at least 1 risk factor and that few of these risk factors are associated with adverse outcomes. This has the potential to inform revisions to the AAP BRUE risk stratification.⁴  For example, an event characteristic cyanosis may be substituted for higher-risk criteria such as age and event duration. A history of a similar event, clusters of events, and prematurity continue to be associated with the risk of a recurrent event that, when observed in the hospital setting, may lead to an explanation, but these factors do not appear to be associated with a diagnosis of a serious underlying medical condition. This information may be useful in guiding families and practitioners on the value of hospital admission.

We identified that there is a heterogeneous spectrum of serious diagnoses after a BRUE and within the first year of life. The most common of these were epilepsy, including infantile spasms, airway abnormalities, such as laryngomalacia, and abusive head trauma. Unfortunately, 6 of 7 patients with abusive head trauma were not diagnosed with abusive head trauma during the index encounter. The rate of abusive head trauma in our study was 344 per 100 000, which is 10 times higher when compared with other children aged <1 year.^11–13  Given that many patients were not diagnosed with the serious explanation until after discharge (45% of all serious diagnoses), it is important that providers understand the risk and broad heterogeneity of conditions that could present as a BRUE and consider observation periods or testing when there is a reasonable index of suspicion.

This study has several limitations. There are potential biases introduced by the retrospective nature of our study, and not all adverse outcomes may have been captured. We may have underclassified some patients discharged from the ED or hospital without a BRUE diagnosis and a subset of those with less common serious diagnoses (eg, child abuse). However, we used a comprehensive case ascertainment strategy to minimize these systematic misclassification errors. It is also unlikely that our population included all patients with BRUE because some caregivers who observe BRUEs in their children do not seek medical care. Finally, we may have been underpowered to detect rare adverse events.

This study provides important evidence to guide the management of patients presenting for emergency care for a common problem during the first year of life. Although AAP BRUE risk criteria accurately identify patients at low risk for event recurrence, readmission, and serious underlying diagnosis, they inaccurately identify many patients as higher risk. This is likely because many AAP risk factors are not associated with these outcomes. Findings from this study have the potential to impact revisions to BRUE guidelines, help develop strategies to improve patient outcomes, and guide shared decision-making with families.

We thank Donald Brand, PhD, and all the members of the BRUE Research and Quality Improvement Network for their invaluable contributions.

AAP

American Academy of Pediatrics

ALTE

apparent life-threatening event

BREE

brief resolved explained event

BRUE

brief resolved unexplained event

CI

confidence interval

CPR

cardiopulmonary resuscitation

ED

emergency department

GER

gastroesophageal reflux

ICD-10

International Classification of Diseases, 10th Revision

NPV

negative predictive value

OR

odds ratio

PHIS

Pediatric Health Information System

PPV

positive predictive value