Applying Safety Science to Child Welfare

The recent explosion of the SpaceX Falcon 9 rocket on the launch pad triggered a series of steps to identify the source of the disaster and ensure it never happens again. Efforts to identify the full range of system forces at work when disaster occurs is the basis of what is known as “safety science.”

Like the space industry, the commercial aviation industry has long used safety science in the creation of mitigation strategies to eliminate identified risks from the system and prevent the occurrence of fatal accidents.

The MITRE Corporation, an independent not-for-profit organization that operates research and development centers for the federal government, facilitated the development of a public-private partnership with key members of the aviation industry and the federal government to facilitate the sharing of safety data. The data is used to conduct proactive analysis, identifying emerging systemic risks and vulnerabilities before the next serious accident or incident occurs.

Our mission, in aviation and other domains such as healthcare—where we have adapted the aviation safety model to improving patient safety—is to advance and apply science, technology, systems engineering, and strategy to enable the government and the private sector to make better decisions and implement solutions to complex challenges of global and national significance.

Since 2007, when MITRE launched its research focus on aviation safety, there have been only two fatal accidents involving U.S. commercial air carriers on domestic soil, a 67 percent reduction compared to the previous ten years.

Could this same level of success be possible in the child welfare system?

That was the question posed to us by the federal Commission to Eliminate Child Abuse and Neglect Fatalities (CECANF) when we presented testimony at their public meeting in New York City in August 2015.

In September 2012, MITRE began operating the Centers for Medicare & Medicaid Services (CMS) Alliance to Modernize Healthcare (CAMH), a federally funded research and development center (FFRDC) dedicated to strengthening our nation’s healthcare system. Since that time the CAMH FFRDC has supported agencies across HHS, including focusing on its child welfare mission. In beginning to examine the barriers to reducing risk in the child welfare system, we identified several common building blocks from the success of our work on aviation safety that could be applied to a child welfare model, starting with the development of trusted relationships among stakeholders.

Trust is a critical issue for a number of reasons. We believe that a partnership which facilitates the aggregation and analysis of child welfare and other types of relevant data holds promise for improving our ability to identify and mitigate risk in the child welfare system. This process would entail the collection and aggregation of vast amounts of event and context data from many of the partnership stakeholders, much of it proprietary and sensitive. In the case of patient safety and child welfare, HIPAA regulations also require complete anonymity that separates identities from the data.

Since MITRE is not subject to federal or state Freedom of Information Act (FOIA) disclosure requests, partners may share data with MITRE that they may not want to share with the government directly or be subject to the risk of a FOIA request. As a result, MITRE is able to receive sensitive data in a secure environment that provides analytic tools and infrastructure to conduct and share statistical analysis, perform data mining, and fuse multiple data sources together.

As we have initially begun to look at some of the barriers to eliminating child abuse and neglect fatalities, we are seeing challenges that appear very similar to those we initially encountered in our aviation and patient safety work.

One of those challenges involves the need for standardized data. For our aviation safety work MITRE developed a system to extract, transform, and convert data from the native airline format to a unified format for use within the bounds of the partnership. This process eliminated the need for all parties to conform to a standard (that did not exist) and accelerated participation by industry. A similar system will be required for child welfare data since much of the data related to fatalities is heterogeneous and complex. Additionally, definitions of what constitutes abuse or neglect are not standardized across states or jurisdictions.

As participation in these types of partnerships is to be voluntary, stakeholders must see real value or benefit in participating. This value is largely created through providing partners with access to aggregate de-identified data, benchmarks against their peers, and analytic tools that help them achieve their objectives, such as improving child safety, in a way that they would not be able to do on their own.

An analytics-focused partnership of child welfare stakeholders would conceivably include federal, state, and local governments and their non-profit partners, along with commercial organizations that provide child welfare solutions and services. Sources of data could include child welfare agency data from hotline calls and case management, electronic health records from emergency rooms and care providers, information from law enforcement, schools, and others.

As the data is aggregated, the partnership could prioritize the types of risk modeling, visualization tools, best practice and lessons learned sharing, and reporting partnership members believe would be of most use. The partnership could also direct specific research studies on critical topics that could then inform policy and decision-making. For example, such studies might seek to identify whether there are factors or practices in the child welfare system that unwittingly create systemic risks to children, or if there are unrecognized environmental, community, or familial factors that influence risk.

This aggregation of data will ultimately allow for the creation of a comprehensive picture and risk model that the government or any single child welfare organization would otherwise never see. As more stakeholders become involved and more data is collected, it may eventually lead to new insights that allow for proactive prevention of harm before it occurs and prioritization of services for those who need them the most.

Based on an analysis we are currently conducting on the behalf of the HHS Assistant Secretary for Planning and Evaluation, we will provide an assessment of the potential benefits and pitfalls of predictive analytics as applied in the field of child welfare, and to identify and assess federal opportunities to contribute to the advancement of predictive analytics approaches to promoting safety, permanency, and well-being for children in or at risk of entering the child welfare system. The evaluation will include reaching out to local agencies and their partners that have already implemented predictive analytics into their programs and understand what impacts they have observed.

While our longer term vision is focused on developing a partnership based around the aggregation of national child welfare and related data, we are already seeing that there are also immediate opportunities to begin improving data, analytics, and insights at the local level.

For example, MITRE recently launched a research collaboration with the County of San Diego’s Health and Human Services Agency to develop a system architecture to review and aggregate child welfare, medical, substance abuse, and self-sufficiency program enrollment data into an analytical framework. Our effort here, while still in the early stages of development, is generating important insights and will likely help to refine our efforts and approach at the national level.

In its report to Congress, issued in March of 2016, the Commission to Eliminate Child Abuse and Neglect Fatalities noted: “Child protection is perhaps the only field where some child deaths are assumed to be inevitable no matter how hard we work to stop them. This is certainly not true in the airline industry, where safety is paramount and commercial airline crashes are never seen as inevitable.”

Our goal is to apply safety science to child welfare in a way that will ultimately reverse the assumption that these tragic deaths are inevitable and unavoidable.

Mark D. Thomas, PhD, MPA, is Senior Principal and Health & Human Services Portfolio Manager for The MITRE Corporation. Edward B. Walsh III, MS, is the Department Head for Aviation Safety Analysis for The MITRE Corporation. Christopher M. Teixeira, MS, is a Lead Data Scientist in Model-Based Analytics for The MITRE Corporation.

Print Friendly, PDF & Email

1 Comment

  1. The comparison between plane crashes and child deaths is spurious. A plane, while massively complex, is still a mechanical system with a finite number of parts that move in predictable ways. To ensure that a plane doesn’t crash, we can set up systems to make sure that all of those finite mechanical parts are operating in the exact manner necessary to safely fly the plane. There are no such mechanistic and fully predictable characteristics of the human behavioral systems that lead to child deaths. The larger point from this article is well taken – data is a powerful tool that we should absolutely be harnessing to create better systems for preventing children from dying due to maltreatment. But likening this process to safely flying a plane falsely compares a complex human system to a complex mechanical one. It’s a better analogy to compare child death prevention to other prevention efforts in human systems… do we expect the police to be able to prevent all crime? Even all murders? Even with advanced data analytics? Of course we don’t. It would be impossible, no matter how sophisticated the data, to think that law enforcement could ever prevent all homicides, because human behavior does not operate on the same predictable principles as those which operate planes.

Comments are closed.