A network analysis of DSM-5 posttraumatic stress disorder symptoms and correlates in U.S. military veterans

Highlights

• We used a nationally representative sample of trauma-exposed military veterans (NHRVS) and selected 221 who reported clinically significant DSM-5 PTSD symptoms.

• We estimated two networks; a DSM-5 PTSD network and a DSM-5 PTSD network with clinically significant covariates.

• The networks revealed that symptoms were positively connected with especially strong connections emerging between several items.

• Incorporation of clinically relevant covariates into the PTSD network revealed paths between PTSD’s symptom of self-destructive behavior and suicidal ideation amongst others.

Abstract

Objective

Recent developments in psychometrics enable the application of network models to analyze psychological disorders, such as PTSD. Instead of understanding symptoms as indicators of an underlying common cause, this approach suggests symptoms co-occur in syndromes due to causal interactions. The current study has two goals: (1) examine the network structure among the 20 DSM-5 PTSD symptoms, and (2) incorporate clinically relevant variables to the network to investigate whether PTSD symptoms exhibit differential relationships with suicidal ideation, depression, anxiety, physical functioning/quality of life (QoL), mental functioning/QoL, age, and sex.

Method

We utilized a nationally representative U.S. military veteran’s sample; and analyzed the data from a subsample of 221 veterans who reported clinically significant DSM-5 PTSD symptoms. Networks were estimated using state-of-the-art regularized partial correlation models. Data and code are published along with the paper.

Results

The 20-item DSM-5 PTSD network revealed that symptoms were positively connected within the network. Especially strong connections emerged between nightmares and flashbacks; blame of self or others and negative trauma-related emotions, detachment and restricted affect; and hypervigilance and exaggerated startle response. The most central symptoms were negative trauma-related emotions, flashbacks, detachment, and physiological cue reactivity. Incorporation of clinically relevant covariates into the network revealed paths between self-destructive behavior and suicidal ideation; concentration difficulties and anxiety, depression, and mental QoL; and depression and restricted affect.

Conclusion

These results demonstrate the utility of a network approach in modeling the structure of DSM-5 PTSD symptoms, and suggest differential associations between specific DSM-5 PTSD symptoms and clinical outcomes in trauma survivors. Implications of these results for informing the assessment and treatment of this disorder, are discussed.

Introduction

The nosology of posttraumatic stress disorder (PTSD) is highly controversial and widely debated (Hoge et al., 2016; Friedman, Kilpatrick, Schnurr, & Weathers, 2016). There have also been several changes to the diagnostic criteria since the initial introduction of PTSD into the Diagnostic and Statistical Manual of Mental Disorders (DSM). These changes include the definition of the criterion A trauma, the number and nature of symptoms (DSM-5 now includes three new diagnostic symptoms of negative belief, distorted blame, and recklessness), the number and nature of symptom groups, and the recent inclusion of a dissociative PTSD subtype. The latest version of the criteria (DSM-5) was released in May 2013. The current study utilizes a novel psychometric approach based on Network Analysis to identify the way in which DSM-5 PTSD symptoms interact at the individual item level and in turn to identify if these interactions are clinically relevant.

PTSD is an impairing disorder that affects a significant proportion of both civilians and military personnel exposed to a traumatic event (Karam et al., 2014, Kessler et al., 2014; Pietrzak, Goldstein, Southwick, & Grant, 2011). The estimated prevalence rates in military veteran populations range from 9 to 30% (Lapierre, Schwegler, & Labauve, 2007; Sundin, Fear, Iversen, Rona, & Wessely, 2010; Tsai et al., 2014). Veterans displaying posttraumatic symptomatology have greater difficulty with re-integration to civilian life post-deployment (Karstoft, Armour, Andersen, Bertelsen, & Madsen, 2015); report a lower quality of life (QoL) (Giacco, Matanov, & Priebe, 2013); and have greater risk for suicidal ideation (SI) (Jakupcak, Cook, Imel, Rosenheck & McFall, 2009; Jakupcak et al., 2011). PTSD is additionally a highly comorbid disorder with the vast majority of individuals also meeting the criteria for at least one additional psychiatric disorder (Brady, Killeen, Brewerton, & Lucerini, 2000; Pietrzak et al., 2011).

The most recent edition of the DSM (DSM-5; APA, 2013) characterizes PTSD as containing 20 individual symptoms that are grouped across four symptom clusters; Intrusions (IN; B1-B5; see Fig. 1), Avoidance (AV; C1-C2), Negative alterations in cognitions and mood (NACM; D1-D7), and Alterations in arousal and reactivity (AAR; E1-E6). Of note, the structure of PTSD in the DSM-5 has been contended with alternative structural models being proposed. In summary, a number of factor analytic studies have demonstrated that models comprising six (Liu et al., 2014 [Anhedonia Model]; Tsai et al., 2014 [Externalizing Behaviors model]) and seven symptom groupings (Armour et al., 2015 [Hybrid model]) may provide better fit than the four clusters outlined in the DSM-5 (reviewed in Armour, Mullerova, & Elhai, 2016).

A diagnosis of PTSD as set by DSM-5 criteria currently requires that trauma survivors endorse a minimum of six symptoms (at least 1 IN, 1 AV, 2 NACM, and 2 AAR), in addition to reporting significant functional impairment and the persistence of symptoms in excess of one month (APA, 2013). Notably, this prescribed six-item diagnostic algorithm has previously been reported as being so amorphous in nature that it results in 636,120 combinations of PTSD symptomatology (Galatzer-Levy & Bryant, 2013). Consequently, it is possible that individuals with a DSM-5 PTSD diagnosis can have remarkably distinct symptom presentations. Another potential drawback of the current criteria is that individuals who fail to meet them (by exhibiting less than six symptoms in the prescribed DSM-5 fashion) may be as impaired by their symptomatology as individuals who do meet the criteria. For example, failing to endorse at least one avoidance item would preclude someone from the diagnosis even if all other items in all other symptom groups were endorsed.

Although the DSM acknowledges that psychiatric disorders are dimensional, its guidelines maintain that a threshold on the dimension is required in order to define who does vs. does not receive a diagnosis; the DSM-5 nosology for PTSD is thus embedded, from a diagnostic point of view, in a categorical framework. Contrasting this categorical approach, many researchers support a dimensional view to PTSD, stating that the distress experienced by an individual post-trauma exists on a continuum in which the high end reflects more severe distress (cf. Ruscio, Ruscio, & Keane, 2002). In this case, symptoms that reflect PTSD are summed together to determine the position of an individual on the continuum. These sum scores are often used by researchers to assess the PTSD construct, particularly as it relates to the longitudinal course of symptoms (Bonanno, Papa, Lalande, Westphal, & Coifman, 2004; Dickstein, Suvak, & Litz, 2010; Karstoft, Armour, Elklit, & Solomon, 2013) and in assessing PTSD treatment response over time (Richardson et al., 2014).

Irrespective of whether the syndrome of PTSD is viewed through a categorical or dimensional lens – which is also not the focus of the current paper – the commonality between these viewpoints is the proposition that symptoms are reflective indicators of an underlying latent construct that is PTSD. This ‘reflective model’ is the prevailing perspective with which psychopathology is currently viewed and understood (Borsboom and Cramer, 2013, Borsboom, 2008; Cramer, Waldorp, van der Maas, & Borsboom, 2010). In both categorical and dimensional systems, PTSD symptoms are interchangeable and equally reflective of the latent construct that is PTSD; this view is known as the ‘common cause hypothesis’ (Fried, 2015, Schmittmann et al., 2013). A challenge is that symptom groupings as outlined in the DSM (detailed above) have been shown to vary in relation to the precipitating traumatic event (Armour & Shevlin, 2010); vary in their comorbidity with alternative disorders (Contractor et al., 2014; Pietrzak, Tsai, Armour, Mota, Harpaz-Rotem, & Southwick, 2015; Roley et al., in press); relate differentially to perceived quality of life (Giacco et al., 2013); and display differential responses to treatment (Asmundson, Stapeleton, & Taylor, 2004). This implies that PTSD symptoms are not (roughly) interchangeable indicators of one underlying reflective latent variable that causes the covariation among symptoms. In addition, cross-sectional studies have revealed that different groups of individuals, displaying different symptom profiles, show differential associations with external variables (Ayer et al., 2011; Au, Dickstein, Comer, Salters-Pedneault, & Litz, 2013; Breslau, Reboussin, Anthony, & Storr, 2005; Alman et al., 2012, Maguen et al., 2013; Naifeh, Richardson, Del Ben, & Elhai, 2010; Rosellini, Coffey, Tracy, & Galea, 2014).

That symptoms do not interact with each other causally is highly implausible, as psychiatric symptoms by their very nature have direct relations to one another. McNally et al. (2015) exemplified this for PTSD, showing that being presented with a reminder of a traumatic event may in turn trigger psychological and physiological reactions, which may activate avoidance behaviors. This has important clinical implications, given that the ability to identify particular central symptoms in a causal system (symptoms that are highly connected and likely to trigger other symptoms), would allow clinicians to focus on such symptoms in assessment, monitoring, and treatment (Cramer et al., 2013; Fried, Epskamp, Nesse, Tuerlinckx, & Borsboom, 2015). From the traditional perspective of reflective latent variables that cause the covariance among symptoms, such specific symptom-focused interventions would generally not be employed.

A new and growing school of thought in which symptoms are related amongst themselves, rather than being equally reflective of an underlying latent construct, has been empirically tested by the development and application of data analytic techniques known as ‘network analysis’ (Borsboom and Cramer, 2013, Boschloo et al., 2015). From this perspective, symptoms are not reflective of an underlying disorder; instead, the associations among symptoms constitute the disorder. Networks are comprised of nodes (e.g., symptoms) and edges (associations among symptoms). Importantly, a network perspective does not assume symptoms to be interchangeable, but instead allows for the examination of the importance or centrality of symptoms empirically. Highly connected items that are likely to spread activation through the symptom network once activated are more central, whereas items with fewer connections lie on the periphery of a network and are less important (Borsboom and Cramer, 2013, Fried et al., 2015). The network approach to psychopathology has received growing attention and recognition in the last years, and a new review paper discusses its application to a wide variety of disorders, including major depression, psychosis, and autism (Fried et al., under revision).

To date, only a few studies have employed a network analysis of PTSD symptoms (Knefel Tran, & Lueger-Schuster, 2016; Afzalia et al., 2016; McNally et al., 2015). Both Knefel et al. (2016) and Afzalia et al. (2016) examined the way in which PTSD symptomatology associates with symptomatology of other disorders; the former assessed borderline personality disorder and the latter depression. McNally et al. (2015) focused solely on PTSD symptomatology (as we did in the current study) in 362 survivors of the Wenchuan earthquake in China. Key findings included: the centrality of the hypervigilance symptom; connections between anger, sleep, irritability, and concentration difficulties; and associations among intrusive thoughts, dreams, and flashbacks. The study analyzed the DSM-IV PTSD symptoms. In the current study, we estimated networks of PTSD symptoms using data from the National Health and Resilience in Veterans Study (NHRVS), which is a contemporary, nationally representative study of U.S military veterans. We extended the McNally et al. (2015) study in four ways: (1) for the first time, we constructed PTSD networks based on the DSM-5 rather than the DSM-IV criterion symptoms; (2) we employed a sub-sample of veterans reporting clinically significant PTSD symptoms from a representative U.S. veteran study; pertinent given the plethora of research detailing generally higher rates of symptomatology and comorbidity in this population (Giacco et al., 2013, Jakupcak et al., 2011; Karstoft et al., 2014; Tsai et al., 2014); (3) we estimated the network structure among PTSD symptoms, and also, for the first time, added a number of clinical covariates (SI, depression, anxiety, physical functioning/QoL, mental functioning/QoL, age, and gender) into the network to examine whether they display particular associations with particular PTSD symptoms; and (4) we analyzed the robustness and stability of the networks.