Acute and Chronic Posttraumatic Stress Symptoms in the Emergence of Posttraumatic Stress Disorder: A Network Analysis | Psychiatry | JAMA Psychiatry | JAMA Network
[Skip to Navigation]
Sign In
Figure 1.  Partial Correlations in Networks From the Acute and 12-Month Assessments
Partial Correlations in Networks From the Acute and 12-Month Assessments

Identical positioning (layout) of nodes has been imposed by taking the mean of the individual layouts that would have occurred. Minimum edge (connection) weight for inclusion was set at 0 (default) for both networks. Edge thickness and darkness are scaled against maximum absolute weight across both networks. We used the graphical Lasso based on an extended Bayesian information criterion (GLASSO) software to construct the networks. AMNES indicates amnesia; ANGER, irritability; AVSIT, avoidance of situations; AVTHT, avoidance of thoughts; CONC, concentration deficits; DISINT, disinterest in activities; DREAM, nightmares; DTACH, detachment from others; FLASH, flashbacks; FUTRE, foreshortened future; HYPER, hypervigilance; INTRU, intrusions; NUMB, emotional numbing; PHY, physiological reactivity; SLEEP, sleep disturbance; STRTL, startle response; and UPSET, upset by reminders.

Figure 2.  Plot of Standardized Centrality Indices for GLASSO Network
Plot of Standardized Centrality Indices for GLASSO Network

Three node centrality measures of betweenness (A), closeness (B), and strength (C) at the acute and 12-month assessments based on the graphical Lasso using an extended Bayesian information criterion (GLASSO). AMNES indicates amnesia; ANGER, irritability; AVSIT, avoidance of situations; AVTHT, avoidance of thoughts; CONC, concentration deficits; DISINT, disinterest in activities; DREAM, nightmares; DTACH, detachment from others; FLASH, flashbacks; FUTRE, foreshortened future; HYPER, hypervigilance; INTRU, intrusions; NUMB, emotional numbing; PHY, physiological reactivity; SLEEP, sleep disturbance; STRTL, startle response; and UPSET, upset by reminders.

Table 1.  Mean Symptom Severity Scores for Acute and 12-Month Assessments
Mean Symptom Severity Scores for Acute and 12-Month Assessments
Table 2.  Differences in Global Strength Values and Effect Sizes of Centrality Indices
Differences in Global Strength Values and Effect Sizes of Centrality Indices
1.
Zatzick  DF, Rivara  FP, Nathens  AB,  et al.  A nationwide US study of post-traumatic stress after hospitalization for physical injury.  Psychol Med. 2007;37(10):1469-1480.PubMedGoogle ScholarCrossref
2.
Stein  MB, Walker  JR, Hazen  AL, Forde  DR.  Full and partial posttraumatic stress disorder: findings from a community survey.  Am J Psychiatry. 1997;154(8):1114-1119.PubMedGoogle ScholarCrossref
3.
Bryant  RA.  Early predictors of posttraumatic stress disorder.  Biol Psychiatry. 2003;53(9):789-795.PubMedGoogle ScholarCrossref
4.
Pitman  RK, Shalev  AY, Orr  SP. Posttraumatic stress disorder: emotion, conditioning and memory. In: Corbetta  MD, Gazzaniga  M, eds.  The New Cognitive Neuroscience. 2nd ed. New York, NY: Plenum Press; 2000:1133-1148.
5.
Friedman  MJ, Resick  PA, Bryant  RA, Brewin  CR.  Considering PTSD for DSM-5 Depress Anxiety. 2011;28(9):750-769.PubMedGoogle ScholarCrossref
6.
Simms  LJ, Watson  D, Doebbeling  BN.  Confirmatory factor analyses of posttraumatic stress symptoms in deployed and nondeployed veterans of the Gulf War.  J Abnorm Psychol. 2002;111(4):637-647.PubMedGoogle ScholarCrossref
7.
Forbes  D, Elhai  JD, Miller  MW, Creamer  M.  Internalizing and externalizing classes in posttraumatic stress disorder: a latent class analysis.  J Trauma Stress. 2010;23(3):340-349.PubMedGoogle Scholar
8.
Yufik  T, Simms  LJ.  A meta-analytic investigation of the structure of posttraumatic stress disorder symptoms.  J Abnorm Psychol. 2010;119(4):764-776.PubMedGoogle ScholarCrossref
9.
Cramer  AO, Waldorp  LJ, van der Maas  HL, Borsboom  D.  Comorbidity: a network perspective.  Behav Brain Sci. 2010;33(2-3):137-150.PubMedGoogle ScholarCrossref
10.
Edwards  JR, Bagozzi  RP.  On the nature and direction of relationships between constructs and measures.  Psychol Methods. 2000;5(2):155-174.PubMedGoogle ScholarCrossref
11.
Borsboom  D, Cramer  AO.  Network analysis: an integrative approach to the structure of psychopathology.  Annu Rev Clin Psychol. 2013;9:91-121.PubMedGoogle ScholarCrossref
12.
McNally  RJ, Robinaugh  DJ, Wu  GWY, Wnag  L, Deserno  MK, Borsboom  D.  Mental disorders as causal systems: a network approach to posttraumatic stress disorder.  Clin Psychol Sci. 2015;3(6):836-849.Google ScholarCrossref
13.
Blake  DD, Weathers  FW, Nagy  LM,  et al.  The development of a Clinician-Administered PTSD Scale.  J Trauma Stress. 1995;8(1):75-90.PubMedGoogle ScholarCrossref
14.
Johnson  JW, LeBreton  JM.  History and use of relative importance indices in organizational research.  Organ Res Methods. 2004;7:238-257.Google ScholarCrossref
15.
Grömping  U.  Relative importance for linear regression in R: the package relaimpo J Stat Softw. 2006;17:1-27.Google ScholarCrossref
16.
van Borkulo  CD. Network comparison test: permutation-based test of differences in strength of networks. https://github.com/cvborkulo/NetworkComparisonTest. Published 2015. Accessed August 4, 2016.
17.
Epskamp  S, Borsboom  D, Fried  EI. Estimating psychological networks and their stability: a tutorial paper. https://arxiv.org/pdf/1604.08462.pdf. Published September 14, 2016. Accessed August 10, 2016.
18.
Pitman  RK.  Post-traumatic stress disorder, hormones, and memory.  Biol Psychiatry. 1989;26(3):221-223.PubMedGoogle ScholarCrossref
19.
Ehlers  A, Clark  DM.  A cognitive model of posttraumatic stress disorder.  Behav Res Ther. 2000;38(4):319-345.PubMedGoogle ScholarCrossref
20.
Liberzon  I, Khan  S, Young  EA, Yehuda  R, Post  RM. Animal models of posttraumatic stress disorder neuroendocrine aspects of PTSD: adjunctive strategies in the treatment of refractory bipolar depression: clinician options in the absence of a systematic database. In: Steckler T, Kalin NH, Reul JMHM, eds.  Handbook of Stress and the Brain, Part 2: Stress: Integrative and Clinical Aspects. Vol 6. 2005:531-546.
21.
Rauch  SL, Drevets  WC. Neuroimaging and neuroanatomy of stress-induced and fear circuitry disorders. In: Andrews  G, Charney  DS, Sirovatka  PJ, Regier  DA, eds.  Stress-Induced and Fear Circuitry Disorders: Refining the Research Agenda for DSM-V. Arlington, VA: American Psychiatric Association; 2009:215-254.
22.
Stam  R.  PTSD and stress sensitisation: a tale of brain and body, part 1: human studies.  Neurosci Biobehav Rev. 2007;31(4):530-557.PubMedGoogle ScholarCrossref
23.
Shalev  AY, Peri  T, Brandes  D, Freedman  S, Orr  SP, Pitman  RK.  Auditory startle response in trauma survivors with posttraumatic stress disorder: a prospective study.  Am J Psychiatry. 2000;157(2):255-261.PubMedGoogle ScholarCrossref
24.
Harvey  AG, Jones  C, Schmidt  DA.  Sleep and posttraumatic stress disorder: a review.  Clin Psychol Rev. 2003;23(3):377-407.PubMedGoogle ScholarCrossref
25.
Forbes  D, Lockwood  E, Elhai  JD,  et al.  An evaluation of the DSM-5 factor structure for posttraumatic stress disorder in survivors of traumatic injury.  J Anxiety Disord. 2015;29:43-51.PubMedGoogle ScholarCrossref
26.
Elzinga  BM, Bremner  JD.  Are the neural substrates of memory the final common pathway in posttraumatic stress disorder (PTSD)?  J Affect Disord. 2002;70(1):1-17.PubMedGoogle ScholarCrossref
27.
Vasterling  JJ, Brailey  K, Constans  JI, Sutker  PB.  Attention and memory dysfunction in posttraumatic stress disorder.  Neuropsychology. 1998;12(1):125-133.PubMedGoogle ScholarCrossref
28.
Williams  JMG, Barnhofer  T, Crane  C,  et al.  Autobiographical memory specificity and emotional disorder.  Psychol Bull. 2007;133(1):122-148.PubMedGoogle ScholarCrossref
29.
Falconer  E, Allen  A, Felmingham  KL, Williams  LM, Bryant  RA.  Inhibitory neural activity predicts response to cognitive-behavioral therapy for posttraumatic stress disorder.  J Clin Psychiatry. 2013;74(9):895-901.PubMedGoogle ScholarCrossref
30.
van Borkulo  CD, Borsboom  D, Epskamp  S,  et al.  A new method for constructing networks from binary data.  Sci Rep. 2014;4:5918.PubMedGoogle ScholarCrossref
31.
Brown  AD, Root  JC, Romano  TA, Chang  LJ, Bryant  RA, Hirst  W.  Overgeneralized autobiographical memory and future thinking in combat veterans with posttraumatic stress disorder.  J Behav Ther Exp Psychiatry. 2013;44(1):129-134.PubMedGoogle ScholarCrossref
32.
Schacter  DL, Addis  DR.  The cognitive neuroscience of constructive memory: remembering the past and imagining the future.  Philos Trans R Soc Lond B Biol Sci. 2007;362(1481):773-786.PubMedGoogle ScholarCrossref
33.
Schacter  DL, Addis  DR, Buckner  RL.  Remembering the past to imagine the future: the prospective brain.  Nat Rev Neurosci. 2007;8(9):657-661.PubMedGoogle ScholarCrossref
34.
American Psychiatric Association.  Diagnostic and Statistical Manual of Mental Disorders. 5th ed. Arlington, VA: American Psychiatric Association; 2013.
35.
Hoge  CW, Yehuda  R, Castro  CA,  et al.  Unintended consequences of changing the definition of posttraumatic stress disorder in DSM-5: critique and call for action.  JAMA Psychiatry. 2016;73(7):750-752.PubMedGoogle ScholarCrossref
36.
Molenaar  PCM, Campbell  CG.  The new person-specific paradigm in psychology.  Curr Dir Psychol Sci. 2009;18(2):112-117.Google ScholarCrossref
37.
O’Donnell  ML, Alkemade  N, Nickerson  A,  et al.  Impact of the diagnostic changes to post-traumatic stress disorder for DSM-5 and the proposed changes to ICD-11 Br J Psychiatry. 2014;205(3):230-235.PubMedGoogle ScholarCrossref
Original Investigation
February 2017

Acute and Chronic Posttraumatic Stress Symptoms in the Emergence of Posttraumatic Stress Disorder: A Network Analysis

Author Affiliations
  • 1School of Psychology, University of New South Wales, Sydney, Australia
  • 2Brain Dynamics Centre, University of Sydney and Westmead Millennium Institute, Sydney, Australia
  • 3Phoenix Australia, Department of Psychiatry, University of Melbourne, Melbourne, Australia
  • 4Centre for Military and Veteran Health, University of Adelaide, Adelaide, Australia
  • 5School of Psychiatry, University of New South Wales, Sydney, Australia
JAMA Psychiatry. 2017;74(2):135-142. doi:10.1001/jamapsychiatry.2016.3470
Key Points

Question  What are different networks of posttraumatic stress symptoms in the acute and chronic phases after trauma?

Findings  In this network analysis of 852 patients with traumatic injury who underwent assessment in the hospital and after 12 months, reexperiencing symptoms constituted a major network in the acute phase. At 12 months, connectivity was significantly stronger than in the acute phase, and fear circuitry and dysphoric symptoms of posttraumatic stress disorder emerged as connected networks.

Meaning  Trauma memories are centrally linked to other symptoms in the acute trauma phase, which highlights possible early intervention strategies.

Abstract

Importance  Little is understood about how the symptoms of posttraumatic stress develop over time into the syndrome of posttraumatic stress disorder (PTSD).

Objective  To use a network analysis approach to identify the nature of the association between PTSD symptoms in the acute phase after trauma and the chronic phase.

Design, Setting, and Participants  A prospective cohort study enrolled 1138 patients recently admitted with traumatic injury to 1 of 4 major trauma hospitals across Australia from March 13, 2004, to February 26, 2006. Participants underwent assessment during hospital admission (n = 1388) and at 12 months after injury (n = 852). Networks of symptom associations were analyzed in the acute and chronic phases using partial correlations, relative importance estimates, and centrality measures of each symptom in terms of its association strengths, closeness to other symptoms, and importance in connecting other symptoms to each other. Data were analyzed from March 3 to September 5, 2016.

Main Outcomes and Measures  Severity of PTSD was assessed at each assessment with the Clinician-Administered PTSD Scale.

Results  Of the 1138 patients undergoing assessment at admission (837 men [73.6%] and 301 women [26.4%]; mean [SD] age, 37.90 [13.62] years), strong connections were found in the acute phase. Reexperiencing symptoms were central to other symptoms in the acute phase, with intrusions and physiological reactivity among the most central symptoms in the networks in terms of the extent to which they occur between other symptoms (mean [SD], 1.2 [0.7] and 1.0 [0.9], respectively), closeness to other symptoms (mean [SD], 0.9 [0.3] and 1.1 [0.9], respectively), and strength of the associations (mean [SD], 1.6 [0.3] and 1.5 [0.3] respectively) among flashbacks, intrusions, and avoidance of thoughts, with moderately strong connections between intrusions and nightmares, being upset by reminders, and physiological reactivity. Intrusions and physiological reactivity were central in the acute phase. Among the 852 patients (73.6%) who completed the 12-month assessment, overall network connectivity was significantly stronger at 12 months than in the acute phase (global strength values, 6.57 vs 7.60; paired difference, 1.03; P < .001). The network associations among the reexperiencing symptoms were strengthened at 12 months, and physiological reactivity was strongly associated with the startle response, which was also associated with hypervigilance. Strong connectivity among emotional numbing, detachment from others, and disinterest in activities as well as moderately strong links among irritability (anger), concentration deficits, and sleep disturbance were found.

Conclusions and Relevance  As time elapses after trauma, fear circuitry and dysphoric PTSD symptoms appear to emerge as connected networks. Intrusive memories and reactivity are centrally associated with other symptoms in the acute phase, potentially pointing to the utility of addressing these symptoms in early intervention strategies.

Introduction

Posttraumatic stress disorder (PTSD) affects as many as 1 in 4 people who experience traumatic events1 and contributes to marked functional disability, health impairment, and social dysfunction.2 Despite the attention this condition has received, the fundamental nature of the syndrome is poorly understood. Although most survivors of trauma experience elevated stress responses in the immediate aftermath of trauma exposure, most of these reactions abate in the ensuing period, and only a minority of survivors develops a long-standing disorder.3 This process potentially involves fear conditioning at the time of trauma, resulting in overconsolidation of trauma memories.4

Quiz Ref IDAccording to the diagnostic criteria of PTSD, symptoms crystalize into the formal syndrome at some period beyond the acute stress response. The DSM has traditionally conceptualized PTSD as consisting of reexperiencing, avoidance, and hyperarousal clusters of symptoms, although DSM-5 added a fourth cluster consisting of alterations in mood and cognition.5 Although intrusions, avoidance, and hyperarousal are more associated with fear disorders, the dysphoria symptoms (eg, social withdrawal, emotional numbing, and disinterest in activities) are more associated with depressive disorders.6-8

Poor evidence exists regarding the relative structures of PTSD symptoms in short- and longer-term phases. One means to understand how stress reactions develop into PTSD is a network approach that posits that symptoms covary because they are linked via causal associations with each other.9,10 The network approach proposes that PTSD represents a causal system of functionally associated symptoms.11 For example, nightmares can contribute to insomnia, which may contribute to fatigue, which may lead to impaired concentration and irritability. One prior study of networks of chronic PTSD symptoms found that hypervigilance and foreshortened future were central to other symptoms and noted associations between irritability and sleep and concentration problems as well as reexperiencing symptoms and concentration problems.12 The network approach provides the opportunity to compare how PTSD symptoms are associated in the immediate and chronic phases of adjustment. In this study, we use a network approach to evaluate PTSD symptoms of patients who experienced a traumatic injury and underwent assessment within 1 week of their trauma and 12 months later. Although network analyses are somewhat exploratory, we hypothesized that, as a reflection of the emerging posttraumatic stress syndrome over time, network strength would be stronger in the chronic than the acute phases.

Methods
Participants

The Australian Injury Vulnerability Study recruited consecutive survivors of injury admitted to 4 level 1 trauma centers across Australia (Westmead Hospital, Sydney; Royal Adelaide Hospital, Adelaide; The Alfred Hospital, Melbourne; and Royal Melbourne Hospital, Melbourne) from March 13, 2004, to February 26, 2006. Inclusion criteria consisted of age of 18 to 70 years, ability to communicate in English proficiently, and hospital admission more than 24 hours after traumatic injury. Exclusion criteria were moderate or severe head injury, a current diagnosis of psychosis or active suicidality, being a temporary visitor to Australia, cognitive impairment, and/or being under police guard. The study was approved by the research and ethics committees at each hospital, and all participants provided written informed consent.

Procedure

Research assistants conducted interviews after 24 hours had elapsed since the traumatic injury. The mean (SD) duration between injury and initial assessment was 7.22 (9.65; range, 1- 21) days. Initial PTSD symptoms were assessed using the Clinician-Administered PTSD Scale (CAPS),13 which is a structured clinical interview for diagnosing PTSD and its severity.

At 12 months after the initial assessment, research psychologists conducted interviews by telephone using CAPS. All assessments were audiorecorded to ensure ongoing adherence to the protocol. For each symptom, a severity score was formed as the sum of CAPS frequency and intensity scores, resulting in items with values ranging from 0 to 8 (with higher scores indicating greater frequency and intensity of symptoms).

Statistical Analysis

The current data analyses were conducted between March 3 and September 5, 2016. Networks consist of nodes (symptoms) and edges (connections, or associations, between nodes). A variety of methods exists for determining the associations between nodes. Simple correlations between 2 symptoms reflect a combination of influences that potentially obscure the magnitude of the functional association, whereas partial correlations control for other symptoms. We used the graphical Lasso based on an extended Bayesian information criterion (EBICglasso; hereinafter referred to as GLASSO) option in the R package qgraph (https://www.rdocumentation.org/packages/qgraph/versions/1.3.5/topics/EBICglasso) to produce such estimates of the partial correlations.

Relative Importance

The association between A and B can also be expressed in terms of (1) how much A contributes to the multiple regression of B on all the other variables and (2) how much B contributes to the multiple regression of A on all the other variables. Measures of a variable’s contribution (or its relative importance) ideally will reflect the univariate effect of the variable and its effect in the multivariate context.14 To estimate relative importance (RELIMP), we used the lmg method from the R package relaimpo.15

Displaying Networks

GLASSO edges are drawn as a single line because the connections are symmetric, whereas RELIMP edges are drawn as 2 arrows of (possibly) differing thickness because the associations are not symmetric. In displays comparing networks, the mean location of the nodes (determined by the spring option in qgraph) was calculated across both networks, whereas the thickness of the edges is relative to the largest correlation overall.

Network Descriptors

Quiz Ref IDWe also report the following 3 measures of node centrality: betweenness (the extent to which a node lies on the shortest path between 2 other nodes and thus its importance in connecting nodes); closeness (the shorter the path between one node and another, the greater the capacity of either node to influence the other); and strength (the sum of the connections coming into and/or out of a node).11 We tested the difference between network connectivity at baseline and at 12 months using the network comparison test, which provides a permutation-based test of the difference in strength between networks.16 Finally, we report a number of indicators of the robustness of the findings, drawing on evolving research into this aspect of network models (eTables 1 and 2 and eFigures 1-10 in the Supplement).17

Results
Posttraumatic Stress Levels

Quiz Ref IDA total of 1477 patients with traumatic injury met inclusion criteria, and 1138 agreed to participate and completed the initial assessment (77.0%) (837 men [73.6%] and 301 women [26.4%]; mean [SD] age, 37.90 [13.62] years). Participants had survived transport crashes (748 [65.7%]), assaults (71 [6.2%]), traumatic falls (185 [16.3%]), work injuries (57 [5.0%]), or other traumatic injuries (77 [6.8%]). Four hundred ninety patients (43.1%) experienced a mild traumatic brain injury (MTBI). At follow-up, 838 participants completed the 12-month assessment (74.9%). Of these, 82 (9.6%) met criteria for PTSD at 12 months. Table 1 presents the mean symptom severity score for each symptom at the acute and 12-month assessments.

Baseline (Acute) Networks

Quiz Ref IDThe partial correlations (GLASSO) indicated that the strongest connections at baseline (Figure 1) were among flashbacks, intrusions, and avoidance of thoughts, with moderately strong connections between intrusions and nightmares, being upset by reminders, and physiological reactivity. The importance of reexperiencing in the acute phase was underscored by the centrality findings (Figure 2), which indicated that intrusions and physiological reactivity were among the strongest symptoms in the betweenness (mean [SD] of bootstrapped standardized centrality index values, 1.2 [0.7] for intrusions and 1.0 [0.9] for physiological reactivity), closeness (mean [SD] of bootstrapped standardized centrality index values, 0.9 [0.3] for intrusions and 1.1 [0.9] for physiological reactivity), and strength (mean [SD] of bootstrapped standardized centrality index values, 1.6 [0.3] for intrusions and 1.5 [0.3] for physiological reactivity) central indices. The relative importance analysis (eFigure 1 in the Supplement) indicates that intrusions and avoidance of thoughts are reciprocally associated, as are physiological reactivity and avoidance of thoughts (although to a lesser extent).

The other strong dyad in the acute phase was between emotional numbing and social detachment. The relative importance analysis (eFigure 1 in the Supplement) indicated that social detachment and numbing had a direct association, and their centrality measures indicated that they were relatively isolated from other symptoms. Amnesia was poorly associated with any other symptoms in the acute phase, and its weak centrality measures indicated its relative isolation from other PTSD symptoms.

12-Month Networks

At the 12-month assessment, the network associations between reexperiencing symptoms were maintained, although they were stronger, and physiological reactivity was strongly associated with the startle response, which was also associated with hypervigilance. Relative to the acute phase, intrusions were less central in terms of betweenness (mean [SD], 0.4 [0.8]), suggesting that whereas intrusions may exert an effect on other symptoms, this symptom lies on fewer paths of other associations relative to the acute phase. The relative importance analysis (eFigure 1 in the Supplement) indicated moderately strong direct associations between each of the reexperiencing symptoms, with physiological reactivity having a path to the startle response.

Concentration deficits were also highly central in terms of betweenness and closeness but not strength, indicating that, although concentration deficits lie on the path of many associations, the strengths of correlations with concentration deficits were weak. The relative importance analysis indicated a direct association between sleep disturbance and concentration deficits, with the pathway being stronger from concentration deficits to sleep disturbance.

At 12 months, a stronger network was found among emotional numbing, detachment from others, and disinterest in activities. This network was supported by the relative importance, which indicated a direct association between emotional numbing and detachment and an effect of emotional numbing on disinterest directly and via detachment. Regarding centrality, these symptoms had low betweenness, indicating that few symptoms passed through them (although the betweenness of emotional numbing was somewhat higher than in the acute phase). Moreover, these symptoms were not close to other symptoms.

Although amnesia was somewhat isolated from other symptoms in the acute phase, at 12 months we found a strong path from foreshortened future to amnesia. We also found a moderately strong link between foreshortened future and both detachment from others and emotional numbing. Further, the centrality of foreshortened future increased markedly from the acute to the 12-month phases. At 12 months, we found moderately strong links among irritability, concentration deficits, and sleep disturbance. The centrality of irritability increased markedly from the acute phase to 12 months on all indices, and foreshortened future and sleep disturbance increased in closeness and strength.

Quiz Ref IDA comparison of network connectivity of PTSD symptoms in the acute phase and at 12 months indicated that connectivity was significantly stronger at 12 months (based on 100 000 iterations using patients with data at both points, with a paired difference in global strength values of 1.03 (6.57 vs 7.60; P < .001). Effect sizes (Table 2) provide further information about differences in connectivity at both times and indicate that foreshortened future, sleep disturbance, anger, social detachment, amnesia, and concentration deficits had markedly stronger closeness and strength at 12 months than in the acute phase.

eTables 1 and 2 in the Supplement give correlations and SDs for items at both assessments. eFigures 2 to 7 in the Supplement show statistics that allow the robustness and stability of network descriptors to be examined.17 eFigures 8 to 10 in the Supplement provide a network-based perspective on the grouping of PTSD symptoms.

Discussion

Reexperiencing symptoms was connected strongly in the acute phase, suggesting that memory-based reactions are linked in the days after trauma. The observation that reexperienced symptoms are connected in the acute phase accords with fear-conditioning models that emphasize the overconsolidation of trauma memories.18 Furthermore, we found a strong direct association between intrusive memories and avoidance of thoughts in the acute phase. Numerous models propose that intrusive memories are avoided with effort because of their distressing content.19 Avoidance of reminders was apparently less associated with reexperiencing symptoms than was avoidance of thoughts; further, at 12 months, avoidance of reminders was more associated with dysphoric symptoms, suggesting this association may be part of the dysphoric response. The other notable association in the acute phase was the very strong linkage between emotional numbing and distancing from other people. One way to understand this strong association is that emotional numbing leads to social withdrawal because it limits the capacity to engage with others.

At 12 months, the associations of reexperiencing symptoms strengthened. The interesting addition to this network was the startle response, which was also associated with hypervigilance. Together with the reexperiencing symptoms, this association may be conceptualized as a fear circuitry syndrome because it includes symptoms associated with fear conditioning, avoidance, and sensitivity to threat.20,21 The startle response had a more central role at 12 months. Sensitization models of PTSD posit that 1 result of traumatic experience is that one is progressively more sensitive to threats, which can lead to stronger startle reactions over time.22 This theory is consistent with evidence that elevated startle responses only emerge in the months after trauma exposure rather than in the acute phase.23

We note that hyperarousal symptoms were also connected at 12 months, with sleep disturbance associated with anger and concentration difficulties. These associations were also reported in the previous network study of chronic PTSD.12 Sleep problems can lead to irritability and impaired concentration.24 We also note that sleep interference may be heightened in the hospital environment, which may introduce other confounding factors into these associations.

In contrast to the fear circuitry symptoms at 12 months, another observable network consisted of more dysphoric responses, including irritability, and concentration deficits, which tended to be associated with emotional numbing, loss of interest in activities, feeling distant from others, and a sense of foreshortened future. Previous factor analytic studies have identified these symptoms as constituting a dysphoric factor that is generally distinct from the fear-oriented symptoms of PTSD.6,25 Although this distinction is not apparent immediately after trauma, it becomes more apparent by 12 months. This finding is consistent with proposals that, although fear may characterize the immediate posttrauma response, stronger associations of dysphoric reactions may emerge over time.7 Within this dysphoric network, sleep, irritability, and concentration deficits remain interconnected, reflecting the pattern observed immediately after trauma. Emotional numbing and distance from others also remain connected and, by 12 months, are also associated with loss of interest in activities. Emotional numbing and distance from others apparently were associated with loss of interest in activities, possibly reflecting the progressive detrimental effect of emotional and social disengagement on motivation that evolves as PTSD symptoms persist over time.

It is worth noting the central role of concentration deficits in the acute and 12-month phases insofar as concentration deficits had strong betweenness, closeness, and (at 12 months) strength. The strong centrality indices of concentration deficits at 12 months accords with the previous study of network associations in PTSD,12 which also noted concentration problems as a symptom with the strongest centrality. The centrality of concentration problems in PTSD can be understood in the context of evidence that PTSD is associated with diminished working memory and attention,26,27 which can diminish the capacity of people to manage memories and solve problems28 and plays a role in recovery from PTSD.29

At 12 months, foreshortened future became a more predominant symptom in that it was more centrally linked, particularly to dysphoric symptoms. This development may be understood in the context of ongoing dysphoric posttraumatic reactions being associated with a sense of helplessness about one’s future. In this context, we note that amnesia was linked strongly to a sense of foreshortened future at 12 months and to a lesser extent to concentration difficulties. The relative importance suggests that the direct associations were between amnesia and both foreshortened future and concentration deficits. Deficits in concentration may be associated with more general cognitive difficulties that may contribute to poor remembering.30 In terms of foreshortened future, cognitive paradigms have demonstrated that imagining one’s future is specifically associated with recalling one’s past,31,32 which compliments evidence that the same neural networks are implicated in recalling the past and imagining the future.33 Survivors of trauma who have difficulty forecasting their future also may have memory deficits because both capacities involve overlapping mechanisms. We qualify all interpretations regarding amnesia with the recognition that nearly half of our sample sustained an MTBI; thus, amnesia may also reflect the effects of neurologic insult that impaired encoding of the trauma rather than dissociative amnesia.

To our knowledge, only 1 prior report has described network associations of PTSD symptoms,12 which showed quite distinct patterns relative to the 12-month networks noted in the present study. Although that prior study’s findings converged with the present findings in identifying concentration deficits, the startle response, and physiological reactivity as central symptoms, it diverged insofar as the prior study found relatively low centrality of emotional numbing and high centrality of hypervigilance. These discrepancies need to be understood in terms of the very different populations. The present study focused on survivors of traumatic injury 12 months after injury, whereas the prior study was based on a 5-year assessment of Chinese victims who lost a child as a result of a large earthquake.

These findings can be considered in relation to recent debates concerning changes to the PTSD definition in the DSM-5.34 Whereas the revision of PTSD in the DSM-5 encompass a wider array of symptoms, commentators have criticized this shift because it resulted in a reduced focus on fear circuitry symptoms.35 The observed networks in this study appear to support the emphasis on fear circuitry symptoms as the core element in PTSD because, in the acute and longer-term phases, fear circuitry appears to play a central role in the stress response. One of the noteworthy changes in the DSM-5 was that the symptom of foreshortened future was reformulated to encompass negative expectations about oneself and the world on the basis that a limited sense of one’s life span was deemed as too narrow5; interestingly, the present findings indicated that foreshortened future was a central symptom at 12 months, especially in regard to the dysphoric network of symptoms. This centrality may suggest that a sense of foreshortened future, as defined by the DSM-IV, is an important symptom, although replication network analyses are needed with the DSM-5 symptoms to provide comparative analyses. Overall, the 12-month data suggest that PTSD symptoms appear to develop into networks that may be described as predominantly dysphoric and fearful. Although not providing direct support for the proposal to provide greater recognition of non–fear-based symptoms in the DSM-5, this pattern suggests that greater attention is warranted to understand how dysphoric reactions to trauma can interact to contribute to posttraumatic adjustment.

Limitations

We recognize several limitations. First, inferences drawn from this study should acknowledge that network analyses do not define causal associations.11 Second, the strength of certain associations needs to be understood cautiously because the CIs of many of the edges between symptoms are large (eFigure 2 in the Supplement), suggesting that replication with other samples is required to further elucidate the strength of these associations. Third, the nature of the analyses incorporates effects occurring between and within participants, and accordingly the findings may mask the possibility that the PTSD syndrome has already developed within some individuals in the acute phase.36 Fourth, the cohort consisted of hospitalized survivors of injury, and so the findings need to be confirmed in samples exposed to different types of trauma. Fifth, initial assessments were conducted from 2 to 15 days after the traumatic event; network associations may differ within this range of time. Finally, we note that associations between symptoms may be influenced by the order in which they are administered.

Network analyses have the potential for understanding the construct of PTSD. In light of recent debate about the nature of PTSD in the context of diagnostic definitions in the DSM-535 and International Classification of Diseases, Eleventh Revision,35,37 network approaches may shed light on how PTSD symptoms can be understood. For example, the DSM-5 locates avoidance of thoughts and reminders in a separate symptom cluster, whereas the present data suggest that avoidance of thoughts is strongly associated with intrusive memories. Although these data do not shed light on causal mechanisms, the centrality indices raise possibilities regarding points of intervention. For example, 1 interpretation of these findings may be that addressing intrusive memories in the acute phase is a preventative strategy.

Conclusions

The network approach to understanding the associations between PTSD symptoms offers new opportunities to understand how initial stress reactions develop into longer-term PTSD problems. The importance of intrusive memories and associated reactivity were centrally related to other PTSD symptoms in the acute phase, which points to the potential for early intervention strategies that target trauma memories as a focus for secondary prevention.

Back to top
Article Information

Corresponding Author: Richard A. Bryant, PhD, School of Psychology, University of New South Wales, Sydney, NSW, Australia 2052 (r.bryant@unsw.edu.au).

Accepted for Publication: October 21, 2016.

Published Online: December 14, 2016. doi:10.1001/jamapsychiatry.2016.3470

Author Contributions: Dr Bryant had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Bryant, Creamer, O’Donnell, Forbes, Silove.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Bryant, O'Donnell, Forbes, Silove, Hadzi-Pavlovic.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Bryant, Forbes, Silove, Hadzi-Pavlovic.

Obtained funding: Bryant, Creamer, Forbes, McFarlane.

Administrative, technical, or material support: Bryant, Silove.

Study supervision: Bryant, Creamer, O’Donnell, Forbes, McFarlane, Silove.

Conflict of Interest Disclosures: None reported.

Funding/Support: This study was supported by program grant 1073041 from the National Health and Medical Research Council, grant V-11 from the Victorian Trauma Foundation, and Australian clinical research fellowship grant 359284 from the National Health and Medical Research Council.

Role of the Funder/Sponsor: The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

References
1.
Zatzick  DF, Rivara  FP, Nathens  AB,  et al.  A nationwide US study of post-traumatic stress after hospitalization for physical injury.  Psychol Med. 2007;37(10):1469-1480.PubMedGoogle ScholarCrossref
2.
Stein  MB, Walker  JR, Hazen  AL, Forde  DR.  Full and partial posttraumatic stress disorder: findings from a community survey.  Am J Psychiatry. 1997;154(8):1114-1119.PubMedGoogle ScholarCrossref
3.
Bryant  RA.  Early predictors of posttraumatic stress disorder.  Biol Psychiatry. 2003;53(9):789-795.PubMedGoogle ScholarCrossref
4.
Pitman  RK, Shalev  AY, Orr  SP. Posttraumatic stress disorder: emotion, conditioning and memory. In: Corbetta  MD, Gazzaniga  M, eds.  The New Cognitive Neuroscience. 2nd ed. New York, NY: Plenum Press; 2000:1133-1148.
5.
Friedman  MJ, Resick  PA, Bryant  RA, Brewin  CR.  Considering PTSD for DSM-5 Depress Anxiety. 2011;28(9):750-769.PubMedGoogle ScholarCrossref
6.
Simms  LJ, Watson  D, Doebbeling  BN.  Confirmatory factor analyses of posttraumatic stress symptoms in deployed and nondeployed veterans of the Gulf War.  J Abnorm Psychol. 2002;111(4):637-647.PubMedGoogle ScholarCrossref
7.
Forbes  D, Elhai  JD, Miller  MW, Creamer  M.  Internalizing and externalizing classes in posttraumatic stress disorder: a latent class analysis.  J Trauma Stress. 2010;23(3):340-349.PubMedGoogle Scholar
8.
Yufik  T, Simms  LJ.  A meta-analytic investigation of the structure of posttraumatic stress disorder symptoms.  J Abnorm Psychol. 2010;119(4):764-776.PubMedGoogle ScholarCrossref
9.
Cramer  AO, Waldorp  LJ, van der Maas  HL, Borsboom  D.  Comorbidity: a network perspective.  Behav Brain Sci. 2010;33(2-3):137-150.PubMedGoogle ScholarCrossref
10.
Edwards  JR, Bagozzi  RP.  On the nature and direction of relationships between constructs and measures.  Psychol Methods. 2000;5(2):155-174.PubMedGoogle ScholarCrossref
11.
Borsboom  D, Cramer  AO.  Network analysis: an integrative approach to the structure of psychopathology.  Annu Rev Clin Psychol. 2013;9:91-121.PubMedGoogle ScholarCrossref
12.
McNally  RJ, Robinaugh  DJ, Wu  GWY, Wnag  L, Deserno  MK, Borsboom  D.  Mental disorders as causal systems: a network approach to posttraumatic stress disorder.  Clin Psychol Sci. 2015;3(6):836-849.Google ScholarCrossref
13.
Blake  DD, Weathers  FW, Nagy  LM,  et al.  The development of a Clinician-Administered PTSD Scale.  J Trauma Stress. 1995;8(1):75-90.PubMedGoogle ScholarCrossref
14.
Johnson  JW, LeBreton  JM.  History and use of relative importance indices in organizational research.  Organ Res Methods. 2004;7:238-257.Google ScholarCrossref
15.
Grömping  U.  Relative importance for linear regression in R: the package relaimpo J Stat Softw. 2006;17:1-27.Google ScholarCrossref
16.
van Borkulo  CD. Network comparison test: permutation-based test of differences in strength of networks. https://github.com/cvborkulo/NetworkComparisonTest. Published 2015. Accessed August 4, 2016.
17.
Epskamp  S, Borsboom  D, Fried  EI. Estimating psychological networks and their stability: a tutorial paper. https://arxiv.org/pdf/1604.08462.pdf. Published September 14, 2016. Accessed August 10, 2016.
18.
Pitman  RK.  Post-traumatic stress disorder, hormones, and memory.  Biol Psychiatry. 1989;26(3):221-223.PubMedGoogle ScholarCrossref
19.
Ehlers  A, Clark  DM.  A cognitive model of posttraumatic stress disorder.  Behav Res Ther. 2000;38(4):319-345.PubMedGoogle ScholarCrossref
20.
Liberzon  I, Khan  S, Young  EA, Yehuda  R, Post  RM. Animal models of posttraumatic stress disorder neuroendocrine aspects of PTSD: adjunctive strategies in the treatment of refractory bipolar depression: clinician options in the absence of a systematic database. In: Steckler T, Kalin NH, Reul JMHM, eds.  Handbook of Stress and the Brain, Part 2: Stress: Integrative and Clinical Aspects. Vol 6. 2005:531-546.
21.
Rauch  SL, Drevets  WC. Neuroimaging and neuroanatomy of stress-induced and fear circuitry disorders. In: Andrews  G, Charney  DS, Sirovatka  PJ, Regier  DA, eds.  Stress-Induced and Fear Circuitry Disorders: Refining the Research Agenda for DSM-V. Arlington, VA: American Psychiatric Association; 2009:215-254.
22.
Stam  R.  PTSD and stress sensitisation: a tale of brain and body, part 1: human studies.  Neurosci Biobehav Rev. 2007;31(4):530-557.PubMedGoogle ScholarCrossref
23.
Shalev  AY, Peri  T, Brandes  D, Freedman  S, Orr  SP, Pitman  RK.  Auditory startle response in trauma survivors with posttraumatic stress disorder: a prospective study.  Am J Psychiatry. 2000;157(2):255-261.PubMedGoogle ScholarCrossref
24.
Harvey  AG, Jones  C, Schmidt  DA.  Sleep and posttraumatic stress disorder: a review.  Clin Psychol Rev. 2003;23(3):377-407.PubMedGoogle ScholarCrossref
25.
Forbes  D, Lockwood  E, Elhai  JD,  et al.  An evaluation of the DSM-5 factor structure for posttraumatic stress disorder in survivors of traumatic injury.  J Anxiety Disord. 2015;29:43-51.PubMedGoogle ScholarCrossref
26.
Elzinga  BM, Bremner  JD.  Are the neural substrates of memory the final common pathway in posttraumatic stress disorder (PTSD)?  J Affect Disord. 2002;70(1):1-17.PubMedGoogle ScholarCrossref
27.
Vasterling  JJ, Brailey  K, Constans  JI, Sutker  PB.  Attention and memory dysfunction in posttraumatic stress disorder.  Neuropsychology. 1998;12(1):125-133.PubMedGoogle ScholarCrossref
28.
Williams  JMG, Barnhofer  T, Crane  C,  et al.  Autobiographical memory specificity and emotional disorder.  Psychol Bull. 2007;133(1):122-148.PubMedGoogle ScholarCrossref
29.
Falconer  E, Allen  A, Felmingham  KL, Williams  LM, Bryant  RA.  Inhibitory neural activity predicts response to cognitive-behavioral therapy for posttraumatic stress disorder.  J Clin Psychiatry. 2013;74(9):895-901.PubMedGoogle ScholarCrossref
30.
van Borkulo  CD, Borsboom  D, Epskamp  S,  et al.  A new method for constructing networks from binary data.  Sci Rep. 2014;4:5918.PubMedGoogle ScholarCrossref
31.
Brown  AD, Root  JC, Romano  TA, Chang  LJ, Bryant  RA, Hirst  W.  Overgeneralized autobiographical memory and future thinking in combat veterans with posttraumatic stress disorder.  J Behav Ther Exp Psychiatry. 2013;44(1):129-134.PubMedGoogle ScholarCrossref
32.
Schacter  DL, Addis  DR.  The cognitive neuroscience of constructive memory: remembering the past and imagining the future.  Philos Trans R Soc Lond B Biol Sci. 2007;362(1481):773-786.PubMedGoogle ScholarCrossref
33.
Schacter  DL, Addis  DR, Buckner  RL.  Remembering the past to imagine the future: the prospective brain.  Nat Rev Neurosci. 2007;8(9):657-661.PubMedGoogle ScholarCrossref
34.
American Psychiatric Association.  Diagnostic and Statistical Manual of Mental Disorders. 5th ed. Arlington, VA: American Psychiatric Association; 2013.
35.
Hoge  CW, Yehuda  R, Castro  CA,  et al.  Unintended consequences of changing the definition of posttraumatic stress disorder in DSM-5: critique and call for action.  JAMA Psychiatry. 2016;73(7):750-752.PubMedGoogle ScholarCrossref
36.
Molenaar  PCM, Campbell  CG.  The new person-specific paradigm in psychology.  Curr Dir Psychol Sci. 2009;18(2):112-117.Google ScholarCrossref
37.
O’Donnell  ML, Alkemade  N, Nickerson  A,  et al.  Impact of the diagnostic changes to post-traumatic stress disorder for DSM-5 and the proposed changes to ICD-11 Br J Psychiatry. 2014;205(3):230-235.PubMedGoogle ScholarCrossref
×