Machine Learning Could Reduce Hospitalizations by 30% During Pandemic

However, these methods often fail to consider the complex interactions between medications and patients, potentially overlooking those who could benefit the most from treatment. Now, new research suggests that machine learning may be a more effective way to allocate scarce treatments to vulnerable patients during public health crises.

The new study by researchers at the University of Colorado Anschutz Medical Campus (Aurora, CO, USA) highlights the potential of machine learning to more efficiently allocate medical treatments in times of shortage, such as during a pandemic. The research demonstrated that machine learning, by analyzing how different patients respond to treatment, can provide more accurate, real-time information to doctors, health systems, and public health officials than traditional allocation methods. Published in JAMA Health Forum, the study revealed that using machine learning to allocate COVID-19 treatments could reduce hospitalizations by about 27% compared to current practices.

The researchers specifically examined the use of a novel method based on Policy Learning Trees (PLTs) to optimize the distribution of COVID-19 neutralizing monoclonal antibodies (mAbs) during periods of limited availability. The PLT approach was designed to prioritize treatments for individuals most at risk of hospitalization, maximizing overall benefit by factoring in variables that influence treatment effectiveness. The machine learning model was compared to real-world allocation decisions and a standard point-based system used during the pandemic. The results showed that the PLT-based model significantly reduced expected hospitalizations compared to both observed allocations and the Monoclonal Antibody Screening Score, a commonly used tool during the pandemic. The researchers hope their findings will encourage public health agencies, policymakers, and disaster management organizations to explore machine learning as a tool for future public health crises, ensuring that treatments are allocated more effectively when resources are limited.

“Existing allocation methods primarily target patients who have a high-risk profile for hospitalizations without treatments. They could overlook patients who benefit most from treatments,” said Mengli Xiao, PhD, an assistant professor in Biostatistics and Informatics, who developed the mAb allocation system based on the machine learning. “We developed a mAb allocation point system based on treatment effect heterogeneity estimates from machine learning. Our allocation prioritizes patient characteristics associated with large causal treatment effects, seeking to optimize overall treatment benefits when resources are limited.”