Peritraumatic Distress Inventory as a predictor of post-traumatic stress disorder after a severe motor vehicle accident
Abstract
Aim: The aim of this study was to examine the utility of the Peritraumatic Distress Inventory (PDI) as a predictor of subsequent post-traumatic stress disorder (PTSD) in severe motor vehicle accident survivors.
Methods: Patients consecutively admitted to the intensive care unit were assessed immediately and 1 month after accidents in this prospective study. The predictive value for post-traumatic stress symptoms at 1 month of the PDI at initial assessment was examined by using multivariate regression analysis. Moreover, the accuracy of the PDI as a predictor of PTSD was determined using receiver operator characteristic curve analysis. Post-traumatic stress symptoms were assessed using the Impact of Event Scale – Revised questionnaire, and PTSD was assessed using the Clinician-Administered PTSD Scale.
Results: Seventy-nine patients completed the Impact of Event Scale – Revised questionnaire, and 64 patients participated in a structured interview. Of 64 patients, 13 met the diagnostic criteria of full or partial PTSD. The PDI was an independent predictor of post-traumatic stress symptoms (P = 0.003). The data indicated that a cut-off score of 23 maximized the balance between sensitivity (77%) and specificity (82%) in this study. Compared with negative predictive value (93%), positive predictive value was not high (53%).
Conclusion: The study suggests the predictive usefulness of the PDI for subsequent PTSD in accident survivors. Its adequate usage should be further elaborated.
MORE THAN 50 million people each year are injured in the road traffic system worldwide,1 and nearly one-third of the injured patients appear to develop trauma-related psychiatric illnesses such as post-traumatic stress disorder (PTSD).2–7 PTSD has been associated with higher psychiatric comorbidity, attempted suicide, physical illnesses such as asthma, hypertension, and peptic ulcer,8 as well as high health-care costs.9,10 Although acute stress disorder has been shown to be a predictor of chronic PTSD in motor vehicle accident (MVA) survivors,11 it is very difficult to arrive at a psychiatric diagnosis based on an interview following the accident, given the large number of MVA that occur and the limited psychiatric resources available in intensive care units (ICU). Thus, it would be ideal to find predictors of PTSD that can be easily assessed, to provide effective secondary preventive strategies to individuals at high risk.
Brunet et al. proposed the concept of peritraumatic distress and conceptualized the Peritraumatic Distress Inventory (PDI) in 2001, which is a concise self-report questionnaire, assessing not only a threat but various emotional responses experienced during and immediately after a critical incident.12 A cross-sectional study suggested that the PDI was promising as a predictor of PTSD.13 However, the results of previous prospective studies using the PDI were partly inconsistent14–16 due to the delay of the timing when the PDI was assessed, or due to the lack of a predictive model based on theoretical considerations.
Elucidating the utility of the PDI as a predictor of PTSD appears important for effectively assessing individuals at high risk. The aims of this study are to examine the utility of the PDI as a predictor of subsequent PTSD in MVA survivors by addressing methodological problems mentioned above.
METHODS
Participants
Participants were selected from the Tachikawa cohort of motor vehicle accidents (TCOM) study conducted at the National Disaster Medical Center (NDMC) in Tokyo, Japan.17 In the center, about 10% of new patients requiring admission have been shown to have severe physical injuries resulting from MVA.18 The inclusion criteria in the present study were as follows: (i) MVA-related severe physical injury causing a life-threatening or critical condition; (ii) age between 18 and 69 years; and (iii) native Japanese-speaking ability. The exclusion criteria were as follows: (i) diffuse axonal injuries, brain contusion, and subdural and subarachnoidal bleeding detected by computed tomography and/or magnetic resonance imaging (with the exception of concussion), because the presence of traumatic brain injury creates considerable difficulties when assessing psychological responses to injury; (ii) cognitive impairment, defined as a score of <24 on the Mini Mental State Examination (MMSE); (iii) currently suffering from schizophrenia, bipolar disorder, drug dependence or abuse, or epilepsy before the MVA; (iv) marked serious symptoms such as suicidal ideation, self-harm behavior, dissociation, or a severe physical condition preventing the patient from tolerating the interview; and (v) living or working at a location more than 40 km from the NDMC.
Patients with MVA-related physical injury in the present study were consecutively admitted to the ICU of the NDMC between 18 August 2005 and 8 January 2008. Of the 221 patients who met the criteria, 189 agreed to participate in the study. Fifty-nine patients were excluded because we could not assess their peritraumatic distress due to memory loss during the MVA. Thus, there were a total of 130 patients who participated. The study protocol was approved by the institutional review board and ethics committee of the NDMC. After providing a complete description of the study to the subjects, written informed consent was obtained.
Procedures
The median number of days between the MVA and the initial assessment was 2 (range, 0–23 days). The initial assessment was conducted after assessments of cognitive function using the MMSE performed by trained research nurses or trained psychiatrists. The PDI is a 13-item self-report questionnaire measuring distress experienced during and immediately after a critical incident (range, 0–52)12 (Table 1) and it was used at initial assessment. The response format is a five-point Likert scale that ranges from 0 to 4 (0 = not at all, 1 = slightly true, 2 = somewhat true, 3 = very true and 4 = extremely true). It should take just several minutes to complete all items of the PDI. We developed the Japanese version of the PDI in cooperation with the original developers, and demonstrated that the Japanese version has good test–retest reliability (test–retest correlation coefficient, 0.61), a high degree of internal consistency (Cronbach's alpha coefficient, 0.83), and a high degree of concurrent validity.19
| 1. I felt helpless to do more |
| 2. I felt sadness and grief |
| 3. I felt frustrated or angry I could not do more |
| 4. I felt afraid for my safety |
| 5. I felt guilt that more was not done |
| 6. I felt ashamed of my emotional reactions |
| 7. I felt worried about the safety of others |
| 8. I had the feeling I was about to lose control of my emotions |
| 9. I had difficulty controlling my bowel and bladder |
| 10. I was horrified by what happened |
| 11. I had physical reactions like sweating, shaking, and pounding heart |
| 12. I felt I might pass out |
| 13. I felt I might die |
Follow-up assessments were performed for an average of 37 days (range, 24–76 days) post-accident. The participants were asked to visit the NDMC or to return their completed self-report questionnaire in a stamped, addressed envelope. After each assessment, all participants were given a gift voucher for their participation (1000 JPY [10 USD]). Seventy-nine (60.8%) of 130 patients completed the self-report questionnaire and only 64 (49.2%) participated in the structured interview.
Predictive value for post-traumatic symptoms
The predictive value of the PDI for post-traumatic stress symptoms assessed using the self-report questionnaire was evaluated using a model with other covariates. To establish a model for predicting outcome, we selected potential variables as covariates based on the following theoretical considerations. As for covariates, age at MVA, being female, history of psychiatric illness, family history of psychopathology, and lower education level are well-established pre-traumatic risk factors across trauma type.20,21 As for educational level, we used graduation from junior high school as a reference (0), and we assigned 1 to graduation from high school, 2 to graduation from junior or technical college, and 3 to graduation from university or higher educational institutions according to the Japanese educational system. Heart rate at admission was selected because some reports in the medical literature about MVA showed its association with PTSD.2,22,23 Injury Severity Score (ISS) per 10 points was assigned as the objective accident-related variable. ISS is a scoring system that provides a total score for patients with multiple injuries, and it correlates with measures of severity such as mortality and hospital stay.24
Post-traumatic stress symptoms as assessed using the Impact of Event Scale – Revised questionnaire (IES-R) at follow-up were considered as the outcome. The IES-R is a 22-item self-report questionnaire for assessing subjective distress caused by a specific traumatic stressor in the past 1 week (range, 0–88).25,26
Furthermore, the predictive value of the each item of the PDI for post-traumatic symptoms was examined for reference.
Accuracy at predicting a PTSD diagnosis
The accuracy of the PDI at predicting a PTSD diagnosis assessed using a structured interview was evaluated using receiver–operator characteristic (ROC) curve analysis.
Trained psychiatrists (D. N., Y. M.) conducted the follow-up face-to-face assessments. MVA-related PTSD was diagnosed using the Clinician-Administered PTSD Scale (CAPS).27 Participants were deemed to have partial PTSD if they fulfilled only two out of the three symptom criteria (B [re-experiencing], C [avoidance], D [hyperarousal]), and criteria A-1 (stressor), E (duration), and F (impairment) of the DSM-IV-Text Revision.28 Marshall et al. showed that the presence of subthreshold PTSD symptoms significantly raised the risk for suicidal ideation and insisted that more efforts are needed to identify subthreshold PTSD symptoms in clinical populations, epidemiological surveys, and treatment studies.29 From this perspective, many recent studies regarding MVA adopted the diagnostic criteria of partial PTSD.
The utility of the PDI as a predictor of PTSD syndrome (full PTSD and partial PTSD) at 1 month were determined using ROC curve analyses. The optimum cut-off score for the PDI was selected based on the a priori decision to maximize in the ROC curve.
Statistical method
Univariate regression analysis was employed to examine the relationship between the total score and each item of the PDI and post-traumatic symptoms. In a model for the predictive value of the PDI, multivariate regression analysis was employed to examine the relationship between the PDI and post-traumatic stress symptoms adjusted for seven other covariates. Any association with independent variables and covariates were regression coefficient (beta) and they were quantified using 95% confidence intervals (95%CI).
In ROC curve analyses for the accuracy of the PDI, we calculated sensitivity, specificity, positive and negative predictive values, and area under the curve (AUC).
All statistical analyses used two-tailed tests. Statistical significance was established at P-value < 0.05. All data analyses were performed using spss statistical software version 14.0J for Windows (spss, Tokyo, Japan).
RESULTS
Predictive value for post-traumatic symptoms
Demographic, medical and psychiatric characteristics are shown in Table 2. Of the 130 participants, 79 (60.8%) completed the IES-R at follow up and 51 refused to participate. The patients who dropped out of the study did not differ significantly from those who participated in terms of variables selected in this study, including total score on the PDI. The relationship between the total score and the each item of the PDI and post-traumatic stress symptoms are shown in Table 3. The PDI was an independent predictor of post-traumatic stress symptoms after adjusting for potential confounders (Table 4).
| Variables | n | % | Mean | SD | Median | Range |
|---|---|---|---|---|---|---|
| Peritraumatic Distress Inventory | – | – | – | – | 15.0 | 0–40 |
| Covariates | ||||||
| Age | – | – | 39.8 | 15.9 | 37.0 | 18–69 |
| Sex, female | 16 | 20.3 | – | – | – | – |
| Family history of psychopathology, yes | 13 | 16.5 | – | – | – | – |
| History of psychiatric illness, yes | 7 | 8.9 | – | – | – | – |
| Highest level of education | ||||||
| Junior high school | 18 | 22.8 | – | – | – | – |
| High school | 27 | 34.2 | – | – | – | – |
| Junior or technical college | 17 | 21.5 | – | – | – | – |
| University or more | 17 | 21.5 | – | – | – | – |
| Heart rate at admission, b.p.m. | – | – | 84.2 | 15.6 | – | – |
| Injury Severity Score | – | – | – | – | 6.0 | 1–41 |
| Outcome | ||||||
| IES-R at follow up | – | – | – | – | 11.0 | 0–53 |
- IES-R, Impact of Event Scale – Revised.
| PDI item | Beta | 95% CI | P-value |
|---|---|---|---|
| 1 | 4.00 | 2.05–5.94 | <0.001 |
| 2 | 3.05 | 1.03–5.06 | 0.004 |
| 3 | 2.99 | 1.11–4.87 | 0.002 |
| 4 | 3.02 | 1.04–5.00 | 0.003 |
| 5 | 1.43 | −0.97–3.83 | 0.24 |
| 6 | 0.76 | −2.47–3.99 | 0.64 |
| 7 | 1.40 | −0.66–3.46 | 0.18 |
| 8 | 2.31 | −0.82–5.45 | 0.15 |
| 9 | 1.39 | −3.89–6.67 | 0.60 |
| 10 | 3.41 | 1.46–5.37 | 0.001 |
| 11 | 4.04 | 2.14–5.93 | <0.001 |
| 12 | 0.90 | −1.12–2.92 | 0.38 |
| 13 | 2.63 | 0.74–4.52 | 0.007 |
| Total | 0.61 | 0.34–0.89 | <0.001 |
- The each item of the PDI ranges from 0 to 4, and the total score of the PDI ranges from 0 to 52.
- CI, confidence intervals; IES-R, Impact of Event Scale – Revised.
| Univariate regression | Multivariate regression (R2 = 0.343) | |||||
|---|---|---|---|---|---|---|
| Beta | 95% CI | P-value | Beta | 95% CI | P-value | |
| PDI per 1 point | 0.61 | 0.34–0.89 | <0.001 | 0.49 | 0.18–0.80 | 0.003 |
| Covariates | ||||||
| Age per 10 years† | 2.81 | 0.89–4.73 | 0.005 | 1.63 | −0.48–3.75 | 0.13 |
| Women | 13.1 | 5.70–20.4 | 0.001 | 7.64 | −1.31–16.6 | 0.09 |
| Family history of psychopathology, yes | −2.57 | −11.3–6.14 | 0.56 | −3.14 | −11.0–4.67 | 0.43 |
| History of psychiatric illness, yes | 0.72 | −10.5–11.9 | 0.90 | 6.69 | −4.88–18.3 | 0.25 |
| Education | ||||||
| 0 (junior high school) | (Reference) | |||||
| 1 (high school) | 1.83 | −4.89–8.55 | 0.59 | 0.78 | −7.30–8.87 | 0.85 |
| 2 (junior or technical college) | 1.17 | −6.60–8.93 | 0.77 | −1.19 | −10.2–7.80 | 0.79 |
| 3 (university or more) | −5.28 | −13.0–2.39 | 0.18 | −4.06 | −13.1–4.98 | 0.37 |
| Heart rate‡ per 10 b.p.m. | 0.31 | −1.75–2.37 | 0.77 | 1.26 | −0.67–3.19 | 0.20 |
| Injury Severity Score§ per 10 points | −0.21 | −2.98–2.56 | 0.88 | −1.15 | −3.81–1.51 | 0.39 |
- † Entered as age divided by 10;
- ‡ Entered as 50–59 b.p.m., 0; 60–69 b.p.m., 1; 70–79 b.p.m., 2; 80–89 b.p.m., 3; 90–99 b.p.m., 4; 100–109 b.p.m., 5; 110–119 b.p.m., 6; 120–129 b.p.m., 7; 130–139 b.p.m., 8; 140–149 b.p.m., 9.
- § entered as 1–9 points, 0; 10–19 points, 1; 20–29 points, 2; 30–39 points, 3; 40–49 points, 4.
- R2, multiple correlation coefficient, index of goodness of fit in the model.
- CI, confidence intervals; IES-R, the Impact of Event Scale – Revised; PDI, the Peritraumatic Distress Inventory.
Accuracy at predicting a PTSD diagnosis
Sixty-four participants (49.2%) completed the CAPS. The participants who dropped out of the study did not differ significantly in terms of age, gender and a total score on the PDI. Of 64 participants, 50 were men and 14 were women, and only one man (60 years old, ISS = 14) did not complete the IES-R. Average age was 40.5 (SD = 16.3). Five participants (7.8%) met the diagnostic criteria of full PTSD, and eight (12.5%) met that of partial PTSD at 1 month after MVA.
ROC curve analyses are presented in Figure 1. The AUC is 0.83, and the optimum predictive cut-off point of the PDI was a score of ≥23. Sensitivity, specificity, positive and negative predictive values were 77%, 82%, 53% and 93%, respectively (Table 5).
Receiver–operator characteristic curve for the (
) Peritraumatic Distress Inventory corresponding to post-traumatic stress disorder at 1 month. (—) reference line.
| The PDI result | PTSD syndrome at 1 month after MVA | |
|---|---|---|
| Negative | Positive | |
| Negative, n (%) | 42 (82.4) | 3 (23.1) |
| Positive, n (%) | 9 (17.6) | 10 (76.9) |
| Value (95% confidence intervals) | ||
| Sensitivity | 0.77 (0.54–1.00) | |
| Specificity | 0.82 (0.72–0.93) | |
| Positive predictive value | 0.53 (0.30–0.75) | |
| Negative predictive value | 0.93 (0.86–1.01) | |
- MVA, motor vehicle accident; PDI, Peritraumatic Distress Inventory; PTSD, post-traumatic stress disorder.
DISCUSSION
This study showed that the PDI could predict post-traumatic stress symptoms and PTSD at follow up in MVA survivors. The predictive value of the PDI for the IES-R remained after adjusting covariates in a multivariate regression analysis. Sensitivity, specificity and AUC were also moderately high.
Some previous prospective studies did not show that PDI is an independent predictor of PTSD. These studies assessed the PDI from 2 weeks16 to several months14,15 following a traumatic event. The time of assessment in this study was several days following the traumatic event in most participants in order to minimize the effects of inaccurate memory over time. It is suggested that the PDI has a better predictive value when used early after a traumatic event.
AUC of the PDI in this study was comparable to AUC of the IES-R (0.76) in a previous larger study.30 The IES-R is one of the most frequently used tools for measuring post-traumatic stress symptoms, but the PDI can be used immediately after MVA whereas the IES-R was intended to assess PTSD symptoms over the previous 7 days. Thus, the PDI may be a better alternative to predict subsequent PTSD in ICU. The potential disadvantage of the PDI is that peritraumatic distress is difficult to assess when patients do not have a memory of the MVA.
Unfortunately, the positive predictive value of the PDI was not high. This is consistent with previous studies showing that the absence of peritraumatic distress is a strong indicator of the absence of PTSD, while the presence of peritraumatic distress is a weak indicator of the presence of PTSD.31,32 As Weathers and Keane previously suggested,33 the potential use of the PDI might be as an early identification of MVA survivors who are unlikely to develop PTSD so that resources can be directed to others who are at greater risk.
Aside from the PDI and the IES-R, various variables have been investigated as predictors for PTSD so far. For example, high heart rate shortly after MVA has been considered a predictor of later PTSD. Although some studies report a significant positive correlation between heart rate and post-traumatic stress symptoms,2,22,23 a recent review indicated that heart rate cannot be accurately used to identify individuals who are at high risk for later PTSD.34 Much the same is true with regards to peritraumatic dissociation. While peritraumatic dissociation had been once considered to be a strong predictor of PTSD,21,35 a recent review showed no general consensus that it qualifies as an important independent predictor of PTSD symptomatology.36 Recently we have reported that a sense of life threat experienced during and immediately after an MVA was a predictor of PTSD,7 but a previous study on the 2004 Tsunami survivors showed that sensitivity of a sense of life threat was very low (27%).37 Taking these findings into consideration, their flexible applications in diverse situations seem to be important for the effective assessment of patients.
This study has several limitations. First, the small sample size can limit the relevance of the study. Second, the attrition rate was relatively high in this study, even though the patients who dropped out were not significantly different from those who participated in the follow-up assessment in terms of the PDI and other covariates. In an earlier publication, we revealed that significant predictors of dropout were: male subjects, unconsciousness during MVA, low cooperativeness, and less severe injuries.38 Participants with less severe injuries did not need to come to the NDMC for treatment after their discharge, which might affect the attrition rate. Also, those with low cooperativeness might be reluctant to continue to participate in this study. Third, potentially traumatic experiences of the participants in the ICU, such as longer ICU stay, delusional memories, and physical restraint with no sedation were not evaluated. Lack of these data makes it difficult to distinguish traumatic experiences during MVA from those in ICU, thus this needs further research for proposing effective intervention. However, this does not undermine the clinical significance of the PDI found in this study because the distress due to medicalization is an inevitable part of the severe MVA. Fourth, the results were obtained from only one teaching hospital in the suburbs of Tokyo, which can also limit the generalizability of this study. Moreover, the utility of the PDI as a predictor of later PTSD should be examined in survivors of other types of traumatic events.
To conclude, the study suggests the predictive usefulness of the PDI when admitted early after the MVA for subsequent PTSD. Its adequate usage should be further elaborated with attention to its low positive predictive value and the limitations mentioned above.
ACKNOWLEDGMENTS
This study was supported in part by grants from the Japanese Ministry of Health, Labor, and Welfare, and Japan Science and Technology Agency, CREST. Dr Nishi has received research support from Toray Industries, Inc. Dr Matsuoka has received research support from the Japanese Ministry of Health, Labor, and Welfare, Japan Science and Technology Agency, CREST. Dr Yonemoto has received research support from the Japan Society for the Promotion of Science and Health Labor Sciences Research Grant, Japan. Dr Kim has received research support or honoraria from the Japanese Ministry of Health, Labor, and Welfare, Japan Science and Technology Agency, CREST, GlaxoSmithKlein, Pfizer, Meiji Pharmaceutical, Yoshitomi and Meiji Yasuda Insurance Co. Dr Kanba has received research support or honoraria from the Japanese Ministry of Health, Labor, and Welfare, the Japanese Ministry of Education, Culture, Sports, Science and Technology, Eli Lilly, GlaxoSmithKline, Pfizer Inc, Asahi-kasei, Janssen, Tsumura, Ajinomoto, Yoshitomi, Meiji, Kyowa-Hakko, Sumitomo, Organon, Otsuka, Astellas, Mitsubishi and Ono.
