Technology(AI and ML) in EMS

[NysEms2117]

Being a tech nerd(and part time-ish EMT)I’m wondering what you full time providers think the benefits to artificial intelligence and machine learning are in the EMS fields. I think huge advancements can be made with artificial intelligence and “common issues” that emt’s and paramedics run into. I think for rural pre hospital providers artificial intelligence will have increased efficacy. Below are 3-4 links to get the conversation started(I do not agree fully with JEMS, but it’s there)... let me know your thoughts/ ideas

Enter Corti..

Now meet the good side of AI - Corti can spot a heart attack over the phone with 95% accuracy

Danish startup has created an AI that listens, and advises, on emergency dispatch calls

www.google.com

News Archives - JEMS: EMS, Emergency Medical Services - Training, Paramedic, EMT News

News Get the best news from the Journal of Emergency Medical Services.

www.jems.com

EMS agency set to implement Amazon's 'Alexa' into ambulances

Brewster Ambulance Services will begin using Amazon Echos to access the Massachusetts Emergency Medical Services Statewide Treatment Protocol

www.ems1.com

New partnership brings artificial intelligence to ambulances, paramedics

MedyMatch Technology and Samsung NeuroLogica have teamed up to help paramedics and EMTs more quickly and accurately assess stroke patients in prehospital environments. By integrating MedyMatch's artificial intelligence technology into ambulance-based mobile stroke units equipped with Samsung...

www.healthcareitnews.com

A data-driven artificial intelligence model for remote triage in the prehospital environment

In a mass casualty incident, the factors that determine the survival rate of injured patients are diverse, but one of the key factors is the time for triage. Additionally, the main factor that determines the time of triage is the number of medical personnel. However, when relying on a small...

journals.plos.org

 

[DrParasite]

I think AI could be a game changer in a lot of things..... but not in the practice of EMS or 911 dispatching, sorry

From a dispatching point of view, the AI system is attempting to mirror a human brain, so why not just use human dispatchers? One of the biggest issues experienced by dispatchers is the callers calmness or panicking often has 0 bearing on the actual level of sickness of the patient. Which is why they ask objective questions from the caller, and the caller is, quite bluntly, wrong in their answers. not always maliciously, just they don't give accurate answers to the questions asked.

The JEMS article references how Jack Stout revolutionized EMS with the suggestion of SSM, and if I saw Jack Stout today, I would punch him in the face for even suggesting such a horrible idea that has done nothing except hurt EMS in general. As every stock warning says, past experience does not guarantee future performance, and if you could see the future, go buy a lottery ticket instead of predicting where someone will get sick, or shot, or get hit by a car. And the author just wants you to check out all the cool things at the convention, i'm not sure he knows what he's talking about...

Integrating Alexa into ambulances (IE, an always listening device that is constantly transmitting information back to amazon) sounds like a patient confidentiality nightmare. Not only that, but if I want to look up a protocol, I want to SEE it, I don't want Alexa to read it off to me. What if Alexa pulls up the wrong protocol, and the paramedic treats off of it? who is wrong? Our county has an app, and our protocols are available by PDF. so if we look something up, it doesn't get transmitted to amazon or some other third party.

The Stroke thing looked like the AI was helping the identify the stroke, which take the place of the neurologist... so maybe? idk much about it

The triage thing was interesting.... program in some variables,, and determine suitability. Sounds very much like START and it's little brother, JumpSTART, but instead of following the flowchart on the card, you are entering it into a computer using different variables.

Where machine learning shines is in data analytics, and identifying trends, and retrieving information that is has learned. IBM's Watson going on jeopardy was a great example of this, competing against really smart people. And he won every time, because he was able to memorize a lot of stuff. But simply memorizing doesn't always cut it; you can find anything on google, but that doesn't mean you want to use WebMD to treat your illness.

Can AI/ML (which are two different things, but for the sake of complexity, I'll combine them) show you were you have chronic holes in your coverage area? yes. Can they identify patterns, and quickly transcribe 911 calls? yep. Can you find issues, and even identify weaknesses based on inputted data? likely. Can you use an AI system to look up protocols? I guess, but do you really want to? It can be a great resource (like using Google to look up a strange patient condition), but I can see it also being used as a crutch by weak providers who don't know their job (sort of like how ACLS final exams are now open book, because in the real world you can use all types of resources to find the solution).

Do you want to walk into the ER, and have an AI doctor assess and treat you, and perform any interventions? Even if they have been pre-loaded with every text book from medical school.....

Do you want your paramedic to have to rely on AI to figure out what is going on with you?

While I will argue that many 911 dispatchers are restricted to cook book dispatching based on scripted protocols, is it really a good idea to remove the human element with AI? Well, how many people like the voice response system when they call their credit card company or bank? Now take those people, and make them panicked because the reason they are calling is a life or death situation..... Can you see any potential problems here?

 

[NysEms2117]

The JEMS article references how Jack Stout revolutionized EMS with the suggestion of SSM, and if I saw Jack Stout today, I would punch him in the face for even suggesting such a horrible idea that has done nothing except hurt EMS in general. As every stock warning says, past experience does not guarantee future performance, and if you could see the future, go buy a lottery ticket instead of predicting where someone will get sick, or shot, or get hit by a car. And the author just wants you to check out all the cool things at the convention, i'm not sure he knows what he's talking about...

I referenced that I disagree with JEMS.

Integrating Alexa into ambulances (IE, an always listening device that is constantly transmitting information back to amazon) sounds like a patient confidentiality nightmare. Not only that, but if I want to look up a protocol, I want to SEE it, I don't want Alexa to read it off to me. What if Alexa pulls up the wrong protocol, and the paramedic treats off of it? who is wrong? Our county has an app, and our protocols are available by PDF. so if we look something up, it doesn't get transmitted to amazon or some other third party.

However where Amazons data center's and web proxies are in fact HIPAA compliant, i'd venture to guess even HITRUST compliant, which means the patient confidentiality thing wouldn't have legal repercussions, just ethical and most companies don't make their ethics known.

Do you want to walk into the ER, and have an AI doctor assess and treat you, and perform any interventions? Even if they have been pre-loaded with every text book from medical school.....

I think the integration of biometrics and AI/ML could be quite useful here. I think a regional database of known issues with a given patient based off of fingerprints could be quite helpful in the case of a patient that cannot describe past medical histories etc.. Police have proven the concept can work with mobile fingerprinting. https://www.bayometric.com/portable-fingerprint-scanners-law-enforcement/

Do you want your paramedic to have to rely on AI to figure out what is going on with you?

If it means saving my life yes. They can ask dora the explorer if it'll help. But if AI/ML will help then i'm kosher with that. Like everybody on this forum has been saying, paramedics are under-educated as it is.

 

[Summit]

I do a lot of work in IT and healthcare. Most AI isn't. You want to make machine learning choke on data hairballs, feed it **** from a healthcare database. Simple pattern matching + marketing department = AI.

 

[DrParasite]

However where Amazons data center's and web proxies are in fact HIPAA compliant, i'd venture to guess even HITRUST compliant, which means the patient confidentiality thing wouldn't have legal repercussions, just ethical and most companies don't make their ethics known.

yeah, they might be HIPAA compliant, and HITRUST compliant..... but I still don't trust them.... or Microsoft Azure.... and i sure don't like the fact that everything I say is recorded, I'm not the only one:

Amazon's Alexa Is Not Even Remotely Secure and I Really Don't Care

Forty-eight Cadbury Creme Eggs are en route to my house and it’s all Alexa’s fault.

gizmodo.com

How to prevent buddies from pranking you and burglars from getting in

Voice-activated smart speakers are gaining in popularity — but a hacker or a mischievous buddy could cause havoc if you haven't secured your device.

www.nbcnews.com

Naked Security – Sophos News

nakedsecurity.sophos.com

7 Ways Alexa and Amazon Echo Pose a Privacy Risk

Amazon's Echo devices are neat, but having a network-connected microphone in your bedroom is a potential privacy nightmare. Here's what we need to consider about Alexa's effect on our privacy.

www.makeuseof.com

Maybe I have more tin foil on my hat that you do (which is surprising), or maybe you just don't have enough. In either case, I'm not sold.

I think the integration of biometrics and AI/ML could be quite useful here. I think a regional database of known issues with a given patient based off of fingerprints could be quite helpful in the case of a patient that cannot describe past medical histories etc.. Police have proven the concept can work with mobile fingerprinting. https://www.bayometric.com/portable-fingerprint-scanners-law-enforcement/

I think you are 100% right in the fingerprint example; but you're also comparing objective information (a finger print) and rapidly searching a database. which AI is great for. There are only two answers, found a match or not found a match. treating a person is MUCH different. a lot more variables. And as any medical professional will tell you, patients don't always present as the book says they should.

If it means saving my life yes. They can ask dora the explorer if it'll help. But if AI/ML will help then i'm kosher with that. Like everybody on this forum has been saying, paramedics are under-educated as it is.

The solution to the under education is not to use AI as a crutch, it's to increase education.

 

[NysEms2117]

yeah, they might be HIPAA compliant, and HITRUST compliant..... but I still don't trust them

You don't have to, Personally I don't trust it either, for various other reasons, but the bottom line is it meets the requirement so technically could still be used.

I think you are 100% right in the fingerprint example; but you're also comparing objective information (a finger print) and rapidly searching a database. which AI is great for. There are only two answers, found a match or not found a match. treating a person is MUCH different. a lot more variables. And as any medical professional will tell you, patients don't always present as the book says they should.

After finding said match from the technological implementation side I don't think that a red banner saying: "Allergic to: penicillin or cephalosporins" is incredibly difficult. It would work very similar to how the Iphone medical ID works, except readily accessible. Again not something that's ready right this second, but I think it'd be handy to know things like chronic health issues, allergies by the scan of a finger.

Not trying to be a pain, just genuinely intriguing to me. I'm very old school with some of my technology, and I think it has it's benefits in certain places, however severe downfalls elsewhere. unrelated to EMS but - https://www.uclahealth.org/robotic-surgery/what-is-robotic-surgery

 

[DrParasite]

After finding said match from the technological implementation side I don't think that a red banner saying: "Allergic to: penicillin or cephalosporins" is incredibly difficult. It would work very similar to how the Iphone medical ID works, except readily accessible. Again not something that's ready right this second, but I think it'd be handy to know things like chronic health issues, allergies by the scan of a finger.

and there is nothing wrong with that; it's like using google (or your favorite search engine) to search a massive database for information. is that really ML or AI? I mean, search engines have been around since the 90s (lycos, yahoo and altavista were my favorites), which is pretty much you are discussing.

Not trying to be a pain, just genuinely intriguing to me. I'm very old school with some of my technology, and I think it has it's benefits in certain places, however severe downfalls elsewhere. unrelated to EMS but - https://www.uclahealth.org/robotic-surgery/what-is-robotic-surgery

Sure, but one important thing needs to be mentioned:

The robot is never, ever making decisions or performing incisions. Rather, your surgeon is telling the robot what to do, and the robot allows for greater precision than the human hand on its own.

the surgeon still needs to know what is being done, and be both accurate and precise in what he or she wants done; the machine is just the tool to implement it, so I'm not sure how much that qualifies as machine learning or AI.

I think in the field of data analytics, machine learning and AI will be ground breaking. research, pattern recognition, and going through big data and drawing conclusions are all things that will be affected. but I don't know if I could consider searching a huge database of PHI for everyone in the world to be a realistic AI possibility.