The value of Machine Learning in genetic risk assessment for primary care

 

We live in an era of precision medicine, where providers are expected to consider a patient’s variability in genes, environment, and lifestyle when providing care. Patients have growing access to “at-home” genetic testing and they are frequently bringing genetic results to their primary care physicians. However, many patients fail to realize that the world of genetic testing is vast and that there are more than 140.000 genetic tests currently available in the market [1] and this number keeps growing every day. 

In addition, there are a number of different guidelines that physicians need to adhere to when considering genetic testing for their patients. With all this data and genetic tests available, it is getting more difficult for physicians to keep track of the right genetic test for each patient. Luckily we are also on the era of Artificial Intelligence and of everything smart, where we can use machine learning to ingest this broad swath of variables and help physicians, genetic counselors and ultimately patients in the genetic risk assessment process.

 

Genetic testing as we know it

The genetic testing market has seen substantial growth in the last couple of years. With about 10 new tests coming to market every day [2] and with the expansion from single-gene tests to multi-gene panels, the world of genetic testing is hard to follow. A  recent study found that only 1 in 10 primary care providers had any confidence in interpreting genetic testing results, moreover, only one third had ever ordered genetic testing in the past [3]. 

In fact, before genetic testing even happens there is a whole process called genetic risk assessment. An individual’s genetic risk refers to the probability of the individual carrying a specific disease-associated mutation or of being affected with a specific genetic disorder [4]. During the genetic risk assessment, providers determine whether or not the patient needs genetic testing and which type of genetic test.

Calculating the genetic risk for a patient involves analyzing a multitude of variables such as ethnic background for the patient and their parents, and overall mutation rate for each ethnicity; and, if possible, the frequency of mutation in the population, as well as other more specific variables such as the presence of an independent risk factor derived from the individual’s medical and family history.
 

Genetic risk assessment: Complicated yet necessary

An individual must go through a genetic risk assessment in order to get genetic testing done.

This is usually trusted to certified genetic counselors, healthcare professionals with specialized training in medical genetics and counseling, who evaluate and understand a family’s risk of an inherited medical condition and who can recommend the appropriate genetic tests [5]. 

However, the surge in genetic testing means that genetic counselors are in high demand, creating a national workforce shortage that continues to grow. [6] There are only 5,000 genetic counselors in the United States, meaning there is 1 genetic counselor for every 65,000 Americans.

In addition providers and genetic counselors need to adhere to increasing professional medical guidelines from the National Comprehensive Cancer Network (NCCN), American College of Medical Genetics and Genomics (ACMG), American College of Obstetrician and Gynecologists (ACOG) and many others. Just this past summer the US. Preventive Services Task Force (USPSTF) updated its guidelines to include that women with a personal or family history of cancer should request their primary care physician to assess them with a familial risk assessment tool [7].

Also, while assessing the individual’s genetic risk and test selection, providers must comply with insurance requirements or they risk leaving patients with hefty bills. In summary, not only there is a vast amount of data providers need to consider but each individual has its own unique set of variables, making the genetic risk assessment even more complex but necessary to order genetic testing that will be properly covered.
 

Machine learning: solving the problem for the provider, genetic counselor, and patient

Machine learning is a subset of artificial intelligence in which algorithms learn from data. The key fact is that by using machine learning the system can ingest large sets of data and learn from the new data it incorporates over time. Machine learning has endless applications in the healthcare industry. It is currently being used for everything from streamlining administrative processes in hospitals to treat infectious diseases in patients [8].

By using machine learning in genetic risk assessment, we eliminate the issue of managing the accelerating number of variables. The proper algorithm can manage the various guidelines, complex genetic risk calculation variables and the growing number of possible genetic tests. 

However, unlocking the full potential of machine learning (ML) requires recognizing that algorithms must be properly trained. An approach to ensure a robust ML platform is to use human-assisted machine learning, which happens when human experts help the algorithms become more accurate over time.

In turn, the ML platform helps human experts be more productive. The ML platform gets “smarter” over time by using human experts’ feedback to increasingly improve its accuracy [9]. The challenge is still there, find the right algorithms and the best experts to train it.

The right machine learning platform:

• Takes into account all the patient’s available information
• Takes the burden off clinicians and clinics
• Allows primary care physicians to assess the genetic risk for their patients
• Identifies high-risk patients so genetic counselors can allocate their time to the patients who need them most
• Simplifies access to genetic testing for patients

 

In conclusion…

The increasing awareness and acceptance of precision medicine have resulted in patients who demand more of primary care providers in a fast-growing, ever-changing genetic market. The variety and complexity of genetic tests, medical guidelines, insurance requirements result in a complex patient data matrix that providers often find overwhelming.

As the shortage of genetic counselors grows, machine learning can address this issue and help clinicians, genetic counselors, and patients in their genetic testing journey. Moreover, machine learning helps democratize the genetic risk assessment process, so no matter the patient’s economic condition, location, they always get access to the best genetic testing recommendations. 

Do you think machine learning can help make genetic testing  “routine” in primary care?

 

________________________________________________________

Sources:

1. Concert genetics, A Coding Solution for Genetic Testing, September 2019. Link
2. Phillips KA, Deverka PA, Hooker GW, Douglas MP. Genetic Test Availability And Spending: Where Are We Now? Where Are We Going?. Health Aff (Millwood). 2018;37(5):710–716. doi:10.1377/hlthaff.2017.1427. Link 
3. Hauser D, Obeng AO, Fei K, Ramos MA, Horowitz CR. Views Of Primary Care Providers On Testing Patients For Genetic Risks For Common Chronic Diseases. Health Aff (Millwood). 2018 May;37(5):793-800. Doi: 10.1377/hlthaff.2017.1548. Link 
4. Baptista PV. Principles in genetic risk assessment. Ther Clin Risk Manag. 2005;1(1):15–20. doi:10.2147/tcrm.1.1.15.53606. Link
5. National Society of Genetic Counselors. Genetic Counseling Prospective Student Frequently Asked Questions. Link
6. Wired magazine. So Much Genetic Testing. So Few People to Explain It to You. October 10, 2018. Link
7. U.S. Preventive Services Task Force. Final Recommendation Statement. BRCA-Related Cancer: Risk Assessment, Genetic Counseling, and Genetic Testing. Link
8. Mike Thomas. Builtin. Ultra-modern medicine: examples of machine learning in healthcare. September 25, 2019. Link
9. Richard Harris. App Developer Magazine. How human-assisted AI may be the future. December 2017. Link