Chatbots Magazine Become a Writer Intro to Chatbots

Bots for better healthcare


(kormilitzin) #1

Hi folks,

I’m wondering if anyone is working on bots for healthcare? Or if you know somebody who works on that topic.

I’m currently involved in an academic research which is related to machine learning in healthcare. We have developed a platform for collection of the self-reported data from patients involved in clinical trials. This is a standard online-fill-in-questionnaires type of data collection.

I’m curious, if bots might be a suitable framework for sensitive (personal/medical) data collection process?
Which bot vendors (messenger, Microsoft, slack, etc) might be a good fit for that purpose? The most important thing is the data privacy and data ownership.

I am sure that digital healthcare may benefit a lot from bots for data collection purposes and/or as assistants. The main question is whether the “standard” bot frameworks can be used for such tasks.

If you are up to this topic, I’m happy to discuss possible applications of bots in healthcare.


(Alyssa Ong) #2

Hey Andrey! There was an AI paper I took at University and my lecturer (who is a guru in health informatics) was interested in the potential applications of chatbots in the therapy space (e.g. a “replacement” counsellor for depression).

I can imagine some health forms can be tedious and having a bot can help guide people through it better - would that be the main reason you’d use a bot for data collection purposes? Or are there other reasons too?

In terms of using FB messenger/social media type of channels for data collection, I think a lot of people have the perception that whatever they enter into these channels will not be private. I’m not too sure what Facebook does with your private message data either, but sometimes I find ads appearing that relate to my private message chats (not sure if it’s coincidental).

I personally think for majority of people to have enough trust to enter data into your bot, you’d probably have to have some sort of direct chatbot on your website, or on a more “official” page. You can use bot frameworks to facilitate this (some like Microsoft Bot Framework may provide a web embed into your website), but again you will have to see how the bot framework you use handles the data (how long do they store it for, do they destroy it after). When I say bot framework I mean things like Chatfuel/Microsoft Bot Framework/API.ai rather than the channels (Messenger, Telegram).


(kormilitzin) #3

Hi Alyssa,

Thanks for your reply. Could you provide me with a link to the paper you mentioned? I am working in the field of mental disorders with special focus on learning temporal patterns of depression. My background is in machine learning and maths, so I think the paper will be of use to me.

I completely share you concerns about personal feeling of people, about sharing/passing their sensitive information over “popular” social media platforms, so this is why we developed our university’s platform, but I think in terms of scalability, we can’t really compete with professional platforms or offer the same usability.

BTW, many thanks for explaining me difference between platforms and channels, I’m super novice to bots and the terminology is still a bit confusing.


(Mark Stephen Meadows) #4

Hey Andrey ~

We’ve launched and/or prototyped a half dozen of these systems (Id Avatars, a project with the Chinese Government, and now iWithin, which is a general wellness bot for stress management, obesity, depression, etc). Our current build is dedicated to collecting specific end-user data and mapping that to specific recommendations. Simply putting the right info in front of the right person at the right time. We aggregate over 30 APIs for this and the build, currently in Alpha, will be available soon.

Though Botanic.io builds for more than wellness / healthcare (games, financial, etc) I’d be happy to jump on the phone for maybe a half-hour or so to lend a hand if I can be of service. email me at mark@botanic.io if you’d like to set up a time.

Oh, and don’t forget to check out Sense.ly and Skip Rizzo’s Ellie, from the ICT, as that’s been treating veterans with PTSD and they’ve done a ton of great research. Hope that’s some help.

  • MSM

(Niraj Swami) #5

Hi Andrey – we’ve been piloting a bot that translates patient info into a medication recommendations – part of an endocrine study to make better drug decisions (minimize side effects, maximize supplementary benefits).

There’s always a concern of privacy, but that’s probably why we’re headed towards a custom Bot UI app – something that allows us in-built encryption, 2-factor authentication and building a bridge of NLP & UX elements so as to limit sending data to a 3rd party api.

The benefits are great — conversational interfaces sit nicely in patient-physician communication – so the win is there. It’s just that the ubiquity of vendor-agnostic bots (work on Slack + Messenger, etc) – has to be compromised.

Hope this helps!

Niraj


(Cristina Santamarina) #6

This may be interesting for you: http://bots4health.com
I’m working on a chatbot to prevent teenage pregnancy in Uruguay and have a couple more ideas that will follow.
Happy to give you more info if you want to reach me at cristina@bots4health.com
I will also be talking about this topic at botscamp.co in December this year


(kormilitzin) #7

Hi Mark,

It sounds interesting, let me check it and I will come back to you.


(kormilitzin) #8

Hi Niraj,

This is what I am after, I think about “an avatar”, or a conversational agent which can interact more naturally (by means of NLP) with patients, rather than “old fashioned” website fill-in questionnaire. We’ve run several studies on compliance of patients (how they feel to routinely responds to prompts), and it looks like (which is pretty obvious), that if the UX is more natural, they prefer to continue with the self-reported system.

The main challenge is how to create “a natural UX” and I think that a proper bot with NLU might be a good starting point.

Let me dig deeper this topic and I may come back to you with further question, given your experience in this field. Thanks!


(kormilitzin) #9

Hi Hola,

Absolutely! I am very keen to learn from your experience and I hope to contribute mine. I think that application of AI/bots to healthcare is very important and we are doing the right thing.


(Sandeep) #10

How are folks thinking about HIPAA compliance with Messenger given that FB is not compliant ?


(kormilitzin) #11

Good point, Sandeep! Moreover, we should be thinking how to overcome this problem given the national healthcare compliance protocols, for example for the US, UK, Canada, Europe, etc. It might be a good point to think of a standardised procedure.


(Sandeep) #12

Agreed! There is some balance with between ease of use and distribution of a platform like Messenger while meeting the regulations. For Example - you can do an app with a conversational interface. That could be made compliant end-to-end but then you’re back to square one with Apps and notifications.

I wonder if SMS is considered a private channel maybe you can acquire users through public channels like Messenger etc. and get a ph# for compliant conversations… (thinking out loud here.)


(Alyssa Ong) #13

Hey Andrey, I don’t think he’s published any papers on it yet unfortunately, but if he does in future I will let you know :slight_smile:


(jonno.boyerdry) #14

Thanks for your post! I’m a cancer patient and built a chatbot to guide other cancer patients, their friends and families through high-quality resources that I wish I knew about when I was first diagnosed.

It’s incredibly inefficient that newly diagnosed cancer patients have to start from scratch with literally everything related to their disease. This inefficiency causes a tremendous amount of stress for cancer patients and their loved ones.

http://m.me/CancerChatbot


(kormilitzin) #15

Hey jonno.boyerdry

What a great idea! I’m sure this will be very helpful for other patients. It might be interesting if you could add some analytical tools for self management (statistics, time series of markers, self reported data, weight, etc).

From my experience working with chronic diseases (mental disorders), the feedback that patients receive from their data is extremely supportive and helpful. Simply by looking at their past data they can feel much better and stable. Of course this is very disease-specific and not everything might be directly translated into different medical fields, but a general idea of self-managements might be very cheerful and important.

If you have any ideas about further development and you would like to discuss it, I am more than happy to hear.


(kormilitzin) #16

Hey Alyssa,

Fab! Let me know if you have any news. I am very keen to learn more about this research.


(kormilitzin) #17

we used the SMS channel for data collection (along with emails and web GUI). In our case it was approved by the ethics committee, but I’m note sure what happens generally. I think it might very interesting (inspired by your post), to think about a standard international protocol for medical data collection via social media channels. It may open a new type of medical surveillance and really help people.


(Cristina Santamarina) #18

Hey all,

That’s a big pain point - though in my case I’m not really collecting any medical information, just giving information and facilitating a way to reach a doctor in case something is not right. Also I’m focusing on an Uruguayan market. As far as I know that means I do not need to be compliant with HIPAA. Any experts in the room?


(kormilitzin) #19

If you see a great success on your local market, why not to push it forward internationally? I am interested in hearing more from experts on legal aspects and HIPAA compliance.


(Cristina Santamarina) #20

I’m trying to reach teenagers at risk of getting pregnant. For this I need to make sure the personality and language of the chatbot really work for this public! The initial idea was that this could be expanded to other countries in Latin America first and globally later on (as there’s the translation work to be done), but I’m realising language variations will make this a bit hard.