Psychometric Properties of the Adulthood Trauma Inventory

Abstract

Objective:

Psychological trauma is an important public health problem, but previous measurement tools have primarily focused on childhood traumatic events while trauma exposure in adulthood (>18 years of age) has received less attention. The purpose of this study was to examine the psychometric properties of an instrument for assessment of psychological trauma in adulthood, the Adulthood Trauma Inventory (ATI).

Method:

Participants (n = 893) completed the ATI, a 33-item questionnaire modeled after the Early Trauma Inventory-Self Report (ETI-SR), assessing traumatic events occurring after 18 years of age. Participants also completed instruments to assess depression, posttraumatic stress disorder (PTSD), and early trauma (ETI-SR). Internal consistency and item response theory metrics were examined.

Results:

ATI total score (number of items endorsed) yielded the greatest internal consistency (Cronbach’s α = .77) and was significantly (p < .0001) correlated with indices of PTSD (ρ = 0.40), depression (ρ = 0.31), and early trauma (ρ = 0.56). Area under the curve and accuracy values ranged from 0.65 and 70% (depression) to 0.75 and 95% (current PTSD).

Conclusions:

The ATI is a valid measure of adult psychological trauma that may be useful for both clinical assessment and research involving the long-term effects on the individual and psychobiology.

Psychological trauma exposure is a public health risk associated with potentially severe adverse outcomes including increased rates and severity of medical and psychiatric illness, health risk behaviors, and neurobiological alterations (N. Krause, Shaw, & Cairney, 2004; Krug, Dahlberg, Mercy, Zwi, & Lozano, 2002; Oseland, Bishop, Gallus, & Randall, 2016; Schnurr & Green, 2004; World Health Organization, 2010). Psychological trauma experienced in adulthood is associated with worsened health, development of mental illness, including depression and posttraumatic stress disorder (PTSD), chronic pain, and sleep disturbances (Kendall-Tackett, 2009; N. Krause et al., 2004). The manifestations of psychological trauma on physical health vary given type of trauma, gender, medical comorbidities, and other factors (Norman et al., 2006). Specific psychological traumas and environmental hazards also contribute to deleterious health outcomes in displaced persons (Steel, Silove, Phan, & Bauman, 2002), war veterans (Bremner et al., 1999), or witnesses of genocide (Brounéus, 2010). Adverse health outcomes appear to increase with greater number of traumas experienced (N. Krause et al., 2004; Steel et al., 2002) and likely result from elevated inflammatory markers, upregulation of the hypothalamic-pituitary-adrenal-axis (HPAA), and other neurobiological factors (Kendall-Tackett, 2009; Schnurr & Green, 2004). Therefore, the phenomenology of traumatic experiences (i.e., type, age of onset, frequency, and severity of occurrence) is important in understanding the potential impact on social, physical, and behavioral health.

A number of instruments have been developed and validated for the assessment of psychological trauma (Bernstein et al., 1994; Bremner, Vermetten, & Mazure, 2000), including many that can be widely accessed from the Stress Measurement Network (SMN, stresscenter.ucsf.edu). The SMN contains the Trauma History Questionnaire (THQ), Stressful Life Events Screening Questionnaire (SLESQ), and Life Events Checklist (LEC). The THQ (Hooper, Stockton, Krupnick, & Green, 2011) is a 24-item questionnaire demonstrating construct and cultural validity, which requires endorsement and follow-up responses (description, age, and frequency) for items pertaining to general, crime-related, and physical and sexual traumas. The SLESQ (Goodman, Corcoran, Turner, Yuan, & Green, 1998) is a 13-item self-report questionnaire with good test–retest reliability and differentiates Criterion A from subthreshold events in participants with PTSD. According to the SMN, the SLESQ is best utilized for general screening and research purposes, specifically in nonclinical samples. The LEC (Gray, Litz, Hsu, & Lombardo, 2004), a 17-item questionnaire distributed with the Clinician-Administered PTSD scale (CAPS), exhibits good test–retest reliability. The LEC was developed to facilitate the diagnosis of PTSD; as such, there is no formal total score (Weathers et al., 2013) and the Questionnaire also asks follow-up questions related to the relationship to the traumatic event. In addition to the instruments in the SMN, multiple other measures of trauma have been developed (but not limited to): Trauma Life Events Questionnaire (TLEQ; Kubany et al., 2000), Life Stressor Checklist-Revised (LSC-R; Norris & Hamblen, 2004; Wolfe et al., 1996), Trauma History Screen (THS; Carlson et al., 2011), Traumatic Events Questionnaire (TEQ; Crawford, Lang, & Laffaye, 2008; Vrana & Lauterbach, 1994), Stress and Adversity Inventory for Adults (Adult STRAIN; Slavich & Shields, 2018), and Life Experiences Survey (LES; Pretorius, 1998; Sarason, Johnson, & Siegel, 1978).

In general, several of these instruments are limited by the lack of differentiation between psychological trauma exposure in childhood versus adulthood and/or incorporation of age of trauma occurrence into scoring profiles. Only the LSC-R and TLEQ include items (genital touching, intercourse) that are differentiated as occurring during or after childhood. Although these instruments appear to have ecological validity with mental health symptom severity and trauma score (Briere, Elliott, Harris, & Cotman, 1995; Carlson et al., 2011; Crawford et al., 2008), age of exposure is largely unaccounted for in the total scale score. Age of exposure is an important consideration, as traumatic events occurring between the ages of 18–64 (N. Krause et al., 2004) or most proximally (Oseland et al., 2016) have been observed to most closely associate with late-life health deteriorations. Therefore, from both research and clinical perspectives, valuable information is gained by differentiating trauma occurring before and after childhood from the different effects on physical and mental health and neurobiology (Bremner & Vermetten, 2001; Bremner, 2006; Rincón-Cortés & Sullivan, 2014).

Bremner et al. (2000) previously developed and psychometrically examined both the Early Childhood Trauma Inventory (ETI) and a subsequent self-report version (ETI-SR; Bremner, Bolus, & Mayer, 2007). The ETI assesses traumatic experiences occurring before 18 years of age and enquires about general trauma and sexual, physical, and emotional abuse. Even though the ETI has been useful in detailing traumatic experiences during childhood, there is a need to examine how traumatic experiences occurring in different neurodevelopmental epochs may impact stress responses for health outcomes and associated psychiatric conditions. The purpose of this study was to examine the psychometric properties of the Adulthood Trauma Inventory (ATI), an instrument assessing similar constructs to the ETI but delimited to events occurring in adulthood along with other stressful exposures (e.g., work as a first responder) typically only occurring in adulthood. We hypothesized that the ATI would demonstrate adequate psychometric validity for psychological trauma assessment.

Method

Study Subjects

The ATI was administered to 1,459 participants (age = 56.0 (9.0) years; 24% female) as part of the MIMS2 (Vaccarino et al., 2018), Emory Twin Study (Rooks, Veledar, Goldberg, Bremner, & Vaccarino, 2012), and Mental Stress Ischemia Prognosis Study (MIPS; Hammadah et al., 2017) cohorts. The MIMS2 sample (n = 374) included individuals aged 18–60 years with myocardial infarction occurring within the previous 8 months and equal representation of men and women. It also included age and sex matched controls without coronary artery disease. Exclusion criteria included: current alcohol or substance abuse/dependence, schizophrenia diagnosis, current usage of psychotropic medications other than antidepressants, history of traumatic brain injury, history of neurological disorder, or other serious medical condition. The MIPS sample (n = 519) included participants aged 30–79 years with stable coronary artery disease. Apart from the broader age range and no requirement for equal representation by sex, the exclusion criteria were identical to MIMS2. The Emory Twin Study sample (n = 566) included male middle-aged twins from the Vietnam Era Twin Registry (Rooks et al., 2012). All participants provided written and informed consent as approved by the Emory University Institutional Review Board.

Adulthood Trauma Inventory

All subjects completed the ATI (Table 1). Subjects from the MIMS2 and MIPS data sets (n = 893) completed the full 33-item ATI while the Emory Twin Study sample (n = 566) completed a shortened 16-item version (Table 1, starred items). The ATI consists of 33 items developed as a self-report questionnaire modeled after the previously validated early childhood trauma inventory self-report (ETI-SR; Bremner et al., 2000, 2007). The ETI-SR has previously demonstrated good agreement with other childhoodtrauma measures, strong correlations with PTSD symptomology(r = .78), and good test–retest reliability (r = .91; Bremner et al., 2000). The ATI includes questions from the domains of general, physical, and sexual trauma, while limiting item endorsement to only events occurring after the age of 18. The ATI shows some overlap with the ETI but includes items relevant to adulthood (e.g., being a combat veteran) while excluding others that are specific to childhood (e.g., did your parents or caretakers fail to understand you or your needs). The ATI can be self-administered in under 30 min and items endorsed are followed up with questions regarding frequency. Frequency was collected on a five-item Likert scale (e.g., were you ever exposed to a natural disaster: 1, 2–3, 4–5, 6–10, > 10 times).

Table 1.

Adulthood Trauma Inventory (ATI) Item Rate of Endorsement, Item-Total Correlation, Cronbach’s α, Item Difficulty, Item Discrimination, and Item Fit p Value in a Sample of 893 Adults

Item	Endorsed (%)	Cronbach’s α (minus item)	Item-total correlation	Difficulty	Discrimination	Fit p
1. Were you ever exposed to a life-threatening natural disaster?a	25.5	0.76	0.41	1.23	1.05	0.05
2. Were you ever involved in a serious accident?a	37.2	0.76	0.44	0.60	1.07	0.07
3. Did you suffer a personal injury?a	53.4	0.76	0.46	‒0.15	1.14	0.84
4. Did you experience the death of a parent or primary caretaker?a	75.3	0.77	0.3	‒2.01	0.59	0.00
5. Did you experience the death of a sibling?a	35.2	0.77	0.31	1.39	0.46	0.93
6. Did you experience the death of a friend?a	70.9	0.76	0.37	‒1.19	0.86	0.67
7. Did you observe the death or serious injury of others?a	54.2	0.76	0.46	‒0.19	1.07	0.63
8. Did anyone in your family suffer from mental or psychiatric illness or have a breakdown?a	26.1	0.77	0.33	1.72	0.66	0.13
9. Did your parents or primary caretakers have a problem with alcoholism?a	22.3	0.77	0.3	2.21	0.61	0.28
10. Were you ever the victim of assault?a	21.3	0.76	0.44	1.35	1.23	0.06
11. Did you ever work in a stressful job?a	61.7	0.76	0.38	‒0.67	0.80	0.51
12. Were you ever a prisoner of war or a hostage?a	0.6	0.77	0.1	5.03	1.17	0.12
13. Were you ever in combat?a	6.4	0.77	0.23	3.52	0.85	0.35
14. Did you ever experience the death of your biological child?a	9.3	0.77	0.2	5.17	0.46	0.61
15. If EVER MARRIED: Did you experience the death of a spouse?a	8.3	0.77	0.2	4.98	0.50	0.28
16. Did you suffer a serious illness?	62.4	0.77	0.35	‒0.85	0.65	0.36
17. Did you experience the serious injury or illness of a parent or primary caretaker?	68.8	0.76	0.4	‒1.04	0.87	0.17
18. Did you experience the separation of your parents?	23.7	0.77	0.33	1.97	0.64	0.03
19. Did you ever experience the serious injury or illness of a sibling?	35.3	0.76	0.39	0.96	0.70	0.68
20. Did you experience the serious injury of a friend?	36.3	0.76	0.59	0.59	1.22	0.38
21. Did you experience the divorce of your parents?a	18.1	0.77	0.28	2.93	0.54	0.06
22. Did you ever witness violence towards others, including family members?	41.5	0.75	0.55	0.31	1.58	0.37
23. Did your parents or primary caretakers have a problem with drug abuse?	4.6	0.77	0.18	4.68	0.69	0.66
24. Were you ever a victim of a major theft, such as having your house broken into or a family car stolen?	50.4	0.77	0.3	‒0.03	0.50	0.98
25. Were you ever the victim of armed robbery?	15.7	0.77	0.28	2.74	0.67	0.07
26. Were you ever a victim of rape?	8.7	0.77	0.24	3.79	0.67	0.75
27. Did you ever see someone murdered?	8.2	0.76	0.34	2.26	1.37	0.25
28. Did you ever see someone close to you being murdered?	2.6	0.77	0.25	2.96	1.61	0.05
29. Did you ever experience someone close to you being the victim of rape?	10.8	0.77	0.29	2.66	0.91	0.25
30. Did you ever experience the miscarriage of your child (or partner)?	25	0.77	0.26	2.81	0.40	0.37
31. Do you believe these events have an effect on you today?	40.2	0.76	0.49	0.41	1.24	0.06
32. Do you believe these events affect your functioning at work (school)?	15.1	0.76	0.34	2.02	1.01	0.13
33. Do you believe these events affect your current social and family relationships?	31.6	0.76	0.46	0.81	1.23	0.45

^aItems were included in shortened ATI.

Psychiatric and Quality of Life Evaluation

Participants completed two other trauma measurements, the perceived stress score (Cohen, Kamarck, & Mermelstein, 1983) and, within the MIPS sample (n = 519), the LES with scoring delimited to only the negative (adverse) impact score. Psychiatric diagnoses were based on the Structured Clinical Interview for DSM–IV (SCID). Depression was diagnosed from the SCID with Beck Depression Inventory total score also collected (Beck, Steer, & Brown, 1996). Additionally, the somatic and cognitive subscales of the Beck Depression Inventory were calculated in accordance with previous research (Bouman & Kok, 1987). Sleep quality was assessed using the Pittsburgh Sleep Quality Index (PSQI; Buysse, Reynolds, Monk, Berman, & Kupfer, 1989) where a greater total score denotes worse sleep. Posttraumatic stress disorder (PTSD) symptom severity was assessed using the total score from the Clinician Administered PTSD Scale (CAPS; Blake et al., 1995). PTSD diagnosis was completed in conjunction with previous guidelines (Weathers, Ruscio, & Keane, 1999) of frequency ≥1 and intensity ≥2. Early childhood trauma was assessed using the ETI-SR total score (Bremner et al., 2007; Bremner et al., 2000). Presence of early childhood trauma was defined as an ETI-SR-SF score greater than one standard deviation above the mean for healthy subjects (3.5 ± 3.3; >7; Bremner et al., 2007).

To examine quality of life, participants completed the Seattle Angina Questionnaire (SAQ; Spertus et al., 1995) and the Duke Activity Score Index (DASI; Hlatky et al., 1989). SAQ frequency subscale was calculated for each participant similar to previous studies in this population (Pimple et al., 2018, 2015) with a range of 0–100 where lower scores indicate greater angina frequency within the previous 4 months. Physical activity was assessed from the Baecke Questionnaire (Baecke, Burema, & Frijters, 1982) where higher indices indicate greater physical activity levels. Lastly, the state and trait anxiety and anger scales were also examined (Spielberger & Sydeman, 1994) where greater scores indicate greater state and trait anxiety or anger.

Statistical Analysis

Relationships between ATI score and depression, PTSD, early childhood trauma, LES, and sleep quality were analyzed using Spearman’s rank correlation coefficient that were Bonferroni corrected for multiple comparisons. Internal reliability of the ATI with both binary (yes/no) and numeric (Likert) scoring was measured via Cronbach’s α coefficient (Cronbach, 1951) from the psych package in R (cran.r-project.org/web/packages/psych). Item response theory (IRT) methods were also implemented using the ltm package in R (cran.r-project.org/web/packages/ltm). IRT allows for the estimation of individual ability level (Θ; ability in this context refers to trauma exposure where greater ability is equivalent to greater exposure) along with item difficulties and discrimination. Item difficulty is equivalent to P(Θ) = 0.5 where P(Θ) is the probability of an individual at ability level Θ endorsing a given item. Item discrimination is defined the ability of an item to differentiate between individuals above and below the item difficulty level (Baker & Kim, 2017). For this analysis, a two-parameter IRT model was employed, as this model allows for independent estimations of item difficulty and discrimination; analyses comparing Akaike Information Criterion (AIC) to one-parameter and Rasch models also supported this selection. Additionally, the two-parameter IRT model has been commonly used within psychiatry and psychology (da Rocha, Chachamovich, de Almeida Fleck, & Tennant, 2013; Rivollier et al., 2015; Toland, 2014). Within the two-parameter model, P(Θ) is estimated using the equation:

$$P\left(\Theta \right)=\frac{1}{1+e^{a\left(\Theta -b\right)}}$$

where a is the item discrimination and b is item difficulty. The two-parameter model was fit across all responses using the ltm function from the ltm package. Item information curves were calculated as: $I_j\left(\Theta \right)=a_j^2{P}_j\left(\Theta \right)Q_j\left(\Theta \right)$ where a_j is the discrimination parameter for item j, P_j(Θ) is the probability of item endorsement for item j, and Q_j(Θ) = 1 ‒ P_j(Θ). The test information was calculated as: $I\left(\Theta \right)=\sum _{j=1}^J{I}_j\left(\Theta \right)$ where J is equivalent to the number of items within the test (Baker & Kim, 2017). Test information is a measure of precision and is the reciprocal of the variance at a given Θ (Baker & Kim, 2017). Item fit statistics were calculated using the item.fit function based on separating the data into 10 groups (default for the ltm package) with p values calculated using 250 Monte Carlo replications. Within this analysis, H₀ = all items fit the model. Individual estimations were calculated using the factor.scores function within the ltm package using the model fit and recorded response patterns. Test of associations between total score, estimated theta level, and endorsement frequency were completed using Spearman’s rank correlation coefficient.

To examine diagnostic accuracy of the ATI, receiver-operator characteristic (ROC) analyses were completed from the ATI and depression, PTSD, and early childhood trauma. First, a logistical regression model was fit to the independent (binary psychological classification) and dependent variable (total ATI score) using leave 10 out cross-validation. Using the logistical regression model, independent variable predictions were generated for the 10 participants left out and applied to a new dataset. Second, a determination threshold was chosen based on predicted “cost” (i.e., more errors = greater cost) with false positives and negatives given equal weight. Each participant score was converted into a predicted binary classification of either presence or absence of psychiatric condition if value was above/below the threshold, respectively. Subsequent sensitivity (True Positives – (True Positives + False Negatives)), specificity (True Negatives – (True Negatives + False Positives)), and accuracy ((True Positives + True Negatives)/(True Positives + False Positives + True Negatives + False Negatives)) metrics were computed. Area under the ROC curve was computed using the auc function within the psych package and visualized (pROC; cran.r-project.org/web/packages/pROC).

Sex differences were assessed using the Wilcox Rank Sum test. All analyses were completed in base R (www.r-project.org) unless otherwise specified. The level of significance was set at p < .05. Data are presented as median (IQR).

Results

Participant Demographics

Internal consistency comparisons indicated that the short ATI (Cronbach’s α = .58) was inferior to the complete ATI (α = .77) and, therefore, additional analysis was restricted to the full inventory (n = 894). Within this sample median age was 57 (15) years, 37.2% were women, 41.7% of participants were Black, median years of education completed was 14.0 (4.0), median number of caffeinated drinks or foods consumed in a day was 3.0 (3), and median number of alcoholic beverages per week was 1.0 (4.0). Presence of ≥1 psychiatric condition (depression, PTSD) occurred in 31.6% of participants, with 29.1% having experienced depression (current depression: 10.3%), 8.4% experienced PTSD (current PTSD: 5.5%), and 5.9% experienced both depression and PTSD. Women experienced both lifetime (women = 36.0, men = 25.0%; p = .001) and current depression (13.6, 8.2%; p = .01) more than men (25.0%, 8.2%) but lifetime (8.8, 8.2%; p = .87) and current PTSD (5.5, 5.6%; p = .99) were not different. Early childhood trauma was experienced in 45.5% of participants but no sex differences were observed (women = 42.0, men = 47.4%; p = .14).

ATI Scoring

Table 1 presents the frequency of endorsement, Cronbach’s α, and the item-total correlation for all ATI items. Frequency of endorsement ranged from 0.6% (prisoner of war) to 75.2% (death of a parent or primary caretaker). Cronbach’s α for the entire scale was 0.77 (95% confidence interval, CI [0.75, 0.79]) with individual items ranging from 0.75 to 0.79. The mean item-total correlation was 0.30 and ranged from 0.10 to 0.55. Cronbach’s α for the ATI measured with total frequency of occurrence (sum of Likert scale values) was 0.70 (95% CI [0.68, 0.73]) with individual items ranging from 0.67 to 0.72. Given the greater internal consistency of binary responses, ATI total score was computed from number of items endorsed.

Item difficulty, discrimination, and fit p values are presented in Table 1. Only three of 33 items (#4, 18, 28; 9.1%) did not fit the IRT model (p < .05). One item, “Did you ever experience the separation of your parents?” aligns closely with “Did you ever experience the divorce of your parents?” However, combing these items also resulted in an item rejecting the null (p = .02) and therefore no changes were made to the ATI. Figure 1 presents the frequency of participant (top) and item (bottom) Θ values. Both participant and item theta distributions were approximately similar, indicating an appropriately targeted scale. Figure 1B presents the test information curve and standard error. The maximum level of information occurred near Θ = 0, indicating most information was gathered from participants with the median trauma level. Individual item characteristic and information curves are available in online Supplemental Materials Figures 1 and 2. As expected, item difficulty was significantly associated with endorsement frequency (ρ = 0.95, p < .0001, Figure 1C). Lastly, ATI total score was significantly associated with estimated Θ (ρ = 0.98, p < .0001, Figure 1D); total score was, therefore, deemed a valid scoring measure for the ATI.

Figure 1.

Item response theory results. (A) Frequency histogram of participant (top) and item (bottom) responses across levels of theta (difficulty). (B) Test information curve for all 33 items. (C) Relationship between item endorsement rate and item difficulty. (D) Relationship between estimated theta and total score (out of 33) for all participants.

Convergent Validity

Table 2 presents the relationship between total ATI score and select psychiatric conditions and potential risk factors. ATI scores were positively corelated (p < .0001) with measures of trauma exposure, LES negative impact scores and PSS total score. Participants with a history of depression, either lifetime or current, had a significantly greater ATI score than participants without (p < .0001 for both). Participants with PTSD, either lifetime or current, also had greater ATI scores than those without (p < .0001 for both). Participants with early childhood trauma reported elevated ATI scores (p < .0001); greater ATI scores were also observed for each ETI subscale (p < .001). ATI total score was positively associated with depression severity (including the somatic and cognitive subscales; p < .0001), PTSD severity (p < .0001), early childhood trauma (p < .001), sleep quality (p < .001), and an inverse correlation was observed with education years completed (ρ = ‒0.18, p = .0005). In addition, more frequent angina (p < .0001) and lower DASI score (p = .02) were significantly associated with ATI total score. State and trait anxiety also exhibited positive associations with ATI total score (p < .001).

Table 2.

Mean (IQR) Adulthood Trauma Inventory Scores for Participants With or Without Depression, Posttraumatic Stress Disorder (PTSD), and Early Childhood Trauma

Psychiatric or risk factor measure	Absence	Presence	p value	Correlation	p value
Psychiatric data
Depression	9.0 (6.0)	12.0 (7.0)	<.0001	ρ = .31a	p < .0001
Depression (current)	9.0 (6.0)	14.0 (7.3)	<.0001
Somatic subscale				ρ = .33	p < .0001
Cognitive subscale				ρ = .30	p < .0001
PTSD	9.0 (6.0)	14.0 (7.5)	<.0001	ρ = .40b	p < .0001
PTSD (current)	9.0 (6.0)	14.5 (5.3)	<.0001
PSQI				ρ = .35	p < .0001
Early childhood trauma	8.0 (4.3)	12.0 (6.0)	<.0001	ρ = .56	p < .0001
General trauma	7.0 (5.0)	10 (7.0)	<.0001	ρ = .53	p < .0001
Physical trauma	8.0 (5.0)	11.0 (7.0)	<.0001	ρ = .38	p < .0001
Emotional trauma	8.0 (6.0)	11.0 (7.0)	<.0001	ρ = .34	p < .0001
Sexual trauma	9.0 (6.0)	11.0 (6.0)	<.0001	ρ = .26	p < .0001
Life experiences surveyc				ρ = .22	p < .0001
State anxiety				ρ = .17	p < .0001
Trait anxiety				ρ = .22	p < .0001
State anger				ρ = .09	p = .09
Trait anger				ρ = .21	p < .0001
Perceived stress score				ρ = .24	p < .0001
Health outcomes data
Duke Activity Status Index				ρ = ‒.08	p = .43
Angina frequency				ρ = ‒.24	p < .0001
Baecke total scored				ρ = ‒.07	p = .99

Note. Spearman’s rank correlation coefficient between ATI total score and total scores for depression, PTSD, early childhood trauma, sleep quality from the Pittsburgh Sleep Quality Index (PSQI), and negative (adverse impact) scores for the life experiences survey are also presented. IQR = interquartile range.

^aSeverity measured using the Beck Depression Inventory Total Score.

^bSeverity measured using the Clinician Administered PTSD Scale (CAPS).

^cData from Mental Stress Ischemia Prognosis Study (n = 519).

^dData from MIMS2 dataset only (n = 387).

Figure 2 presents the ROC curves for the association of depression, PTSD, current PTSD, and early childhood trauma using the ATI total score. For depression, model accuracy was 0.70 with a sensitivity of 0.37, specificity of 0.83, and area under the curve of 0.65. For lifetime PTSD, accuracy was 0.92 with a sensitivity of 0.13, specificity of 0.99, and area under the curve of 0.73. For current PTSD, accuracy was 0.95 with a sensitivity of 0.10, specificity of 0.99, and area under the curve of 0.75. Lastly, for early childhood trauma, accuracy was 0.69 with a specificity of 0.58, sensitivity of 0.83, and area under the curve of 0.77.

Figure 2.

Receiver operator characteristic curves for the association of depression, posttraumatic stress disorder (PTSD), current PTSD, and early childhood trauma (ETI) by the Adulthood Trauma Inventory total score. TPR = true positive rate (True positives/(True Positives + False Negatives)); FPR = false positive rate (False Positives/(False Positives + True Negatives)).

Divergent Validity

To examine divergent validity of the ATI, data regarding behavioral outcomes (physical activity) and other psychological states (anger) were assessed. In a subset (MIMS2), Baecke total score was not significantly associated with ATI total score (p = .20); however, the subscales of work-related (ρ = ‒0.14, p = .007) and sport-related activity (ρ = ‒0.11, p = .03) were inversely associated. State anger was also not associated with ATI total score (p = .09) despite participants general level of anger, trait anger, being significantly associated (p < .0001).

Discussion

The results presented indicate the ATI is a valid instrument for adult psychological trauma measurement. The ATI demonstrated good levels of internal consistency, item difficulty levels which coincided with response frequencies, convergent validity with other items of trauma, and divergent validity for measurements not involved with trauma exposure such as state anger and activity status. Additionally, ATI score was correlated with years of education, a known protective factor for the effects of psychological trauma (Brewin, Andrews, & Valentine, 2000). Comparison of alternate scoring methods, including severity/magnitude of trauma and truncated item count yielded poorer internal consistency. The ATI, using only item endorsements, is a valid measure of adult psychological trauma and complements assessments of early trauma.

The primary finding of this study is the acceptable internal validity and constructs validity of the ATI. Cronbach’s alpha across all ATI items demonstrated acceptable internal consistency (α > .7), similar to the ETI-SR (α = .78–.90; Bremner et al., 2007). Furthermore, item-total correlations were of similar range (~0.3) with combat exposure being lowest for both scales, possibly related to low endorsement frequency. Death of a spouse also had a low item-total correlation, which on face value may be because of this event representing a random occurrence versus a causal association with other items but could also result from the rare endorsement frequency (6%). Similar to the ETI-SR (Bremner et al., 2007), utilizing quantitative descriptors of frequency or severity for endorsed items did not improve ATI performance. Most trauma inventories (Goodman et al., 1998; Hooper et al., 2011; Kubany et al., 2000) include subsequent prompts capturing details for endorsed items. The contribution of frequency or severity items to the psychometric properties of these scales is unclear given their open-ended nature and/or omission from the validation. However, additional information related to traumatic experiences, specifically the age during exposure, may provide valuable information regarding the impact on the individual, potentially altering psychiatric symptomology (van der Kolk, Roth, Pelcovitz, Sunday, & Spinazzola, 2005). To address this, the ATI is designed to mirror early childhood trauma inventories and comprehensively examine multiple dimensions of psychiatric trauma across the life span.

The IRT analysis demonstrated a good fit using the two-parameter model. The similar distributions of estimated participant Θ and item difficulty indicate the questions were appropriately targeted to the participants. The test information curve, a measure of precision (Baker & Kim, 2017), also followed the general participant distribution. Therefore, the ATI provided better estimations at levels with more participants; this characteristic allows for a maximum number of correctly categorized individuals. The strong associations between item endorsement rate and item difficulty along with estimated trauma severity and ATI total score further support the suitability of the ATI to adequately measure a wide range of adulthood trauma exposure. Furthermore, only three items did not satisfy the item fit criteria when attempting to group participants along a continuum of trauma severity which supports the ATI as being conducive to quantifying adult trauma exposure (Toland, 2014). The low difficulty items (#4, 6, 17, Table 1) could be considered occurring as a programed life event with a sense of inevitability (N. Krause et al., 2004) and, as such, may not contribute to future traumas. We recognize the cohort assessed in the current study is potentially older than the most at-risk demographics for neurobiological effects of trauma exposure (N. Krause et al., 2004). Future studies should aim to assess the ATI within younger cohorts such as young adults (18–25 years).

In addition to demonstrating good internal validity, the ATI also fulfilled the requirement of convergent validity. ATI total score was positively associated the negative event subscale of the LES, a validated measure of traumatic experiences (Sarason et al., 1978). The LES presents a greater number of potentially traumatic events, including standalone events (e.g., “marriage”), which may explain the lower correlation than may be expected with a similar measure. ATI total score was also positively correlated to the PSS. The PSS has also been previously associated with another measure of adult trauma, the Adult STRAIN (Slavich & Shields, 2018), to a similar magnitude. The previous authors indicated the weak association between the PSS and adults STRAIN is likely because of less discriminant validity (Slavich & Shields, 2018). ATI scores were significantly greater for participants with PTSD, depression, and early childhood trauma along with being positively associated with symptom severity. The area under the curve values of ≥0.68 all exceed a “medium” effect with three of the four variables yielding “large” effects when compared with an equivalent Cohen’s d (Rice & Harris, 2005). The ATI exhibited positive associations with some psychiatric conditions (66 to 93%) but the ability to identify absence of the condition (specificity) far exceeded the performance to determine presence (sensitivity). However, the ATI is not a diagnostic measure and performs similarly to other validated measures of psychological trauma (Crawford et al., 2008; Hooper et al., 2011; Kubany et al., 2000). It is not surprising that ATI total score correlated with severity of PTSD and depression as other trauma measures routinely demonstrate this relationship (Goodman et al., 1998; Hooper et al., 2011; Kubany et al., 2000; Norris & Hamblen, 2004) and these psychiatric conditions likely result from adverse biological and psychological adaptions after trauma exposure (Schnurr & Green, 2004). A greater presence of psychiatric conditions with greater trauma exposure may also be explained through other mechanisms such as breakdown of “resource caravans” (Hobfoll, 2014), or coordinated resources (e.g., family) aimed at preserving domains of: safety, calmness, attachment, hope, and efficacy. Individual traumatic events, such as death of a sibling, may result in a cascade of resource losses within these domains and leaving the individual susceptible to further traumas (Hobfoll, 2014) and worsened quality of life as assessed by more associations with greater angina frequency, less functional capacity, and greater somatic subscale scores of the Beck Depression Inventory. These findings suggest a potential utility for measurement in behavioral medicine settings.

Few trauma indices have examined the convergent validity between early childhood and adult trauma. First, many scales of trauma include early childhood events (Carlson et al., 2011; Goodman et al., 1998; Kubany et al., 2000) that preclude comparison. However, differentiating between developmental epochs is important as psychological trauma occurring at different developmental stages is known to have varying effects on neurobiology (Andersen et al., 2008; Bremner, 2006; Pechtel, Lyons-Ruth, Anderson, & Teicher, 2014; Rincón-Cortés & Sullivan, 2014; Teicher et al., 2003). Early childhood trauma also is predictive of adult trauma (Söchting, Fairbrother, & Koch, 2004), especially for sexual abuse (Krug et al., 2002; Messman-Moore & Long, 2003; World Health Organization, 2010). Early childhood trauma also appears to increase risk for and symptom complexity of PTSD (Bremner, Southwick, Johnson, Yehuda, & Charney, 1993; Briere, Kaltman, & Green, 2008) and other disorders of extreme stress (van der Kolk et al., 2005), potentially increasing treatment-seeking behavior. In contrast, trauma experienced in adulthood is more related to adverse health outcomes (A. J. Krause et al., 2017; Oseland et al., 2016).

The ATI exhibits important differences compared with existing measures of previous trauma exposure in the determination of general adult traumatic load. Of the trauma measures within the SMN, none of which delimit exposure timeframe to adulthood, the ATI appears most closely associated with the THQ (Hooper et al., 2011). However, the ATI enquires about combat exposure (prisoner of war or combat) that, with difficulty values of Θ = 5.0 and 3.5, respectively, are important for classifying individuals exposed to high levels of previous trauma and the potential neurobiological consequences. This limitation is also present the SLESQ which, with only 13 items, does not assess many high-difficulty traumatic events. The SLESQ enquires for more follow-up information compared with the ATI (Goodman et al., 1998), although the psychometric properties of that information are unclear. The LEC is targeted toward clinical populations without an agreed-upon measure of trauma exposure (Weathers et al., 2013) and, thus, is not likely suitable for similar assessment protocols as the ATI. Lastly, one measure of adult trauma exposure (Adult STRAIN; Slavich & Shields, 2018) also appears similar to the ATI; however, data for individual items have not been published that precludes further comparison.

We recognize limitations of the current study and ATI. First, the main limitation is the retrospective nature of self-reported event recall required for the ATI. While limitation is present in many trauma questionnaires, we cannot exclude the potential influence on the results. As others have demonstrated (Roemer, Litz, Orsillo, Ehlich, & Friedman, 1998), reporting of traumatic events is susceptible to error such that individuals with greater trauma exposure may disproportionately endorse items. Other sources of trauma recall bias have been presented previously such as the nature of item and psychological state during event recall. Sexual abuse victims may develop a selective amnesia as a defense mechanism regarding the event (Maughan & Rutter, 1997) that may depress endorsement rate. Psychological state during recall may alter willingness and ability to endorse specific items; however, previous studies have also observed similar endorsement rates with altered psychological state (Paivio, 2001). The presence of recall bias also supports the usage of binary item endorsement to create an index of trauma exposure as potential error may increase with more information requested from participant (e.g., age, event frequency; N. Krause et al., 2004). Second, the subject sample was drawn from a clinical setting in the United States. Individuals in other countries may experience greater rates of different traumas such as natural disasters (Mason et al., 2014) or genocide (Brounéus, 2010) and additional items or translation may be required. Population-based studies may also be required with the ATI to establish cultural validity (Hooper et al., 2011).

Conclusions

The ATI is a valid measure of adult psychological trauma. The ATI exhibits convergent validity with measures of PTSD, depression, and early childhood trauma. The ATI, when paired with other measures of childhood trauma, allows for systemic investigation into traumatic exposure occurrence within two distinct development epochs, childhood and adulthood. Therefore, the ATI is an instrument that can be used by clinicians and researchers to understand general traumatic load on an individual that could have potential neurobiological and psychological consequences.