Before becoming Vice President of the United States, Mike Pence served as governor of the State of Indiana. He was sworn in on January 14, 2013, with several policy goals, including improving the health and well-being of Indiana citizens. During his term in office, the state faced a number of endemic problems, including a high rate of infant mortality, a growing opioid epidemic involving both prescription and illegal drugs, and an ongoing problem with criminal recidivism.
Chris Atkins, who had been Governor Pence’s policy director during his gubernatorial campaign and subsequently served as chief financial officer for the state of Indiana, explained in a February 2018 interview, “…early in his administration, Governor Pence challenged his team to come up with big, audacious ideas to implement his desired policy outcomes. As we were looking at solutions to these big problems, we realized we needed a new way to approach problem-solving – a strategy for tackling big issues like infant mortality. That strategy became data-driven government.”
The United States’ infant mortality rate has been in steady decline for decades, dropping from 9.4 deaths per 1,000 live births in 1990 to 5.6 in 2016. Indiana’s rate, however, has been persistently high, and with 7.5 deaths per 1,000 live births, the State has one of the highest rates in the country. Black children are particularly at risk with 14.4 deaths per 1,000 live births – more than double the rate for white children (6.4). Given that infant mortality is often used as an indicator of the level of health in a country, a startling statistic is that if Indiana were a country, its infant mortality rate would be higher than that of Kuwait (7.0), and for black children it would be higher than Colombia (13.6), the West Bank (14.1), and Jordan (14.2).
As is often the case for governments dealing with complex and wicked problems, Indiana’s policymakers were having difficulty seeing through the fog of Big Data and situational complexity to get to the root cause of the state’s infant mortality problem. Because of this, Mr. Atkins explains that the state “…had to go with their well-intentioned gut – in this case, beliefs that parental smoking, drinking and doing drugs were the root cause of high rates of infant mortality.” Unfortunately, efforts to reduce these behaviors had no discernable impact on the rate. Mr. Atkins and his team realized that they would need deeper insights to solve the problem – specifically, they needed to know more about the mothers and their infants.
Fortunately, Governor Pence’s predecessor, Mitch Daniels, had instituted a strong culture of data-driven decision-making within the State’s agencies, so the Pence administration only had to figure out how to scale that culture across agencies. The team theorized that predictive analytics technology could be applied to the agencies’ Big Data holdings to study the root causes of infant mortality and develop a roadmap for how to address the problem. This was the inception of Indiana’s Management and Performance Hub (MPH).
The MPH was established in response to an ambitious vision: to create the most effective, efficient and transparent state government in the country by developing an industry-leading comprehensive enterprise-wide data-driven management system. Technically, the MPH is a real-time predictive analytics platform based on SAP technologies. It enables data scientists to surface and statistically quantify the importance of known and previously unknown risk factors buried in Big Data and to establish correlations between those risk factors to gain actionable insights into wicked problems like infant mortality.
For its first project, the state applied predictive analytics to 9 billion rows of data across 50+ datasets to identify subpopulations with underlying drivers for infant mortality. Essentially, they matched mothers and infants between the data sets (while maintaining confidentiality in a secure environment) and started looking for patterns. The data told them that there were a number of statistically important risk factors that had never been considered by health professionals; correlations between some known and previously unknown risk factors, which in combination increase the overall risk; and particular subpopulations and regions with quantifiable and predictable drivers for infant mortality.
But most importantly, the data told them that their gut feelings were wrong: Access to prenatal health care – not substance abuse – surfaced as being the most significant determinant of infant mortality. Mr. Atkins describes this as “…a transformative insight because it told us that we had been pulling the wrong policy levers.” They used this newfound understanding to secure a $13.5 million budget appropriation for more targeted programming. And having demonstrated the effectiveness of predictive analytics, ongoing funding was allocated to the state’s Management and Performance Hub to tackle other wicked problems.
The opioid epidemic
The United States is in the throes of an opioid epidemic, with more than 2 million Americans addicted to prescription pain medications, illicitly made fentanyl, and illegal drugs such as heroin. A staggering statistic is that although the United States accounts for only 5% of the world’s population, its citizens consume 80% of the world’s painkillers. This nationally declared public health emergency is very much an ongoing and increasingly severe problem, with drug overdoses involving synthetic opioids other than methadone accounting for 42,249 deaths in 2016, having doubled from the previous year. The opioid epidemic has hit Indiana particularly hard, with 1 in 20 citizens reporting non-medical use of opioid pain relievers in 2016. In 2015, intravenous drug use put the state at the epicenter of the nation’s worst HIV outbreak in two decades. And alarmingly, 80% of Indiana employers have observed misuse of prescription drugs in the workplace.
Like infant mortality, opioid abuse is a wicked problem, influenced by complex and interrelated social and economic factors. Indiana’s State Police had evaluated a significant amount of data around positive drug tests and pharmacy thefts, but they’d been struggling to gain a holistic view of the problem due to the siloed nature of their data sets. Prompted by an HIV and Hepatitis C outbreak in Scott County, Governor Pence issued an executive order creating a task force on drug enforcement, treatment, and prevention. The MPH team partnered with the task force to combine and analyze the data sets, in the hope that future such events could be predicted and averted.
Public safety information was combined with health data to identify opioid abuse hotspots and gaps in treatment – exactly the types of actionable insights that had been lacking. For example, more than 40 opioid-related deaths had been reported in Tippecanoe County – more than 60 minutes’ drive from the nearest opioid treatment program. The state leveraged these new insights to deploy 5 additional treatment centers in the locations of greatest need, where previously they would have been placed solely on the basis of population density. The MPH’s ability to drill down into the data also provides the potential to model addiction risk and recommend targeted interventions.
Each year, more than 700,000 individuals are released from prisons in the United States. The Indiana Department of Correction defines recidivism as a return to incarceration within 3 years of the offender’s date of release from a State correctional institution. In 2016, the state’s recidivism rate was 37%, with nearly half of returning prisoners having been convicted for a new crime. Studies have shown that young males are more likely to re-offend, while those who participate in work release programs are less likely. But what’s more difficult to determine is how to reduce recidivism by targeting the right programs, at the right time, to the right offenders – this was the task set for the MPH team.
Where previous studies had focused on the overall effectiveness of programs and facilities, the MPH enabled analysis to be focused on the individual, considering personal characteristics and criminal history. No longer limited to studying general program effectiveness for cohorts of offenders, data scientists could ask: if this program is applied at this time to this offender, how much of a decrease in their probability of recidivism can we expect? Consequently, the state has been able to personalize work release plans and more effectively match programs with individuals to optimize outcomes. The MPH has also been leveraged to inform policy decisions on important issues like sentencing reform.
A data-driven approach to combatting wicked problems
What’s apparent from the scenarios addressed by Indiana’s Management and Performance Hub is that at the center of every wicked problem is a complex multi-dimensional human being. Considering that 75% of the state’s recidivist offenders are diagnosed with substance use disorder, many individuals have undoubtedly been identified as belonging to the high-risk cohort across multiple MPH studies. Indeed, this type of cross-correlation could be at the core of why Indiana has a high rate of pharmacy robberies.
Therein lies the true value of the MPH – not as a solution to a single problem – but as a data-driven platform that can be leveraged to provide actionable insights for any given problem; the interdependencies between problems; and even to predict where new problems may arise. To quote Mr. Atkins: “Big problems typically don’t respect system, program and state borders… but many agencies have data in their custody that could help the government gain actionable insights [that can] help to better calibrate policy and planning toward strategies that are data-driven instead of gut-driven… Data-driven government is not about technology. It’s about leveraging data as an asset to deliver on government’s mission to improve citizen’s lives.”
For more on how technology can address some of society’s toughest problems, see Mapping A New Strategy To Fight Opioid Addiction.
A wicked problem is a social or cultural problem that is difficult or impossible to solve due to incomplete or contradictory knowledge; the number of people and opinions involved; the large economic burden; and the interconnected nature of these problems with other problems. The terminology was originally proposed by Rittel and Webber (University of California, Berkeley, 1973) in a landmark article, in which the authors observed that there is a whole realm of social problems that cannot be successfully treated with traditional analytical approaches.