When the body’s immune system has an extreme response to an infection, it can cause organ dysfunction that leads to shock, multiple organ failure, and death. This life-threatening condition is known as sepsis, and is the leading cause of death among kids in low- and middle-income countries (LMICs), such as Uganda. To help prevent more children from dying of this condition, the Institute for Global Health (IGH) at BC Children’s Hospital and BC Women’s Hospital + Health Centre is inviting researchers to help design an open-source algorithm that predicts in-hospital mortality in LMICs. The 2024 Pediatric Sepsis Data Challenge will accept registrations until December 11. If you’re new to, or a veteran of, data science, register now.
This is a global collaboration to enhance sepsis recognition — SMART sepsis care is the future.
– Dr. Mark Ansermino
“This is a global collaboration to enhance sepsis recognition — SMART sepsis care is the future,” says Dr. Mark Ansermino, the co-executive medical director at IGH and the data challenge project lead. Dr. Ansermino is also an investigator at BC Children’s Hospital Research Institute (BCCHR) and a professor in the Department of Anesthesiology, Pharmacology and Therapeutics at the University of British Columbia (UBC).
“We’ll provide participants with an opportunity to learn, develop skills, and use their knowledge to create a model that could eventually be used clinically, reducing the burden of pediatric sepsis in LMICs where, currently, there’s no recommended model for clinical use,” says Charly Huxford, research assistant at IGH. “We hope to see a mix of data experts and trainees, LMIC representatives, and anyone interested in global health.” Participants can be anywhere in the world, but they’re encouraged to embrace international, cross-discipline collaboration within their teams.
We hope to see a mix of data experts and trainees, LMIC representatives, and anyone interested in global health.
– Charly Huxford
There will be two phases in the challenge. First, participants will become familiar with the generated training dataset and build preliminary models. Nothing done at this stage will impact their score, and they’ll be able to receive feedback. In the second phase, participants will submit their models for scoring. “We’ll run the algorithms against the original dataset and announce the best-performing model,” Huxford says. There’s no monetary prize, but all models will be open source so that anyone in the world can access them.
To protect privacy and prevent patient identification, the training dataset provided to participants was generated from a subset of the original clinical dataset. “Participants should consider this incredible learning opportunity because artificially generated data is a hot topic, as it’s expected that, in the future, 50 per cent of all data-driven prediction models will be generated this way,” says Huxford.
The original data comes from a clinical study in Uganda through IGH’s Smart Discharges, a precision public health research program in East Africa run by Dr. Matthew Wiens, investigator at BCCHR and assistant professor in the Department of Anesthesiology, Pharmacology and Therapeutics at UBC. This program aims to improve post-discharge health outcomes in children with severe infections by using scientifically rigorous, data-driven prediction models to identify at-risk individuals.
Timely and appropriate care of pediatric patients with suspected sepsis is key to reducing the risk of death or morbidity.
Here’s how a model developed in this challenge could work in a clinical setting: as soon as a child is admitted to a hospital, their variables would be added to the model to generate a risk score and determine their risk of death. If the child is at a high risk, care would be escalated to deploy resources quicker, provide treatments faster, and improve outcomes. “Timely and appropriate care of patients with suspected sepsis is key to reducing the risk of death or morbidity,” Huxford says, pointing to evidence that every hour of delay in care increases the risk of death.
Published in 2020 in The Lancet, a global study estimated that roughly 49 million cases and 11 million deaths related to sepsis (20 per cent of all-cause deaths in the world) were recorded in 2017. Eighty-five per cent of those cases and deaths occurred in developing countries, mostly in sub-Saharan Africa and Southeast Asia. Patients aged 19 and under accounted for more than half of those cases, and children under the age of five represented 41 per cent of all cases and 26 per cent of all sepsis-related deaths.
Risk-stratified approaches have potential to improve care in any health-care system, by making the most efficient use of limited resources.
– Charly Huxford
The federal funding agency for health research, Canadian Institutes of Health Research, has adopted a framework that encourages research that advances health equity for all. The purpose is to accelerate health improvement around the world, creating an opportunity for Canada to address historical wrongs and achieve transformative impacts for everyone. “We’re aiming to build collaboration and capacity among researchers across the globe who could address these problems in the future,” says Huxford. “Risk-stratified approaches have potential to improve care in any health-care system, by making the most efficient use of limited resources.”
The 2024 Pediatric Sepsis Data Challenge welcomes teams of 2–5 members, and is open to participants of all experience levels — from data science beginners to seasoned professionals. If you’re interested in participating in the challenge and don’t have a team, matchmaking services are available to solo challengers. Register now!
Top photo by Aditya Romansa on Unsplash.
