PC2062: Identifying Risk Factors for Severe Clinical Outcomes in Pediatric COVID-19 Patients in a National Electronic Health Record Repository

Background: Over 12.8 million pediatric COVID-19 cases have been reported in the US, representing 19.0% of all cases. However, inconsistent reporting of severity outcomes in children clouds the impact of COVID-19 on children in the US and child vaccination rates vary widely across states. Hence, we used a national electronic health record data pipeline to assess the effect of vaccination, clinical and demographic factors on severe outcomes in children with COVID-19¬.

Methods: This retrospective observational study examined clinical, demographic factors, and vaccination on COVID-19 outcomes among children less than 18 years using the Optum® COVID-19 low-latency data pipeline (Optum Inc, Eden Prairie, MN). We included patients who tested positive by polymerase chain reaction or antigen from March 12, 2020 through January 20, 2022. Our primary outcome was illness severity (non-hospitalization, hospitalization, ICU or worse). We performed multivariable logistic regression (R version 4.1.2; α = 0.05) accounting for age, BMI, validated pediatric comorbidity index (PCI) score indicating pre-test comorbidities, race and ethnicity, gender, region, and predominant COVID-19 variant at time of diagnosis for the patient's geographic region. Three models were developed bucketing children by age (0-2, 3-10, and 11-18 years) because of clinical differences between ages, data set coding limitations, and vaccine eligibility (BMI excluded for 0-2 years, vaccination status included for 11-18 years).

Results: Among pediatric patients, 218,759 (18.7%) tested positive for COVID-19. 8,717 (4.0%) were hospitalized, required ICU care or ventilation, or died. By far, the strongest predictors of severe outcome in all groups were from higher PCI scores (representing pre-COVID-19 comorbidities, ORs 2.27 [2.02-2.56], 4.21 [3.62-4.91], 5.33 [4.75-5.97]) or from residency in the US South region (ORs for the Midwest compared to the South: 0.28 [0.24-0.32], 0.18 [0.16-0.21], 0.18 [1.70-0.20]) . In all groups, "unknown insurance status", Alpha variant, and Black race were statistically significant and borderline clinically significant predictors of severe outcome (p < 0.0001). Of note, the odds ratio of severe disease comparing at least one vaccination to no known vaccination is 0.55 (0.48 - 0.63) in the vaccine-eligible cohort. Odds ratios for all models are depicted in Figure 1. Individual feature contributions to each model via Wald score minus df are depicted in Figure 2, where increased Wald score minus df indicates increased significance of the parameter to predict severe outcomes.

Conclusion: Pre-existing illness was the strongest independent predictor of severe COVID-19 outcomes among pediatric patients, and vaccination was significantly protective. Use of a large-scale, national data pipeline allowed for up-to-date, centralized assessment of outcomes across a diverse patient population, revealing persistent health disparities across racial and geographic lines throughout the pandemic. Our study demonstrates that many easily assessed parameters can reliably predict adverse outcomes in pediatric COVID-19, indicating populations in need of outreach for vaccination and other preventative efforts.