How one state’s data office improves government services

Josh Martin, the chief data officer for the state of Indiana, sees state government from a perspective that very few other people do. With access to everything from hunting licenses to prisoner rolls, Martin can provide answers to state officials that would be nearly impossible to glean for other experts.

“Indiana is really unique in that we have an agency that’s focused on leveraging data as a strategic asset,” he said.

During the early days of the pandemic, Martin told Route Fifty at a Data and Analytics Summit last week, the governor’s office tasked his team with helping a public health school determine how prevalent COVID-19 was in the population. It was before tests were widely available, so the biostatisticians would have to use stratified random sampling to send postcards to residents to get tested. But for that probability-based approach to work, they needed a database of all the residents of Indiana.

“I don’t have a master citizen index,” Martin recalled thinking. “I don’t know where everybody lives right now. We can’t do this.” But the governor’s office insisted that he figure out a way. So, in four days, Martin and his team pulled data from the state revenue department and bureau of motor vehicles, along with records from other agencies, to create the list.

That list proved useful as vaccines were rolled out. Indiana used an age-based approach to distribute the first rounds of vaccines. Martin’s team was able to notify residents when shots were available, starting with postcards for people over 80 and eventually sending text messages to younger residents when they were eligible.

In one instance, the public health agency reached out asking for help in getting people to a Saturday vaccine clinic in northeast Indiana. The Wednesday before it was scheduled, only 10% of the slots were taken, Martin said. The team found people in neighboring ZIP codes who were eligible for the vaccines, and sent notices to those patients via email and texts. “Within four hours,” Martin recalled, “90% of the spots were filled.”

One of the reasons Martin and his team were able to perform those tasks is that the Indiana Management Performance Hub uses probabilistic record linkage to help connect data from the state’s many agencies. “We run well over a billion records from 71 different source systems on a weekly basis to bring that stuff together so we can solve problems,” he said.

Martin’s team realized that linking different datasets was one of the most common tasks that agencies asked for, and the state was paying contractors to do it “over and over again.” So the tool for connecting different databases was one of the first things the data office built out, he said.

Martin cautioned, though, that the connections aren’t reliable enough to determine whether someone is eligible for certain benefits, because they will generate false positives and negatives. While the percentage of errors might be small, it can affect a lot of people when multiplied over millions of records, Martin explained.

Even with the work the data office has been able to do, Martin said there are still many people in state government that have never heard of the hub and what it does. So Martin and his colleagues have had to pitch their services to other agencies, often showing how they can help an agency solve a business problem or just eliminate busy work, so its staff can focus on more meaningful tasks.

It’s an easier sell than when he approached departments in his former role as a performance auditor. Then, agencies wanted to get rid of him. “Now, I come in and say, ‘Hey, I’m bringing tools to bear, I don’t want any money and I have free consulting services … to help you figure out these problems.”

The data office is also helping build data literacy among front-line workers. The pandemic taught Martin that data literacy among the public was low, so he wanted to make sure public-sector employees understood how to use it. Now his office helps ensure that government workers know how to navigate data dashboards and how to determine what measurements matter the most to the work they do.

Many of the commercially available data training programs focus on business uses for data tools, which help people analyze sales and losses. So the hub created its own program with more relevant lessons for state employees. The lessons are short and are delivered by email. More than 1,800 state employees have earned a “badge” for completing the first level of the program, and 700 have gone on to the second level.

Of course, the state also has to hire more data scientists to incorporate data analysis into its everyday missions.

That can be trickier now than it used to be, Martin said. Government agencies used to be able to tell applicants that, even though the pay wasn’t great compared to the private sector, the benefits were often far better. But now tech companies offer unlimited vacation and other benefits that governments don’t.

So Martin reminds candidates of the type of work they can do with the state.

“[Data scientists] want to do things that are meaningful and impactful. They want to do things that bring value to their life, and I am loaded with that stuff,” he said. “I’ve got the most interesting data you’re going to find. It’s going to impact yourself; it’s going to impact your neighbors. You can make the state a much better place.”

“Do you want to go and work for Google or Facebook and do profits and losses, marketing analyses and those types of things?” he asked. “Or do you want to come and figure out how we reduce infant mortality, how we better rehabilitate people post incarceration, how we build a better educational workforce pipeline? … Let’s do the more exciting things, the hands-on things that can bring a lot of value to you.”