skip to Main Content
  • This page as PDF

AI in Healthcare: 5 Areas in Which Artificial Intelligence Is Disrupting the Status Quo

Summary

The rapid expansion of artificial intelligence (AI) across healthcare holds the promise of dramatically altering not only diagnosis and treatment but also research, risk assessment, drug development, care management, and even insurance and payment systems.

AI’s ability to approximate human outcomes without direct or explicit human involvement is opening the door to a range of disruptive applications capable of boosting care quality and access while saving money and time.

Confidence in the technology’s potential is reflected in the projected growth of the healthcare AI market worldwide: analysts expect it will jump to more than $19 billion by 2026 from $1 billion in 2017. In the US alone, AI could generate healthcare savings of $150 billion annually by 2026.

AI Defined

AI generally is defined as “computer systems able to perform tasks normally requiring human intelligence, such as visual perception, speech recognition, decision-making, and translation between languages.” Put another way, AI is an array of technologies that “allows machines to sense, comprehend, act and learn.”

Acquiring these capabilities requires that AI systems be trained via machine learning to identify patterns in large volumes of data, then programmed to draw inferences or conclusions based on their cumulative experience. The objective is to spot anomalies that humans might otherwise miss while achieving ever-greater levels of accuracy as the underlying algorithm self-corrects and the data set expands.

AI also can utilize natural language processing to extract relevant information from unstructured data like clinical notes and medical journals. Deep learning utilizes layered, nonlinear algorithms to mimic biological, neural networks in order to capture and act upon a wide array of non-structured or non-labeled information. Compared to machine learning, deep learning offers the potential for faster and more accurate conclusions based, in part, on its ability to recognize errors.

Origins

Today’s explosion of AI applications—in healthcare and beyond—marks the culmination of concepts first proposed in the 1950s and 1960s. The term artificial intelligence was coined in 1955 by John McCarthy, a visionary cognitive scientist who organized a landmark workshop the following year at Dartmouth College to explore the challenge of replicating human thought with computers.

Ten years later at Stanford University, the DENDRAL Project was developed to help organic chemists identify unknown molecules. It is considered the first expert system

capable of automating decision-making and problem-solving. The project spawned a range of other programs that advanced the science of machine-learning algorithms.

With the rise of personal computers in the 1980s and 1990s and the vast data collection they enabled, AI concepts migrated from the research lab to clinical applications. Accelerating this transition was the deployment of electronic health records (EHRs) and the wealth of clinical information they provided. Recent advances in computers’ ability to quickly analyze enormous amounts of data have further extended the reach of AI.

Applications

AI is clearly in its infancy relative to the technology’s ultimate potential, but multiple applications are already poised to disrupt healthcare’s status quo—including some that rely on patient wearables to generate real-time biometrics. Here are 5 key areas in which AI is showing significant promise:

1. Preventive Health and Risk Assessment

Chronic, non-communicable diseases, such as cancer, diabetes, cardiovascular, chronic respiratory, and mental health conditions, represent 90% of US healthcare spending annually, according to the CDC. As such, harnessing AI to prevent and predict chronic disease and other illnesses has become a national priority. A diverse array of applications has emerged, including:

  • An automated tool to accurately identify a common marker of heart disease in patients receiving chest computerized tomography scans for lung cancer screening
  • Retinal scans of smokers to spot biomarkers or physiological changes associated with increased cardiovascular risk
  • AI assessment of structured and unstructured patient data, including demographic and prescription drug information, to identify patients at risk for hospital-acquired infections, sepsis, and readmissions

2. Diagnosis

Diagnostic tools represent one of the largest areas of current AI activity. Some notable examples:

  • AI is being applied to diagnostic imaging in a wide variety of ways, with studies showing AI can equal or surpass human capabilities in the identification of imaging features; a 2017 study, for example, found that a deep learning network can recognize specific forms of breast cancer in images with 100% accuracy;
  • Molecular biology expertise and AI are being combined to provide early cancer detection from blood tests through the recognition of disease-associated patterns in DNA, RNA, proteins, and other biomarkers
  • A recent study found that ordinary wearables like Fitbit and Apple watches linked to an AI application were able to detect diabetes with 85% accuracy, high blood pressure with 80% accuracy, and sleep apnea with 83% accuracy; the study included 200 million sensor measures from more than 14,000 participants worldwide

3. Precision Medicine

Though still largely in the conceptual phase, precision medicine—the refinement of care to meet individuals’ specific needs by accounting for their gene variability, environment, and lifestyle—is being driven almost exclusively by AI.

  • Efforts are underway to harness AI to provide clinical decision support for genetic test results at the point of care in support of data-driven, personalized care plans
  • Microsoft computer scientists have partnered with the Jackson Laboratory, an independent biomedical research organization, to curate via AI existing medical research about cancer-related genetic mutations and drugs that target them to help clinicians identify the most appropriate treatments for specific cases
  • The UCLA Institute of Precision Health is bringing together multiple disciplines to make the use genetic and genomic data more practical for patient care. UCLA researchers are also synthesizing huge amounts of both structured and unstructured clinical data via cloud computing and AI to provide physicians and researchers with more readily available clinical insights and knowledge.

4. Drug Development

AI applications are gaining traction across the drug development and testing environments to help refine and accelerate compound development, speed products to market, and reduce costs.

  • Berg is using AI to analyze patient biological and outcomes data to identify distinctions between cancerous and healthy cells and then applying the knowledge to drug development to create an improved understanding of how drugs work
  • A company called Antidote is more efficiently matching patients and clinical trials to accelerate the trials process and improve patient access
  • Atomwise has harnessed supercomputers linked to a database of molecular structures to speed determinations about which drug compounds can be effective; in 2015, the company flagged 2 potential drugs for reducing Ebola infectivity in less than a day; the process traditionally would have taken months or even years

5. Administration and Care Delivery

As important as AI is in the clinical environment, it also holds major promise in healthcare’s administrative and care delivery domains. Applying AI to many of the processes and functions that support care delivery and payment is expected to produce a substantially more streamlined and efficient healthcare system in the years ahead.

  • An avatar-based mobile phone chatbot app, linked with patient self-monitoring, has produced a 75% decrease in 30-day hospital readmission rates for patients recently discharged with chronic heart disease and a 66% reduction in patient monitoring costs
  • Qventus’s automated AI platform helps hospitals more effectively predict and manage a range of functions, including discharge planning, patient prioritization, wait times, and staffing
  • Robotic process automation is being applied to repetitive, high-volume tasks in revenue cycle management, benefit verification, prior authorization management, and vendor contract management to decrease days in accounts receivable, increase cash acceleration, and increase collections.

Avalere has expertise in marketplace and regulatory trends, data collection and analytics, and strategy and communications. Many years of experience working with clients and partners in these areas enable us to continue to facilitate key partnerships and further advance AI opportunities in healthcare.

To receive Avalere updates, connect with us.

Find out the top 2020 healthcare trends to watch.

From beginning to end, our team synergy
produces measurable results. Let's work together.
Back To Top