How AI and machine learning can predict illness and boost health equity
Photo: Brett Furst
Artificial intelligence and machine learning are key to unlocking patient data and solving some of healthcare's most complex problems. Even as the U.S. seeks to put the COVID-19 pandemic in the rearview mirror, many who survive the initial illness suffer debilitating long-term health impacts, especially those with underlying health conditions.
Technology allows easier access to disparate data sources without compromising data privacy or integrity. In addition, advanced analytics deliver real-time insights, enabling providers to predict outcomes and diagnose illness early to intervene with patients at risk of developing long-term COVID-19 and other chronic diseases.
To delve deeper into these technologies and their ramifications in healthcare, Healthcare IT News spoke with Brett Furst, president of HHS Tech Group.
Q. How can health system IT leaders leverage AI and machine learning technologies to unlock and explore patient data for the accurate, early detection and treatment of chronic disease?
A. Health system IT leaders face two principal challenges specific to healthcare data: It's very messy, and there's a considerable amount of it.
For example, much of the valuable patient information related to social determinants of health is buried as unstructured data in the notes sections of electronic health records, making it difficult to capture and analyze. Further, in addition to patient data in EHRs, a wealth of information can be gleaned from other sources, such as surveys, games, retail and social media, which in many cases no one has ever combined.
When researchers and providers combine diverse data sets, the healthcare industry can acquire new insights at the population health level. However, due to the messiness and volume of healthcare data, manual or human efforts to draw value from it are not sufficient, and that's where artificial intelligence and machine learning enter the picture.
AI, machine learning and other related technologies enable providers and researchers to obtain a wider breadth of insights in a dramatically shorter time frame than traditional data extraction, capture and analysis methods.
AI and machine learning are essential tools to solve complex healthcare data problems, such as identifying patients exhibiting signs of a yet-to-be-diagnosed chronic disease or analyzing wastewater samples to pinpoint the next pandemic.
Q. How are the nation's largest COVID-19 research database and related analytics helping health system leaders and other organizations such as medical associations better understand the impact of COVID-19 on behavioral health and populations with underlying chronic conditions?
A. The COVID-19 Research Database features a wide array of data about most of the U.S. population, including claims, encounters, social determinants of health and various other types of clinical data. The database effectively pulls these vast data sets together in one place, links them and drives discoveries.
For example, about the impact that pre-existing chronic conditions, such as diabetes, a chronic obstructive pulmonary disorder and chronic kidney disease, have on patients infected with COVID-19.
By leveraging one of the world's most comprehensive cross-linked data sets, researchers can draw new conclusions about how COVID-19 affects vulnerable populations. For example, researchers can study data about patient access to healthcare, nutrition and economic opportunity, and see how COVID-19 infections affected individuals across these groups.
Ultimately, the goal is to establish cause-effect relationships between multiple variables to determine how they interact and affect people with COVID-19. But before we can get there, we need to draw out correlations from billions of health records, which requires heavy use of AI, machine learning and advanced analytics.
Q. How are AI and machine learning predicting health outcomes and promoting the adoption of evidence-based strategies to improve treatment quality and safety?
A. Providers are using AI and machine learning to identify at-risk populations developing chronic conditions such as chronic kidney disease and hypertension before these patients are diagnosed and their health conditions grow more serious.
Although patients in these populations are undiagnosed, predictive modeling can account for many data factors to pinpoint the patients who require interventions. Consider chronic kidney disease, for example, a common, serious, costly and often preventable disease. However, many people in the early stages of the disease don't even realize they have it because patients typically exhibit few symptoms in the early going.
At the same time, providers can use AI and machine learning to measure the efficacy of available interventions for patients at various stages of chronic kidney disease. Coupled with real-time analytics that delivers insights that identify patients most at risk of developing chronic kidney disease, these technologies empower medical providers to serve their patients best with more timely diagnoses and treatments.
Q. How can AI and machine learning be used to close health equity gaps and improve access to care for underserved populations?
A. The first step to closing health equity gaps and improving care access for underserved populations is identifying patients who would benefit from these initiatives.
For example, we can pinpoint many different groups of patients, such as those who lack access to transportation, those who live in food deserts or those who live more than 10 miles away from the nearest healthcare facility. Once identified, we can refer patients to community-based organizations that address their specific SDOH needs.
Alternatively, local governments can make changes to infrastructure, such as using empty parking lots to set up mobile testing laboratories in areas where residents have difficulty accessing care. Similarly, governments can work with local retailers, such as discount stores, once we identify food deserts to stock fresh food options.
By combining geographic-mapping data with patient SDOH and demographic data, providers can discover where disparities exist and work with local stakeholders to design programs to address those disparities.