COVID-19 caught the world off-guard, sending governments, populations, and the entire healthcare community — physicians, nurses, scientists, pharmaceutical companies, and medical device manufacturers included — scrambling to discover and deliver rapid solutions as millions of lives hanging in the balance.
The heroic efforts of everyone involved have at last brought the end into view; now, as we begin to emerge from the pandemic and are able to assess its long-term impact, one important development becomes clear: the dramatic rise in the use of artificial intelligence (AI) to help doctors, scientists, and other healthcare practitioners achieve new insights and reveal unseen relationships between infection, rapid treatment, and long-term results faster than ever before.
We are now at an inflection point. More technology companies than ever are working to develop novel AI-enabled algorithms and devices to aid biopharmaceutical firms in drug discovery and label extension. More medical centers are turning to AI to analyze their EMRs in an effort to improve patient care. New hopes abound; so do unprecedented questions for the industry to consider.
The union of machine intelligence and human health
Technology and artificial intelligence have the opportunity to dramatically improve healthcare research and the delivery of clinical care.
For example, Mayo Clinic recently launched Anumana, a new partnership that develops and delivers ECG algorithms to enable early diagnosis and interventions for hidden and/or undiagnosed cardiovascular diseases. Anumana applies machine intelligence to the enormous repository of data in Mayo Clinic health records, enabling AI algorithms to see patterns in ECGs that humans cannot, thus unlocking previously hidden correlations between symptoms, vital signs, and disease. This kind of innovation empowers healthcare providers to diagnose otherwise hidden conditions sooner than ever before, allowing for earlier intervention and improved patient outcomes.
Focus on patient privacy
Such technological advances rely heavily on in-depth analysis of patient health records, and the opportunity to transform this massive amount of data into actionable insights is the future of healthcare. But at the same time, peoples’ demands for privacy are growing. The amount of health data being generated is growing at almost 50% per year, and rapid technological advancements are creating new data privacy challenges as a result.
De-identification of patient data, which removes and/or alters any information that can put a name to a record, is a vital part of the role AI plays in processing health records. Natural Language Processing (NLP) technology in this area is evolving dramatically as leading health technology companies and medical centers focus on developing stronger, safer, and more effective mechanisms by which they generate and preserve anonymous patient data.
Better data for more confident results
The most potentially valuable information in patient health records exists as what is called “unstructured” data, including physicians notes, lab reports, scientific papers, pathology images, and other knowledge that is in natural language or diverse formats that have not been readily “computable” except by human intervention. Such unstructured data is rich in biological context and therapeutic outcomes, yet even given years, teams of experts would not be able to interpret such vast quantities of information, much less derive meaningful insights and translate them into better diagnosis, care, and therapeutics for patients in need.
Today, advanced NLP technology is able to speedily process such unstructured data, synthesizing and harmonizing it with other, siloed biomedical datasets. This gives researchers the power to ask questions of massive amounts of aggregated data and receive answers more rapidly and accurately than ever before possible. Scientists can now produce real-world evidence in real-time, quickly converting the world’s biomedical knowledge into deep insights that advance the discovery and development of diagnostics and therapeutics.
Such rapid advances naturally come with questions. Chief among these are concerns that research performed by AI-enabled algorithms is sometimes based on flawed methodology and limited and/or low-quality data, leading to skewed, inaccurate results. While this may be true in some cases, many of the leading healthcare technology companies are committed to partnering only with premier academic medical centers, conducting research on large, diverse, and deep datasets, and submitting results to rigorously peer-reviewed journals.
The power to address racial disparities
Unfortunately, racial disparities are well documented across a range of healthcare settings and diseases. It is vital to ensure that future research and clinical care enabled by AI algorithms represents all populations and that unconscious biases from the practice of clinical medicine don’t unwittingly influence future algorithm development.
Fortunately, when harnessed properly, insights provided by artificial intelligence can help clinicians better understand treatments for various racial and ethnic groups free of any biases. If these AI algorithms are trained with data sets drawn from diverse populations, they will be a vital tool in eradicating racial bias from healthcare and focusing on what matters: caring for everyone equally.
A new frontier for treating disease
Over the last 15 years, the world has seen an explosion in the quantity of healthcare data available, one so massive it has far outpaced the human ability to consume and make sense of it. At the same time, advances in machine learning and computing power began to allow us to extract and process this staggering volume of knowledge information in ways never before possible. By bringing this to the world’s attention, the COVID-19 pandemic showed us just how far and how fast we can push the envelope when it comes to leveraging healthcare data to uncover real-world, life-saving, and time-saving insights.
