Hello Doctor, I'm sick! I think I have ...

Hello Doctor, I'm sick! I think I have ...

The journey of the symptom checker

The other day I woke up with a stomach pain, so I googled it and used a symptom-checker; well at least that’s how it started and within the next half hour the internet and my brain convinced me that I might have anything from gastritis to kidney stones to … any guesses? of course it’s cancer ..
Oh, my bad! it’s “ovarian cancer”; you can’t be vague about a medical diagnosis after all, can you!?

The Age of the Cyberchondriacs:

If you are imagining an age of super cool cyborgs back flipping across the television screen, you couldn’t be more wrong. Rather, imagine a world where people are frightened of contracting every disease they can ever imagine, and how their lives would be (this can still be an interesting movie!)

Cyberchondria is the new internet age term for people who get anxious about their health after trying to gain medical advice from the internet.

While there is no denying that the internet is an infinite repertoire of information, and symptom checkers are the heralds of how technology can contribute to our health,  they lack the medical sagacity required to provide a more accurate diagnosis or take better decisions.


A doctor's dru(d)gery:

Scenario 1:
You have a stomach ache. You googled all possible conditions that can cause stomach ache and you started relating to it. You enter the doctor’s clinic. You tell him you’ve got stomach ache and a set of whole other symptoms that the internet associated with stomach ache.

Now your doctor has to:

  1. Identify the actual issue and solve for it
  2. Distinguish the actual issues from the internet-fed, fear-ignited symptoms
  3. Allay your fears and treat your anxiety too!

You are being treated for 2 at the cost of 1, victory? No. Instead of working with what you have and reaching a probable diagnosis, the doctor now has a lot of false positives to deal with and eliminate, each affecting his decision.

Moreover, a doctor who could be treating other patients is now misspending his precious time counselling you on what your mind has perceived as an acute condition. All of us would have been in that situation sometime.

Scenario 2:
You have a stomach ache. You didn’t google it (Amazing!). You enter the doctor’s clinic. You tell him you’ve got stomach ache and he asks you a few related questions and prescribes you medicines. Everything's going well. Then, you get back home and realise that you forgot to mention you were on medications some time ago. Maybe it’s irrelevant? Maybe not? Oh you are lazy to follow up... But then, why aren't the meds working as expected !!?

In this case,

  1. Your doctor would have made the decision based on your presenting and related symptoms but not based on all the medical history you missed out on summing up.
  2. You would either have to carry all your past medical documents or remember each and every medically relevant information at the time of your appointment and even then, there could be chances of error.
-

The solution to both the scenarios mentioned above is to collect and summarise any relevant medical information from a patient before their appointment and aid & assist a Doctor to make the right diagnoses, and these notorious symptom checkers present us with an opportunity here.

According to a study by Harvard students published in the British Medical Journal, which compared the top 23 symptom checkers that are using different algorithms across U.S, U.K and Europe, the correct diagnosis came up on top only 34% of the time, irrespective of the demographics. For the same set of symptoms, doctors were able to scale up to 71% accuracy.

Based on various journals published on Jama internal medicine and by Stanford researchers, experiments were conducted on physicians diagnostic accuracy on a wide scope of symptoms and across large demographics, which resulted in ~55% accuracy for initial findings.

Contrary to that, when doctors were armed with a patient’s medical history and symptoms, and then combined it with diagnostic decision support tools, the initial diagnosis efficacy increased to ~80%.


The EKA(only) way to do it:

The key to getting the right answers is asking the right questions, and doctors do that with great precision. They also know how to evaluate your answers and ask you the right follow-up questions so as to provide you the best treatment available.

Taking our learning from the points above, we developed an in-house health assessment product at Eka, using the powers of symptom-checkers, that focuses on:

  1. Asking the right questions and gathering as much medically relevant info (medical history, relevant symptoms, timelines) from a patient.
  2. Communicating with the patient in a language of his convenience. We have multiple regional languages to opt from.
  3. Providing a valuable summary of the data collected from the assessment, which when combined with other medical information that the patient has stored on our platform (like vitals, records, previous diagnosis etc), can assist doctors across various domains to provide the right treatment to their patient.

Our AI is built on continuous and ongoing input from doctors and tries to bake their daily practical experiences into it, and broadens the scope of medically relevant data gathered from patients instead of a theoretically heavy approach that has a sole purpose of providing diagnoses to patients.

Our assessment platform solves the following concerns:

  1. We save a doctor’s precious time by giving them a summarised collection of a patients symptoms and background clinical information.
    We took feedback from doctors using our platform; one of them quoted
“It saves at least 2 minutes of my examination time”

     In a clinic, that caters to 25 patients a day with each patient taking 15–20 mins of consultation time, thats 25*2 = 50 minutes of doctor extra time, to cater to the in-depth needs of his patients.

A Doctor armed with your symptoms and medical history and can help you better.

2.    A Doctor can now spend less time documenting and spend more time on the analysis of the information collected, thus making a more effective diagnosis based on sound medical data of the patient.
       A couple of doctors quoted,

“I can spend less time documenting and more time understanding the patient’s concern”
“In at least 3 out of 10 cases my drug choices have changed basis the additional information provided by patient”

A beauty with brains:

A little bit from the engineering perspective now, after all I’m an engineer. I won’t delve into the nitty-gritty of the product but give you an overall view. Our objective was to build a product that was:
1. User Friendly, so people could use it intuitively without racking their brains which was already troubled for their health
2. Highly customisable, which would enable us to collect as many different kinds of medical information (records, choices, text, date etc) from users as possible
3. An AI-based intelligent and medically aware platform, that would help ask relevant questions to users and aid a doctors diagnosis

We decided to build a backend driven, highly customisable input and information gathering platform that is driven by different engines (rule-based and AI) to efficiently gather medically relevant information from patients.

The Client

This feature is available only on android and IOS platforms. This decision helps us provide the most user friendly and smooth experience to users by utilising the native components of these platforms to enhance the experience.

The client breaks down the entire input gathering platform into components and uses set protocols to communicate with the backend which allows dynamic rendering of a component (eg: camera, numeric, date, files, choices etc) based on the question to be asked/ information to be gathered.

The Platform

The platform itself encompasses a collection of components, their interactions, actions (move between pages) and integrations with the Engines which decide on the question to be asked.

The platform has been devised in such a way as to have a minimum dependency on the client and to keep it oblivious to all logic (except rendering or redirecting) thus making it customisable and reusable by componentizing the presentation layer.

Secondly, It provides a vast generic interface that allows internal users to create various different kinds of flows for information to be collected from the user without involving code changes. It has the capability to be customised to the level of an individual doctor/practice.

The Medical Knowledge-base:

It is a repertoire of medical terms and jargon replenished by internationally certified sources like:
SNOMED (An internationally approved knowledge-base and ontology of healthcare concepts)
LOINC (international standard for labs & procedures)
and highly-qualified DOCTORS themselves, who are continuously reviewing and adding content based on their medical expertise.

This data is ingested by our AI to ask patients for relevant medical information and assist doctors in their diagnosis


Ok, so “what’s in it for me,” you say?

Enough of me blabbering about us already! I won’t answer this question for you, rather I would request you to use the app, try out the experience and give us your feedback. It matters to us! You matter to us! We are new, we are growing and we genuinely want to be the first stop for all your healthcare needs. So this is not a bragging session, rather a humble request to help us reach out to each and everyone out there, be their partner for all and any medical-related concerns and create awareness on how health is one of the most important aspects for a developing nation like ours. Enjoy the video !! Cheers to your health! :)


Sources & References:

  1. https://www.bmj.com/content/351/bmj.h3480
  2. https://www.researchgate.net/publication/341449810_Doctors'_diagnostic_accuracy_with_and_without_diagnostic_decision_support
  3. https://jamanetwork.com/journals/jamainternalmedicine/fullarticle/2565684 -Comparison of Physician and Computer Diagnostic Accuracy
  4. https://jamanetwork.com/journals/jamainternalmedicine/fullarticle/1731967