The overall objective of this study is to improve the standard of care of critically ill pediatric patients. The specific aims are to describe the clinical profile and outcomes of all admissions to Asian pediatric ICUs, determine the risk factors associated with poor outcomes, determine quality indicators for benchmarking ICU performance across sites and develop and train artificial intelligence algorithms to predict mortality, length of ICU stay and resource utilization.

Critical illness is associated with high mortality and morbidity. The mortality rate in pediatric ICUs globally may vary from 2% in high resource settings to a 28% in low resource settings. This variable mortality outcome is due to several factors. There are differing patient populations (e.g. medical, surgical, oncology) who get admitted to the ICU and at different severity thresholds. Patients in the region are ethnically and genetically diverse. And lastly, ICU management itself is variable and dependent on expertise and resources. This study will adopt a multicentered prospective observational design. Data on patient demographics, clinical characteristics, ICU therapies, ICU quality indicators and outcomes will be recorded prospectively. Statistical analysis will include descriptive statistics, multivariable analysis, area under the receiver operating curve analysis and a variety of machine learning algorithms to achieve its aims. Existing PICU severity scores (including but not limited to the PIM3, PRISM3/4, PELOD2, PSS, pSOFA) will be evaluated and if necessary, new variables/scores developed which perform better in the regional setting.

Though the main methodology is recruitment of all pediatric ICU admissions, a pre-determined random sampling protocol is allowed for sites with limited in resources for recruitment and data collection. This protocol may involve recruitment of all admissions for 1year (52 weeks), all admissions for 1 month per quarter over 1 year (16 weeks), all admissions for 1 week per month over 1 year (12 weeks), all admissions for 2 weeks per quarter over 1 year (8 weeks) or all admissions for 1 week for quarter over 1 year (4 weeks) - this will be declared at the start of the study by each participating site before recruitment begins. This strategy will allow us to include sites with and without sufficient resources to be represented in this study. As ethics approval will take time, each site may enter the study at differing time points and recruit for 1 year. Patients who have previously consented under the Singapore Pediatric Intensive Care Registry (SG-PedIC) under a similar single-center pilot study protocol which started from 2020 may be included. In the statistical analysis, random down sampling may be performed to avoid over-representation from high recruitment sites.

A planned subgroup study will be conducted for patients with and without pediatric chronic complex conditions (PCCC). PCCC represent a significant and growing subset of patients admitted to PICU. Although PCCCs make up only 10% to 17% of pediatric hospital admissions, they account for more than 50% of PICU admissions and use more than 75% of PICU resources. These conditions often involve multi-system involvement and prolonged hospitalizations, leading to substantial healthcare resource utilization and posing considerable challenges for clinical management. Understanding the characteristics, outcomes, and risk factors associated with these patients is crucial for improving clinical care and resource allocation. In this subgroup study, we will characterise patients with PCCCs, determine outcomes and identify risk factors for poor outcomes.

Inclusion Criteria:

• All patients admitted to the pediatric ICUs of participant hospitals

Exclusion Criteria:

• No consent

Though the main methodology is recruitment of all pediatric ICU admissions, a pre-determined random sampling protocol is allowed for sites with limited in resources for recruitment and data collection. This protocol may involve recruitment of all admissions for 1year (52 weeks), all admissions for 1 month per quarter over 1 year (16 weeks), all admissions for 1 week per month over 1 year (12 weeks), all admissions for 2 weeks per quarter over 1 year (8 weeks) or all admissions for 1 week for quarter over 1 year (4 weeks) - this will be declared at the start of the study by each participating site before recruitment begins. This strategy will allow us to include sites with and without sufficient resources to be represented in this study.

At study completion, data may be shared with other researchers provided patients had consented for use of data for future research and requestor obtains ethics approval and data sharing agreement.