Research

Unreported past child maltreatment may contaminate research results

UNIVERSITY PARK, Pa. — Researchers may unknowingly be overlooking a critical factor that could aid in more consistent results in studies involving child abuse. Penn State Network on Child Protection and Well-Being faculty member Chad Shenk is opening the door to more precise results in the realm of child abuse and maltreatment and is hoping to use his findings to strengthen the research process as a whole.

Shenk, assistant professor of human development and family studies, used data from a five-year study to determine whether research contamination leads scientists to reach different conclusions about the effects of child maltreatment. The study was published in the March 2015 issue of the Journal of Pediatric Psychology.

Contamination occurs when the comparison group includes — unknown to the researcher — subjects who have come into contact with the variable that is being studied. In this case, that variable is abuse and maltreatment. The presence of contamination can be a serious issue regarding the validity of any study.

“Contamination can limit your ability to detect a difference between two groups,” Shenk said.

He found that when contamination is present and left uncontrolled, the effects of maltreatment are diminished, making the results appear closer to those of the experimental group. This can lead researchers to believe there is no connection between the cause and effect when — without this contamination — there may be.

Shenk’s study observed the health status of female adolescents, ages 14 to 19, following an instance of child maltreatment. He focused on four outcomes — teenage pregnancy, obesity, major depression and cigarette use within the past month — with the goal of discerning how contamination can impact research results. Participants were recruited into two groups. One group, the experimental group, was referred by Child Protective Services (CPS) and had been exposed to substantiated child maltreatment in the past year. The comparison group reported not having any CPS involvement.

Shenk ran the data in two different ways. The first time around, he analyzed the data based solely on how the two groups were recruited. This is considered the more “common” research approach.

Shenk then screened the comparison group a second time using two different methods to ensure no prior history of maltreatment. In addition to the preliminary screening, participants in the comparison group were asked to self-report any maltreatment and CPS was consulted for records of abuse or maltreatment. Shenk found that almost 45 percent of participants in the comparison group did, in fact, have a history of child abuse or maltreatment, providing a striking example of the prevalence of contamination in research.

“Approximately 40 percent of people who experience substantiated child maltreatment fail to report this information during a research study,” Shenk noted.

He also said that there are a number of reported CPS cases that are never confirmed, leaving a child record-free, but with a history of maltreatment. To add to this, some abuse cases are never reported and some children are being abused, but never receive any official aid. Without careful analysis, this contamination could alter a researcher’s end results.

Once Shenk controlled contamination in the comparison group, the results changed measurably.

With contamination in the comparison group, Shenk found child maltreatment in adolescent females only was able to significantly predict the prevalence of teenage births and past-month cigarette use. When contamination was identified and controlled, he discovered that child maltreatment not only significantly predicted teenage births and past-month cigarette use, but also obesity and major depression. With contamination restricted, the rates for these health outcomes more closely matched national averages.

These results are only true when safeguarding for contamination in the form of both CPS reports and self-report methods, he said. When screening for only one or the other, the results appeared similar to the contaminated analysis, highlighting the importance of a thorough contamination check.

This study is among the first of its kind to address the effect of contamination on research results. Shenk’s findings could mean that child maltreatment has stronger and more far-reaching effects than researchers previously thought.

Shenk will continue his quest for an explanation with graduate students from Penn State. They plan to run models from a five-site longitudinal study in hopes of replicating Shenk’s previous work. This time, however, the group of adolescents is more representative of the national population, with more variation in race, background and socioeconomic status. His overall goal, Shenk said, is to make the research process more reliable and receive results that can be easily and accurately replicated.

The framework for Shenk’s new study is in place and he hopes to make the results public by the end of the year.

Last Updated July 28, 2017

Contact