Screening for Risk of Persistent Posttraumatic Stress in Injured Children and Their Parents FREE

Posttraumatic stress disorder (PTSD) is a constellation of psychological symptoms (reexperiencing a trauma, avoiding reminders of it, and hyperarousal)¹ that can occur in children or adults in response to catastrophic events, such as disasters^2- 3 and terrorism,^4- 5 or, more commonly, after everyday catastrophes, such as injuries from motor vehicle crashes and other causes.^6- 11 Given the high frequency of childhood injury^12- 13 and rates of PTSD in injured children ranging from 13% to 45%,^9,14- 15 injury is an important cause of traumatic stress in childhood. In adults, the lifetime incidence of PTSD is about 1 in 12,¹⁶ with motor vehicle crashes alone resulting in PTSD in 28 per 1000 adults in the United States.¹¹ Untreated traumatic stress is a key determinant in poor health outcomes after injury,^17- 18 highlighting the importance of identifying and addressing psychological needs of injured children and their parents.

The diagnosis of PTSD requires symptom persistence for at least 1 month,¹ a period beyond when most children are discharged from trauma care. Stress responses resolve spontaneously in the majority of injured children¹⁹; however, most who develop PTSD and not diagnosed or treated.⁶ Clinicians lack tools to identify children at risk of traumatic stress, who require closer monitoring and possible psychological intervention. Empirically supported triage for psychological referral and intervention is consistent with current conceptualizations of prevention practice²⁰ and has proved useful for other disorders (eg, the CAGE questionnaire used to screen for alcoholism²¹).

The goal of this research was to develop an extremely brief, stand-alone screening tool feasible for use by clinicians during acute trauma care, before discharge to home, to assist in identification of injured children and their parents who are at risk of development of clinically significant, persistent posttraumatic stress symptoms. We hypothesized that such a tool could be developed based on preexisting and acute factors identifiable at the time of acute care (eg, prior trauma exposure, severity of exposure to the index traumatic event, type and severity of early stress responses).

The Screening Tool for Early Predictors of PTSD (STEPP) was derived from a longer survey composed of items that covered key evidence-based and theoretical domains^22- 23 associated with development of PTSD in children and parents. The survey was administered within 1 month of injury, and posttraumatic stress symptoms were assessed at least 3 months after injury. The most parsimonious set of items predicting posttraumatic symptomatology with high sensitivity and reasonable specificity were used to comprise the final version of the STEPP for children and for parents.

The current study was part of a prospective research program concerning posttraumatic stress in children and their parents after hospital admission for pediatric injuries sustained as a pedestrian, passenger, or bicyclist between July 1999 and October 2001. All children between the ages of 8 and 17 years admitted for treatment of traffic-related injuries to a large, urban, pediatric level I trauma center were eligible for inclusion. Children or parents unable to participate in an interview because of cognitive limitations or insufficient spoken English skills or whose family lived more than 2 hours away from the hospital were excluded from the study. Institutional review board–approved parental consent and child assent forms were signed in the hospital or at home. The first available parent (or guardian) who was a custodial parent, living with the index child, and who had (or was sharing) primary responsibility for the child's care was interviewed.

At the earliest possible time after consent was obtained but no later than 1 month after the index injury event, a trained research assistant administered the risk factor survey. (To assess test-retest reliability, we administered the survey twice to participants whom we were able to enroll during their hospital stay; the first administration took place in the hospital at the time of consent and the second administration took place in the home within the first month for this convenience subset of participants.) At least 3 and no more than 13 months after the injury (mean, 6.5 months; 93% within 9 months), child and parent posttraumatic stress symptoms were assessed in an interview conducted in the child's home. Children and parents were interviewed separately in nearly every case (in 7 cases, the parent chose to be present for all or part of the child's interview). Study interviewers completed a 2-day training session. Adherence to standardized interview protocols was assessed by a doctoral-level clinical psychologist (N.K.-A.) via ongoing review of audiotaped interviews.

A thorough review of the literature concerning the etiology of child and adult PTSD guided the selection of items for the risk factor survey: 20 questions asked directly of the injured child, 25 of the parent, and 5 items easily obtainable from the child's medical record. Among these 50 items, 32 were hypothesized to predict child posttraumatic stress. A distinct but overlapping set of 32 items was hypothesized to predict parent posttraumatic stress. Table 1 shows the domains covered in the risk factor survey.

Children's posttraumatic stress symptoms were assessed via a validated interview measure, the Clinician-Administered PTSD Scale for Children and Adolescents (CAPS-CA),³⁶ while parents' posttraumatic stress symptoms were assessed via a validated 17-item self-report measure, the PTSD Checklist (PCL).³⁷ The CAPS-CA interview includes assessment of impairment due to PTSD symptoms of the child's social, academic, or developmental functioning. Parent impairment due to PTSD symptoms was assessed via parent responses to a single item at the end of the PCL. For purposes of analysis, a positive case was defined as meeting criteria for PTSD or for subsyndromal PTSD (defined as children or parents who reported at follow-up at least 1 moderate to severe posttraumatic stress symptom in each of the 3 categories defining PTSD [reexperiencing, avoidance, and hyperarousal] and who had impairment from these symptoms in 1 or more areas of functioning). This case definition (referred to subsequently as persistent traumatic stress) was selected as an appropriate target for prediction because it denotes a group of children or parents likely to be in need of assistance beyond normal medical follow-up.³⁸

Children completed 3 additional validated brief checklists at follow-up to assess posttraumatic stress symptoms (symptom scale of the Child and Adolescent Trauma Survey [CATS]),³⁹ generalized anxiety symptoms (the short form of the Multidimensional Anxiety Scale for Children [MASC-10]),⁴⁰ and depression symptoms (the short form of the Children's Depression Inventory [CDI-S]).⁴¹

Relevant demographic (eg, age, race/ethnicity, sex) and clinical (eg, mechanism of injury, date of injury, details of hospitalization) information was abstracted from each child's medical record. International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes,⁴² Abbreviated Injury Scale scores,⁴³ Injury Severity Score,⁴⁴ and the child's heart rate at triage in the emergency department were gathered from the hospital's trauma registry. The Abbreviated Injury Scale and Injury Severity Score have been proved reliable and valid measures of injury severity in children.^43,45 The ICD-9-CM codes for each child were used to generate 2 dichotomous scores: presence or absence of loss of consciousness and presence or absence of a long bone fracture. A dichotomous variable was created for age (8-11 years vs 12-17 years). Heart rate at triage was dichotomized, taking into account variations in heart rate with child age: for children aged 8 to 11 years, the cut point was 104/min and for children aged 12 to 17 years, the cut point was 97/min.

Demographic and injury characteristics of participants completing follow-up assessment were compared with those lost to follow-up, with t tests or Wilcoxon tests used for continuous variables and χ² or Fisher exact tests used for categorical variables.

Data analysis proceeded in 3 stages. First, the frequency and variability of each potential predictor item and its association with the outcome were assessed; items associated with child or parent persistent traumatic stress outcomes at the P<.20 level were retained as candidate predictors for further analyses.

Second, multiple logistic regression analyses were performed separately to model persistent traumatic stress outcomes for parents and children. We used a best-subset approach, in which the best models containing 1, 2, or 3 variables and so on are selected at each step. A best model is identified based on comparison of the log likelihood estimates. No more variables are added to models when addition does not produce a significant increase in the log likelihood. In this approach, it is possible to have more than 1 best model with comparable likelihood. In selecting between models with equivalent performance, we examined these quantitative results in light of theoretical considerations (coverage of key domains for PTSD etiology) and practical considerations (eg, the generalizability of item wording to various types of pediatric injury beyond traffic-related injuries).

Once the best models were identified, we constructed the receiver operating characteristic curves, examined the area under the curves, and determined the cutoff scores that achieved maximum sensitivity while maintaining reasonable specificity, accounting for positive and negative predictive values. A scoring rule was generated for clinical use of the final child and parent screening tools.

Third, we examined test-retest reliability (κ statistic) for each item of the final STEPP questionnaire and for the presence of a positive screening test. To assess the breadth vs specificity of the distress predicted by the STEPP (for children), we examined the association of a positive screen with continuous measures of posttraumatic symptoms (on a second measure not used in modeling PTSD outcome), general anxiety symptoms, and depressive symptoms.

Two hundred sixty-nine parent-child dyads were enrolled in the study and completed a risk factor survey assessing potential predictors of PTSD outcome. Of these families, 171 (63%) completed a follow-up assessment for either the child or the parent. Complete data were available on 147 children (55%), 162 parents (60%), and 140 parent-child pairs (52%). No differences were found in demographic characteristics, mechanism of injury, admission to the intensive care unit, or severity of injury between those completing the study and those lost to follow-up. Table 2 presents demographic and clinical characteristics of the sample.

Twenty-five children (16%) and 25 parents (15%) were classified as positive for persistent traumatic stress. In 8 (5%) of the parent-child pairs with follow-up data available, both parent and child were classified as positive for persistent traumatic stress. Five children (3%) had clinically significant total scores on the CATS (PTSD symptoms), 6 children (4%) on the MASC-10 (general anxiety symptoms), and 5 children (3%) on the CDI-S (depression symptoms).

The best logistic model for children consisted of 8 predictors (area under the curve, 0.81). The best model for parents had 6 predictors (area under the curve, 0.81). Figure 1 shows the final STEPP screening tool and Table 3 shows odds ratios and P values for each final STEPP item as a single predictor of outcome and as part of the multivariable logistic regression analyses predicting outcome. Table 4 shows the performance of the STEPP in predicting persistent traumatic stress outcomes, based on the final cutoff scores for the STEPP (4 or more positive items for children and 3 or more positive items for parents). The odds ratio for a positive STEPP screening result in predicting persistent traumatic stress was 6.5 (95% confidence interval [CI], 1.8-22.8) in children and was 26.6 (95% CI, 3.5-202.1) in parents. One hundred fifty-nine (59%) of the children and 151 (56%) of the parents had a positive STEPP screening test result. In 102 (38%) of the families, both parent and child had positive STEPP screening test results.

For children, a positive screen was associated with higher levels of later posttraumatic stress symptoms on a second, independent measure (CATS symptom scale score median [range], 11 [0-36] vs 6 [0-25]; P = .008 by Wilcoxon test), as well as higher levels of later general anxiety symptoms (MASC-10 median [range], 9 [0-26] vs 7 [0-24]; P = .005 by Wilcoxon test), but was not associated with later depressive symptoms (CDI-S score median [range], 1 [0-14] vs 0 [0-6]; P = .11 by Wilcoxon test).

The risk factor survey was administered twice to 56 children and 60 parents (in 50 parent-child dyads, the survey was administered twice to both parent and child). Test-retest analyses for the 8 STEPP items requiring child or parent response showed that responses to the 5 items denoting factual events (eg, "Were you with the child in the ambulance?" "Was anyone else hurt or killed?") were consistent over time, with κ statistics higher than 0.60. Responses to the 3 items representing emotional states (eg, helplessness, feeling afraid) were much less consistent, with κ statistics less than 0.25. However, test-retest reliability for the presence of a positive screen was excellent for children (κ = 0.86; 95% CI, 0.68-1.0) and very good for parents (κ = 0.67; 95% CI, 0.43-0.91).

A theoretically derived, empirically validated, stand-alone screening tool, the STEPP was developed for use in the acute care setting, before discharge to home, to alert clinicians to injured children and their parents who are at high risk of posttraumatic stress months after the child's injury. The brevity (4 dichotomous questions asked each of the child and a parent and 4 readily available pieces of information from the medical record), simple scoring rule, and excellent screening test performance properties suggest that this new screening tool can be easily administered in the acute care setting by health clinicians. Similar to the CAGE questionnaire,²¹ which is used in triage of services for alcoholism, this tool can be used to guide clinicians in making evidence-based decisions for the allocation of scarce mental health resources for traumatic stress.

The busy acute care setting necessitated the brevity of the tool and the simple wording. As such, the STEPP should not be viewed as a comprehensive evaluation of child and parent risk factors. Many items among the initial 50 candidate items demonstrated association with the future development of posttraumatic stress symptoms and have demonstrated roles in the etiology of the disorder, but the STEPP represents the most parsimonious group of these items necessary for prediction. There has been debate about the relative contribution of early psychological vs physiological indicators for prediction of PTSD risk among adults⁴⁶; these results suggest that a combination of event-related factors, early physiological reactions (heart rate), and early psychological responses may serve well as acute predictors for child PTSD development.

The STEPP should be viewed as a triage tool to identify from among all injured children and their parents those for whom a subsequent mental health evaluation is required. During that evaluation, it would be appropriate to conduct a more comprehensive assessment of domains represented in the STEPP (such as the individual's exposure to particularly frightening or horrifying elements of the trauma, separation from loved ones during the event, and the child's preexisting behavioral functioning). It would also be appropriate to assess domains linked to PTSD risk (eg, the individual's history of exposure to other traumatic events, the impact of the parent's acute emotional responses on the child) that are not represented in the items included in the STEPP.

The high sensitivity of the STEPP with modest specificity supports its role as a screening tool, not a diagnostic test. Screening for PTSD risk factors is appropriate in the acute trauma care setting because the diagnosis of PTSD cannot be made until symptoms have persisted for at least 1 month,¹ a course well beyond the length of stay for most pediatric trauma admissions. Our research and that of others^6,9 has pointed to the high number of children and their parents with PTSD who receive no care, and lack of care has been associated with poor medical and functional outcomes.^17- 18 Therefore, while acute care clinicians cannot make a diagnosis of PTSD, it is vital that they identify those at risk for the development of PTSD in efforts to ensure that appropriate psychological care is delivered after acute medical treatment is completed.

It may not be necessary to refer all injured children and their parents for mental health services. The majority have acute symptoms, and a minority go on to develop PTSD.^6,19 Implementing evidence-based triage tools in the health care setting to identify and refer those at greater risk for psychological difficulties is consistent with both National Institute of Mental Health models for best practice in preventive care²⁰ and the surgeon general's call for integration of child mental health promotion into medical care.⁴⁷ Furthermore, recent evidence suggests that universal debriefing after trauma may be ineffective or harmful.⁴⁸ Therefore, a screening tool like the STEPP is needed to guide appropriate delivery of mental health services that ensure appropriate psychological support for injured children and their parents. The very high negative predictive value suggests that the tool would be most useful in triage, by screening out those who are least likely to be in need of further psychological intervention. Also, a positive STEPP screen appears to be predictive of later child posttraumatic and general anxiety symptoms rather than depressive symptoms, further aiding in appropriate psychological triage.

The STEPP is not a diagnostic measure of traumatic stress disorders, but rather, it can serve as a screening tool to predict risk of future and persistent posttraumatic distress. Assessment of traumatic stress disorders, whether for the purpose of estimating the burden of mental disorder associated with trauma or for clinical management, requires the use of validated tools that provide an appropriate Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition diagnosis.

The STEPP was developed in a population of children who had traffic-related injures and their parents. However, its wording is general and may apply to children with other traumatic injuries. The results have implications for other populations of children and adults who were injured, both intentionally and unintentionally; however, the results are not directly generalizable to these populations. The STEPP screening tool and the entire study were conducted in English, precluding generalizability of the results to native speakers of other languages. The next steps in research would be to validate and, possibly, refine this screening test in other patient populations and to modify and test the utility and validity of a similar screening tool for injured adults.

The items in the STEPP were administered as part of a research study by study personnel from one children's hospital in the context of a longer assessment of injured children and their parents. Further assessment of the utility of the STEPP requires testing within the context of multiple busy trauma care settings serving diverse populations. In addition, protocols using the STEPP to guide subsequent assessment and intervention will need to be developed and evaluated.

In summary, the STEPP represents a new method to alert clinicians to those patients and their parents in need of closer monitoring and supervision for PTSD. Incorporation of a tool such as the STEPP in trauma management protocols could aid in triage for the delivery of psychological care, thereby reducing the unmet psychological needs of injured patients.