[00:00:05]
>> The I've been. Around since I have to do my introduction. I've been trying to talk for more than 15 years. Professor in industrial system engineering. And the past 201312 I've worked primarily in health care delivery and doing all kind of modeling related to health care delivery and you can see my slide really as this is what really what I'm doing policy help I'm going to x. with really massive data.

[00:00:42]
And I'll give you a kind of an idea of what kind of data us as I have in case we have any any anyone interested to work with with those data sets. My collaborative of children home from the whole of Emory University t c Do you see all the.

[00:01:06]
I've covered quite quite a bit the. Collaborators outside of Georgia. Books one as related to the complexity of health care the other one is hope your axis are right main areas of interest. And here is a snapshot of the things I'm doing. Really coming from really my focus is fundamentals but really when I when I referred to fundamentals really I caused boundaries across multiple areas including dusty engineering the lunatics and all that spills through teaching advisement and so on this is the website of.

[00:01:49]
My group. And I will mention later one of the areas that I've been working on it's also dissemination I do have data up to date upwards right now I'm working on the 3rd one. So I'm joined you're bringing a lot of. To make it available to make it.

[00:02:11]
Packet easier to actually use it. All my research is on this website. All the information about who is in my group who is not in my group. This is kind of a laundry list of data. That I've been working on is just to give you an idea of the of the data and are really only used in the.

[00:02:38]
Health care delivery help policy so the main the main data set I've worked with. A very large numbers of medical claims data but I really pitched in with electronic health records. A lot of publicly available data off so this is not often when we we when we when we start a project and procure delivery help policy health on these and general really we need to bring data for many many sources through its they'd as a very fragmented in health care and anything that we need even if we actually start with electronic or her records those have information coded that we need to bring to make sense out of those codes so usually takes several months to a half a year or even more to actually get started on a project before we really have the data before we do any analytics before we get anything anything going so getting to that 1st getting to figure out what the right data is what the chord that we're interested how we can pull information for that cohort that takes a very long time and you can see there is a lot of data up within it is.

[00:03:57]
Some of them are useful more useful than others. And I want to bring up not sure how many of you who are here from here about the Medicaid claims data I've been working with this for 718 years I would hope more people know about it but a part of the actually not a lot of people on campus they know that we have this data opportunity so what back inches doesn't 13 we brought I was the.

[00:04:24]
Project. The we brought the Medicaid claims data those are all the many words were maybe Katie sure and patients not only Ga we have with. Us. So all the patients that ensured everyone who is insured on Medicaid we have every single claim that was submitted for a patient and when we say a claim well that minces.

[00:04:53]
Anytime you go to a doctor or the doctor can submit 101-010-2108 sow's and claims for one visit for one visit so what we have here really I would say it's a the level of. Of billions of those records and the record it's not a number as actually a medical record and information about the patient information by the provider information of other diagnosis in for much of a procedure so is a medical record so you can imagine the size of the challenges that come with that and that's a very challenging data to work with like all claims even electronic health records are challenging us well because there's a lot of information that needs to be piled in from other sources for example when you look back most of.

[00:05:44]
There are those diagnosis codes numbers not like we're we're feeding these order they will come with the specific code and we need to make sense out of those codes is not always true for to say Well I'm looking for kids with fitting these are. Order and we're going to get that information for them as there are several different combinations of codes they'll with need to pull in to identify the correct word so always that's been actually it's always as I mentioned it's a challenging 1st step and there is a lot of other information and data needs to be on the Stude the data dictionary is more than 400 pages and we're literally half week we know that in and out those all 400 pages by now in 70 years I know where things are.

[00:06:34]
So. So the challenge is with with with Tokyo data Rick words words it's very heterogeneous has a lot of information and it's hard to make sense out of a Obviously it's very large We dedicated servers we have a dedicated team to be able to pull all the information we need everything takes for ever to with more looking like for a needle needle in a haystack it takes it takes a long time imagine that if we need to pull on I'm going to give one of the again the same example the fees are made to pull data for all the patients that all the kids that had a feeding disorder severe feeding disorder from all was.

[00:07:18]
From all the. Billions of records. That takes an enormous time is literally looking for $10000.00 kids. Millions and millions and millions of tens and hundreds of millions are ship ations so is a. Very it's complicity although we may look at smaller problems but looking into that data and digging into that data actually makes it challenging and it's even more challenging the fact that those are patient identifiable files which means we have identifiers of those patients and that means we have we have several levels of save carts of of for patient privacy and so we access the data in a very concise.

[00:08:01]
Reigned system so that makes it even more challenging because we cannot build there we were bossed databases around that often when you're outside in a world you can put it in on a cloud you can work on a different environment this has to be a static this house this has a lot of what kind of tools it obvious tools and other tools we can use.

[00:08:26]
But it's available for anyone all the faculty members and students a Georgia Tech Not so if anyone is interested to work with they contact me and if faculty member then the faculty member in is to go through our b. c.m.s. approval but actually that's pretty quick and students will have to work with their advisor with a faculty member but then once the advisor and faculty member is approved then the students can work elsewhere so is accessible on campus and we have it right now with i Pad We're going to move it coda hopefully.

[00:09:06]
This is kind of another laundry list is a list of various projects I've worked on this not I have worked on many more of the things that's so later earlier things that I've been interested to see what the most recent book I wrote is the Health Care Act as this is one big area of research.

[00:09:28]
But also right now the 2 areas that I mean the focus. Is a policy related to or a hold of mental health services and particularly. The behavior of whole services to help policies around that because I'm part of the commission governance commission on reform and so I have the ear of the state representatives so I can actually really bring a lot of the a lot of the.

[00:09:57]
Decision to the table where of those people that actually really make make to make a policy in our state. And I was sure you just snapshots of various projects I've been doing. In the past and questions that I've addressed. For example were to intervene were the gaps relate to each axis and I've looked at what people areas this is the primary care but also look for other for a primer for adult primer and here and other areas and so on and so what we do we're doing here we apply a lot of analytics behind us map there is a lot of analytics a lot of data a lot of modeling then we're looking.

[00:10:47]
In France and identified what are the areas that have significant needs. For intervention. This is related to looking differences in again an axis between children on Medicaid and those that are have other types of availability. And the differences are striking and it's actually not only for Georgia through I have a large grant on this actually to expand it across the entire us to look at various interventions.

[00:11:21]
So this is this is another area of interest I've done some someone else's I was scolded return on investment analysis this is particularly important for Medicaid. And. Maybe catering burst Simmons and Medicaid policy to the Medicaid programs Willie they look. They want to return a return on investment analysis and yes we can make that case for better outcomes and better health and better quality of care but.

[00:11:56]
Really what we really moves the needle is actually a return on investment type of analysis and most of one else's are really looking very is from a health economics or various analysis economic analysis could be cost effective those cost benefit analysis but I'm looking usually cost savings what how much money actually make it will get back literally from providing a service or not providing a service so but it's more for writing a service so this is actually was analysis of return on investment for actually providing preventive dental care for children and I could show that Jack and Willie make millions of dollars back to their pocket.

[00:12:36]
That will they will provide preventive that the cure through for children young children. And I've got the return of us one analysis this as we've looked and this is a project is still kind of ongoing it's it's very challenging looking at the whole current cost. There's been a lie.

[00:12:55]
Out of conversation around what has not been done much as around health care costs for children and I started looking into it and what we see for adults for of the things are very different to the cost structure and cost expenditure looks different for children for example there's a lot of investment preventive preventive care so that's actually Primary Care to us is what we want to see this is kind of what you are seeing here is preventive care and this is the more severe conditions.

[00:13:27]
Other things I've done this is a simulation study looking to merge an acquisition of hospitals in New York City looking the implications of those. More like to we. Not to a but the still. More similar type of analysis is this is. This approach and I work to. Looking at predictions and I see you volume volume.

[00:13:56]
Looking. At this and. This project done it a long time ago but I think it's really relevant right about to wrap it up for our area so we can go telemedicine the importance of the cost effectiveness of telemedicine in many areas. And another area with health care visualization is really big and this is done by one of my Ph d. students several years ago she pull a lot of data into.

[00:14:30]
Into this. Visualization. Visual. Of the health care system in the u.s. and what he can actually use this. Map quite of what he can see what are the players in health care system in the u.s.. Plans for us to go companies hospitals physician organizations very little very little very little.

[00:15:00]
Presence for patients for government and so those sort of 3 players for 44 players. In our system. And as I mentioned earlier dissemination this is another area. Focusing on sharing the data sharing knowledge sharing maps. So this is the 1st one if you go into my website you'll see the 1st one.

[00:15:30]
This was funded by health care. And. What we have here is a lot of data we pull from many many sort from many sources on various aspects of health care social determinants and so on we are actually working on a similar visualization Mefford diabete is we have another visualization of that is there prevalence of various conditions so I constantly.

[00:15:56]
Various. Constantly add on. Because we have a lot of data to share. And just a few words on. All of the oldest things I showed you before those. This research they did to help policyholder the lever it really is. It's funded on the ground and fundamental research this is where I've started my career back in 2005 when I started here in Georgia Tech I was more of.

[00:16:32]
The flavor my research was more fundamental research than I shifted to what I'm doing right now but I have not lost track of rigor that's one of the things that is very important whether I submit a paper to a public health join or to health services research yes you know a lot of those appendix but that appendix is really based on a lot of rigorous research and the area that I've been mostly interested in is closing the gap.

[00:17:04]
In. The musician modeling in a looking. For optimization models. And. The the framework. As the idea of incorporating early on in a model in a modeling techniques and when we have the output of those models then we can actually quantify how uncertain the output is and how uncertain the decision making there is the argument if you want we may want to do 2 to have to make one decision one of the decision and that is that's important but also it's important to understand to make a decision in a certain environment we see what happened yesterday so making decisions is not certainly very much like we have an economy for now is going to make environment is actually very important because sometimes I've done that.

[00:18:01]
Paper and I've looked at various aspects relating to enterprise transformation sometimes it's environment it's so uncertain that you may don't you may not want to do make a decision at all just wait so uncertainty in this is you making is very important because another layer of of information a lot another layer of good decision making.

[00:18:27]
And this is my other recent area that I'm interested taking a lot of modeling techniques that are an optimization statistics and figure out figure out how to break those down to a proxy made them because we don't have independence is very actually hurt to break those down used to use distributed computing and this is really inspired by the fact that we're using very large models high dimensional models with large.

[00:18:53]
And when especially in a context where we want to come for us to do it in a t. when we were early rounds models are more complex competition it's a very big component so this is where we're looking at very young and prove the complications so we can actually run those models so we can really make sense out of those models so this is another area that they start picking up and and I'm not looking there are different ways to look at distributed computing looking on the model excite how to break down the models in order to be able to run those on multiple cores.

[00:19:30]
And this is my favorite book. I've read this on and on to my son and I think I realized why I like it because that's how my life goes and if you read it and I have it's like you had this piling up all those streaks falls apart at one point but the very optimistically picks up a little over so this is kind of my life has been.

[00:19:55]
Picking up the beginning up and coming back and doing more. And we're very optimistic that things will take a different turn and this is my contact information. Yes. Yes. So because the project I'm working on a very complex I have a large team and really my challenge is that I'm not doing research anymore I can do research but I'm not I feel like that I then read as many papers I don't do the implementation I always love to do that so I would like to that's the challenge that if I become a manager and I'm not I'm not sure how good of a good a manager out does my main job in the getting funding and I just can't keep this this rolling keep this rolling does my challenge and I think that's kind of what you all face to.

[00:21:00]
I can do the research. That's not a problem I can talk to the decision makers I love going to hearings and being connected. And quite well connected with decision makers representatives in the state I've been doing very good on that but as just managing all that that is my challenge.

[00:21:27]
Right. Now we don't is this Those are claims data we only know what tests they have to for example if they had a lab we know from the lab but we don't know the outcome of that lab which was not clinic on information. You could pull that off but it takes it's always hard because you have to get the information about a specific condition of course and then the more rigid is that associated with that so yeah you can get that information of course but.

[00:22:03]
Yes it is available as I said you can actually let me know I can add you to the list of faculty members to work with that data set yes. Absolutely I'm always fit you know I'm Bob hoping more and more faculty members actually will be working with that dataset now probably.

[00:22:33]
Yes. Yeah yeah we number of yes because we have an i.d. has the same id and we've done a lot of studies were reluctant for example kids with asthma will want to see how their treatment the are under the care were looked over time because you want to see what happens when you have dental preventive care here what happens down the road right downstream.

[00:22:55]
You can follow across everything else in the for children is easier because children are. There more they're there to roll money and Medicaid is more stable for a dollar as they are in and out so the so I kind of for this kind of an else have probably focused on like that research anyways but yeah.

[00:23:15]
Yeah you lose that patient at 18 Exactly yes. Yeah well not 65 they can pick up because Medicaid is not Medicare they made 22 they made before you may see them back in a system. Now I don't know whether they would have the same right that's something I have not investigated have not looked so once there when the specific.

[00:23:43]
Problem they may have the same I don't know about that but then they become adults and it's a different. Care system they have different kind of services they have it's it's a very very different Medicaid for adults yes. We've tried to do that with. As you know we figure out how to pair of a good deal of my kids we have a good system for that we can do it we can do it we cannot do it for all the kids and all the mums good for good through good amount of c.s. we were able to pair them.

[00:24:25]
Yeah. Yes. Because of the mothers when they give birth they are on Medicaid many of them have Medicaid programs so you can see there they are very short by Medicaid that's. You know yes. I don't even know I I know what you said correctly so it's it's such a complex system even with with all the the tools and expertise I have around me they don't need her is an issue because there are a lot of things and I worked was practitioners and of course I don't understand it's a complex I I cannot blame them blame them if they don't I can because it's very very complex then and I've been hoping that all the students I'm training will go out that will do the right thing you will do they will need whoever would work with a God this level will need a lot of training really a lot of training and really working directly for many years and on gone through various projects to get their head around so it's not an easy task.

[00:26:09]
Anything else. Have another talk. So I get Jeff down here.