Mayo Clinic research shows AI asthma risk detection in children.

AI helps detect Caribbean children with asthma at risk of severe infections

A new frontier in paediatric asthma care

Asthma has long been recognised as one of the most common chronic childhood diseases worldwide, yet it remains deeply complex, unpredictable and uneven in how it affects young patients.

Some children experience mild, intermittent symptoms that are easily controlled, while others suffer repeated, severe attacks that disrupt schooling, strain families and, in the worst cases, become life-threatening. New research from the Mayo Clinic, published in the Journal of Allergy and Clinical Immunology, signals a meaningful shift in how clinicians may soon identify which children are most at risk, long before those severe outcomes emerge.

Using artificial intelligence tools designed to analyse large volumes of health data, Mayo Clinic researchers have demonstrated that it is possible to identify high-risk asthma profiles in children as young as three years old.

These tools can flag which children with asthma are more likely to develop serious respiratory infections such as pneumonia and influenza, as well as severe asthma exacerbations requiring emergency treatment or hospitalisation. For regions such as the Caribbean, where asthma prevalence is high and access to specialised paediatric care can be uneven, the implications are profound.

Study Zone Big Kid Books

(5 book series)

Are you tired of endless flashcards that don’t translate into real-world reading success? For many children and adults, traditional methods like flashcards and dry word lists simply do not work. It is easy to feel stuck and defeated when words do not sound the way they look.

Asthma as a global and regional health burden

Asthma affects an estimated 262 million people globally and remains one of the leading causes of school absenteeism according to health statistics, emergency department visits and paediatric hospital admissions. In Caribbean nations, asthma prevalence among children is among the highest in the world, influenced by a mix of genetic factors, environmental exposures, air quality, allergens, viral infections and socio-economic conditions. Despite advances in inhaled therapies and clinical guidelines, asthma outcomes vary widely between individuals, even when they share similar diagnoses and treatment plans.

One of the greatest challenges in paediatric asthma care is predicting which children will progress to severe disease. Respiratory infections are the most common triggers for asthma attacks, yet not every child with asthma responds the same way to viral or bacterial exposure. Symptoms fluctuate over time, making early risk assessment difficult. As a result, many children only receive intensified care after repeated severe episodes, rather than before those episodes occur.

Precision medicine meets artificial intelligence

The Mayo Clinic study sits at the intersection of precision medicine and artificial intelligence. It forms part of the institution’s broader “precure” strategy, a prevention-focused model of care aimed at identifying disease risk early and intervening before serious illness develops. Rather than waiting for asthma to worsen, precure seeks to anticipate which patients are likely to follow a more dangerous trajectory and tailor care accordingly.

Artificial intelligence is particularly well suited to this task. Modern health systems generate enormous amounts of data, much of it buried within electronic health records and unstructured clinical notes. Traditional statistical methods struggle to process this information at scale. Machine learning and natural language processing, however, can extract patterns and relationships that are invisible to the human eye.

In this study, researchers applied multiple AI tools to electronic health records from more than 22,000 children born between 1997 and 2016 in south-eastern Minnesota. While the population itself was geographically specific, the size and depth of the dataset allowed for robust analysis of early-life asthma patterns and outcomes.

How the AI tools work

The AI systems developed by the Mayo Clinic team were designed to replicate and enhance existing clinical reasoning rather than replace it. Using machine learning algorithms and natural language processing, the tools scanned clinical notes for key information such as wheezing episodes, coughing patterns, allergic conditions and family history of asthma or atopic disease.

Crucially, the AI tools applied two well-established diagnostic frameworks used in paediatric asthma assessment: the Predetermined Asthma Criteria and the Asthma Predictive Index. These checklists are commonly used by clinicians to assess asthma risk in young children who may be too young for formal lung function testing. By automating the application of these criteria across thousands of patient records, the AI tools could consistently identify children who met one or both sets of criteria.

The most striking findings emerged when researchers examined children who met both diagnostic benchmarks. This subgroup, identified through AI-driven analysis, represented a distinct asthma phenotype with significantly higher risks of infection and severe disease.

Clear differences in early childhood outcomes

The differences between this high-risk subgroup and other children with asthma were both statistically and clinically significant. By the age of three, children in the high-risk group experienced pneumonia at more than twice the rate of their peers. Influenza infections were nearly three times more common. These children also had the highest rates of severe asthma exacerbations requiring systemic steroids, emergency department visits or hospital admission.

Respiratory syncytial virus, a major cause of lower respiratory tract infections in infants and young children, was also more prevalent in this group during the first three years of life. RSV is already known to play a role in the development and worsening of asthma, and its higher incidence in this subgroup further supports the idea that certain children have an underlying vulnerability to both infection and airway inflammation.

Importantly, these differences were detectable years before many children would typically be labelled as having severe asthma. This early signal is what makes the AI tools so clinically valuable.

Improve Spelling and Reading Skills

(10 books)

This is a series of fun spelling books available in both e-book and paperback formats. These books are designed for various educational purposes, including story time, spelling improvement classes, poetry sessions, and reading intervention programs, as they help improve phonological and phonemic awareness. The books contain stories filled with words that have rimes containing digraphs, trigraphs, and 4-letter graphemes.

Available on Kindle and Paperback

The biological signature of high-risk asthma

Beyond infection rates and clinical outcomes, the study also identified biological characteristics that help explain why this subgroup is more vulnerable. Children in the high-risk group were more likely to have a family history of asthma, eczema, allergic rhinitis or food allergies, pointing to a strong genetic and atopic component.

Laboratory data from earlier research revealed elevated markers of allergic inflammation in these children. These included higher eosinophil counts, increased allergen-specific IgE levels and elevated periostin, a protein associated with type 2 immune inflammation. Impaired lung function was also noted, even at a young age.

Taken together, these features describe a specific asthma subtype driven by heightened immune reactivity and allergic inflammation. This biological profile appears to predispose children not only to asthma exacerbations but also to more severe responses to respiratory infections.

Why early identification matters

From a medical standpoint, the ability to identify this high-risk asthma subtype early opens the door to more proactive and personalised care. Instead of applying a one-size-fits-all approach, clinicians could intensify monitoring, vaccination strategies, environmental interventions and preventive therapies for children flagged by AI as high risk.

For families, this could mean fewer emergency visits, reduced exposure to systemic steroids and better long-term lung health. For health systems, particularly in resource-limited settings, early risk stratification could help prioritise care for those most likely to need it, improving outcomes while managing costs.

“This study brings us one step closer to precision medicine in pediatric asthma, where care shifts from reacting to advanced severe asthma to focusing on prevention and early identification of high-risk patients,” says Young Juhn, MD, MPH, professor of pediatrics at Mayo Clinic and lead author of the study. Dr Juhn leads several Mayo Clinic research programs, including the Mayo Clinic Children’s AI Program, the Precision Population Science Laboratory, and the HOUSES socioeconomic health program.

Online urgent care

Make an appointment today with our online urgent care providers for non-emergency, urgent care needs via video visit — around the clock, from wherever you are. If we cannot help you or provide care, we’ll offer a full refund.

Urgent care appointments available online within 15 minutes day and night, every day
1M+ members treated
We accept most major insurances although insurance is not required

Implications for the Caribbean

The Caribbean context gives this research additional urgency. Asthma prevalence among Caribbean children is high, and respiratory infections remain a major cause of hospitalisation. Environmental factors such as Saharan dust, urban air pollution, humidity and allergen exposure can exacerbate symptoms. Socio-economic disparities may also limit access to consistent preventive care.

AI-driven tools, once validated across diverse populations, could help Caribbean health systems identify vulnerable children earlier, even in primary care settings. Because these tools rely on electronic health records and clinical notes rather than expensive new tests, they could be adapted to existing digital health infrastructures as they continue to develop across the region.

Early identification also aligns with public health goals, allowing for targeted education, vaccination campaigns and family support for those most at risk.

Expanding and refining the technology

The Mayo Clinic research team is not stopping with this initial study. Plans are underway to test the AI tools in broader clinical environments and across more diverse populations and health systems. This step is essential to ensure that the models perform accurately beyond the original study setting and do not inadvertently reflect regional or demographic bias.

The researchers also aim to integrate biological data more deeply into the AI models. By combining clinical history with biomarkers and genetic information, future tools could refine asthma subtypes even further, allowing for more precise prevention and treatment strategies.

In parallel, the team is exploring therapeutic innovation. One planned study will investigate a compound designed to dampen hyperactive immune responses associated with asthma. Using laboratory-grown cellular models known as organoids, researchers hope to better understand how asthma develops at the cellular level and how it might be prevented before symptoms fully emerge.

A cautious but hopeful path forward

While the findings are compelling, they also highlight the need for careful implementation. AI tools must be transparent, clinically validated and used to support, not replace, medical judgement. Data privacy, equitable access and clinician training will be essential considerations as such technologies move closer to routine care.

Nevertheless, the study represents a meaningful advance in paediatric asthma research. It shows that with the right tools, it is possible to see risk earlier, understand disease more deeply and intervene more intelligently.

Redefining prevention in childhood asthma

The Mayo Clinic’s work underscores a broader shift in medicine, one that prioritises prediction and prevention over late-stage treatment. For children with asthma, this shift could be transformative. Identifying high-risk profiles at age three rather than age ten or fifteen changes the entire arc of care.

For parents, clinicians and health policymakers in the Caribbean and beyond, the message is clear. The future of asthma care lies not in waiting for the next severe attack, but in recognising vulnerability early and acting decisively. Artificial intelligence, when grounded in sound clinical science and ethical practice, may become one of the most powerful tools in making that future a reality.

____________________