Why AI could be key for rare diseases
Dr Peter Fish, head of clinical partnerships at Mendelian discusses why artificial intelligence (AI) could be the key to diagnosing rare diseases.
Increasingly, technology seems to be at the heart of daily life, from hailing cabs and fitness trackers to online marketplaces. As with many other industries in the last few years, digital innovation is increasingly reshaping access to healthcare and how clinicians operate, with GPs available for consultations over video chat and digitalisation named as a major priority in the NHS’ longterm plan. Yet as advanced technology continues to help the majority of people lead healthier lives, it’s important to consider how it can be used to diagnose and treat more complex cases like rare diseases.
Many think rare diseases only affect a small number of people and that you are unlikely to ever meet someone with a rare disease, much less be afflicted yourself. In a sense, that’s true, with some rare diseases, like Progeria, so unique in number that they may only affect a handful of people per country. However, it is estimated that rare diseases affect 350 to 400 million people globally and around 1 in 17 people will have a rare disease at some point in their lives, meaning that as a group, rare diseases are actually far from rare and have wide-reaching repercussions.
Currently in the UK, it takes on average 5.6 years to even be diagnosed with a rare condition, many of which may be chronic, life-changing or fatal. During these years, patients undergo a distressing diagnostic odyssey of sorts, on average visiting eight physicians, including four specialists. Patients endure many tests, from imagery or blood tests to more invasive practices, and often receive misdiagnoses - sometimes multiple times - before finding an answer to their condition.
While frightful and frustrating for individuals, it’s not only patients who feel the consequences of such inconclusive hospital visits, but also clinicians, healthcare systems worldwide and ultimately, taxpayers. Last year, a Mendelian-commissioned report looked at the cost and resource impact in the last 10 years and found that, while undiagnosed, rare disease patients have cost the NHS significantly more than £3.4 billion, as existing diagnostic pathways for rare diseases are inefficient and large amounts are spent on inconclusive tests.
However, with the advancement of digital innovations in the healthtech space, AI could be the answer to this growing issue. While technology is being used to discover new drugs, increase genomic testing and improve quality of life, it can also be used to filter and flag possible rare diseases to physicians much earlier in the diagnostic journey, benefiting patients, providing information, and reducing unnecessary costs along the way.
Diagnoses through AI
Artificial intelligence solutions can conduct an audit of lengthy research in a fraction of the time of doctors, scanning the latest discourse and updating local knowledge on a daily basis. For example, technology can source information from a multitude of different resources globally to give clinicians and patients a clearer picture of symptoms and possible causes in relation to rare diseases. Generally, clinicians use pattern matching with regards to patient data (height, weight, medical history, etc.) to form a diagnosis, but can only measure hundreds, or at most thousands, on their own. Using Mendelian, clinicians can computationally, and in seconds, use tens of thousands of data patterns to generate a ranked list of diseases that can be focused on to find a diagnosis. In many cases, clinicians may never have heard of or seen cases of these diseases before and may not naturally have considered them in the patient in question.
Implications for widespread use
It’s good news that there is progress in the right direction - currently, AI solutions are being trialled by the NHS to help GPs identify patients with rare or hard to diagnose conditions. Already, a grant from Innovate UK allowed for the implementation of Mendelian’s specialised screening system in select clinics, providing augmented intelligence with data analysis to flag those that match potentially unique conditions. Once the technology has analysed a patient’s symptoms, they are then flagged to the GP who has various options, including referring the patient to a specialist or recommending further analysis and testing.
Flagging patients with certain criteria or symptom patterns may also aid drug trials down the line, as only 5% of rare diseases currently have a drug treatment or cure. As clinical trials require a minimum number of patients, at times there often aren’t enough diagnosed patients for trials to be initiated. Instead, such studies must collect a sample size by reversing the process: searching for symptoms rather than already diagnosed patients, a process which can be expedited through such AI identification solutions.
As with anything, the first step to solving a problem is to name the issue. With the help of AI solutions that expedite diagnoses, around 400 million people across the world may begin to fight back against their rare condition faster than ever before, providing research, awareness, and efficient drug therapy for rare disease patients to come.