AI's Rare Disease Odyssey - Swathi Balaji

Note: The author utilized Midjourney, a generative AI program, to generate the above image from natural language descriptions.

In honor of Rare Disease Day (February 29, 2024) this month, welcome to the rollercoaster of medical mysteries, where patients embark on a journey fraught with uncertainty—the "diagnostic odyssey." For those facing rare diseases, the road to an accurate diagnosis can be a years-long saga, filled with countless doctor visits, unnecessary tests, and, unfortunately, often a misdiagnosis. This protracted timeline not only results in ineffective care but also leads to irreversible damage as the disease progresses. It's like a medical scavenger hunt, but the stakes are far from trivial.

 

Navigating Uncharted Territories

In the United States, diseases affecting less than 200,000 Americans are classified as rare, encompassing over 6,000 conditions worldwide (Cohen & Felix, 2014). These diseases, often chronic and disabling, pose a significant public health challenge. According to the National Organization of Rare Diseases (NORD), the average diagnostic journey for rare diseases spans 5-7 years (NORD Undiagnosed Rare Diseases Registry | NORD, 2022). This prolonged process not only delays treatment initiation but also inflicts considerable psychological distress on patients and their families. Living with a rare disease often subjects the patient to a lifetime of complex care, profoundly affecting their education, physical mobility, and financial stability.

Patients often find themselves on a "diagnostic odyssey," a term reflecting the feeling that no single practitioner comprehensively considers their condition. The current workflow for providers is to take a medical/family history, perform a physical exam, order laboratory tests, conduct imaging tests if needed, and then refer the patient to a specialist. The similarity of rare diseases to other conditions, combined with the no one-size-fits-all approach to diagnosing rare diseases, contributes to diagnostic delays or misdiagnosis. Even good doctors fail to recognize conditions that are right in front of them. The consequence of an early or late diagnosis can lead to worsened symptoms along with the development of other health problems, ultimately resulting in a decline in patient well‐being. These challenges not only affect more than 350 million people worldwide but also create a substantial economic burden on the healthcare system. The overarching need for a solution is clear: a way to streamline the rare disease diagnostic odyssey and support healthcare providers in their quest for accurate and timely diagnoses.

 

Is Artificial Intelligence a Buzzword?

Genomic technologies are technologies used to manipulate and analyze genetic information. The diagnostic landscape has evolved from using cytogenetic techniques using FISH and Karyotype, gene sequencing, and DNA microarrays which are still used in today’s practice. These are powerful tools that providers such as genetic counselors use to convey accurate diagnosis for patients and their families. This could help patients take the best medications and treatments for their disease indications.

While genetic research has come a long way since its original discovery, there is still room for more advancements and developments. However, focusing solely on genetic tools isn't enough if providers across different healthcare systems or countries cannot standardize rare disease findings. The collaborative sharing of sequencing data among clinicians, patients, and organizations is essential to build a robust worldwide network and raise awareness about rare diseases. For providers to practice at the top of their scope and avoid timely case-prep, it is vital to have a searchable or conversational platform to streamline the diagnosis process.

GeneMatcher is a freely accessible web site developed with support from the Baylor-Hopkins Center for Mendelian Genomics as part of the Centers for Mendelian Genomics network. It was designed to connect patients, their families, clinicians and researchers from around the world who share an interest in similar genes. The goal for making GeneMatcher available was to help solve “unsolved” exomes (Sobreira et al., 2015). This is done through cases from research and clinical sources.

While this platform has made significant strides in connecting patients, providers, and researchers worldwide, there is still room for improvement to make the process of searching for relevant information more streamlined. Artificial intelligence (AI) enables machines to perform operations requiring human intelligence, encompassing learning, while analyzing vast amounts of information to identify trends and make decisions with unprecedented speed and precision, emulating human intelligence (Wojtara et al., 2023). AI has the potential to revolutionize the diagnostic process by enabling doctors to analyze extensive datasets, including medical images, genetic data, and electronic health records, identifying intricate patterns difficult for humans to discern, ultimately offering an efficient solution for providers and shortening the diagnostic journey for patients.

Leveraging AI opens the door for healthcare providers to crowdsource crucial differentials specific to rare diseases, encompassing phenotypic characterization, specific biomarkers, historical data, pathology reports, and other factors, considering the inherent heterogeneity in the presentation of these conditions. Within the realm of AI, machine learning (ML) serves as a subset that aids diagnosis through various algorithms, including pattern identification and classification based on past examples. Given that 80% of rare diseases are genetic, AI holds significant potential in analyzing data to provide accurate diagnoses (Rare Genetic Diseases, n.d.). By constructing an ML algorithm, individual cases become puzzle pieces systematically pooled to create a comprehensive population-based dataset. AI then meticulously analyzes patterns within populations that share similar differentials, offering guidance to decode the diagnostic puzzle of rare diseases. An additional asset of AI, natural language processing (NLP), adds predictive analysis capabilities, particularly beneficial when extracting critical data from electronic health records (Wojtara et al., 2023). The overarching objective of using AI in genetic healthcare is to decode the diagnostic journey for individuals grappling with rare diseases, ultimately delivering the sought-after answers to patients through the systematic utilization of crowdsourced data and advanced AI analysis.

 

It's Only The Beginning

Embracing the transformative potential of AI in the rare disease diagnostic landscape not only enhances diagnostic efficiency for providers and offers hope to patients on their diagnostic odysseys but also symbolizes a significant journey for AI itself— an odyssey into the uncharted territories of rare diseases. The fusion of technology and compassion emerges as a powerful catalyst capable of positively reshaping the trajectory of medical diagnoses for rare diseases. The collaborative synergy of AI and crowdsourced data not only enriches our comprehension of individual rare diseases but also unfurls avenues for discerning shared patterns across diverse populations. This data-driven approach holds the key to decoding the diagnostic odyssey for individuals grappling with the complexities of rare diseases, providing a much-needed ray of hope for more accurate and timely diagnoses.

References

Cohen, J. P., & Felix, A. (2014). Are payers treating orphan drugs differently? Journal of Market Access & Health Policy, 2(1), 23513. https://doi.org/10.3402/jmahp.v2.23513

NORD Undiagnosed Rare Diseases Registry | NORD. (2022, August 5). https://rarediseases.org/living-with-a-rare-disease/nord-undiagnosed-rare-diseases-registry/

Rare Genetic Diseases. (n.d.). Retrieved January 30, 2024, from https://www.genome.gov/dna-day/15-ways/rare-genetic-diseases

Sobreira, N., Schiettecatte, F., Valle, D., & Hamosh, A. (2015). GeneMatcher: A matching tool for connecting investigators with an interest in the same gene. Human Mutation, 36(10), 928–930. https://doi.org/10.1002/humu.22844

Wojtara, M., Rana, E., Rahman, T., Khanna, P., & Singh, H. (2023). Artificial intelligence in rare disease diagnosis and treatment. Clinical and Translational Science, 16(11), 2106–2111. https://doi.org/10.1111/cts.13619

Show, Don't Tell: Empathy in Practice - Alex Stauff

Many of the skills I learned as a resident assistant (RA) have been helpful so far in grad school. A lesson that has really stuck with me came from a three minute video (link below) we watched during one of our summer training sessions. It was an animated short on empathy, with delightful anthropomorphic animals and a voiceover by Brené Brown.

In her words, “Rarely can a response make something better. What makes something better is connection.”

This little clip assuaged so much of the anxiety I had about supporting students through a difficult time. Fears like—what if I say the wrong thing? What if the other person can’t tell that I genuinely care?  

Words certainly matter, especially in a field like genetic counseling, where precision in language is critical. But without connection, words roll off like rain. There’s nothing anchoring them.

To me, connection means first and foremost: presence. This feels like an achievable starting point, requiring nothing more than focusing on the person in front of you. No matter how a conversation unfolds, I can be present.

Dr. Vellody, director of the Down Syndrome Center of Western PA, gave a really beautiful example of emotionally responsive communication in a lecture to first year students last semester.

We asked for advice about disclosing a Down syndrome diagnosis in the prenatal setting. I think many of my classmates and I were getting hung up on choosing the right words.

But Dr. Vellody’s focus didn’t start with the words. He said, provide the diagnosis, and then wait for an emotion to come up. Validate it. Wait for another emotion … validate it …. And keep doing this until a person is ready to talk about anything other than how they feel.

When I started my position as a GCA at Magee-Womens Hospital, I was able to sit in on weekly case conferences. This was a really helpful way to “peek behind the curtain” and start getting a feel for clinical and psychosocial considerations before I applied to school. I remember one particular situation a student shared, and the advice she got …

This student and her supervisor recently saw a patient who was newly diagnosed with breast cancer. The patient was extremely upset throughout the session and finally burst out at the student, “Have YOU ever had cancer?! Has anyone in your family even had cancer? Then, how can you have ANY idea what you’re talking about?” The student was distraught and had been at a loss for what to say – she posed the question to the group.

Truthfully, I was glad to be a fly on the wall during the discussion. It feels awful when someone’s anger feels—fairly or unfairly – directed at us. It can be so easy to slip into deflections and defensiveness.     

A seasoned nurse navigator gave advice that immediately reminded me of that short video I love so much.

She said, “You’re young, but I bet you’ve felt fear, or grief. And if you haven’t – you will. Those are likely some of the emotions underlying this woman’s anger. Acknowledge them and connect to those.”

I’m planning to keep that invitation to vulnerability and openness at the front of my mind as I prepare for patient simulations later this semester, and oh-so-quickly approaching rotations this summer.

Video: https://www.youtube.com/watch?v=1Evwgu369Jw