Identifying child mistreatment is one of the most difficult aspects of child protection. Often caseworkers have little to go on: Perhaps a neighbor reports excessive shouting in a child’s household. Or perhaps a child arrives late to school, looking disheveled, and tells a teacher that her parents have been fighting. Assessing the situation becomes even murkier if the child is very young. If a four-year-old says his parents are mad at each other, what should an adult make of that?
“Once a report comes in, a team of caseworkers has to figure out whether or not to go out into the field and investigate,” says Maria D. Fitzpatrick, Policy Analysis and Management. “It’s very labor intensive, and child protective agencies are extremely understaffed. The process they use to decide whether to investigate often leads them to spend a lot of energy investigating very low-risk cases while at the same time underinvestigating cases that are actually high risk.”
Algorithms Pinpoint Children at Risk
Fitzpatrick, who focuses much of her research in the areas of child and family policy, was asked by two county agencies in Douglas County, Colorado, to help them change the way they operate. The agencies were interested in using predictive computer modeling to assist them in pinpointing which child maltreatment reports to investigate. Fitzpatrick joined with departmental colleague Christopher Wildeman, now at Duke University, to set up a trial and evaluation of a predictive machine-learning model created by other researchers.
The model generated a risk score for each child suspected of being maltreated, in order to identify those with the highest risk of future removal from the home over a two-year period, Fitzpatrick explains. To come up with the score, the algorithm processed hundreds of variables concerning the family in question.
“Even though caseworkers had access to most of the same records that the machine-learning model did, they were trying to analyze that data with a human brain, which is much less accurate than the model,” Fitzpatrick says. “The risk score is one more piece of information that can help bring a little clarity to the process.”
The researchers found that use of the risk score information changed caseworkers’ decisions for the better, making them more likely to investigate the riskiest cases and slightly less likely to investigate the least risky cases, Fitzpatrick says.
The Importance of Teachers for Child Protection
In another project centered on child wellbeing, Fitzpatrick took a look at the role of teachers in identifying and reporting child maltreatment. Together with Cassandra Benson, PhD ’19 Economics, now at the United States Air Force Academy, and Samuel R. Bondurant, then a PhD student at Texas A&M who has since graduated and is now at the United States Census Bureau, Fitzpatrick used economic and statistical tools to analyze data about child maltreatment reporting. The researchers looked for reporting patterns, as well as changes in reporting, connected to whether children were in or out of school, for instance, during summer vacation.
“We should realize that teachers play this role of reporting child maltreatment, and we should give them the tools to better identify it.”
Fitzpatrick and her collaborators found that child maltreatment identification and reporting drop off significantly when children aren’t in school. “That means there’s probably a lot of maltreatment that’s going unidentified when children aren’t in front of teachers,” Fitzpatrick says. “This suggests we might want to take that into account when we’re designing supports for children and families. We should realize that teachers play this role of reporting child maltreatment, and we should give them the tools to better identify it so they don’t confuse maltreatment with other things, like poverty.”
Although teachers are mandated to report child maltreatment, very few people know that, Fitzpatrick points out, and even the teachers themselves aren’t always clear about their role as reporters. “Right now we have a very hodgepodge system of maltreatment reporting,” she says. “Teachers don’t get much training in it, and they don’t know what services are available to families. We need to help them be better participants in the system.”
Can Retirement Negatively Impact Health?
Along with her focus on child and family policy, Fitzpatrick researches various aspects of retirement. Her study exploring the connection between retirement and health has attracted broad media and news attention. “There’s an enormous amount of literature looking at whether retirement is good or bad for people’s health,” she says. “It’s a really complex question. The classic idea is that someone who works in a physically demanding job, such as a railroad worker, finds their health improves after they retire because they are no longer doing such backbreaking work. But not many of us do that sort of physical work these days. Many people find their work provides them with engagement, mental and social stimulation, and reasons to get out of the house. For those people, retirement might have a negative impact on health.”
Fitzpatrick joined with Timothy J. Moore, now at Purdue University, to address the question of retirement’s effect on health. They analyzed national death certificates in the United States, specifically focusing on people who were 62 years old at the time of death, since that is when most first become eligible for social security benefits. “Because of eligibility, a lot of people retire right at age 62,” Fitzpatrick explains. “Among men, 10 percent retire at that age. If retirement affects health, we thought we might see a change in mortality at that age.”
The researchers, indeed, found a two percent increase in male mortality at age 62. Single men with lower levels of education were affected the most, and the most common causes of death were lung-related conditions such as chronic obstructive pulmonary disease (COPD). “That suggests that, for men, on average, early retirement at 62 has pretty severe health consequences,” Fitzpatrick says.
Other research has shown that upon retirement, people become more sedentary and increase negative health behaviors such as smoking. Fitzpatrick and Moore hypothesize that the retired men were not in good health to begin with, and when they indulged in these negative behaviors the adverse health effects hastened their deaths. “Our findings connecting retirement and mortality aren’t true for everyone,” Fitzpatrick says. “The effect on women, for example, was negligible. But our research shows that some groups should take care in thinking about the transition to retirement. A lot of people prepare for retirement financially and in other ways, but very few think about how they are going to take care of their health.”
More Than Finance and Widgets
Fitzpatrick’s training is in economics, and her varied research interests illustrate the usefulness of the discipline, she says. “Most people think economics is either about finance or widgets, but it’s really a paradigm for understanding how people behave, why they make the decisions they do, how they allocate scarce resources like their time and money,” she explains. “Economics gives us rigorous methods for analysis and for thinking about the decisions we make on an individual and societal level, and it can help us figure out how to spend our time and resources in ways that are most effective.”