Statistics sometimes get a bad rap, as being somehow divorced from the real world of complex events and relationships. But in several notable cases, statistics helped provide a useful view of seemingly diverse and sporadic events.
Back in the 1980s for example Jack Maple, a cop in New York City's subways, got tired of responding to crimes after the fact, and decided to put together information that would predict where crimes would occur. He had no computers or fancy analytic technology -- just crayons and butcher paper -- but Maple's analysis of crime statistics superimposed on maps of the subway, revolutionized police work, and comstat, as it is now called, has enlisted the help of computerized analytical tools and has spread to police departments around the world.
The South Carolina Office of Research and Statistics (ORS) is also breaking new ground in the use of statistical data. ORS crunches the numbers to help analyze a broad spectrum of social services programs -- from health to justice, education and corrections -- to provide a sort of "information dashboard" for some 20 state agencies and private health-care providers, in order to help the state assess the effectiveness of various programs and focus social services money and attention where it will make the biggest difference in the lives of those being served.
"One of the projects we did," said Pete Bailey, health and demographics section chief, "was to look at what happened to children that aged out of the juvenile justice system, what proportion of them were incarcerated later, and so on. The Department of Juvenile Justice itself didn't have any data on adult arrests or incarceration, but we do, because we receive that from state law enforcement. So with permission, we conducted a study."
Bailey says ORS is also doing a study for the Department of Education. "Unfortunately, in South Carolina we did not always have the ability to track a kid from year to year." Bailey said the study will tie educational data to Medicaid system data for low-income children, and to the social services and juvenile justice systems. "And what that means is that ... you would be able to do analysis to see how Medicaid children are doing in school versus food stamp children, foster care, or protective services cases that weren't removed from the home. You could look at the impact of all of those. And the next step we're going to is ... to be able to look individually at each of those kids with a tracking number and -- without knowing who they are -- look at their history in terms of how did they get where they were in the educational system, in health, or with social services, or with law enforcement ... What caused their blocks and their breaks and their successes?
"That's an awesome capability. Government has a responsibility to use all the information that's sitting in every computer they can get their hands on, to better understand and evaluate why our programs work or don't work and how to come up with better outcomes. If we do that, you add to government a volume way of work per employee and who gets the best outcomes. We can use that to improve those that aren't doing so well."
Connecting human services data to elected officials responsible for funding programs is one of the possibilities said Bailey. "Tell them the problems people have in their districts ... How many people are on food stamps or in foster care? How are kids doing in school? continued Bailey. "And once you do this, when they are elected, you can evaluate every year how things have gone. It sort of feels like democracy."
So if this is such a great idea, why isn't every state doing it? David Patterson, deputy chief of health and demographics said that while South Carolina probably has the largest and most comprehensive state level warehouse in existence, other states are also looking at doing something similar. For example, they have had some contact with Arizona, Arkansas and Maryland in that regard. But as every government agency knows, sharing data, protecting privacy and knocking down stovepipes to get an enterprise view is not always easy. There are technological as well as human barriers. So how did ORS build its data-sharing system, and what advice do they have for others?
Workable Data Sharing
ORS collects data from over 20 state agencies said Patterson, "and we don't release anything without prior approval of the originating source. We apply algorithms to the data that allow us to maintain entity relationships at the person level across all these data."
"We're like a data Switzerland," he said. "We develop a lot of applications for customers, everything we do is customer-driven and requires their approval. The [memo of understanding] process resulted in some statutes to clarify and extend ORS' authority."
When it comes to sharing data, said Bailey, there are some wrong ways to go about it. "A lot of states tend to try to have one agency grab another agency's data, and people might be willing to share their data, but they're not willing to have you put your data in their computer, because we live in a world where whoever has the most data or the biggest computer is thought to be the winner ... And if you put your data in my computer, you're likely to get up the next morning and see in the paper that I've done an analysis showing you did some stupid stuff."
Instead, ORS has an agreement with data-providing agencies so that those agencies have full control of their data. "They allow us to run through the unique IDs," said Bailey, "and build a tracking number so we can link data across all of these systems without using the identifiers. Once we have the tracking number, that's what's linked to the statistic, so we can link a massive amount of data from some 27 blocks of agencies and God only knows how many programs, and [in this way] they love to share data and do research together, because that's where the great answers are. So you see, that's different from saying 'I want you to give me your data'"
Patterson said that preceding the agreements "is the issue of privacy protection, consensus on the mission, and transparency on what we do. If we weren't neutral, then none of these other things would happen."
ORS is not competitive for budget with any of these agencies, and has it has the trust of the private sector, which enables the collection of data on hospitalizations, emergency room visits, outpatient surgeries, even home health and free clinic visits, etc., and that in turn can be linked with state agency data.
So ORS crosses the boundaries between private, public, not for profit, health, social services, criminal justice and education systems -- without making anybody unhappy or upsetting the balance of power between organizations.
However, said Bailey, each state agency has a "federal godfather," and if those federal agencies don't get along well at the national level, it can make it difficult at the state level.
Connecting social services data to geographical location gives it additional import. "Panorama has helped us to build the mapping application," said Randy Rambo, IT/DBA manager of health
and demographics., "using ESRI mapping to drill down to any type of grouping that you want to have mapped -- legislative boundaries, census tracts, virtual neighborhoods, etc."
The enthusiasm for data use at this level is evident as ORS staff suggested combining data -- such as people moving in or out of a community, crime data, kids' progress in schools, emergency room injuries or violence -- in such a way as to help isolate the actual causes of community decay or other problems that may develop.
"We have massive data sitting in government computers that represent pieces of the puzzle," said Bailey. "If you put it together we can better understand our children, and our parents and us as humans, so that we could better make a substantial difference between the world we have versus the world we could have. And the world we could have is awesome."
NEW ON THE PODCAST