How AI is Cracking the Code on Earlier Diagnosis of Rare Disease

By Komodo Health Editorial Team

There are rare diseases, and then there are diseases so rare that they often go undiagnosed for several years, even as morbidity and mortality risks mount 1, 2. Homozygous familial hypercholesterolemia (HoFH) is one of those underdiagnosed, undertreated, and largely unrecognized diseases1, 2. For those who have this rare disease, the prognosis is typically early coronary artery disease and valvular heart disease, resulting in a high risk for myocardial infarction (heart attack), ischemic stroke, or sudden death1, 2.

Caused by genetic mutations that allow a build-up of low-density lipoprotein (LDL) cholesterol, 3, 4 also known as “bad” cholesterol, HoFH affects approximately one in 250,000 people5. Left untreated, patients develop extremely high cholesterol levels at an early age; thus, this rapid increase in LDL can cause atherosclerotic cardiovascular disease (ASCVD) in the first decades of life6.

When caught in time, however, the disease can be managed with diet and medication. With very little awareness of HoFH among providers, however, and no ICD-10 code for accurately identifying and tracking existing patients, identifying patients early has historically been difficult3,7. As a result, most patients are diagnosed once their symptoms have become very serious2.

AI Finds Its Calling in Healthcare
This is precisely the type of scenario that proponents of AI in healthcare have been holding out as the ultimate use case for machine learning algorithms that have the ability to scour through reams of unstructured data to find the needles in the haystack.

Nature Paper-BUTTONThey were right.

In fact, Komodo Health researchers, working in collaboration with Regeneron, recently published a study in Nature that illustrates how AI-powered algorithms, when paired with a robust database of real-world patient journeys, can be used to identify patients with HoFH based on the pattern of symptoms, diagnoses, and treatment they receive prior to diagnosis.

The study, “A machine-learning algorithm using claims data to identify patients with homozygous familial hypercholesterolemia” utilized claims data from the Komodo Healthcare Map™ to develop a machine-learning model to identify potential HoFH patients. By identifying true HoFH patients (confirmed via physician or prescription medicine used to treat HoFH), the research team was able to identify 331 patients with HoFH in the Komodo Healthcare Map.

From there, the team was able to analyze the complete healthcare journeys of those patients, compare them against a cohort of non-HoFH patients with risk of cardiovascular disease, and start testing AI-powered machine learning algorithms to zero in on the specific patient characteristics of the true positive HoFH cohort. These included a range of common comorbidities, age, gender, and therapeutic profiles – even commonalities in lab results data.

Identifying Patterns to Trigger Earlier Intervention
The end result was a predictive algorithm that is capable of identifying patients who have a high likelihood of having HoFH, and in some cases, before they are even formally diagnosed.

While there are some limitations acknowledged in the study – notably that the data used in the analysis only goes back to 2015, which could exclude some important information from early in the patient journey from the predictive model – the findings are enormously encouraging.

By using AI to scour through layers of detailed data that humans could never reliably parse and to find unexpected connections between those disparate data points, researchers are breaking new ground in understanding the trajectory of ultra-rare diseases like HoFH. It’s important to remember that AI models like this one are only as good as the data upon which they’re trained. Ensuring that the foundational data underpinning these algorithms is comprehensive, robust, and accurate is, in no small part, critical to addressing the challenges in identifying HoFH patients. Armed with this information, providers, Life Sciences teams, public health officials, advocacy groups, and payers will be able to improve screening and diagnosis of high-risk individuals earlier.

For more information about how Komodo Health is using AI to support rare disease research, check out “Predictive Alerting: Leveraging Novel AI Approaches for the Early Detection of Patients With Rare Disease”.

To see more articles like this, follow Komodo Health on Twitter, LinkedIn, or YouTube, and visit Insights on our website.

References
[1]
Gidding, S. S., Ballantyne, C. M., Cuchel, M., et.al. (2024). It is Time to Screen for Homozygous Familial Hypercholesterolemia in the United States. Global heart19(1), 43. https://doi.org/10.5334/gh.1316
[2] Cuchel, M. Bruckert, E., Ginsberg, H.N., et al. Homozygous familial hypercholesterolaemia: new insights and guidance for clinicians to improve detection and clinical management.  A position paper from the Consensus Panel on Familial Hypercholesterolaemia of the European Atheroschlerosis Society, European Heart Journal (2014), 35, 2146 – 2157, doi:10.1093/eurheartj/ehu274.
[3] Cuchel, M., Raal, F.J., Hegele, et al. 2023 Update on European Atherosclerosis Society Consensus Statement on Homozygous Familial Hypercholesterolaemia: new Treatments and Clinical Guidance, European Heart Journal., 2023, 00, 1-15,  https://doi.org/10.1093/eurheartj/ehad197.
[4] Raal F.J, Hovingh G.K, Catapano A.L. Atherosclerosis. 2018; 277, 483-492.
[5] National Organization for Rare Disorders, Understanding Rare Disease, Homozygous Familial Hypercholesterolemia, Homozygous Familial Hypercholesterolemia – Symptoms, Causes, Treatment | NORD  accessed October 25, 2024.
[6] Kayikcioglu, M, Tokgozuglu, L., Current Treatment Options in Homozygous Familial Hypercholesterolemia, Pharmaceuticals, https://www.mdpi.com/1424-8247/16/1/64 . 16 (1), 1-14. https://doi.org/10.3390/ph16010064

[7] Centers for Disease Control and Prevention. National Center for Health Statistics. Classification of Diseases, Functioning, and Disability, ICD-10-CM. ICD-10-CM | Classification of Diseases, Functioning, and Disability | CDC   Accessed October 25, 2024. 

Explore Related Posts