Robert Pynoos (University of California, Los Angeles) discussed ways of operationalizing the assessment of exposure to potentially traumatic events and the assessment of posttraumatic stress reactions in children. He began by saying that the approach to collecting data on these topics in children has evolved differently from the data collection approaches in adults. Unlike for adults and adolescents, there is no checklist of potentially traumatic events for children. In addition, the literature of trauma in children is much more nuanced and focused on details, such as age of onset, duration, and serial or sequential occurrence.
Pynoos described the SAMHSA-supported National Child Traumatic Stress Network (NCTSN), coordinated by the University of California, Los Angeles (UCLA) and Duke University National Center for Child Traumatic Stress. The NCTSN uses the UCLA PTSD Reaction Index for the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5), including the Trauma History Profile, as part of their core dataset. The scale includes 23 individual traumatic event types. Pynoos noted that clinical studies of children show that witnessing a parent’s rape produces levels of PTSD that are essentially equivalent to being sexually abused. A threat to a parent or sibling is considered to be one of the elements of feeling life threat among children. Because of this, the scales for children include a category for direct victimization and a separate category for being a witness.
Pynoos said that the literature on trauma in children also differs from
the literature on trauma in adults in the way it addresses issues related to bereavement and the interplay of trauma and grief reactions. A relatively higher proportion of the deaths experienced by children occur under traumatic circumstances compared to the deaths experienced by adults. One example is the sudden death of a primary caregiver among young children.
One of the points underscored by Pynoos was that in psychiatric epidemiological studies it may be important to oversample children with comparatively rare, but high magnitude exposures. This could include children who witnessed homicide or the rape of a parent in order to evaluate severity of impact and outcome. This approach is similar to the study of rare medical conditions among populations.
Pynoos said that it is important to be mindful of the developmental epidemiology of exposure. Exposure to certain types of traumatic events is more likely at certain ages, and the profile of the event changes depending on age. Some events are more likely to co-occur among children, especially in early childhood. For example witnessing domestic violence, physical abuse, psychological maltreatment, neglect, and impaired care-giving can form a constellation of early childhood exposures. In addition these same conditions often are associated with lack of supervision and increased risk for dog bites, serious burns, and near drownings. The literature focused on trauma in adults rarely takes account of this co-occurrence when discussing early childhood exposure. These items are important to include in order to understand the full context of trauma history.
The exposure configuration changes in adolescence. Being a driver or passenger in a fatal car accident, witnessing gang rape, criminal victimization, and trafficking become more relevant. In addition, the adolescent experience is not the same as that of a younger child either. For example, being an adolescent driver or passenger in a car accident is different from being in a car accident while being driven to school by a parent.
Pynoos said that the risk of exposure specific to different events increases at different points over the life course. Thus, it is useful to think about the developmental epidemiology of exposure, rather than just thinking of a list of events. Researchers have also observed a “risk caravan,” meaning what additional risks are accrued with the accumulation of different types of exposures.
Figure 4-1 shows the differences in the pattern of trauma types in early childhood, school age, and adolescence, based on data collected by the NCTSN. Pynoos said that the data are not from a nationally representative sample, but they show informative differences based on a large-scale (N = 19,088) database of children and adolescents receiving services in the United States.
Based on the NCTSN data, Pynoos and his colleagues have been able
to isolate cases of only emotional abuse in early childhood, and examine its effects in relation to its own characteristics and as a component of the early childhood constellation. They learned that emotional abuse under age 6 produces similar levels of PTSD as other traumatic events, perhaps because emotional abuse, such as threats of abandonment, is experienced by children as a life threat.
Pynoos said that the data also show how exposures to several different events work together. Emotional abuse has an additive effect when co-occurring with physical abuse, neglect, and witnessing domestic violence. In addition, symptom profiles may change as exposures occur across developmental periods. For example, when childhood sexual abuse is added to other early exposures, the symptom profile is dominated by posttraumatic stress relations related to the sexual abuse, perhaps masking some of the other trauma-related reactions. The data also show a cascade of effects for exposure: sexual abuse at age 6 increases the risk of sexual assault by age 9. Through childhood and into adolescence, the risk for other issues that SAMHSA is interested in also tends to accumulate, including drug abuse, HIV, and various risky behaviors, representing a caravan of risk. Pynoos commented that this finding also means that it is possible to identify the many different points where one can intervene and possibly prevent the emergence of the next risk factor.
In terms of the debate about the advantages of a systematic review of trauma exposure in comparison with asking gateway questions, Pynoos said that in his work he benchmarks exposures against developmental periods, rather than asking about lifetime exposure. His method involves providing blocks of time that respondents can use to reference their experiences: for example, when you were little, before you went to school, in elementary school, in junior high school, or in high school. This approach can increase the reliability of responses in children and adolescents.
In the case of adolescents with exposure to multiple events, Pynoos said that he and his colleagues ask respondents to construct a hierarchy of events by ranking them and then indicating which ones are the most disturbing to them in their current lives. Sometimes they ask respondents to construct two hierarchies, one for childhood and one for adolescence, because research shows that adolescents mentally split off their adolescent experiences from their younger experiences. The responses are often unexpected: for example, when an adolescent ranks standing at a bus stop when he was 13 and seeing a man brutally beating his wife as more intrusive than a recent experience of being in a shooting.
Pynoos said that children down to the age of 8 can reliably self-report and provide comparisons to evaluations using structured interviews, such as the Clinician-Administered PTSD Scale (CAPS) and the child version of the Schedule for Affective Disorders and Schizophrenia. Chil-
dren can be accurate reporters if the questions are phrased carefully with developmentally appropriate wording, and if they are adequately tested. However, there are certain types of data that children are not very good at reporting. For example, some of the typical gateway questions about upsetting memories and flashbacks do not work well in children. Some of those data, such as reports of restless, agitated sleep, can be collected with better accuracy from the parents. In addition, the age 6 and younger criteria for PTSD in DSM-5 notes that children can have repetitive play, re-enactment behavior, and intrusions without overt signs of distress.
In terms of criteria C, Pynoos said that the lack of endorsement of avoidance symptoms is one of the main reasons why children do not meet diagnostic criteria for PTSD. For example, children are typically unable to describe “feeling numb.” The challenge with asking about avoidance is that children do not often have a choice for physical avoidance. In the DSM-5, the wording was changed to “efforts to avoid” and associated behaviors are included, such as a child throwing a tantrum when the parents want to take her or him somewhere that might serve as a reminder of a traumatic event. When it comes to avoidance, children are more likely to endorse the “do not want to talk about it” response option. Among category E symptoms, sleep disturbance is important, especially because in a young child it can have an enormous impact on learning.
Pynoos reiterated that the symptom profile can change as children become older. For example, in some of the studies of New York City school children, conducted in the aftermath of 9/11, school-age children tended to report efforts to avoid, while adolescents did so less frequently, instead describing other problematic behaviors.
In terms of the transition to the DSM-5, Pynoos noted that the UCLA PTSD Reaction Index Trauma History Profile and the CAPS for children and adolescents are available. These now include wording for the new symptom items D and E (see Chapter 2), including negative emotions, such as guilt and shame, which require developmentally appropriate wording. They also include child-specific items for other trauma-related expectations, and child and adolescent worded questions about irritable and aggressive behavior, and reckless or self-destructive behavior. Pynoos emphasized that it is very important to thoroughly test these types of items.
One of the challenges raised by Pynoos is related to formulating questions for adolescents about current PTSD when the traumatic event happened before the age of 6. The criteria for children 6 and under are different from the criteria for those who are older. Deciding which criteria to use is not immediately obvious. Pynoos said that asking about dissociative subtype is particularly difficult, but evidence suggests its importance even among young children.
In order to establish symptom presence, Pynoos and his colleagues use pictorial tools as anchors. For example, to get reliable frequency in days per month, they use a calendar that illustrates each answer option. To collect data on degree of intensity and determine how much the symptoms bother respondents, they use pictures of glasses filled to various levels. Pynoos believes that this technique leads to more reliable reporting in the case of children and adolescents than using verbal labels alone.
A question that has not yet been settled in the context of the DSM-5 is that of the cutoffs for counting a symptom as present. A cutoff is needed even in the case of a continuous scale if the goal is to arrive at a conclusion that is a diagnostic probability. Pynoos said that he and his colleagues have a study in progress to help answer this question. Another outstanding question noted by Pynoos is the extent to which a proxy symptom question for some level of lifetime PTSD would work in children and adolescents.
Pynoos also discussed the concept of functional impairment, a key criterion of PTSD in the DSM-5. The text of the DSM-5 provides a developmental framework regarding functional impairment, including in school and among peers. In young children, avoidant behavior may lead to restricted play or exploratory behavior; in adolescents, it may lead to reduced participation in new activities or missed developmental opportunities, such as dating and learning to drive. Pynoos emphasized that developmental outcomes need to be considered along with what has typically been considered to be functional impairment. For example, studies have shown that sexual abuse in childhood can lead to diminished self-care in adulthood. Such behavior is not a functional outcome the way it is normally defined, but as a developmental outcome it has profound influences on health behavior. In contrast to developmental delays, adolescents may show developmental accelerations as an outcome of traumatic experiences that increase the risk of further exposure.
NCTSN data show that among adolescents that have had multiple traumas earlier in childhood, there is a substantial subgroup that has subclinical levels of PTSD that are associated with major functional impairments. In addition, children and adolescents who meet only criteria B and D can have significant functional impairment, and different clusters of symptoms may have different causal relations to outcomes (for example, risk behavior, health consequences). He noted that if a study design calls for skipping some items, it could mean skipping the ones that would otherwise be the most highly endorsed by respondents.
Research on comorbidity has shown some interesting patterns in children and adolescents. For example, studies on the aftermath of disasters and terrorist attacks, such as 9/11, have found increased separation anxiety disorder in adolescents, which is not typically expected in that
age group. Pynoos emphasized that in examining issues such as substance abuse in adolescents, it is important not to overlook exposure to death as a possible contributing factor. When bereavement leads to substance abuse, the associated behaviors are better understood in adults, and they need to be further studied in adolescents. He and his colleagues developed the Persistent Complex Bereavement Disorder Checklist for use in clinical research.
Finally, Pynoos said, another particularly important issue is multiple comorbid conditions among adolescents with complex trauma histories. A new diagnosis that has been proposed by a collaborative group of the NCTSN is developmental trauma disorder that gives priority to disturbances in development.
Benjamin Saunders (Medical University of South Carolina) discussed the measurement of potentially traumatic events and PTSD in children, with specific focus on implementation considerations. He agreed with Pynoos that the most difficult cases to measure and treat involve children who have been exposed to multiple traumatic events. He added that there are events that can be potentially traumatic to children, but would not be similarly traumatic to adults, or even adolescents, so the developmental aspects of what may or may not be traumatic based on age is something that is important to consider when deciding what needs to be measured. In addition, asking an adult about things that are meaningfully important to them at the present that were potentially traumatic when they were children could result in a list that does not correspond to the types of events that are included among the DSM-5 criterion A events (see Chapter 2).
Starting with the premise that no single study can measure everything related to potentially traumatic events, PTSD, and related outcomes, Saunders discussed several strategies for narrowing down the list of items to those that are key to include in a particular study. Starting with reviewing the prevalence rates in the population for specific traumatic events would be a reasonable approach. Another useful initial step would be reviewing existing data on impact, in other words, the percentage of people with a certain type of experience who develop PTSD or the percentage of people who have PTSD as a result of the experience. He noted that some events, such as sexual assault, are included in almost all data collection instruments on the topic because of the broad agreement about their potentially traumatic nature.
In some cases, a particular topic may be of interest for a specific study
or become more relevant due to current events generally. An example is sexually exploited children: 15 years ago, it was not a topic that was typically assessed in data collections on trauma, but it is now almost always included because of the increased visibility of the issue.
Saunders commented that Schell’s discussion (see Chapter 3) about formative scales and the idea of identifying the outcomes of interest before the relevant traumatic events was useful. That approach could reduce the likelihood of items being introduced simply because they happen to be of interest to someone at a particular moment or are subjectively considered potentially more traumatic than others by particular researchers.
Beyond measuring potentially traumatic events, understanding the incident characteristics, the context of the event or events, and other background information about respondents can also be critical because they are often associated with the development of PTSD. Robert Ursano (Uniformed Services University of the Health Sciences) mentioned the importance of understanding the community context, and Saunders pointed out that geocoding may be useful to add to data collection. Other data that are typically collected as part of studies on the topic of trauma include whether the event was a single event or part of a series of events, the duration of the event, and the respondent’s age when the event first happened and when it stopped. In the case of children, in particular, traumatic events are often repeated incidents.
Saunders underscored the importance of understanding trajectories and the sequence of exposure that leads to increasingly more risky behavior, a topic that was also discussed by other speakers. He said that understanding the neurobiological and sociological processes involved and the reasons why some children develop difficulties and others do not are currently the most promising areas of research in the field of trauma. He pointed out that the complex interactions among events, outcomes, mediators, and moderators can be particularly difficult to tease apart with data from large national surveys, and it is not clear to what extent is it possible for SAMHSA to undertake a large-scale project, but he argued that examining these issues would move the field forward. Nonetheless, he warned that even a hypothetical study that had unlimited resources would be challenged in developing an approach that would come close to fully capturing all the relevant information. The nature of the topic is such that there will always be a river of possible alternative explanations for outcomes running below the data.
An important consideration when collecting data about trauma in children is that if children are interviewed, permission from their parents is required. And for some age groups, parent interviews need to be substituted for the interviews with children. Researchers have to decide when it makes sense to interview a child, said Saunders. In some cases,
parent interviews can produce reasonably good information, and interviewing parents may be more efficient if they need to be contacted to obtain permission.
Saunders summarized the characteristics of a good screening approach:
- includes multiple questions covering the range of experiences within type;
- assures that items are behaviorally specific to reduce interpretation;
- uses language level consistent with the target age group;
- cues retrievable memories of past events; and
- matches respondents’ interpretations and labeling of experiences.
He pointed out that the last item on the list (assuring that the questions match a respondent’s interpretation and labeling of experiences) is the most challenging in the context of interviewing children. Researchers need to develop questions with language that corresponds to the schema used by children and their views of the experiences.
Saunders said he agreed with Terrence Keane (Boston University School of Medicine and U.S. Department of Veterans Affairs National Center for Posttraumatic Stress Disorder) that questions on this topic are very susceptible to order effects and that the sequence of the sections also deserves careful attention. In his research, Saunders said, he likes to begin with easy questions, followed by the sensitive questions, and then another set of easy questions, which may be followed by a debriefing.
Some of the common errors he noticed in instruments on this topic include
- not asking key questions;
- “gate” questions and single screening items;
- undefined terms that are open to significant interpretation by respondents (e.g., physically abused, sexually abused, fondled, bullied, raped, molested, attempted, domestic violence);
- double- (or more) barreled questions;
- lengthy or overly wordy questions; and
- asking follow-up questions after each screening hit.
Questions with the shortcomings highlighted above can be especially difficult for children and adolescents and can lead to higher error rates in some age groups. For example, questions that are open to interpretation or are lengthy can present more challenges for children than for adults. Asking follow-up questions after each screening hit can lead to response bias in any age group, if it cues respondents that a “yes” answer will lead
to more follow-up questions, and they begin altering their responses as a result.
Saunders also listed several factors that can affect case detection:
- level of perceived confidentiality offered, concerns of getting self or others in trouble, fear of retribution;
- context of the screening setting
- – location of respondent (home, school, other)
- – method (in-person, group, telephone, paper, computer)
- – who is present? (interviewer, parents, teacher, peers, siblings);
- recall of events by respondents
- – experiences not recalled, forgotten, and not accessible
- – experiences partially forgotten, but retrievable with the right cuing
- – remembered experiences, but not defined by the respondent in the same way as the screening question is worded
- – remembered experiences that are willfully withheld; and
- willful nondisclosure.
Saunders said that perceived confidentiality is a particularly important consideration when interviewing children because they tend not to understand or believe that the information they provide will be kept confidential. A related issue is willful nondisclosure. While adults can also be reluctant to talk about traumatic events, such reluctance is more common among children and adolescents. The reasons for this may be in part that, for children, the questions are more likely to be about something that happened in the recent past rather than an event that happened decades ago in the case of adults. Children may have had less time to process the event and develop a perspective on it.
Other reasons for willful nondisclosure include
- sense of stigma, shame, guilt, self-blame;
- threats or instructions by a parents;
- fear of punishment, “getting into trouble”;
- fear of consequences to family and family members;
- cultural and familial beliefs about privacy;
- psychological distress about events;
- fear of retribution by assailant; and
- history of negative outcomes from prior disclosures (disclosure inoculation).
Saunders said that a history of negative outcomes from prior disclosure can be particularly challenging to overcome. Some children may
be “inoculated” against talking about what happened to them because they had tried to talk about it before and bad things happened, or nothing happened. Careful question wording can help reduce some of these challenges.
Saunders noted that there are several ethical questions that arise in the context of research on trauma about children. One question is whether the interviews will be overly distressing. He said that he and his colleagues, as well as other groups of researchers, have conducted studies to examine this issue and concluded that the interviews do not appear to be overly distressing. However, less is known about the reactions of younger children than about adolescents.
Another question that comes up is whether the parents get upset when they learn about the types of questions that are being asked of their children. Saunders said that some parents do have objections, and it is important to think through the concerns they might have prior to contacting them.
A related issue is whether asking the questions could place some children at risk from their parents. Saunders said that their longitudinal studies seem to suggest that this is not the case because they found that children with trauma histories are more likely to participate in the follow-up waves of the studies, after the initial interview, than children without trauma histories. This finding could be an indication that these children did not experience any repercussions after participating and that they found the explanation and information provided to them as part of the study helpful.
Saunders also pointed out that collecting data about trauma in children means collecting data that can have legal implications. He said that it is important to carefully consider how the identifying information is stored and who has access to it, as well as whether the data can be subpoenaed. There are also mandatory reporting laws that may apply, and these can be different by state, so a plan is needed for how to manage situations in which this issue may arise. Saunders said that he and his colleagues also use a “child in danger” protocol, similar to what SAMHSA used in the Mental Health Surveillance Study, and it seems to work well.
Graham Kalton (Westat) asked whether there are ways to deal with situations in which a parent is abusing the child and so does not grant permission for the interview. Saunders said that this is likely happening and that the most one can do is to develop survey materials that reduce this problem as much as possible. He acknowledged that it is likely that this leads to underestimates of child trauma in all surveys. However, he noted that in his studies usually less than 10 percent of parents decline to have their children interviewed after the parent interview is completed. Schell noted that some of the phenomena that are being measured are very rare
in the population, so refusals can make a big difference. Pynoos added that an additional issue with the proliferation of cell phones is that survey researchers are less likely to be calling a landline and then being able to continue the interview with the child, on the same line, after obtaining permission from the parent. The fact that most people and many children have their own cell phones complicates the data collection process.
Kilpatrick commented that if the survey is about a variety of topics, researchers do not need to begin the conversation by saying that they would like to ask children about whether they had been abused. For example, one of their studies, the National Survey of Adolescents, was about a range of topics that are important to parents and families, such as community violence and alcohol and drug use. He said that it is important to provide an accurate description of what the study is about, but providing too many specifics can increase nonresponse bias.
Kalton said that it has been noted that child reports often differ from parent reports and that some studies that include teacher reports find that the teacher reports are also very different. Some researchers argue that multiple reports are necessary to measure issues of this type. Saunders agreed that there are typically significant differences in what is reported, depending on who is providing the information, and that this is generally the case with topics of this type. He noted that, in some sense, all of these reports may be accurate from the perspective of the person who is reporting. There is typically more convergence in data about child behavior than about internalizing problems, such as depression or PTSD, which definitely represents an analytic challenge. Pynoos commented that the topic of trauma presents special challenges in this regard because the link between traumatic exposure and behavior is rarely identified by parents and even less often by the schools.
James Jackson (University of Michigan) asked Saunders to clarify why he thinks that children and adolescents are more skeptical of promises of confidentiality. Saunders said that many children are afraid of consequences, such as getting into trouble or getting someone else into trouble. Convincing them that what they say will be kept confidential is especially difficult if, as part of the informed consent process, they are also told that in some cases what they say may have to be reported (e.g., in mandated reporting situations). He reiterated that very careful wording is crucial. Jackson said that his own research with adolescents leads him to think that adolescents are skeptical about adults’ ability to “keep secrets,” which is a small, but important distinction, and that understanding these nuances is crucial in order to be able to address the concerns. Pynoos commented that his research indicates that adolescents are more likely to disclose exposure to traumatic events when the questions are administered by computer rather than in person.