4 ways artificial intelligence can speed the adoption of precision medicine
How to take steps toward improving the use of data to help keep patients healthy.
Artificial intelligence has an important role to play in the data-driven science of precision medicine.
“The future of personalized care is about how we can take advantage of all this data being generated inside and outside the healthcare enterprise to help providers and patients make decisions about how to stay healthy,” says Tom Andriola, chief digital officer and co-director at The Institute for Precision Health at UCI Health in Irvine, Calif. “AI tools like machine learning, computer vision and natural language processing are enablers for us to achieve this end.”
AI can help speed up adoption of precision medicine by breaking down data silos, helping physicians pinpoint treatments, improving interaction with patients and helping to make new genetic discoveries.
Breaking down data silos
So far, precision medicine has focused on obtaining and analyzing genetic data to help identify patients with an inherited risk of cancer and other diseases, as well as developing and prescribing targeted treatments based on genetic insights.
Many experts now recognize that genetic data must be combined with other data, including information on a patient’s medical history, lifestyle and environment, to truly personalize care for individuals.
“Genetic data has gotten its own silo in healthcare,” says Peter Hulick, M.D., medical director at the Mark R. Neaman Center for Personalized Medicine at NorthShore University HealthSystem in Evanston, Ill. “That’s not doing anybody any good. We have to break down the data silos and work on interoperability across the organization for whatever data we’re collecting.”
To address this issue, NorthShore is storing the genetic data it obtains from testing patients as discrete data in electronic health records (see also: At the Frontlines of Precision Medicine: Data and IT Challenges). This allows the health system to use some of this data in clinical support tools.
One early win is a machine learning algorithm that warns hospital staff when a patient is at risk for long QT syndrome, a heart rhythm condition that can be deadly. Some people are born with genetic changes that cause this syndrome. Other causes include low potassium or magnesium imbalances and the use of certain medications (e.g., diuretics). The machine learning algorithm is trained to search a patient’s medical record for both genetic and nongenetic risk factors and issue a warning when a patient is at risk for prolongation of the QT interval, which can lead to a serious arrhythmia.
Helping physicians pinpoint treatments
To deploy precision medicine, The Christ Hospital Health Network in Cincinnati has embedded a number of AI-supported clinical decision support tools across the organization.
One machine learning tool is helping providers at the health network prevent adverse drug events by encouraging pharmacogenomic testing when recommended. The U.S. Food and Drug Administration has identified about 100 drugs that have gene-drug interactions. Pharmacogenomic testing determines whether a patient carries a gene or gene mutation that would interact with one of these drugs. When gene-drug interactions are present, the medicine may cause adverse side effects or may not work well in the patient.
The AI-supported tool reviews all the medications a patient is taking that are documented in the medical record. A gene-to-drug interaction probability score is then calculated based on whether the patient is taking one of the medicines that the FDA has identified as having potential gene-drug interactions. From this, the tool estimates the likelihood that a patient’s medicine would change if the prescribing provider ordered pharmacogenomics testing.
Researchers at The Christ Hospital Health Network determined that a patient’s gene-drug interaction risk score could predict an emergency department visit or hospitalization better than simply focusing on whether the patient has a major condition, such as kidney failure, heart failure and diabetes.
In addition to preventing adverse drug events, pharmacogenomic testing can help identify the best medicine for patients, says Burns C. Blaxall, Ph.D., executive director of precision medicine at the health network.
Improving interaction with patients
The Christ Hospital Health Network is also encouraging patients to get genetic testing to identify whether they are at risk for breast, ovarian, colon and other inherited cancers.
A pilot project targets women who are scheduled to get a mammogram. Five days prior to their appointment, women receive a risk evaluation survey, deployed as an interactive chatbot, via text or email. The survey asks questions about the patient’s medical and family history to identify whether they are at a higher risk of developing cancer and would benefit from genetic testing. When appropriate, the medical record creates an automated referral to the health network’s genetic counselors or High-Risk Cancer Screening Clinic.
see also: At the frontlines of precision medicine: Data and IT challenges
David Jones, Ph.D., offers a vision of how precision medicine will evolve in the next 20 years to transform healthcare.
By 2042, many babies will have their genome sequenced soon after birth, revealing their risk for hundreds of diseases and conditions, both rare and common, says Jones, chief scientist at Intermountain Health in Salt Lake City. Healthcare providers will have this genetic data at their fingertips to help them develop customized prevention and treatment plans for each patient, he predicts...READ MORE
The chatbot survey uses natural language processing. Patients can opt to answer the survey using their voices, and the NLP function automatically records the answers in the medical record.
Over one year, more than 20,000 women filled out the chatbot survey. Of those patients, 5,100 were referred for genetic counseling and testing and 2,400 were referred to the High Risk Cancer Screening Clinic, Blaxall says.
The provider organization expects to expand this preventive genetic testing approach to all primary care patients, Blaxall says. It’s also developing a cardiovascular disease risk chatbot that will soon be ready for testing.
Helping to make new genetic discoveries
HerediGene, one of the largest DNA studies in the United States, is underway at Salt Lake City’s Intermountain Health. The goal is to identify hundreds of yet-to-be-discovered genes and gene mutations that cause various diseases.
David Jones, Ph.D., Intermountain’s chief scientist, says AI technologies could help clean up gene sequencing data and then identify gene mutations from that data.
“Right now, it takes a human scientist to look at a genomic data set and conclude this person has a mutation in gene X,” Jones says. “This is a big data and artificial intelligence challenge. A computer should be doing all the analyzing and then just tell us what the answer is.”
Decades from now, many people will have genetic profiles as part of their medical records, indicating what genes or gene mutations they have that raise their risk of various diseases, Jones predicts.
Medical science will continue to advance, uncovering better ways to prevent and treat diseases with genetic underpinnings, he says. “There will be a big role for AI in analyzing the data in the background and then reporting back to providers and patients in a timely way,” Jones says. “For instance, patients may get a notification that says, ‘Here’s some new information about a genetic disease that you have and here’s a website where you can talk to a virtual genetic counselor about what you should do.”
This content is is part of a series of articles and related material highlighting the people, processes, and technologies currently advancing Genomics and Precision Medicine. See related content