Using an AI-driven decision support system to predict the severity of COVID-19 and identify best interventions, researchers analyzed electronic health record data from 5,371 patients admitted to a South Florida hospital.

Seasonal influenza, respiratory syncytial virus (RSV), and COVID-19 are actively circulating throughout the United States. These respiratory illnesses are contributing to widespread health concerns, with cases being reported in various regions nationwide.

Using artificial intelligence and machine learning, researchers from Florida Atlantic University’s Christine E. Lynn College of Nursing and College of Engineering and Computer Science, in collaboration with Memorial Healthcare System , are pushing the boundaries in health care to foster innovation, enhance decision-making, and ultimately improve health outcomes for individuals and populations.  

To predict the severity of COVID-19 and best therapeutic interventions during the pandemic, researchers established an AI-driven decision support system by identifying critical features influencing the severity of disease outcomes in patients hospitalized with COVID-19 in a South Florida hospital.

Specifically, the study focused on predicting the need for intensive care unit (ICU) admission with or without mechanical ventilation and intermediate care unit (IMCU) admission. The goal was to leverage these features to enable faster and more accurate forecasting of treatment plans, potentially preventing critical conditions from worsening.

For the study, researchers analyzed electronic health record (eHR) data from 5,371 patients admitted to a South Florida hospital with COVID-19 between March 2020 and January 2021. They trained three Random Forest models to predict mechanical ventilation, ICU, and IMCU admission using 24 variables, including sociodemographics, comorbidities, and medications. The analysis centered on data collected at the time of hospital admission.

Results of the study, published in the journal Diagnostics, in early fall 2024, show that the models for ICU with mechanical ventilation, ICU, and IMCU admission identified the following factors overlapping as the most important predictors among the three outcomes: age, race, sex, body mass index (BMI), diarrhea, diabetes, hypertension, early stages of kidney disease, and pneumonia.

Researchers also found that individuals 65 and older (‘older adults’), males, current smokers, and BMI classified as ‘overweight’ and ‘obese’ were at greater risk of severity of illness. The study also explored the severity of the disease under the co-occurrence of risk factors.

“This is one of the very few studies that explored such interactions among risk factors using machine learning interpretability approaches. For example, pneumonia combined with diabetes increased mechanical ventilation risk, while diarrhea interacted strongly with diabetes for ICU admissions,” said Debarshi Datta, Ph.D., senior author and an assistant professor in FAU’s Christine E. Lynn College of Nursing. “IMCU severity was linked to combinations like diarrhea with pneumonia and hypertension in older adults. Additionally, medications such as angiotensin II receptor blockers and ACE inhibitors appeared to lower disease severity, aligning with prior research on their protective effects.”

The top features identified by the models’ interpretability were from the ‘sociodemographic characteristics,’ ‘pre-hospital comorbidities,’ and ‘medications’ categories. However, ‘pre-hospital comorbidities’ played a vital role in different critical conditions. In addition to individual feature importance, the feature interactions also provide crucial information for predicting the most likely outcome of patients’ conditions when urgent treatment plans are needed during the surge of patients during the pandemic.

Compared to earlier studies, this novel approach stands out by using readily accessible eHR data and combining machine learning interpretability techniques with traditional statistical methods. This method enabled a deeper understanding of features like age, sex, BMI, and comorbidities such as diabetes and hypertension across different severity levels.

“While biomarkers have been used in other studies, their limited clinical accessibility makes our findings more practical for real-world health care applications,” said David Newman, Ph.D., co-author, professor, and statistician, FAU Christine E. Lynn College of Nursing. “By identifying critical factors and interactions influencing COVID-19 outcomes, our study provides actionable insights for improving patient care and supporting health care systems during high-demand conditions.”

Importantly, the application of AI/machine learning in health care extends beyond the COVID-19 disease, holding promise for improving diagnosis, treatment selection, disease surveillance, and patient outcomes across various medical specialties and health care settings. This knowledge empowers public health authorities to proactively plan and implement targeted interventions, mitigating the impact of disease outbreaks and optimizing health care delivery.

“Developing an AI-driven decision support system to predict critical clinical events in COVID-19 in-patients not only meets the urgent demands of a pandemic but also breaks new ground in AI and machine learning in health care,” said Datta. “By utilizing advanced technologies and algorithms, such as machine learning, researchers and clinicians can harness the power of data-driven insights to revolutionize patient care.”

Study co-authors are Subhosit Ray, Ph.D., a postdoctoral fellow; Laurie Martinez, Ph.D., an assistant professor; Safiya George Dalmida, Ph.D., former dean; all with FAU’s Christine E. Lynn College of Nursing; Javad Hashemi, Ph.D., inaugural chair and professor of the Department of Biomedical Engineering and associate dean for research, FAU College of Engineering and Computer Science; Candice Sareli, M.D., vice president and chief medical research officer, Memorial Healthcare System; and Paul Eckardt, M.D., chief, Memorial Division of Infectious Disease, Memorial Healthcare System.

-FAU-