Multisite development and validation of machine learning models to predict severe outcomes and guide decision-making for emergency department patients with influenza.

Objective: Millions of Americans are infected by influenza annually. A minority seek care in the emergency department (ED) and, of those, only a limited number experience severe disease or death. ED clinicians must distinguish those at risk for deterioration from those who can be safely discharged. Methods: We developed random forest machine learning (ML) models to estimate needs for critical care within 24 h and inpatient care within 72 h in ED patients with influenza. Predictor data were limited to those recorded prior to ED disposition decision: demographics, ED complaint, medical problems, vital signs, supplemental oxygen use, and laboratory results. Our study population was comprised of adults diagnosed with influenza at one of five EDs in our university health system between January 1, 2017 and May 18, 2022; visits were divided into two cohorts to facilitate model development and validation. Prediction performance was assessed by the area under the receiver operating characteristic curve (AUC) and the Brier score. Results: Among 8032 patients with laboratory-confirmed influenza, incidence of critical care needs was 6.3% and incidence of inpatient care needs was 19.6%. The most common reasons for ED visit were symptoms of respiratory tract infection, fever, and shortness of breath. Model AUCs were 0.89 (95% CI 0.86-0.93) for prediction of critical care and 0.90 (95% CI 0.88-0.93) for inpatient care needs; Brier scores were 0.026 and 0.042, respectively. Importantpredictors included shortness of breath, increasing respiratory rate, and a high number of comorbid diseases. Conclusions: ML methods can be used to accurately predict clinical deterioration in ED patients with influenza and have potential to support ED disposition decision-making.

Acute care utilization for ambulatory care-sensitive conditions among publicly insured children.

BACKGROUND: Although characteristics of preventable hospitalizations for ambulatory care-sensitive conditions (ACSCs) have been described, less is known about patterns of emergency and other acute care utilization for ACSCs among children who are not hospitalized. We sought to describe patterns of utilization for ACSCs according to the initial site of care and to determine characteristics associated with seeking initial care in an acute care setting rather than in an office. A better understanding of the sequence of health care utilization for ACSCs may inform efforts to shift care for these common conditions to the medical home. METHODS: We performed a retrospective analysis of pediatric encounters for ACSCs between 2017 and 2019 using data from the IBM Watson MarketScan Medicaid database. The database includes insurance claims for Medicaid-insured children in 10 anonymized states. We assessed the initial sites of care for ACSC encounters, which were defined as either acute care settings (emergency or urgent care) or office-based settings. We used generalized estimating equations clustered on patient to identify associations between encounter characteristics and the initial site of care. RESULTS: Among 7,128,515 encounters for ACSCs, acute care settings were the initial site of care in 27.9%. Diagnoses with the greatest proportion of episodes presenting to acute care settings were urinary tract infection (52.0% of episodes) and pneumonia (44.6%). Encounters on the weekend (adjusted odds ratio [aOR] 6.30, 95% confidence interval [CI] 6.27-6.34 compared with weekday) and among children with capitated insurance (aOR 1.55, 95% CI 1.54-1.56 compared with fee for service) were associated with increased odds of seeking care first in an acute care setting. CONCLUSIONS: Acute care settings are the initial sites of care for more than one in four encounters for ACSCs among publicly insured children. Expanded access to primary care on weekends may shift care for ACSCs to the medical home.

Racial and Socioeconomic Disparities Evident in Inappropriate Antibiotic Prescribing in the Emergency Department.

STUDY OBJECTIVE: Inappropriate antibiotic prescribing for acute respiratory tract infections is a common source of low-value care in the emergency department (ED). Racial and socioeconomic disparities have been noted in episodes of low-value care, particularly in children. We evaluated whether prescribing rates for acute respiratory tract infections when antibiotics would be inappropriate by guidelines differed by race and socioeconomics. METHODS: A retrospective cross-sectional analysis of adult and pediatric patient encounters in the emergency department (ED) between 2015 and 2023 at 5 hospitals for acute respiratory tract infections that did not require antibiotics by guidelines. Multivariable regression was used to calculate the risk ratio between race, ethnicity, and area deprivation index and inappropriate antibiotic prescribing, controlling for patient age, sex, and relevant comorbidities. RESULTS: A total of 147,401 ED encounters (55% pediatric, 45% adult) were included. At arrival, 4% patients identified as Asian, 50% as Black, 5% as Hispanic, and 23% as White. Inappropriate prescribing was noted in 7.6% of overall encounters, 8% for Asian patients, 6% for Black patients, 5% for Hispanic patients, and 12% for White patients. After adjusting for age, sex, comorbidities, and area deprivation index, White patients had a 1.32 (95% confidence interval, 1.26 to 1.38) higher likelihood of receiving a prescription compared with Black patients. Patients residing in areas of greater socioeconomic deprivation, regardless of race and ethnicity, had a 0.74 (95% confidence interval, 0.70 to 0.78) lower likelihood of receiving a prescription. CONCLUSION: Our results suggest that although overall inappropriate prescribing was relatively low, White patients and patients from wealthier areas were more likely to receive an inappropriate antibiotic prescription.

Exploring the impact of missingness on racial disparities in predictive performance of a machine learning model for emergency department triage.

Objective: To investigate how missing data in the patient problem list may impact racial disparities in the predictive performance of a machine learning (ML) model for emergency department (ED) triage. Materials and Methods: Racial disparities may exist in the missingness of EHR data (eg, systematic differences in access, testing, and/or treatment) that can impact model predictions across racialized patient groups. We use an ML model that predicts patients' risk for adverse events to produce triage-level recommendations, patterned after a clinical decision support tool deployed at multiple EDs. We compared the model's predictive performance on sets of observed (problem list data at the point of triage) versus manipulated (updated to the more complete problem list at the end of the encounter) test data. These differences were compared between Black and non-Hispanic White patient groups using multiple performance measures relevant to health equity. Results: There were modest, but significant, changes in predictive performance comparing the observed to manipulated models across both Black and non-Hispanic White patient groups; c-statistic improvement ranged between 0.027 and 0.058. The manipulation produced no between-group differences in c-statistic by race. However, there were small between-group differences in other performance measures, with greater change for non-Hispanic White patients. Discussion: Problem list missingness impacted model performance for both patient groups, with marginal differences detected by race. Conclusion: Further exploration is needed to examine how missingness may contribute to racial disparities in clinical model predictions across settings. The novel manipulation method demonstrated may aid future research.