Abstract
Introduction: Greater than one in four children with sickle cell disease (SCD) treated for pain in the emergency department (ED) return within 14 days of ED discharge. Despite disproportionately high return visit rates, scant data exist evaluating the factors driving ED revisits making targeting interventions to higher risk children difficult. In this multicenter study, we identified patient- and visit-level predictors of ED revisits in children with SCD pain.
Methods: This is a multicenter retrospective cohort study of ED visits by children <19 years old with SCD presenting to 1 of 12 Pediatric Emergency Care Applied Research Network (PECARN) Registry sites between 07/2019 and 12/2024 with an index visit for uncomplicated SCD pain and discharged home. Index visits were defined by no ED visits within the previous 14 days. We studied the association between return visits and visit level factors (e.g., site, age, sex, ED opioid doses, ED pain scores, receipt of an opioid prescription at discharge, Childhood Opportunity Index, primary payer) and prior ED utilization patterns (e.g., visits in last six months, distance from the ED, etc.) via chi-square tests, two-sample t-tests, or univariable logistic regression as appropriate. Data from 07/2019-12/2023 were included in the derivation set; the validation set consisted of visits 01/2024-12/2024. Classification and Regression Trees (CART) were used to predict children most likely to return within 14 days.
Results: We included 5,920 ED visits for acute SCD pain resulting in discharge. Children <12 years old comprised 45.7% of visits; 50% of visits were by males, and 95% by Black and non-Hispanic children. There were 1,702 (28.8%) ED return visits within 14 days, with variation by site (22.6% - 34.7%). The following variables demonstrated a relationship with having a return visit in univariable analyses: age, distance from home to ED, number of opioid doses administered in the ED, initial ED pain score, number of visits to the ED with uncomplicated SCD in the previous 6 months, and return visit following an ED visit for uncomplicated SCD pain in the previous 6 months. Our CART analysis on the training dataset identified the number of ED visits in the prior six months for uncomplicated SCD pain as the primary splitting variable. The 537 children with 4 or more ED visits in the past six months had a 51% return visit rate compared to 25.7% in the 4315 children with fewer than 4. There were 123 visits in the validation set with 4 or more prior visits, 76 (61.8%) had a return visit within 14 days. Using the validation set, model performance yielded a sensitivity of 77.6% and specificity of 32.8%. A repeat CART analysis with variables only available at the current visit revealed that the 3473 children who received 3 or more parenteral opioid doses in the ED, had an initial ED pain score of 9 or 10 or lived within 14.7 km of the ED had a 31.5% return visit rate compared to 20.7% in the 1379 children without those risk factors. There were 788 visits with at least one of these risk factors in the validation set, 258 (32.7%) had a return visit within 14 days. Sensitivity and specificity on the validation set were 80.4% and 29.0%, respectively. The primary splitting variable in the second CART, receipt of three or more parenteral opioid doses, yielded 1425 visits and a return visits rate of 35.4% compared to 25.6% in visits with only 1 or 2 parenteral opioids. Combining the primary splitting variables, 1634 children with 3 or more parenteral opioids in the ED or 4 or more ED visits in the past 6 months resulted in a 36.8% return visit rate within 14 days versus a 24.2% return visit rate in children not meeting the specified criteria.
Conclusion: Our prediction model revealed that the number of ED visits in the prior 6 months is the most important determinant of return visits within 14 days. This highest risk group, with an over 50% return visit rate, comprised only 20.6% of all return visits in the entire dataset. Sites with additional resources may wish to expand to a larger but lower risk group of risk factors identified in this analysis to more significantly reduce total return visits. Future research will utilize the models to support clinical decision making and the design of an intervention to decrease return visits for high-risk patients identified. Incorporating predictors is essential to optimize care and support the development of outpatient SCD acute pain management guidelines.
