Short answer: Arun Kalaiselvan says Arya now builds care teams around patients instead of filling shifts. Anand Chandrasekaran says a heuristic eligibility system scores caregivers in tiers, and Arun says they are studying drive time as a traveling salesman problem.
Watch this answer from Episode 24 (starts at 23:20) · Episode 24 notes and transcript
From Episode 24: This AI Fills Healthcare Shifts in Minutes, starting at 23:20.
Arun Kalaiselvan [23:20]: Absolutely. That's that's kinda what we are hoping for as well. In fact, we had a recent shift internally on how we think about shifts. So we've stopped talking about shifts. We're talking about people now. So everything is about the patient. We don't say, hey, this shift is available. We say, this patient is available to be taken care of. Right? And that subtle change is the difference between, oh, I'm going to go service a shift versus I'm going to go take care of a patient. But also, it's a shift in thought process of how do you think about schedules, right? Instead of saying, hey, I have this schedule that's open today and I'm going to take it, we say, there's this patient who needs to be taken care of, say, every Monday of the week. And now a caregiver can go, okay, I would like to take out this patient every Monday of the week. I'm going to assign myself to, like, the rest of their care time period or whatnot. So, like, we are now building care teams rather than just filling shifts.
Arun Kalaiselvan [24:15]: And that's the big sort of the change in ideology and methodology here entirely.
Anand Chandrasekaran [24:21]: And one thing that's not spoken enough in this chapter of how we do it is we have our own, like, really trick, crude eligibility system that allows for all of this to happen. We try to get information from all of our customers, but we kind of also have our own heuristic based eligibility system that helps us get this system going. And we've been very, very successful in matching caregivers only because of that. And that is something that we built, like, one of the first things that we built when we decided that we're gonna do scheduling. And from using that, we've gotten to, like, onboard all of our clients into that because we were able to understand. I mean, Kunal, our CEO, worked with them and tried to understand what the process is and came up with a solution which we implemented. And we got it to a point where it's like you can take any customer. Most of the systems in that is going to be the same. And we kind of get the data plus add some more heuristics and that we go. We get you the results that you need.
Anand Chandrasekaran [25:20]: That's actually been one of the biggest differentiators when it comes to today, how they are seeing and how are they benefiting is is what you're seeing in this. Right? Like, the more impact is because of that set of calculations and heuristic analysis that we do and that's that's been a very, very important aspect.
Arun Kalaiselvan [25:35]: And to add a color to that, the output of that is a tiered system. Basically, we say for this patient, here is the top tier of caregivers that are eligible and these are the scores per caregiver at that pace. And then we will go from the highest scoring caregiver and, like, pull up the first tier, call them if we need to urgently fill a shift, for instance. They get the first phone calls, and then we'd go down the tiers and whatnot. So that, we're like, how do we maximize the best care possible for the patient while also maximizing returns for our customer and making sure that schedules are streamlined. Something that we're thinking about is how do you help caregivers manage their schedules in a fashion where they aren't spending a lot of time behind the wheel, for instance. So we're, like, we're looking at the traveling salesman problem where we're playing
Jordan Metzner [26:26]: Classic greedy algorithm. Right?
Arun Kalaiselvan [26:28]: Yeah. How you like how do you schedule visits for a patient where they start from home, they drive, and then the last shift is this them being pretty close.
Jordan Metzner [26:39]: You have a traveling salesman problem with time windows. I know all about this problem. Oh, man. You have no idea, guys.
Sam Nadler [26:46]: We can talk about this
Jordan Metzner [26:48]: A whole another episode.
Anand Chandrasekaran [26:50]: Keep saying that it's not a traveling salesman problem because there is a time window and known params.
Jordan Metzner [26:56]: No. I'll tell you the book is on Amazon. I can tell you the author who wrote it. And there's a bunch of traveling salesman problems you have with time windows with different types of constraints in your particular circumstance. My guess is you have like a known start time but an unknown end time of care. So you have some type of time window that you have to guarantee against. But I'd love to talk about it more offline. We could talk about traveling salesman indefinitely. Okay, let's move on to the news because this is probably maybe even getting a little too nerdy for our audience.
Sam Nadler [27:26]: Yeah. Let me kick it off.
