Getting the Right Data to the Right People at the Right Time

Ryan Ko:

Now, I’d like to quickly introduce our moderator and hand it over to Lynn Overmann. Lynn currently is the Senior Data Strategist at Opportunity Insights, a research and policy institute based at Harvard University that focuses on improving economic opportunity. Prior to Opportunity Insights, Lynn served as a Vice President of Criminal Justice at Arnold Ventures, where she developed and executed a multimillion-dollar grant portfolio focused on leveraging data and technology to improve policing in the United States, with a specific focus on reducing low-level arrests. Prior to that, Lynn has also held key positions as a Senior Policy Advisor to the US Chief Technology Officer in the White House OSTP during the Obama administration, where she created the Data-Driven Justice Initiative, a bipartisan coalition of more than 148 local and state governments publicly committed to combining data across health, social services, and criminal justice systems. Lynn has also had several senior policy positions at the US DOJ and the Department of Commerce as well.

Ryan Ko:

Prior to the Obama administration, Lynn was a civil rights and criminal defense attorney in Miami, Florida, starting her career by serving five years as a public defender, litigating dozens of cases to jury trial, and successfully challenging unconstitutional practices in police departments and jails. Lynn graduated from the NYU School of Law and received her bachelor of arts from Bryn Mawr College. Just on a personal note, Lynn and I have been Twitter buddies for quite some time now, and this is my first time actually in a live event with her, so, personally, I am very excited to hand it over to Lynn. So, without further ado, all yours.

Lynn Overmann:

Thank you so much, Ryan, and I hope you put up with … I have kind of weird Twitter content that goes public service, hockey, and rescue dogs, so if any of that is of interesting to you, please feel free to follow me. Thank you so much to Code for America for hosting us. Like all of us, I really miss the Summit because it was one of the greatest opportunities every year to hang out with folks that are deeply geeky like we are. But it’s really wonderful, I think there’s some really great opportunities with these virtual sessions to have more folks participate than would otherwise be able to do so. Before we dive into the panel, I just wanted to provide a quick overview substantively of where we’re going to be focusing and then give you a little bit of a run of show.

Lynn Overmann:

As Ryan mentioned, the focus for the panel was getting the right data to the right people at the right time, and as we all know, that actually can have multiple different use cases. So we talked through the different use cases that we wanted to tackle. One of those was getting the right data to frontline responders so that they can help determine how to connect folks to services when they’re in crisis. It can include getting the right data to service providers to help them get a more comprehensive view of an individual’s needs, both reducing burdens on those individuals to repeatedly share the same information over and over again as well as protecting privacy. It can also include getting the right data to researchers to ask critical questions about the impact and effectiveness of programs or policies.

Lynn Overmann:

While our panelists can all answer questions related to any of those use cases, given the budget challenges that we know are coming down the pike or are already hitting for many state and local governments as a result of the economic and health impacts of COVID-19, we really wanted to focus today’s panel on how our panelists are able to use data to better targeted resources to the people who most need them. I think for all of us who’ve worked in public service the reality, unfortunately, most of the time is that some of the most critical services are often resource-constrained, so while we are certainly focused on limited budgets in the near-term, I think these are lessons that are going to be valuable moving forward in these social service spaces for some time to come. So, in terms of the structure, I just want to give a quick overview of our panelists beyond what you may have been able to read in their bios because they each bring different perspectives and experiences to the table as well as being at different stages of being able to leverage data and how long they’ve been able to do that.

Lynn Overmann:

So Erin Dalton is the Deputy Director of Health and Human Services for Allegheny County, Pennsylvania, which includes Pittsburgh. I think I saw somebody sign in and say hello from Pittsburgh, so you’re neighbors. Allegheny County is actually a national model for cross-system data, having built a data warehouse many moons ago that is capable of linking social services data with a range of other data sources on housing, criminal justice, child welfare, healthcare, and more, and that long-term access to integrated data has really helped influence and drive many of their policies and programs towards improvement.

Lynn Overmann:

David Schwindt ran the Data-Driven Justice Project for Johnson County, Iowa, which includes Iowa City and is also an Iowa City police officer. Johnson County, Iowa is newer to cross-system data sharing, and Dave can certainly share how they were able to bring a number of county and city agencies on board to share data across criminal justice, health, homelessness, and emergency services for the first time ever. Dave also has the unique distinction of standing up the first-ever Code for America Brigade in Iowa, which I think we actually bullied him into doing at the last Code for America Summit.

Lynn Overmann:

Then, just to show how much we love all Johnson Counties in the country, we have Chris Schneweis, who also is from Johnson County, Kansas, who oversees their My Resource Connection program, which leverages Johnson County’s data connections between criminal justice, health, and behavioral health systems to allow service providers to receive alerts or information when their clients interact with other key systems and then safely communicate with other service providers to ensure that the clients are receiving the services that they need in the moment. Johnson County, for those of you who don’t know, is the most populous county in Kansas, and its largest city is Olathe, and, Chris, correct me if I pronounce that incorrectly.

Lynn Overmann:

I also just want to give a quick overview of how we’re planning to run the panel. Our goal is to make this as conversational as possible and to maximize the amount of time for questions. So, to encourage that, each of our panelists are going to speak once. I’m going to ask them each one set of questions to share more about where they are in achieving cross-system data sharing and specific examples of how they’ve used data to better target resources. After that initial presentation, we will jump into questions and answers. As Ryan mentioned, Matthew Tamayo-Rios, who’s the founder and CEO of OpenLattice, will be moderating both the chat and Q&A function throughout the panel to surface questions for the panelists. So if there’s something that comes up that feels particularly relevant, we’ve asked Matthew to go ahead and interrupt and interject with a key question. But, otherwise, we’ll tackle those questions after we’ve gotten through these initial presentations.

Lynn Overmann:

And then, finally, if we were in-person at the Code for America Summit, for any of you who have been to any workshops that I’ve ever participated in, we almost always do some type of completely ridiculous icebreaker at the beginning of a workshop to get to know each other, which is obviously not feasible given the volume of folks in the medium. So, in lieu of that, we’re going to do two things. First, Matthew’s going to share a brief three-question survey with all of you so that we can a better sense of who has joined us today and where your interests lie. Then I’m going to share a couple of key fun facts about each of our panelists that they would probably not share themselves because some of them are a little embarrassing.

Lynn Overmann:

So, first, Chris, his last name is Schneweis, which actually means Snow White in German. Even though he looks a little different from the Disney character, he is still beloved. Erin has, I think, the best election day plan that I’ve heard to date. She and her family are going to be adopting a dog, which I can’t think of a better way to distract yourself after voting than settling a new dog into the home. Continuing with the dog theme, early in his career, Dave actually had a certified search and rescue dog named Sarah Ann, who he named after his elementary school crush. A few years later, Dave met and married his wife, who is also named Sarah Ann, so he has three Sarah Anns as the loves of his life. Matthew actually has the best how he got involved in technology story that I think I’ve heard. When he was eight years old, his parents owned a corner store, and Matthew figured out how to hack into their inventory system to place mass orders for ice cream. Not surprisingly, he got caught pretty quickly, but in the interim, he had a lot of great ice cream to eat.

Lynn Overmann:

And then, as Ryan mentioned, I actually started my career as a public defender, and one time in the middle of a closing argument, the court reporter clocked me speaking at more than 300 words per minute. That is both a fun fact and a warning that when I get excited, I talk fast, so if any of you are having any problems hearing me or if I’m going too quickly, please just tell me to slow down, and I promise I will do my best. So I would love to just go ahead and dive right in with our panelists, and starting with Erin. Erin, Allegheny County is known for having access to robust combined data from many years. How are you able to leverage that data to test new ideas or policies when innovations arise, and how have you leveraged data to more effectively target resources?

Erin Dalton:

Thank you, Ann. I’m going to go ahead and share my screen, and hopefully I’m not as good at Zoom as I am at Teams. Can you guys see my screen or no? I think that’s a no.

Ryan Ko:

Not yet, no.

Erin Dalton:

All right, sorry. I think I did that wrong. Come on. I can do it. Just did it in the practice, didn’t I?

Lynn Overmann:

You did. It was seamless in the practice.

Erin Dalton:

Of course it was. Why?

Lynn Overmann:

And they were very pretty slides.

Erin Dalton:

All right. So this worked right? You can see my screen now?

Lynn Overmann:

Yes.

Ryan Ko:

Yep.

Erin Dalton:

Okay. And you can still hopefully see Matthew’s survey, right? So people can still participate in that. I don’t think I’m ruining that. All right, I’ll pretend I’m not. I might be able to beat Lynn on the speaking fast. I do have a ton to cover because I wanted to share so much, so I am going to go super warp speed. But feel free to ask questions at the end. So thanks for having me, super excited, wish we were in-person. Now, it won’t click through. Sorry. Oh my god. Now, I have no access to my computer, which is good. Hmm.

Lynn Overmann:

Erin, would it help if we jumped to Dave and then came back to you?

Erin Dalton:

Sure, go ahead.

Lynn Overmann:

All right. So, Dave, aside from hoping that somewhere out there, there’s a picture of your former Sarah Ann search and rescue dog, Dave, your jurisdiction is earlier on in the process of sharing data across systems, but you’ve been able to make good progress in a relatively short period of time. What’s your advice for jurisdictions who are looking to start or expand cross-system data sharing on how to overcome reluctance or risk aversion, and how did you use data in Johnson County, Iowa to help identify the resources you need as you all were standing up new services and facilities to address your homeless and populations with mental illness and substance use issues?

David Schwindt:

Quite the couple of questions, isn’t it?

Lynn Overmann:

All the easy stuff.

David Schwindt:

I guess lessons I learned and we learned the hard way when it comes to the data sharing, first, think operational, not just conceptual. By that, I mean if you want to reach out to local government, local service providers, to ask them about participating in data sharing, there’s a couple of complexities there. One is if it’s not a specific project that’s relevant to their organization, it’s really difficult to get them to dedicate the time and resources to investigate that and how they can do that. So try and find something that they can operationalize. Most every one of those organizations, at least locally here, is going to have some type of legal and technology costs if they’re going to participate in data sharing for setting it up and making sure they’re meeting whatever compliance and security they need. So they really need to have something they can grab ahold of to see the real value to their clients as soon as possible.

David Schwindt:

When I first started, I reached out to the head of organizations, because in our area, I knew and had worked with a lot of them and learned pretty quickly that was a problem. So I found I had a lot better effort if I started with the good relationships I had lower in an organization to find out what are the current pain points that organization’s trying to deal with today and then work with them to understand how could data sharing help their organization solve that problem and what other organizations could they possibly share data with to help solve that problem. From there, just tailor a pitch to the higher up in that organization that’s relevant to a pain point right now. So a little bit of homework in the front end can really go a long way down the road.

David Schwindt:

I also learned the term data sharing is really a double-edged sword because a lot of people who don’t have any familiarity with data sharing immediately think of it the same as open data, and they think, “Well, I can’t just give my data out to anybody to do whatever they want with it.” Since data sharing isn’t that, I actually stopped using data sharing when I talked about it and tried to use an analogy of a data bank. People are pretty familiar with banks and depositing money and having it be secure, and so I talked to them about that and that they could then share their data, any part of it, with them. It’s not a black or white, all data or no data, everybody or nobody. They can actually tailor that to the needs of their organization or whatever project they’re working on.

David Schwindt:

Then the way we worked on it here is, with the help of OpenLattice, we started with just what qualified as public records dispatch data, and we got that exported out of the proprietary system we use and into OpenLattice’s platform, and that opened up a whole new world for us. One, we weren’t stuck with just the canned reports that so many places have, and that’s really a pain point, especially if you’re looking to work with law enforcement in your local area. The records management systems and computer-aided dispatch systems generally have canned reports that don’t meet the needs of the agency, especially if they have something come up and they want a dedicated report. So by getting the data out into OpenLattice’s platform, we could then use third-party tools, and I happen to use Tableau because it’s what I had access to. That just opened up a whole new world for us, both in terms of being able to visualize our own data but also being able to do it almost immediately whenever we needed it.

David Schwindt:

So we did it for trying to identify high utilizers and understanding the majority of the time for our chronically homeless high utilizers, they were being arrested for public intoxication. Another group of individuals that are frequently arrested for public intoxication in our area are college students from the University of Iowa. With that information, it can help the planning process for a non-criminal sobering unit that we’re looking to open here in January or February of 2021, so that if somebody is just in a position where they’ve had too much to drink but they’re not a real public safety problem, we have someplace other than a jail that we can take them where they don’t end up in the criminal justice system.

David Schwindt:

Then, more recently, we’re looking at, the same as everyone else, re-imagining our police department. How can we change the dispatch protocol to get more appropriate service providers responding to calls for service instead of just armed police officers? By having that data out in OpenLattice, we can do as complex of queries as we want without really just bogging down our production system that’s being used live for dispatching officers, fire trucks, and ambulances out to calls. We can hit that data as hard as we want. OpenLattice handles all of the storage and computations of those queries, and we can just try and learn as much as we can about how we can divert those resources and just inform our city council and city management about what we’ve found.

Lynn Overmann:

Great. Thanks so much, Dave. And, Erin, have you figured out your technical issues?

Erin Dalton:

I think I’m back. Can you guys see my screen and hear me? No? You guys see my screen?

Ryan Ko:

Yeah, we just saw you switch it from slide one to two.

Erin Dalton:

Okay. And you can hear me? Yes. Okay.

Ryan Ko:

Yep.

Erin Dalton:

Perfect. All right, thank you. Sorry for the technical difficulties, and good always to follow Dave and the work that they’re doing there. So we aren’t a place that lacks in integrated data. I’m not going to go through all these data systems that we’ve integrated. These are integrated at the person level. We’ve been doing this work for over 20 years in Allegheny County, and so we’ve got a big advantage there. Lots of you are thinking about, “How do I integrate all of this physical health and criminal justice and education and basic needs data,” and that’s a huge set of work, but it’s also work to then think about how do you best use that data. So we’ve been also thinking about that for a long time. I’ll show you just one slide of how we use data at the person level. I’m excited to hear that another panelist is going to talk about alerts because I think that can be super useful. We do some of that, too. The folks on this panel have worked together, too, on frequent utilizer work, so how we’re thinking about frequent utilizers and how we’re putting in place tools that use these integrated data to help make resource decisions.

Erin Dalton:

This is just the one slide I’m going to show you, and it’s clearly not a real person, but this system allows not only our workforce and our provider network but also our clients themselves to get their records back. So if we’re going to take all this data and integrate it, how do we return that, thinking of who needs data the most, the people themselves to manage their own care. We want to get that back to them. We want them to be able identify their caseworkers, update their information, and look at all their records. So I wanted to just mention that as something we’re doing in Allegheny County. I also just want to point you to our website, Allegheny County Analytics. There are a number of data tools. A lot of them, like Dave, are in Tableau, but some of these, they’re tools that allow you to query and get cross-tabs of data as well, so you can show your folks in your jurisdiction that other people are doing this and making it available publicly.

Erin Dalton:

So just jumping into the frequent utilizer analysis, a lot of folks want to integrate data to identify those people who maybe they should most be serving. I think it’s really important to be thinking about why you want to analyze this. There’s non-trivial questions about how you’ll analyze it, and then, really importantly, what will you do when you’ve identified the results? So just quickly, we’ve looked at all of our crisis systems or at least most of our crisis systems, arrests and jail bookings, emergency room, at least Medicaid-paid emergency room stays, emergency shelter, homeless shelter, and mental health crisis.

Erin Dalton:

There’s lots of different ways you might want to look at this and think about the cut-offs, right? If you choose, a lot of people in criminal justice, I hear them choosing cut-offs of three. Why would you do that? Why would you focus on this side of the equation? You could be thinking about, “I want to be focusing on the top five percent,” sort of the way we have around emergency rooms or mental health crisis. You might be thinking about, “Well, we want to have the most bang for our buck, so if we serve these 12%, how many jail bookings or arrests might we prevent if we serve those folks well?” I just want to point out, there’s a lot of decisions to make about that. And then, third, looking at the total. If you’re going to be thinking about individual intervention, if you’re saying, “We really just want to help these people who have the most services,” some of these cut-offs, depending on the decisions you make, present you with a ton of people to serve, right? And so how could you serve 5000 people or something with your intervention? Maybe you want to look at different kinds of cut-offs.

Erin Dalton:

Just thinking about that one step further, I chose the mental health crisis group. So, here, we’ve got what I think is maybe the best way to be thinking about it for our purposes, the top 5% of the cohort. They have, if you will, a big bang for their buck. They use 37% of the mental health crisis services. Then you have to think about who are they, right? I’m just showing you a couple things here, age, but we also-

Matthew Tamayo-Rios:

Hey, Erin.

Erin Dalton:

Yeah?

Matthew Tamayo-Rios:

Sorry to interrupt real quick, but we got two good, quick questions, well, one of them may not be so quick, about the analysis. The short one is can you explain what you mean by cut-off?

Erin Dalton:

Yeah, sure. Sorry. We use the words cut-offs or thresholds. When people are thinking about frequent utilizers, how would you define frequent utilizer? I think a lot of times I hear in discussions that people just pick a number and say, “Anyone with three or more arrests or three or more bookings or 10 or more emergency room visits in a year or in two years is our definition of frequent utilizer.” So that’s what I meant by cut-off there.

Matthew Tamayo-Rios:

Great. Then the other question was around data standardization. One of the folks in the audience is looking at a human service data specification for standardizing referral data fields. What do you use to standardize this data before analysis?

Erin Dalton:

Mm-hmm. Yeah, I’m not sure it’s a quick question. I think we’ve essentially created a groomed layer for each of these services, and then we have the other data that we bring in depending on the analysis. So I liken it to services rendered records in healthcare or billing records, like, “This person was arrested on this day for this thing,” and there’s a start and end date associated with that, or, “This person had a mental health crisis service this date and time, this diagnosis, and this was the name of that service.” So I don’t know if that helps. There’s probably a lot of discussion to have about that, Matthew.

Matthew Tamayo-Rios:

Yep. And I will add one quick thing. There’s also a standard that’s being developed, I think, called Open Referral that has a bunch of the fields, so they’re attempting to build out and standardize this, and so that might also be a good option.

Erin Dalton:

Yeah, for sure. We’re absolutely looking at that and a number of other things around our referral data, which is particularly challenging. I don’t want to ruin all of the time. I don’t want to take the whole panel. But the point of looking at who these folks are, so people tend to think, “Okay, I’ve identified my frequent utilizers of mental health crisis services in this case. They’re all the same. These are one group of people. We can serve them perhaps with the same type of intervention.” And just to point out, on this chart, you can see the age groupings are really bimodal here. The blue line is the frequent utilizers. So you’ve got pretty young folks, and then you’ve got a persistent group of older folks. I doubt the same intervention’s really going to work for both groups.

Erin Dalton:

So I think you really have to look at who these folks are and then look at the opportunity, right? In the year after these frequent utilizers of mental health crisis services had frequent use, they had a ton of other service connections. These are places you might find these folks in order to do something different, offer better services to them, but it’s also, perhaps to Dave’s point, why would the emergency room folks care about this? Well, if we can do a better job serving them, 75% of them have had an emergency room visit in the year after frequent mental health crisis services. So that’s just some thoughts on that.

Erin Dalton:

So where are we putting some tools? One of them, we’re putting in the homeless system. If we’re already making decisions about who gets, for example, permanent supportive housing or any kind of supportive housing beds, let’s make sure we’re using the best data we can to make those decisions so that we are serving high-risk, high-need folks in those services. One of the use cases here is previously, prior to implementing some of these data-driven tools, about 40% of our permanent supportive housing beds were going to folks who were relatively low risk of harm if left unhoused. That’s going to be true in all of your communities probably, unless you’re using significant data-driven approaches.

Erin Dalton:

We did essentially the same thing that I just showed you with the frequent utilizer analysis, slightly different outcomes. But we predicted, for all people who come into homeless housing, which is about 30,000 calls per year, not people but calls, and we looked at, if left unhoused, what is their likelihood of a mental health inpatient service, a jail booking, or four or more emergency room visits in the 12 months after, and we validated on mortality. We found that using a previous tool, which is an actuarial tool that a lot of folks are using, it wasn’t doing a very good job of predicting those kinds of outcomes, and if we implemented a different kind of tool, we would serve, on this chart on the left, the nines and tens here, 64% of whom, for example, had four or more ER visits in the year after asking for homeless housing services.

Erin Dalton:

That’s super important because, as I mentioned, without a tool like that, we were giving some of those permanent supportive housing beds to relatively low-risk folks. It turns out, and not super surprising, if you give that kind of housing, like a high-end housing, to lower-risk individuals, they’re going to have less outcomes, right? These kind of housing services, not to say that they don’t work well for lower-risk folks, relatively speaking, but you will see more bang for your buck with higher-risk folks. What this chart shows is a 6.5% reduction in ER visits for high-risk folks and essentially no effect for folks that are relatively speaking low risk. So it’s dually important to provide those services.

Erin Dalton:

That’s one of the ways we’re using data to help prioritize homeless services and permanent supportive housing. We implemented this tool in September, and there’s a bunch of papers on our website about it. You can imagine, just to turn it back to the mental health crisis services, if we put a tool at mental health crisis, how might we better serve those people that end up being just super high risk of adverse outcomes, and what might we design to support them there? So thank you.

Lynn Overmann:

Awesome. Thank you so much, Erin. It’s always great to see what you all are able to do in Allegheny, and hopefully we can all get there one day, too. So, Chris, similar to Allegheny, Johnson County, Kansas has had several decades of access to combined criminal justice data, which you’ve leveraged to build broader cross-system data sharing and to build the My Resource Connection program. Can you talk a little bit about how you’ve leveraged your combined your data to surface and address differences in approaches across the towns that make up your county, which I think is a circumstance that a lot of our government partners find themselves in?

Lynn Overmann:

Then anything that you want to mention about the My Resource Connection to Erin’s point about how valuable alerts can be to flag for folks that people may be in need of services more immediately than their traditional check-ins might require. I hate to throw you under the bus this way, Chris, but for folks who are participating, Chris is also an expert in HIPAA, so he brings a really strong privacy lens to all of the work that he does, and I know he gets five million questions on how to navigate HIPAA. But if we could start with the data sharing, Chris, that would be fantastic.

Chris Schneweis:

Oh, absolutely. First off, I’d love to just send a shout-out and thank Code for America and OpenLattice for having us here today. Definitely, it’s hard to follow Dave and Erin because of the great work that both those jurisdictions do. But specifically to Johnson County, to those who aren’t familiar, Johnson County’s part of the Kansas City metro area, which is a 2.3 million population. Johnson County itself’s a little over 600,000. So with that said, that encompasses 20 different cities, 17 different municipal and county law enforcement agencies. Johnson County has two jails that encompass about 1100 beds. What we have been looking at doing is how do we take this data that we have, we’ve been working with since the ’90s with what was called our Justice Information Management System, which is now transitioning to a system called Niche that Johnson County, the county, operates on behalf of the local law enforcement. So when you’re arrested, everything you do with Johnson County or with our local cities goes through the same database. Because of that, it follows you all the way from arrest through adjudication.

Chris Schneweis:

If you end up on probation, again, same data system is used, and that allows us to look at what are people being arrested for. When you have people from local cities that are continuously being arrested, it allows us to take that data, look at it, break it down, and see, “Okay, these frequent utilizers or these people that are coming in and out of our jail quite frequently, what are they being arrested for?” That allows us then to go back to those municipalities and say, “Do you realize this individual, just one individual, keeps getting arrested for disorderly conduct?” Maybe it’s not really disorderly conduct. What are the things going on with this individual? Is it that they’re homeless and they just happen to always be loitering in the same area? We have nothing to do with them. In Johnson County, we don’t really have a homeless shelter, so they don’t really know where to take these people. I don’t know what to do with you, so I’m going to arrest you for disorderly conduct. You’re going to end up in jail.

Chris Schneweis:

One of the other things we’ve done with integrating our criminal justice as well as our behavioral health data is when you’re booked into the Johnson County Jail, everyone that gets booked in gets a brief jail mental health screening that is administered by our deputies. Every deputy is trained on how to do this. It’s a simple eight-question assessment that’s provided, and it shows if you have potential for a behavioral health issue that maybe is not diagnosed. Because another thing about Johnson County is we are a fairly affluent county, so must people do not stay in our jail. Most people are in and out of our jail within 32 hours.

Chris Schneweis:

So because of that, you may have a behavioral health issue, but you never actually meet one of our nurses and actually truly get screened before you bond back out and now you’re back out on the street. So by able to mix that data together, we can then go back to our local municipalities and say, “By the way, this person you keep arresting for disorderly conduct over and over again, they also appear to potentially have a behavioral health issue that maybe hasn’t diagnosed and they haven’t received any treatment for.” Then it gives these local municipalities a different way to look at when I go and deal with this individual, maybe we get our CIT officers involved, maybe we look at potential different options.

Chris Schneweis:

The other thing that we also do is we work with University of Chicago to actually develop a business intelligence model where we actually use machine learning to look at can we identify those individuals who have a propensity of having an adverse interaction with local law enforcement that’s going to end up with them being booked into jail. So what you’re actually talking about doing is using data to identify people before they ever have an interaction, before they ever meet with that law enforcement officer the first time, so that what we can do then, Lynn mentioned us providing notifications, because of our data systems in Johnson County, we are able to notify our Johnson County Mental Health system. We have a list of 100 people every month that this BI, this machine learning model, identifies and that we then can notify them, they can do the outreach on, and see, “Hey, we realize you’re not in services right now. How are things going? Would you be open to talking with us because we just want to follow up, make sure everything’s going all right with you and your world.”

Chris Schneweis:

What we’ve found is we have tremendous success with people saying, “You know what? No, I’m not really doing real well.” We end up getting them in. If we can’t provide the services, Johnson County Mental Health, it allows us the ability to refer this person to another service provider that maybe can meet their needs. They end up getting into services, they end up not having that adverse interaction, they end up not getting booked into our jail because as most of us on this call probably realize, those people that get booked into those jails, those frequent utilizers, that people that have potential behavioral health issues, those are the ones that are going to stay in your jail longer. They’re not going to bond out. They’re going to be the ones stuck in your jail longer, and they’re not going to get their needs met there, and eventually they are going to get out, they are going to be back out on the street, and they’re going to be in the exact same position they were.

Chris Schneweis:

So the last thing that I will just touch on before I hand it back over to you, Lynn, is one of the things we’ve also been able to do. Now, this, when you talk about providing protected health data and notifying, this is where it gets a little touchy. But one of the things we are able to do when you mix that data together is if you have people that are involved with our Johnson County Mental Health, our behavioral health provider in our community, if they have a case manager and, guess what, that client or consumer has an ambulance called tonight that takes them to a local hospital because they don’t know how else to get their needs met other than to call 911, we do because that data all resides in the same location. It comes into the same location every night.

Chris Schneweis:

We can then see that, the system identifies it, it matches up the case manager, and it fires off a notification to that case manager that, “Hey, you may not be aware of this, but your client just had an ambulance call last night. They ended up at the hospital.” Because when you talk about interventions, as we all know, interventions work much better if they’re timely, and it doesn’t do you any good to find out from your client or consumer a week later that, “Oh, last week, oh yeah, I took the ambulance and I ended up in the hospital because I was in crisis, I was off my medication,” so on and so forth. Well, now, a whole week has passed before you ever got that information.

Chris Schneweis:

To where it is allowable under HIPAA between two covered entities to be able to share that information, with our data capacity, we have built that into something that we are allowed and are able to do with our case managers here. What it’s done is it’s streamlined that human service delivery and it’s made things a lot more efficient and effective. It’s cutting down on ambulance runs, which, for us, we estimate’s about $1000 just to run that ambulance. So, again, the primary goal with this, as with everybody that’s trying to do this, it’s client needs. It’s making sure your clients are getting the services they need in the most timely manner possible. But there is a cost associated with that, and there’s cost savings when you can share that data and do things in a more efficient and effective manner.

Lynn Overmann:

Fantastic. Thank you so much to Chris and Dave and Erin for sharing the great work that you all have underway. We’re going to open it up now for questions. I think that I’ve already seen a few questions come flying across the Q&A, so, Matthew, what have you seen, and how would you like us to start prioritizing answers?

Matthew Tamayo-Rios:

Thanks, Lynn. Oh, looks like the next question is around geographic challenges. So, Chris, I think you alluded to this a little bit, about how you are representing more than one jurisdiction. So someone was interested, is how do you think about data sharing challenges when the individuals you’re interested in are crossing geographic either county or state boundaries, either college students or homeless individuals that may move around?

Chris Schneweis:

What we’ve actually been able to do is … So our My Resource Connection, which Lynn alluded to at the start of this, which actually takes our criminal justice data as well as our human service data, similar to Allegheny County. What we’ve started doing is working with our local jurisdictions, so Wyandotte County, which is Kansas City, Kansas, the city of Kansas City, Missouri, which is across the state line on the Missouri side, we’ve started going to them saying, “Look, we can host this for you. We will only charge you what it costs us, just full cost recovery, to get you on board in the same kind of a platform,” very similar to what OpenLattice does, and say, “Okay, if everybody can get on the same platform, then those notifications become more seamless,” because if you’re receiving mental health services on the Jackson County side but you get booked into Johnson County Jail, that booking is public record, and if it’s all compatible and it’s all mixed together in the same data system, we can make that notification possible.

Chris Schneweis:

So what I think you have to look for is I think most jurisdictions out there are open to getting on a platform and doing something similar. What you have to do is you have to find that sweet spot with elected officials or people that hold the purse strings to where it can’t be cost prohibited, so much like what you guys do with OpenLattice. If you can make it cost effective for them to do that, I think what you find is it’s very open, people are open to doing it, and then what you have to do is you have to look for the low-hanging fruit, which is how is it legally possible to share data, because a lot of things people hide behind as the reasons they can’t share data, they misunderstand what the statutes do and do not allow.

Lynn Overmann:

Yeah, to add to Chris’s response, which I think is super helpful and would love Erin and Dave to jump in if they have thoughts, too, I think a lot of the work that we targeted at the outset when we started working on DDJ and how I’ve been able to make friends with these wonderful people was really focused more at the local and county level. I think there’s a pretty significant role that states could play in the data sharing space and, to a certain extent, the feds as well. But I think there are opportunities to get data-mapping, where you say, “Who has the best data and the most comprehensive data? Where does it sit, and what coverage would that provide me?” So, for example, a lot of states run Medicaid programs. If there was a state partnership with Medicaid programs or state hospital systems, you might be able to get more comprehensive coverage to address that, people who are moving across jurisdictional lines.

Lynn Overmann:

Thinking a little bit more proactively, and I think in the next iteration, potentially in a new administration, trying to figure out what are the different levels of government, where is it most helpful to look at the folks who are moving, because if you move … We actually had a project with Middlesex, Massachusetts, which has, I think, upwards of 54 police departments in one county. We had a frequent utilizer that was showing up on one police data. He had 38 calls for service in about two weeks. He got a restraining order, and he ended up moving to another police department, and he was brand new to that police department. So that lack of data sharing can really decrease visibility, and it can make it very difficult to do what Erin was talking about, which is target folks who most need services if you don’t have that cross-system sharing at different levels of government.

Erin Dalton:

Yeah, just to add there, yeah, we have crazy fractured government in Allegheny County, too. We have 43 school districts, 110 police departments, and so on. But I think … Not but. And I think if we’re going to do that, and I think it’s a good idea, we just have to make sure that then states do the right thing and share it with the people who need it. So, yeah, states hold a lot of these data. I’d love for them to integrate themselves, but then, just like we have at these county and city levels, share them with the people who need them most. If the counties are the ones doing the actual operations, they need to share that back with the counties, the counties need to share with our providers, we need to share with the people we’re serving, and so on. I just feel like the higher up in government you get, the more reluctant they are to do anything outside of their boundaries. So if we can get them to integrate data, that’s amazing, and then they do need to continue to share it with who’s using it.

Lynn Overmann:

Great. Matthew, are there other questions? And then I think maybe while we’re answering more targeted questions, if folks are up for dropping into the chat what was the thing that most motivated them in participating in this, just so we have a general sense beyond the survey, or questions that have arisen for themselves. Obviously, we’re not going to be able to tackle everything, but to the extent that we’re seeing common themes that are motivating folks, that would be helpful to hear.

Matthew Tamayo-Rios:

Yep. So I think a next question, we got a good one about compliance and concerns with shared data. I’m going to combine two of them. So one is what forms do you use, and then the other one is what type of activities are stored in the audit log and how frequently those are reviewed. What are the rules around or what are the processes or best practices around clients opting to release their shared data? Is it as simple as them opting into a release? Some people say that, “Hey, HIPAA actually doesn’t let them do that.” What’s the actual story on releasing client data, and can you provide any examples? I know Chris is jumping at the bit for that one, but I’ll open up the floor for everybody.

Chris Schneweis:

I mean, one of the things I would encourage, Matthew, with any jurisdiction that’s listening here today is that one of the best things you can do is get a third-party HIPAA or healthcare compliance law firm on board, which I actually contract with within the county, because if you’re relying on … And this is not a knock on anyone’s personal city or county, legal department, but for the most part, those legal departments, their responsibility is to cover the jurisdiction, so it’s easier to say no in almost every circumstance than it ever is to investigate how is it possible, and that’s understandable. It’s kind of like risk management. When you get a third-party jurisdiction, so in our case, we utilize a law firm named Spencer Fane, they are involved in every data sharing agreement that we do, they’re involved in every involvement that we have related to how we share and communicate about data.

Chris Schneweis:

But one of the key things is, is under HIPAA, if you have two covered entities serving the same individual and it falls under one of the three approvable reasons that you can share data, there is no need for a release. It’s just not needed. If you’re providing behavioral health services to this individual and an ambulance call is run on that individual, they’re both covered entities. They both bill electronically to Medicare and Medicaid for the services they render. That data can flow seamlessly. Now, if you ask five different compliance officers or five different attorneys, you’re probably still going to get five different responses to that. But what we have been told is there’s no need for that. There’s no need to go to the client and say, “Can we release your information?” The statute already allows for it.

Chris Schneweis:

Now, the other thing that I’ll just throw out there is HIPAA’s also one of those laws where … Actually, it’s one of the very few laws where state law can trump federal law. So if you have a state law that actually prohibits the sharing of that information, even though HIPAA may allow it, doesn’t matter that HIPAA allows it. If your state law is more restricting, it’s going to override HIPAA, so you still have to keep that in mind. That’s something else that people have to make sure they keep in mind when they’re looking at it. It’s not always just a HIPAA issue.

Erin Dalton:

Yeah, just to add one thing, amen to all that and we’ve done the same thing, we have outside counsel, agree with all the reasons, and then I linked somewhere … I don’t know. It’s on our website under … There’s a paper on our data warehouse, and there are links in that, including policy documents we’ve written. In addition to what Chris said, we also, just like he said, we take the right to re-disclose en masse, and we’ve written up our rationale for that, so that may also be helpful. We did that just like Chris, with private counsel. That may also be useful to people as one approach to how these data are shared.

David Schwindt:

And this is right in line with what I mentioned as to one of the hurdles we saw, is that a lot of our services are provided by local nonprofits who don’t have the funding set aside or the in-house expertise to provide this type of guidance. I’ve found it really difficult for them to be able to find the money and justify spending that to, number one, find that appropriate outside counsel and then pay for that expertise, which we did have one entity do, and the attorney got back with them, just saying, “Well, yeah, you can probably do it, but it’s going to be difficult and it’s probably not worth the money you’re going to pay me.” For those nonprofits and others, that legal question, I’ve found to be a hurdle.

Lynn Overmann:

Great. Really quickly, I shared in the chat, at the end of the Obama administration, we were able to get HHS to provide a frequently asked question around you can share across criminal justice and service providers such as homeless providers. So I added the link into the chat in the event that that’s helpful to anyone. I think what we discovered, although it’s painful and I think this just reminds me how helpful it would be if there was better and more comprehensive guidance from the federal government about the different ways we could achieve the data sharing that you all are talking about … In the meanwhile, certainly, these connections across communities who have been able to do it, I know folks are always willing to help other folks who are trying to figure out how to do these things. It is imminently achievable, and in a dream state, we’re going to get to a point where it doesn’t have to be every community either figuring out who they can ask community by community. But, in the meanwhile, I would definitely encourage you to look to Matthew, Chris, Erin, and Dave as resources to help you figure out what’s already been done, so you’re not reinventing the wheel.

Matthew Tamayo-Rios:

Yep. And I’m going to try to squeeze a few more questions in here. One simple one should be there’s a question about what platforms folks are using to capture and track their data sharing [inaudible 00:49:20], privately developed, Microsoft stack, or other. I think that should be a quick one for folks to run through.

David Schwindt:

Johnson County, Iowa’s using OpenLattice.

Erin Dalton:

Matthew, you probably can answer that question about ours better than I can. I could do it, but I feel like you know a lot about this stuff, so you might characterize it better.

Matthew Tamayo-Rios:

Okay, so my understanding is Allegheny County has basically built a data warehouse that’s … I don’t know whether it’s a SQL server, but I think they’re just running it in a standard SQL database on the backend with a bunch of batch jobs that actually end up doing all the linking. So it’s basically privately developed in-house. They have a really sophisticated rules engine for actually doing the matching, so they can really fine tune how they match individuals, and then they have a notifications base and eventing system that actually pushes data from events out across boundaries. Sometimes, interfaces, it’s not as simple as it’s all one system because sometimes the courts have to … Some people pass the data directly. Some people are like, “Here’s a CSV thing you’re going to have to pick up and then process.” So we did on-sites at both Allegheny and Johnson County to get a better understanding and learn from what they’re doing. Chris, if you want, happy to take yours as well.

Chris Schneweis:

Go ahead because based on what you’ve said thus far, I know you pretty much outlined what my RC is, which is built on a SQL platform, but, yeah, take it away.

Matthew Tamayo-Rios:

It’s similar. So my RC is similarly built on the SQL platform. I think one of the cool, unique aspects of my RC is that it also has a public-facing portal that people can go to to basically look at services that are available, as well as internally, what we saw is that the behavioral health officers and other folks at the county had an internal portal that automatically filtered down the information that was relevant to them. It was sort of like a one-stop shop where folks could go in and tap into it as they needed to make more informed decisions. Again, these systems were really incredibly ahead of their time. I think my questions would be how did you guys get this built 20 years ago and get the resources. I think that might be a broader audience question as well.

Chris Schneweis:

The one thing I will say, and this is one thing I would encourage other jurisdictions hopefully to not end up in this position before they have to do this, is our justice information management system got started just because the DA and the judges wanted a platform that was more centralized to the work they did and they had the pull back in the ’90s to really make that push. My resource connection actually came about because of a consulting work that was done that showed that we were doing a poor job in our human service delivery in the county. Then, unfortunately, we had a very tragic situation where we had a behavior health case manager killed in the line of duty by one of her consumers, and when they went and actually investigated the situation, what they found was that this individual and his family had had some interaction with different departments, offices, and agencies and people had concerns and questions and things like that, none of which ever got relayed. So then that raised the bar that, you know what, we need to be doing a better job of the way we communicate and share data from just a sheer safety standpoint.

Lynn Overmann:

I’m just keeping an eye on the time, and I know I’ve already failed in my duties because I was supposed to help us hit a stop around 1:55, but, as you can tell, we all love answering all of these questions and many more. I think we could stay on forever if we were allowed to. But I just wanted to throw it back to Matthew really quickly. I think Erin and I have both shared our contact information in the chat. We certainly welcome folks reaching out. I’m happy to connect folks with anyone else on the panel who you’re interested in talking to who hasn’t been able to add their information in. Then, Matthew, do you want to just touch really quickly on the open source community that OpenLattice is putting together?

Matthew Tamayo-Rios:

Yes. So OpenLattice itself is a completely open source product, so you can actually go and find the code online and play with it directly yourself. But we’re starting to be more thoughtful about how we start building communities so that other folks, when they go build integrated data systems, they have something that they can start with, that they can actually contribute to in how we do that. So we’re going to be launching something more formally on October 28th, and if you’re interested on it, please feel free to follow up for more information and details.

Lynn Overmann:

Great. Ryan, I think we barely clocked in under 2:00. Is there anything we need to-

Ryan Ko:

This is lovely. Just wow.

Lynn Overmann:

Thank you all so much-

Ryan Ko:

This was-

Lynn Overmann:

… for such a great conversation and for all your questions.

Ryan Ko:

Yeah. Thank you, Lynn, David, Erin, Chris, Matthew. Just thank you, everybody. And thank you to all of our participants who joined and OpenLattice, again, for sponsoring us. Please do continue the conversation online, on Twitter @codeforamerica, on our blog at codeforamerica.org/news. We, as always, welcome contributions at /donate, and do check out our other events as well at /events. Thank you, everybody. I think we’re only a minute over. I would hope that this is the start of a conversation. So have a good one.

David Schwindt:

Thank you.

Lynn Overmann:

Bye, everyone.

Matthew Tamayo-Rios:

Thank you.

Ryan Ko:

Thank you, everybody. Bye.