Elizabeth R. Alpern

Objective: To evaluate interrater reliability and validity of the Australian Mental Health Triage Scale (AMHTS) for children presenting to the emergency department (ED) with behavioral health symptoms. Methods: We conducted a prospective observational study of ED visits by children aged 3–17 with a primary behavioral health symptom at two US children's hospitals. Nurses assigned Emergency Severity Index (ESI) and AMHTS Scores. Interrater reliability was assessed using intraclass correlation coefficients (ICCs) on visits with two independently assigned AMHTS Scores. To evaluate validity, quasi‐Poisson regression examined the association of AMHTS Scores and ED resource utilization, adjusted for sociodemographic and clinical factors. Resource use was categorized using the standard ESI definition and an expanded definition incorporating behavioral health‐specific resources. Results: We included 454 visits (45.4% aged 10–14, 56.2% female) with at least one AMHTS Score, of which 94.1% were ESI 2 and 26.9% were AMHTS 2. Among 221 visits with two AMHTS Scores, ICC was 0.48 (95% CI 0.37, 0.58). Higher‐acuity AMHTS Scores were associated with greater resource utilization. Standard resource use was higher for AMHTS Score 1 (Incidence Rate Ratio [IRR] 1.82, 95% CI 1.12, 2.96) and Score 2 (IRR 1.64, 95% CI 1.15, 2.40) versus AMHTS Score 5. When applying the expanded definition, resource use was higher for AMHTS Score 1–4 versus 5. Conclusion: Interrater reliability of AMHTS was fair, and higher‐acuity scores were associated with greater ED resource utilization. AMHTS Scores were more widely distributed than ESI Scores, suggesting a more responsive approach to pediatric behavioral health triage.

Objectives: Firearm injuries are a leading cause of pediatric mortality, but may not be adequately identified by International Classification of Diseases (ICD) codes. Natural language processing (NLP) could improve identification. We derived an NLP model from medical narratives and compared the sensitivity and specificity of ICD codes and the NLP models to identify pediatric acute firearm injury (AFI). Methods: Retrospective study of ED visits by patients (aged 0 to 17 y) presenting to 7 pediatric emergency departments (ED) with a potential AFI between 2011 and 2019. A medical narrative review was performed to identify AFI narratives. These narratives, plus randomly selected narratives, underwent language feature extraction with logistic regression to identify topics significantly associated with the categories. Two Random Forest models were created: model 1 on all narratives associated with a potential AFI and model 2 using the confirmed AFI narratives. Model 2 was then applied to a new, unreviewed selection of random narratives. A second round of medical narrative review was performed on the random narratives, including those identified by model 2 as AFI. Sensitivity and specificity of ICD and NLP for the identification of AFI were calculated. Results: A total of 2407 ED visits for potential AFI were identified. Of those, 1183 narratives were annotated by narrative review as AFI. In the randomly selected narratives (n=1150), 47 additional AFI were identified, 16 by NLP and 31 by narrative review, for a total of 1230 AFI. ICD codes correctly identified 1171 (95.2%) AFI, with a sensitivity of 95.2% (93.8% to 96.3%) and specificity of 93.5% (92.6% to 94.4%). NLP correctly identified 1041 (84.6%) AFIs in the final data set with a sensitivity of 84.6% (82.5% to 86.6%) and specificity of 94.7% (93.8% to 95.5%). Conclusions: ICD and NLP both had acceptable sensitivity and specificity for identifying pediatric AFI. Future directions include evaluating newer NLP models to address false positives and the identification of the intent of injury.

The reported impacts of the COVID-19 pandemic on child maltreatment in the United States have been mixed. Encounter trends for child physical abuse within pediatric emergency departments may provide insights. Thus, this study sought to determine the change in the rate of emergency department encounters related to child physical abuse. A retrospective study within the Pediatric Emergency Care Applied Research Network Registry. Encounters related to child physical abuse were identified by 3 methods: child physical abuse diagnoses among all ages, age-restricted high-risk injury, or age-restricted skeletal survey completion. The primary outcomes were encounter rates per day and clinical severity before (January 2018–March 2020) and during the COVID-19 pandemic (April 2020–March 2021). Multivariable Poisson regression models were fit to estimate rate ratios with marginal estimation methods. Encounter rates decreased significantly during the pandemic for 2 of 3 identification methods. In fully adjusted models, encounter rates were reduced by 19% in the diagnosis-code cohort (adjusted rate ratio: 0.81 [99% confidence interval: 0.75–0.88], P <.001), with the greatest reduction among preschool and school-aged children. Encounter rates decreased 10% in the injury cohort (adjusted rate ratio: 0.90 [confidence interval: 0.82–0.98], P = .002). For all 3 methods, rates for lower-severity encounters were significantly reduced whereas higher-severity encounters were not. Encounter rates for child physical abuse were reduced or unchanged. Reductions were greatest for lower-severity encounters and preschool and school-aged children. This pattern calls for critical assessment to clarify whether pandemic changes led to true reductions versus decreased recognition of child physical abuse.

Objective: This study aimed to compare the performance of a pediatric decision support algorithm to detect severe sepsis between high-risk pediatric and adult patients in a pediatric emergency department (PED). Methods: This is a retrospective cohort study of patients presenting from March 2017 to February 2018 to a tertiary care PED. Patients were identified as high risk for sepsis based on a priori defined criteria and were considered adult if 18 years or older. The 2-step decision support algorithm consists of (1) an electronic health record best-practice alert (BPA) with age-adjusted vital sign ranges, and (2) physician screen. The difference in test characteristics of the intervention for the detection of severe sepsis between pediatric and adult patients was assessed at 0.05 statistical significance. Results: The 2358 enrolled subjects included 2125 children (90.1%) and 233 adults (9.9%). The median ages for children and adults were 3.8 (interquartile range, 1.2–8.6) and 20.1 (interquartile range, 18.2–22.0) years, respectively. In adults, compared with children, the BPA alone had significantly higher sensitivity (0.83 [95% confidence interval {CI}, 0.74–0.89] vs 0.72 [95% CI, 0.69–0.75]; P = 0.02) and lower specificity (0.11 [95% CI, 0.07–0.19] vs 0.48 [95% CI, 0.45–0.51; P < 0.001). With the addition of provider screen, sensitivity and specificity were comparable across age groups, with a lower negative predictive value in adults compared with children (0.66 [95% CI, 0.58–0.74] vs 0.77 [95% CI, 0.75–0.79]; P = 0.005). Conclusions: The BPA was less specific in adults compared with children. With the addition of provider screen, specificity improved; however, the lower negative predictive value suggests that providers may be less likely to suspect sepsis even after automated screen in adult patients. This study invites further research aimed at improving screening algorithms, particularly across the diverse age spectrum presenting to a PED.

Objectives: Wait time for emergency care is a quality measure that affects clinical outcomes and patient satisfaction. It is unknown if there is racial/ethnic variability in this quality measure in pediatric emergency departments (PEDs). We aim to determine whether racial/ethnic differences exist in wait times for children presenting to PEDs and examine between-site and within-site differences. Methods: We conducted a retrospective cohort study for PED encounters in 2016 using the Pediatric Emergency Care Applied Research Network Registry, an aggregated deidentified electronic health registry comprising 7 PEDs. Patient encounters were included among all patients 18 years or younger at the time of the ED visit. We evaluated differences in emergency department wait time (time from arrival to first medical evaluation) considering patient race/ethnicity as the exposure. Results: Of 448,563 visits, median wait time was 35 minutes (interquartile range, 17–71 minutes). Compared with non-Hispanic White (NHW) children, non-Hispanic Black (NHB), Hispanic, and other race children waited 27%, 33%, and 12% longer, respectively. These differences were attenuated after adjusting for triage acuity level, mode of arrival, sex, age, insurance, time of day, and month [adjusted median wait time ratios (95% confidence intervals): 1.11 (1.10–1.12) for NHB, 1.12 (1.11–1.13) for Hispanic, and 1.05 (1.03–1.06) for other race children compared with NHW children]. Differences in wait time for NHB and other race children were no longer significant after adjusting for clinical site. Fully adjusted median wait times among Hispanic children were longer compared with NHW children [1.04 (1.03–1.05)]. Conclusions: In unadjusted analyses, non-White children experienced longer PED wait times than NHW children. After adjusting for illness severity, patient demographics, and overcrowding measures, wait times for NHB and other race children were largely determined by site of care. Hispanic children experienced longer within-site and between-site wait times compared with NHW children. Additional research is needed to understand structures and processes of care contributing to wait time differences between sites that disproportionately impact non-White patients.

OBJECTIVES: To determine the time to positivity for bacterial pathogens and contaminants in blood and cerebrospinal fluid (CSF) cultures in a cohort of febrile infants ≤60 days of age. METHODS: This was a secondary analysis of prospective observational multicenter study of noncritically ill infants ≤60 days of age with temperatures ≥38°C and blood cultures (December 2008 to May 2013). The main outcome was time to positivity for bacterial pathogens and contaminants. RESULTS: A total of 256 of 303 (84.49%) patients with positive blood cultures, and 73 of 88 (82.95%) with positive CSF cultures met inclusion criteria. Median time (interquartile range [IQR]) to positivity for blood cultures was 16.6 hours (IQR 12.6–21.9) for bacterial pathogens (n = 74) and 25.1 hours (IQR 19.8–33.0) for contaminants (n = 182); P < .001. Time to bacterial pathogen positivity was similar in infants 0 to 28 days of age (15.8 hours [IQR 12.6–21.0]) and 29 to 60 days of age (17.2 [IQR 12.9–24.3]; P = .328). Median time to positivity for CSF was 14.0 hours (IQR 1.5–21.0) for bacterial pathogens (n = 22) and 40.5 hours (IQR 21.2–62.6) for contaminants (n = 51); P < .001. A total of 82.4% (95% confidence interval, 71.8–90.3) and 81.8% (95% confidence interval, 59.7%–94.8%) of blood and CSF cultures showed bacterial pathogen positivity within 24 hours. CONCLUSIONS: Among febrile infants ≤60 days of age, time to blood and CSF positivity was significantly shorter for bacterial pathogens than contaminants. Most blood and CSF cultures for bacterial pathogens were positive within 24 hours. With our findings, there is potential to reduce duration of hospitalization and avoid unnecessary antibiotics.