IE 11 Not Supported

For optimal browsing, we recommend Chrome, Firefox or Safari browsers.

Data Analytics Helps Indiana Change its Approach to Infant Mortality

According to a new report, the risk condition most predictive of adverse birth outcomes was the number of prenatal visits -- nearly 65 percent of infant deaths were to mothers with 10 or fewer visits.

Since the spring of 2014, Indiana has been using data analytics to reduce the state's historically dismal infant mortality rate -- and the results of those efforts have been announced in a new report. The primary discovery? There is a subgroup of mothers whose infants face a steep risk for dying and birth-related issues. 

Indeed, the state's youngest mothers on Medicaid, who are not getting the recommended number of prenatal visits, make up 1.6 percent of all births -- but they account for nearly 50 percent of all infant deaths, according to the report. 

The state commissioned Indianapolis-based KSM Consulting to help it unify and crunch data from across state executive branch agencies to address the infant mortality issue. The project is Indiana's first effort using its new data analytics infrastructure, but the state will soon use it to tackle recidivism and child abuse prevention as well. 

These focused data analytics projects are a part of a larger effort propelled by Gov. Mike Pence beginning early last year called the Management and Performance Hub, which provides centralized data sharing, correlation and analysis for the state; this is a collaborative effort among the governor's office, the Office of Management and Budget and the Indiana Office of Technology (IOT).

Pence and Indiana's State Department of Health identified the state's consistently high infant mortality rate -- which in 2011 was the sixth highest rate in the country -- as a good issue for technology to help solve. 

"One of the ways to do [that] was to look at data analytics and look at what the real problems are and not just what folks thought," said Indiana CIO Paul Baltzell. 

Although the research literature provides an understanding of general risks for infant mortality, the state's goal was to get a focused picture of how these and other variables affect Indiana's population in order to identify high-risk subpopulations, whom the State Department of Health can target with interventions, according to Cris Johnston, practice director for organizational improvement at KSM Consulting.

Whittling Down the Data

Having first whittled information from mothers in the state and some 50 data sets, data scientists examined 17 integrated data sets from five agencies and four public sources. The data sets included countless information related to the health of infants and mothers, including financial, criminal history, health-care and maltreatment information and were combined with medical factors and demographic data. The results were mapped out at the city, county and ZIP code levels. 

In doing this work, Baltzell said staff from the state and KSM Consulting worked together as one project team, which used the real-time SAP analytics tool HANA to analyze the datasets. KSM was also hired to work with the IOT to prepare for the project, as well as to build algorithms and analyze and make use of the results. 

Data analytics was used to examine all information related to infant mortality in the state, and also to infants with adverse health outcomes highly correlated with mortality. A bonus of this widened scope -- which includes preterm infants and those with low birth weight -- is that preventing these conditions leads not only to better outcomes, but also to health-care savings, said Johnston.

"It's not only the infant deaths that we want to attack," he said, "but there's also a cost factor in treating the babies that were born and live that still have these health problems."

Data Challenges and Successes

To gain insight into the risk profiles of certain subpopulations, the team used its analysis to develop a dynamic risk assessment tool -- a tool that gives information about the risk of having a low birth-weight infant based on user inputs like ZIP Code and number of prenatal visits. "We really took that funnel and starting looking at the high-risk populations," Johnston said. 

In taking on a project of this size and scope, the IOT encountered its share challenges, Baltzell explained, including ensuring the data was acquired in a legal way and creating protocols for sharing it, coordinating the incoming data around production schedules, and establishing connectors to bring the data in. 

A bonus of doing this work was that the state addressed data quality, protocol and organizational issues for the data sets and applied an even stricter security standard to the resulting analytics environment, he said. 

Also, all this data digging laid the foundation for the next analytics projects, since there will be less staging work. This also puts the project's estimated $1 million price tag in perspective since it includes the set-up of a core infrastructure that will serve multiple use cases, Baltzell explained. 

In addition to cleansing and zeroing in on data sets, Baltzell noted that early digging led to the discovery of additional information that proved especially useful for Indiana's findings.  

"Pretty early on when we started the data science, they recognized Medicaid records were a wealth of data to help solve the problem," he said. "So we were very aggressive in going and adding those."

According to the state's findings, mothers 15 to 20 years of age and those enrolled in Medicaid have a higher risk of poor outcomes for their infants. Although young mothers on the whole typically have better outcomes, they are also 50 percent likelier to receive Medicaid, with which they are not getting the appropriate prenatal care, according to the report.  

But the risk condition most predictive of adverse birth outcomes was the number of prenatal visits. According to the report, nearly 65 percent of infant deaths were to mothers with 10 or fewer prenatal visits. 

Infants born to the highest risk, 1.6 percent of mothers with these combined factors tended to face many health-related challenges. Although these infants made up 5 percent of all Medicaid births, they accounted for 35 percent of its birth-related expenses, according to the report.       

Classically related factors affecting infants -- smoking, maltreatment and risks present during infant sleep -- appeared in the analytics data, but were not correlated with any particular subgroup, and thus are not good candidates for targeted interventions, according to Johnston. 

Making Good Use of Detailed Data

Although the initial analytics work is over, the wider effort to address the state's infant mortality rate has just begun; the State Department of Health is using the new information to find ways to address health-care enrollment and to encourage women to attend the full complement of prenatal visits, Johnston said.  

One thing Indiana has found after identifying the significance of prenatal visits, Baltzell said, is that there are some easily remedied reasons women don't go, such as not having a ride. In these cases, he said the state can find ways to help connect expecting mothers with public transportation options. 

"That's one of the simple things we were like, 'Wow,'" Baltzell said. "In some cases, it's just a matter of getting the mother to the appointment."

The new information is driving funding and programs for infant mortality in Indiana's next biennial budget, which will be haggled over in April. Pence is recommending $2.5 million be used for the creation an application for the public, which can assess whether a mother is at risk and, if so, connect her with appropriate resources. 

The proposed budget also includes $5 million for infant mortality spending for programs created from the new analytics information.

"That's an exciting thing to have data to really make those decisions with," Baltzell said.

Another exciting part of the project, Johnston added, will be the ease of evaluating these new intervention efforts going forward, as researchers can use the dynamic risk assessment tool to understand the affects of intervention and trends, identify new high-risk populations, and suggest policies.     

"Once you do the hard work of getting everything set up, then you have the ability to make different decisions and do different things," he said.

Through the process, both Baltzell and Johnston learned the importance of incorporating the state's business units in building the core analytics infrastructure and also regarding results from the research.    

Next, the state will apply these same analytics techniques to recidivism, and then to child fatality prevention. The goal of child fatality prevention, Baltzell said, will be to empower case workers with information in the field to be successful.

And moving forward, Johnston added, "We believe the success in doing these initial use cases will lead to more collaborative efforts between state agencies, OMB, IOT and the governor’s office to solve complex problems facing Indiana citizens."