Log in

Login to your account

Username *
Password *
Remember Me
Delay in the diagnostic process of rare diseases- lack of knowledge or leaking method?

Delay in the diagnostic process of rare diseases- lack of knowledge or lacking method?

Dr. Shmuel Prints, MD & MPH, an Internal Medicine specialist. Founder of NDC Medicine- a startup for diagnosing medical mysteries. Dr. Prints presented NDC Medicine for the first time in the "2nd World Congress of Rare Diseases and Orphan Drugs 2017" in London, UK.

Following my previous blog, I would like to explain what truly causes the delay in the diagnostic process of rare diseases. let's start from the basics.

As you know, the journey towards a rare diseases diagnosis (also known as the diagnostic odyssey) is long and full of difficulties. The average patient spends nearly 5 years, meeting 7 different doctors, & gets 2-3 misdiagnosis until they're accurately diagnosed. 

Studies in the past decade conclude that the weak spot mainly delaying the diagnostic process is outpatient care. The reason for it is considered to be a lack of awareness and knowledge about rare disease amongst primary and secondary care providers. That is why many efforts are invested in their training, in building online databases for rare diseases and so on. But little has changed. 

So what is truly the reason doctors worldwide, even in developed countries from Canada to Australia miss the possibility of a rare disease? 

In my opinion, the route of this problem lies in the diagnostic approach dominating the contemporary health systems, and guiding physicians throughout their career. Or as the famous saying describes it best: "If you hear the hoofbeats, think of horses, not zebras". 

The scientific base for this approach lies in the data mining for categorical items. that is called classification. It groups objects in categories or classes. The goal is to accurately predict the class for each case in the data.

Classification is not an exclusive method for medicine. It doesn’t matter whether you want to identify who is the artist of this portrait, or what disease does your patient with chest pains suffer from. In all these cases, we run the same investigation in order to classify the object. 

That is why I will use a simpler example from everyday life:

Let's say I just returned from the farmers market with a big fruit basket. Inside it I have oranges, apples, pears, and peaches. I left the fruit basket on the kitchen table. I kindly asked my grandson, who doesn’t recognize fruit types yet, to bring me some apples.

So I told him to bring me a fruit that is green, round and smooth.

The classification can be done in several ways:

 

The first is to group all the round ones, so we excluse the pears. Then we take out all the green ones and we find the green apples. To complete classification of all 4 fruit types, I must differentiate also the rough core fruits from the smooth ones and find the orange and peach.

The second way to classify is by grouping all the smooth fruits, so we exclude the oranges. Then by taking out all the round ones, we exclude the pears. And from the round smooth fruit separating the green ones, thereby finding the apples I wanted.

Eventually, to classify all 4 fruit types we can build 6 different "classification trees"- algorithms.

A differential diagnosis for a patient with chest pains can derive from 80 different reasons. (N-1)!=1X2X3… X79=  From this simple fruit example we can deduce, that the number of classification trees for this patient will be almost endless!

But what does the number of algorithms matter when the end result is the same? In most cases, it doesn't matter. But what happens in unusual cases?

Let's go back to the kitchen. Let's say this time we picked up at the framers market a new type of green oranges.

Using the first classification tree will cause a mix up between the apples and green oranges, and mismatch will occur.

But using the other classification tree will keep the division correctly.

In the medical world, symptoms & signs recognition is almost never perfect due to: problems in communication with the patient, bias lab testing or imaging tests results and so on.

Therefore we get imperfect classification algorithms, and we face the problem of choosing the best classification tree that will bring us as doctors to the right diagnosis.

Naturally, modern medicine chooses the common route. Doctors tend to give a diagnosis more common, as those illnesses occur often. They follow the Maximum Probability Concept and this way excel in classifying common diseases. However, physicians are more prone to be indifferent towards mismatches, like our green oranges, our zebras- like rare diseases. 

Rare diseases are not the only ones in the mismatching community. There can be diseases that are more prevalent in some countries but are rare in other areas. For example, tropical disease patients in developed countries also suffer from late diagnosis and go through a diagnostic odyssey. There are underdiagnosed diseases. They are more common diseases with unusual medical symptoms. Like most rare diseases, for a physician they will appear as "medical mysteries" and inevitably be neglected by the common diagnostic technique. Therefore, I group rare and underdiagnosed diseases under the same category – Neglected Diseases.

The important conclusion from all this is that in order to succeed in the diagnostic process of rare and neglected diseases we need a unique diagnostic technique for difficult medical cases.

Luckily, we already have it in the pattern recognition toolbox. We all know it and use it. Try to read this:

How is that possible if all the words are scrambled?

Well, the important element here is previous experience in English reading. What is so unique about this method? We correctly classify the object even though most of the features are mixed up or even false. We don’t know exactly how we do it, but clearly the success derives from our personal experience with the object.

These specific characteristics that we sometimes can't even describe are carved into the human memory and allow us to identify our relatives in childhood photos, to read texts written in hand writing and to brilliantly diagnose a rare disease we once handled before.

Scientifically, this approach is similar to another data mining method. It's called network learning. However, in our case, it is not the computer making conclusions and defining classifying rules based on experience. It is the human brain.

Implementing this approach is the basis of a revolutionary medical start up I founded to accelerate the diagnostic process of neglected diseases from several years- to several days. It's called NDC Medicine. To learn more about it click here.

Cookies make it easier for us to provide you with our services. With the usage of our services you permit us to use cookies.
More information Ok