COVID-19 Network of Networks Expanding Clinical and Translational approaches to Predict Severe Illness in Children (CONNECT to Predict SIck Children)

The SARS-CoV-2 pandemic has manifested in children with a wide spectrum of clinical presentations rangingfrom asymptomatic infection to devastating acute respiratory symptoms, appendicitis (often with rupture), andMultisystem Inflammatory Syndrome in Children (MIS-C), a serious inflammatory condition presenting severalweeks after exposure to or infection with the virus.

These presentations overlap in their clinical severity whilemaintaining distinct clinical profiles.

Public health and clinical approaches will benefit from an improvedunderstanding of the spectrum of illness associated with SARS CoV-2 and from the capacity to integrate data toachieve two goals: (i) to identify the clinical, social, and biological variables that predict severe COVID-19 andMIS-C, and (ii) to target those populations and individuals at greatest risk for harm from the virus.

We proposethe COVID-19 Network of Networks Expanding Clinical and Translational approaches to Predict Severe Illnessin Children (CONNECT to Predict SIck Children) comprising eight partners providing access to data on >15million children.

Our network will systematically integrate social, epidemiological, genetic, immunological, andcomputational approaches to identify both population- and individual-level risk factors for severe illness.

Ourunderlying hypothesis is that a combination of multidimensional data – clinical, sociodemographic, epidemiologic,and biological -- can be integrated to predict which children are at greatest risk to have severe consequencesfrom SARS-CoV-2 infection.

To test our hypothesis, we will develop CONNECT to Predict SIck Children, anetwork of networks that leverages inpatient, outpatient, community, and epidemiological data resources tosupport the analysis of large data using machine learning and model-based analyses. For the R61 phase, wewill develop and refine predictive models using data from our network of networks (Aim 1).

We will also recruitparticipants previously diagnosed with either COVID-19 or MIS-C (along with appropriate controls who have hadmild or asymptomatic infections with SARS-CoV2), who will provide survey data (including social determinants)and saliva and blood samples to identify persisting biological factors associated with severe disease (Aim 2).

Wewill iteratively assess our models using a knowledge management framework that considers the marginal valueof data for improving models' predictive capacity over time.

In the R33 phase, we will validate and further refinepredictive models incorporating data from additional participants recruited throughout our network of networks,including newly infected children with severe COVID-19 or MIS-C identified through real-time surveillance (Aim3).

We seek to develop predictive models for children and adolescents that are useful, sensitive to communityand environmental contexts, and informed by the REASSURED framework specified by the RFA.

The modelsand biomarkers developed through our nationwide network of networks will produce generalizable knowledgethat will improve our ability to predict which children are at greatest risk for severe complications of SARS-CoV-2 infection. This knowledge will facilitate interventions to prevent and treat severe pediatric illness.