Many health settings use predictive analytics – a branch of data science aimed at analyzing health information and predicting future outcomes – to better manage chronic diseases, reduce hospital readmissions, and more.
However, that all hinges on the quality of available data, the interoperability of its sources, and the confidence clinicians have in its accuracy. Without those three elements, information derived from predictive analytics loses value and credibility.
Ashley Beecy, MD, FACC, Chief AI Officer at Sutter Health, has seen the value of predictive analytics – equipped with accurate, high-quality data – firsthand:
“We are developing tools to identify advanced heart failure by predicting abnormal peak VO₂ from echocardiograms, information that would normally require complex cardiopulmonary exercise testing,” Beecy shared. “We also apply predictive analytics to anticipate patient outcomes, such as clinical deterioration, malnutrition, or postpartum depression, which supports timely, proactive care.”
The excerpt below from our recent eBook digs into some of the ways AI is being implemented to improve patient and population health.
Predictive analytics
Predictive analytics at-a-glance
Predictive analytics – the use of data, statistical modeling, and machine learning algorithms to analyze health information and predict future health outcomes – is applied across care settings to help with initiatives like chronic disease management and reducing hospital readmissions. But making predictive analytics useful can pose significant challenges depending on the quality of available data, the interoperability of data sources, and the confidence clinicians have in the accuracy of analytics.
AI in Action
IMO Health, in collaboration with Harvard Medical School, developed an AI-driven predictive model to detect cognitive decline by analyzing the clinical notes of patients, aged 50 and older, who had been diagnosed with mild cognitive impairment. The study evaluated large language models (LLMs), including GPT-4 and Llama 2, using advanced prompting techniques and retrieval- augmented generation (RAG). To improve accuracy, researchers combined LLMs with traditional machine learning models, creating a hybrid approach that increased diagnostic precision from 70–79% to over 90%. This AI-enhanced method can enable earlier identification of cognitive impairment, supporting timely interventions for Alzheimer’s disease and related dementias.
How AI is being applied
AI is being implemented in several ways to forecast and improve patient and
population health, including:
Risk stratification
AI algorithms are used to analyze data in order to identify patients at high risk for adverse outcomes, such as readmissions or complications.
Early disease detection
Predictive models are employed to spot early warning signs of diseases, creating opportunities for timely intervention.
Resource optimization
AI predicts patient volumes and care needs, helping hospitals allocate staff, beds, and resources more effectively.
Personalized care plans
Predictive analytics can be leveraged to customize care plans by forecasting how patients might respond to specific treatments.
