Introduction
Post-traumatic stress disorder (PTSD) is an anxiety disorder whose symptoms include re-experiencing the original traumatic event, avoidance of trauma-related stimuli, and hyperarousal (American Psychiatric Association, 2000). A large body of research has addressed the issue of PTSD in the child or younger adult (Shalev, Reference Shalev2009). In view of demographic changes in the world, increased scientific and clinical knowledge is required to better understand PTSD as it occurs in the elderly person (Averill and Beck, Reference Averill and Beck2000). Research on PTSD in the elderly has focused on disorders related to early-life trauma (Böttche et al., Reference Böttche, Kuwert and Knaevelsrud2012). There are few data on persons who have undergone trauma as an elderly. Little is known of post-traumatic clinical presentation, impact of aging on the onset of the disorder, risk factors, and course of the symptoms (Lapp et al., Reference Lapp, Agbokou and Ferreri2011). It is debated whether age may make the elderly more vulnerable to development of PTSD. For some authors, elderly persons have diminished physical resources, less effective social support, and a tendency to underuse aid resources making recovery after traumatic events more difficult (Jia et al., Reference Jia, Tian, Liu, Cao, Yan and Shun2010). Other authors consider that age may be a protective factor against PTSD if the person had successfully resolved a previous trauma with more effective coping strategies and better social support (Chung et al., Reference Chung, Werrett, Easthope and Farmer2004).
A recent meta-analysis conducted by our group suggests that the intensity of the immediate responses to a traumatic event, i.e. peri-traumatic distress and peri-traumatic dissociation, are major risk factors for the development or persistence of PTSD symptoms among children and adults (Thomas et al., Reference Thomas, Saumier and Brunet2012). The concept of dissociation refers to an experience that combines depersonalization, altered perception of reality, and amnesia (van der Hart et al., Reference van der Hart, Nijenhuis and Steele2005). Peri-traumatic distress is a concept relating to the feelings experienced by the subject at the time of the traumatic event which overlaps in part with the A2 criterion (i.e. fear, helplessness, and horror) for PTSD according to the DSM-IV-R (Brunet et al., Reference Brunet2000). The severity of peri-traumatic distress and dissociation were shown to be robust predictors of later development of PTSD symptoms and diagnosis in both adults and children (Birmes et al., Reference Birmes2005b; Bui et al., Reference Bui2010). These findings have yet to be confirmed in the elderly subject. It is important to identify predictors of PTSD so that subjects at high risk may be identified and referred in a timely manner to a specialized program. The aim of our study was thus to assess the utility of peri-traumatic dissociation and distress as predictors of PTSD symptom onset and diagnosis in the elderly.
Methods
Subjects and procedure
We carried out a prospective longitudinal study in an open cohort of male and female subjects aged 65 years and over, consecutively admitted between January 2008 and December 2010 to the emergency departments of the University Hospitals of Toulouse and Bordeaux, France, after a potentially traumatic event: either physical assaults or road traffic accidents that met the PTSD A1 criterion (i.e. life-threat) defining trauma exposure. One hundred seventy-six individuals meeting these criteria were screened. The exclusion criteria were a head injury with loss of consciousness, current life-threatening injury, a history of PTSD or of chronic abuse, a history of mental disorder that had required continuous care or hospital admission, known neurodegenerative disease, or neurological problems as defined by a score of <28/30 on the Mini-Mental State Examination. One hundred thirty-four (76%) subjects were contacted and 29% of them accepted to participate and fulfilled inclusion criteria (59 refused and 36 were excluded).
Within seven days after the traumatic event, sociodemographic data were collected as well as medical, surgical, and psychiatric history (Sheehan et al., Reference Sheehan1998), including peri-traumatic responses using the self-report Peritraumatic Distress Inventory (PDI) and Peritraumatic Dissociative Experiences Questionnaire (PDEQ). The PDI evaluates the level of distress experienced during and immediately after, in the minutes and hours that follow the traumatic event (Brunet et al., Reference Brunet2001). It is a 13-item self-report questionnaire scored on a Likert-type scale from 0 (not at all true) to 4 (extremely true). Scores range from 0 to 52. A score or 3 or more on one of the items 1 (fear), 4 (helplessness), or 10 (horror) indicates that the person qualifies as having been exposed to a traumatic event according to the DSM-IV. The PDI has been validated in several languages, including French (Jehel et al., Reference Jehel, Brunet, Paterniti and Guelfi2005). The PDEQ measures immediate dissociative reactions observed during a traumatic event and in the minutes and hours that followed (Marmar, Reference Marmar, Weiss and Metzler1998). The subjects score the intensity of each of the ten items on a graded scale from 1 (not at all true) to 5 (extremely true). The total score is the sum of all items and is between 10 and 50. In the adult, a score >22 corresponds to elevated peri-traumatic dissociation (Birmes et al., Reference Birmes, Carreras, Charlet, Warner, Lauque and Schmitt2001). The French version of this instrument has been validated in two groups of adults (Birmes et al., Reference Birmes2005a).
Study participants underwent three visits, 1, 6, and 12 months after their admission to the emergency department as a result of trauma exposure. At each follow-up point, PTSD symptoms were assessed using the Clinician-administered PTSD scale (CAPS), a semistructured interview administered by a trained clinician. The CAPS yields a symptom severity score (0–136) as well as a diagnosis for the previous month (Weathers et al., Reference Weathers, Keane and Davidson2001). Every month, between months 1–6, participants also returned by mail the PTSD Checklist-Specific (PCL-S), a self-report measure for civilians yielding a PTSD symptom severity score ranging from 17 (no symptom) to the maximum score of 85 (Weathers et al., Reference Weathers, Litz, Herman, Huska and Keane1993). The study was approved by the Ethics Committee of Toulouse University Hospital and all participants gave written informed consent.
Data and analyses
A total of 39 eligible participants were recruited. Baseline measures were obtained for all participants. The semistructured PTSD interview was successfully completed for 31, 25, and 22 participants at months 1, 6, and 12, respectively. The self-report PTSD measure was completed monthly by 21–26 participants between months 2 and 6. Noncompletion of the interviews and questionnaires was mainly due to participants dropping out. Dropouts had slightly higher PTSD (CAPS) scores. The missing values pattern was monotonous and data was considered as “missing at random” (conditional on trauma type and missingness at earlier time points; Harrell, Reference Harrell2001).
Descriptive statistics were obtained for all participants. Basic measures of association between baseline variables and between baseline and follow-up variables were obtained. Significance, strength, and direction of the association between peri-traumatic responses and PTSD measures were estimated fitting generalized linear mixed models (Brown and Prescott, Reference Brown and Prescott2001) using a two-sided alpha level of 0.05. Mixed models account for the correlation between repeated measures over time. One model was fit per type of PTSD measure (CAPS score, PCL-S score, and CAPS diagnosis). Age and sex were selected as control variables in order to avoid model overfitting; the sample size (39 subjects with measures repeated at 3 or 4 time points) allowed 4 degrees of freedom to fixed parameters in the mixed model. The random effects consisted of patient-level intercept and slope.
Mixed models estimation methods are robust when dependent variable values are missing at random. However, dropouts will not have as much influence in the model-fitting process compared to participants with complete data (Brown and Prescott, Reference Brown and Prescott2001). Therefore, multiple imputation was used (Harel and Zhou, Reference Harel and Zhou2007). The SAS software and the R software were used to impute missing PTSD scores and diagnoses from observed variables creating ten data sets. Analyses were repeated over each data set. Mixed model parameter estimates were combined using Rubin's rules (Harel and Zhou, Reference Harel and Zhou2007). Statistical values for tests, averaged over the ten analyses, are reported in the section below. Residual analyses were performed to assess the appropriateness of both the mixed models and the imputation models (Harrell, Reference Harrell2001). Square-root transformations were applied to CAPS and PCL-S scores as the untransformed variables did not distribute normally.
Results
Independent variables distribution are shown in Table 1. The average age of participants was 72.4 years old (SD = 5.3), with a range of 65–84. Most participants were women (64.1%). Most participants were living with their spouse (56.4%) or alone (33.3%). Participants were married (56.4%), single (7.7%), separated (10.3%), or widowed (25.6%). Participants' education levels were as follows: elementary school diploma (46.2%), high school diploma (20.5%), or some higher education (33.3%).
Independent variables distribution
MVA = motor vehicle accident; PDEQ = Peritraumatic Dissociative Experiences Questionnaire (range: 10–50); PDI = Peritraumatic Distress Inventory (range: 0–52).
The participants involved in a motor vehicle accident consisted of 59.0% of the sample. Within the 41.0% that were victims of an aggression, 81.3% were women. Overall, the average peri-traumatic distress score was 15.8 (SD = 9.9; range = 0–40), while the average PDEQ score was 20.2 (SD = 8.8; range = 10–42). The severity of PTSD symptoms was higher among those having experienced an aggression: the mean CAPS score taken over time and imputation was of 36.1 (95% CI; 22.8–49.3) for patients having experienced an aggression, while it was of 15.2 (95% CI; 10.8–19.7) for patients involved in a motor vehicle accident. The difference was statistically significant (t = 2.23, df = 77, p = 0.03). Patients involved in aggression reported more severe peri-traumatic distress with mean scores of 22.8 (95% CI; 17.7–28.0) versus 10.9 (95% CI; 8.0–13.7). The same conclusion applies for peri-traumatic dissociation, with mean scores of 25.0 (95% CI; 19.6–30.4) versus 16.9 (95% CI; 14.3–19.5). Both results were statistically significant (t = 4.59, df = 37, p = 0.001 and t = 3.04, df = 37, p = 0.004, respectively).
Results from mixed model regression did not detect a significant effect of age, sex, nor time. Significant associations were detected between peri-traumatic distress and all three PTSD measures, with p-values of 0.008 (PCL, t = 2.71, df = 116), 0.016 (CAPS score, t = 2.47, df = 77), and 0.03 (CAPS diagnosis, t = 2.20, df = 77), respectively. Analyses conducted with nonimputed values yielded similar results. Table 2 presents population averaged means and percentages, estimated using the predicted values from the mixed model regression and its parameter estimates, for visualizing the strength and direction of the detected associations. Results are applicable at every time point (the modeled effect of time was not significant). No association was detected between peri-traumatic dissociation scores and the three PTSD measures. Cohen's f 2 was used to estimate effect sizes. Results suggested a moderate effect size for the distress variable (f 2 = 0.18), and a small effect size (f 2 = 0.05) for the dissociation variable.
Predicted mean PTSD symptom scores and diagnosis, adjusted for age and sex, as a function of self-reported peri-traumatic distress symptoms in a sample of 39 elderly
PDI = Peritraumatic Distress Inventory (range: 0–52); PCL = PTSD checklist (range: 17–85); CAPS = Clinician-administered PTSD Scale (range: 0–136).
Discussion
Our main objective was to evaluate the predictive value of peri-traumatic reactions in the development of PTSD symptoms and diagnosis in the elderly subject. The first finding is that the severity of PTSD in the elderly subject is greater in the case of physical assaults than in the case of road traffic accidents. This is in line with previous data confirming that man-made violence is more traumatogenic than other types of trauma (Wahlstrom et al., Reference Wahlstrom, Michelsen, Schulman and Backheden2008), although other factors are also involved (Seedat and Stein, Reference Seedat and Stein2001).
The second finding is that peri-traumatic distress but not dissociation predicted the severity of PTSD symptoms and diagnosis, irrespective of how and when PTSD was measured. Elevated peri-traumatic distress (i.e. a score equal or above 20 on the PDI) was associated with a diagnosis of PTSD on the CAPS at 1, 6, and 12 months in more than 50% of the cases thereby suggesting the allocation of clinical resources to that group of invididuals. Peri-traumatic distress is recognized as a predictor of the development of PTSD in children and adults. The same appears to be true of the elderly individuals after physical assault or a road traffic accident.
The absence of significant effect between peri-traumatic dissociation and the PTSD measures has also been noted in other studies. According to the meta-analysis of Thomas et al. (Reference Thomas, Saumier and Brunet2012), peri-traumatic distress would predict the development of the disorder while peri-traumatic dissociation would predict the lack of integration of the traumatic memory and thus, the persistence of chronic-enduring PTSD. It is also possible that this negative result was due to the lack of statistical power and small effect sizes.
Risk factors for PTSD and peri-traumatic symptoms are rarely evaluated in the elderly. We found no prospective studies on PTSD after physical assault or road traffic accidents in this population. In a context of bereavement, lack of expressive ability, fear of one's own death or serious illness, and helplessness experienced at the time of the death are peri-traumatic variables that can predict PTSD (Elklit and O'Connor, Reference Elklit and O'Connor2005), as can peri-traumatic distress (O'Connor, Reference O'Connor2010).
The association between PTSD and sociodemographic variables has already been evaluated in the elderly individual in cross-sectional studies. After a natural disaster such as a hurricane, certain variables generally recognized as risk factors for psychopathology such as low social support, prior exposure to severe traumatic events, and poor health status are present in elderly but also in young individuals (Acierno et al., Reference Acierno, Ruggiero, Kilpatrick, Resnick and Galea2006). Chen et al. have shown that after a typhoon, the principal factors associated with PTSD were female gender, poorer self-perceived health, relocation, family death, and depression (Chen et al., Reference Chen2011).
Study limitations
A notable limitation of our study is the small number of individuals included and the possible lack of representativeness of the sample. However, this sample was recruited consecutively at the emergency department of two regional university hospitals, thereby decreasing somewhat the sampling bias. Another limitation is that we did not assess the repercussions of trauma on mental health and cognitive functions. In the elderly, anxiety disorders find expression in this way (O'Hara, Reference O'Hara2006), which were not evaluated here. Finally, we did not assess earlier trauma in life.
Conclusion
Studies of PTSD in the elderly refer almost exclusively to combat, disasters, and the Holocaust. They are often epidemiological and cross-sectional. Future research should concentrate on situations of recent trauma in civilians and assess the impact of aging on the vulnerability of elderly individuals to developing PTSD.
Conflict of interest
None.
Description of authors' roles
A. Brunet wrote the paper; S. Sanche performed the statistical analyses and assisted in writing the paper; Aude Manetti, Bruno Aouizerate, Régis Ribéreau-Gayon, and Sandrine Charpentier participated in subjects recruitment, inclusion, and assisted in writing the paper; P. Birmes assisted in writing the paper; C. Arbus designed the study and supervised the writing of the paper.
Acknowledgments
This work was supported by a grant from the Clinical Research Hospital Program from the French Ministry of Health (PHRC 2006). Alain Brunet acknowledges receiving a salary award from the Fonds de recherche du Québec (FRQ-S) while working on this manuscript.
