Patterns of comorbidity associated with ICD-11 PTSD among older adults in the United States

Little research has been conducted on posttraumatic stress disorder (PTSD) comorbidity among older adults regarding the description of PTSD in the 11th version of the International Classification of Diseases (ICD-11). This study sought to provide evidence of a dimensional model of psychopathology using the 'Hierarchical Taxonomy of Psychopathology' (HiTOP) model as a theoretical framework to explain patterns of ICD-11 PTSD comorbidity. Distinct patterns of ICD-11 PTSD comorbidity among a nationally representative sample (n = 530) of adults aged 60 years and older from the United States were examined using latent class analysis (LCA). Covariates associated with comorbidity classes were assessed through multinomial logistic regression. ICD-11 PTSD was highly comorbid with other psychopathologies. LCA results favoured a two-class solution. Class 1 (71.7%) was characterised by moderate probabilities for major depressive disorder and alcohol use disorder; Class 2 (28.3%) was characterised by a moderate-high probability of general psychopathology and was associated with lower social support, spousal/partner physical abuse, and history of attempted suicide. PTSD was highly comorbid with other disorders among older adults. Distinct patterns of PTSD comorbidity exist among this cohort and these findings can aid clinicians and researchers in understanding and predicting maladaptive responses to trauma and associated psychopathology.

Patterns of comorbidity associated with ICD-11 PTSD among older adults in the United States

Introduction
Posttraumatic stress disorder (PTSD) frequently co-occurs with other psychiatric disorders including major depressive disorder (MDD; Rytwinski et al., 2013), generalised anxiety disorder (GAD; Gallagher and Brown, 2015), eating disorders (Brewerton, 2007), substance use disorders (SUD; Driessen et al., 2008;Pietrzak et al., 2011), psychotic disorders (Seow et al., 2016), borderline personality disorder (BPD; Frías and Palma, 2015), and also with suicidal ideation (Panagioti et al., 2012). High rates of comorbidity are unsurprising given that trauma exposure is a common risk factor for PTSD as well as all other forms of psychopathology (Bendall et al, 2008;Brewerton, 2007;Copeland et al., 2018;Gilbert et al., 2009;Kessler et al., 2010;Najavits et al., 2017;Varese et al., 2012). One goal of the new model of PTSD presented in the recently published 11 th version of the International Classification of Diseases (ICD-11;World Health Organization [WHO], 2018) was to reduce diagnostic comorbidity by including only a small number of core symptoms (Maercker et al., 2013). Retaining core symptomsand excluding transdiagnostic onesshould reduce diagnostic comorbidity if psychiatric disorders are orthogonal (although orthogonality in this context is not without enduring controversy; see Marshall, 2020).
However, recent findings indicate that psychopathological/ psychiatric comorbidity rates remain extremely high for ICD-11 PTSD Karatzias et al., 2019;. The ICD-11 also contains a sibling disorder to PTSD, Complex PTSD (CPTSD; WHO, 2018). In order to meet the diagnostic requirements for CPTSD, an individual must meet the requirements for PTSD and additional symptoms reflecting 'disturbances in selforganization'. Akin to PTSD, CPTSD has been found to be highly comorbid with other forms of psychopathology (Karatzias et al., 2019).
A wealth of data shows that diagnostic comorbidity is extremely common (Caspi and Moffitt, 2018;Kotov et al., 2017). Diagnostic comorbidity can be explained and understood within a dimensional model of psychopathology. The 'Hierarchical Taxonomy of Psychopathology' (HiTOP; Kotov et al., 2017) is a prominent dimensional model of psychopathology which proposes that variation and covariation in psychopathology is explainable in terms of a small number of superordinate correlated latent dimensions (e.g. 'internalising', 'externalising', and 'thought disorder'). These dimensions are divided into 'subfactors' that explain covariation between specific clusters of mental health problems that cut across traditional psychiatric diagnoses. In the HiTOP model, PTSD (symptoms) sits within the 'distress' subfactor of the 'internalising' dimension. Thus, the HiTOP model predicts that PTSD should be most strongly correlated with disorders within the same subfactor such as MDD, GAD, and BPD. Moreover, because the internalising dimension is correlated with all other dimensions, PTSD would also be expected to correlate with disorders in other dimensions, albeit to a lesser extent. Thus, HiTOP states that comorbidity is unavoidable because psychiatric disorders are manifestations of the same, or related, latent variable(s). Furthermore, focusing on the core symptoms of PTSD should reduce measurement error, thereby increasing covariation with other psychiatric disorders. As such, the parameters of the HiTOP model predict that reducing PTSD to its core symptoms will not only not lead to a reduced rate of diagnostic comorbidity, but it may in fact lead to an increased rate of comorbidity. Miller and colleagues (2012) examined the latent structure of PTSD comorbidity in a sample of military veterans with high rates of PTSD and found that three latent factors ('fear', 'distress', and 'externalising') adequately represented the structure of PTSD comorbidity. Consistent with the HiTOP model, these factors shared similarities such as the 'distress' factor encompassing the same disorders (MDD, GAD, dysthymia, and BPD) that has been observed as a common factor found among community samples (see Beesdo-baum et al., 2009;Kotov et al., 2017). Identifying discernible patterns of comorbidity, along with the risk factors for these different patterns of comorbidity, is important as this information may aid clinicians in preventing individuals from progressing to highly dysfunctional multimorbid trajectories associated with adverse outcomes such as increased social and occupational impairment, suicidal ideation, and physical morbidity (Calabrese et al., 2011;Hefner and Rosenheck, 2019;Momartin et al., 2004).
Variable-oriented approaches to modelling comorbidity, such as those used by Miller et al. (2012), presuppose, and function optimally with, sample homogeneity. When examining a potentially heterogenous population where distinct subgroups may exist, such as individual responses to trauma, person-oriented approaches such as latent class analysis (LCA) -which can account for sample heterogeneity -may be more appropriate (von Eye and Bogat, 2006). Prior studies examining patterns of PTSD comorbidity using LCA across a range of internalising and externalising disorders have revealed three discrete classes (Galatzer-Levy et al., 2013;Müller et al., 2014). One class reflected individuals with a low probability of comorbidity across all disorders (except for a moderate probability of major depression); another was characterised by a high probability of comorbid mood and anxiety disorders; and the third was characterised by a high probability for SUDs, mood, and anxiety disorders. Suicidal ideation was found to be a characteristic of all high comorbidity classes.
This finding is in line with research suggesting that psychiatric comorbidity is a risk factor for suicidal behaviour in relation to both PTSD comorbidity and general psychopathology (Calabrese et al., 2011;Gili et al., 2019;Kavalidou et al., 2019;Kavalidou et al., 2017;Norman et al., 2018;Turecki and Brent, 2016).
The literature pertaining to PTSD comorbidity specifically among older adults is relatively underdeveloped; the available data suggests that PTSD in older adults is highly comorbid with a range of internalising and externalising disorders (Averill and Beck, 2000;Chopra et al., 2014;Glück et al., 2016;Pietrzak et al., 2012;Pless Kaiser et al., 2019;Spitzer et al., 2008). Epidemiological evidence indicates that psychiatric morbidity (including PTSD) and comorbidity is significantly lower among older adults in comparison to their younger counterparts (Gum et al., 2009;Kessler et al., 2005;Thomas et al., 2016). For example, Kessler and colleagues (2005) found that 11.6% of adults aged 60 years and older reported having two or more psychiatric disorders, which was substantially lower than those aged 18-29 (33.9%), 30-44 (34.0%), and 45-59 (27.0%). Furthermore, it has been found that older age is associated with decreased odds of psychiatric comorbidity among individuals who met the criteria for a past-year diagnosis of PTSD (Reynolds et al., 2016). Therefore, as rates of diagnostic psychiatric comorbidity may be substantially lower among older adults, it is important to determine if patterns of covariation among psychiatric disorders observed from previous studies among the general population are generalisable to this population.
In this study, we investigated patterns of comorbidity for ICD-11 PTSD among a nationally representative sample of older adults (60 years and above) from the United States.
Based on the predictions of the HiTOP model, we formulated four hypotheses. First, we hypothesised that rates of diagnostic psychiatric comorbidity involving ICD-11 PTSD and a range of other psychiatric disorders would be high. Second, that the highest comorbidity rates would be found for 'distress' related disorders including MDD, GAD, and BPD. Third, based on prior findings regarding the latent structure of PTSD comorbidity, we hypothesised that multiple latent classes would be identified, including classes characterised by (i) low comorbidity, (ii) comorbidity with internalising disorders, and (iii) comorbidity with externalising and/or psychotic disorders. Finally, we hypothesised that the latent classes characterised by the highest levels of diagnostic comorbidity would be associated with a history of suicidal behaviour. Moreover, we included additional covariates that have been found to play an important role in predicting positive/negative mental health outcomes, such as social support (Wang et al., 2018).

Participants and Recruitment Strategy
Participants in this study were drawn from the National Epidemiologic Survey on Alcohol and Related Conditions-III (NESARC-III) study which is a nationally representative sample of non-institutionalised adults from the U.S. aged 18 years and older (N = 36,309).
Information on the NESARC-III data are available elsewhere (Grant et al., 2014). Protocols of the NESARC-III project received ethical approval from the institutional review boards of the National Institutes of Health and Westat, and all participants provided their informed consent. Approval for secondary analysis was granted by the ethical review board of the first author's affiliated institution.
The current sample (n = 530) were selected from the full NESARC-III dataset based on several inclusion criteria: (a) were aged 60 years or older, (b) reported experiencing or witnessing at least one traumatic event in their lifetime, and (c) met the symptomatic requirements for an ICD-11 PTSD diagnosis. Data were adjusted for oversampling (of ethnic/racial minorities) and non-responses and were weighted to reflect the U.S. civilian population as per the 2012 American Community Survey (Bureau of the Census, 2013). All parameter estimates were adjusted for the complex survey design of the NESARC-III based on the stratification, clustering, and weighting of the study population, whereas sample size is based on the unweighted data. The sample included a higher proportion of females (59.9%, n = 330) than males (40.1%, n = 200), and the average age was 67.65 years (SD = 6.67) (see Supplementary Table 1 for other sample characteristics).

Measures
All data were gathered using the Alcohol Use Disorder and Associated Disabilities Interview Schedule-5 (AUDADIS-5; Grant et al., 2011). The AUDADIS-5 is a structured, diagnostic interview which assesses participants for symptoms associated with an array of psychiatric disorders.

Traumatic exposure:
Participants were presented with two lists of different traumatic events that they may have experienced. One list consisted of 19 traumatic events that they may have personally experienced (e.g. childhood sexual abuse). The second list consisted of 13 traumatic events that they may have witnessed/learned about (e.g., witnessing/learning about another person's childhood sexual abuse). Participants were then asked to select, from these lists, the traumatic event(s) that they have personally experienced or witnessed/learned about. Participants could also select "other" as an additional option if the type of traumatic event that they experienced was not specified on either list. However, in order to reduce response burden, respondents could only report experiencing a maximum of four different types of traumatic events from these two lists and were instructed to specify their most stressful traumatic event. If the respondent had experienced more than four traumatic events, only the four most severe events were recorded. All PTSD items were responded to in relation to their most distressing event.
For the purposes of the current study, several of the direct traumatic events were used as covariates of latent class membership. These included being sexually abused before age of 18 years, being sexually assaulted as an adult, being physically abused before age of 18 years, being beaten up by a spouse/romantic partner, being beaten up by someone else, experiencing other forms of interpersonal trauma (kidnapped, stalked, or mugged), experiencing a noninterpersonal trauma (serious or life-threatening injury/illness, saw a dead body or body parts, being admitted to a juvenile detention/jail, or experiencing a natural disaster), and experiencing a war-related trauma (being injured in a terrorist attack, experiencing active military combat, being a peacekeeper/relief worker, being a civilian in war zone/place of terror, being a refugee, being a prisoner of war).

ICD-11 PTSD symptoms:
Items were extracted from the AUDADIS-5 that corresponded to the six ICD-11 symptoms. See Supplementary Table 2 for a comparison between the items extracted from the AUDADIS-5 and their corresponding items from the International Trauma Questionnaire (ITQ; Cloitre et al., 2018), the only available and psychometrically supported measure of ICD-11 PTSD and CPTSD. Symptoms were answered using a dichotomous response format ('yes' = 1, 'no' = 0). The internal consistency among the sample of older adults was satisfactory (Cronbach's alpha = .77). In addition, the ICD-11 model of PTSD, using the same items as the current study, has previously been found to provide excellent statistical fit among adults aged 60 years and above (Fox et al., 2020).
An ICD-11 PTSD diagnosis requires the presence of at least one of two 'Reexperiencing' symptoms, one of two 'Avoidance' symptoms, and one of two 'Sense of Threat' symptoms. Diagnosis also requires that these symptoms cause functional impairment, however, this criterion was not screened for with all participants. As such, participants Interview for Substance and Mental Disorders, DSM-5 version]), and indicated that the concordance of DSM-5 diagnoses between the AUDADIS-5 and PRISM-5 were fair-tomoderate (Hasin et al., 2015). Furthermore, test-retest reliability of past-year, prior-to-pastyear, and lifetime diagnoses were generally fair-to-good (Grant et al., 2015), with Kappa statistics ranging from .35 -.87 for the diagnostic variables, and intraclass correlation coefficients ranging from .45 -.85 for their respective continuous scales.

Attempted suicide:
To assess history of suicidal attempts, participants were asked "in your entire life did you ever attempt suicide?". Responses were scored using a dichotomous response format ('yes' = 1, 'no' = 0). This item has been used in previous versions of the NESARC and was found to have a moderate test-retest reliability across NESARC Wave 1 and Wave 2 (Palmetto and Link, 2010).

Perceived social support (PSS):
PSS was measured using the general population version of the 12-item Interpersonal Support Evaluation List (ISEL-12; Cohen et al., 1985). Half of the items are positive statements (e.g. "if I wanted to have lunch with someone, I could easily find someone to join me"), and the other half are negatively phrased (e.g. "if I were sick, I know I would find someone to help me with my daily chore"). The items are measured on a four-point Likert-scale ('definitely false' = 1, 'definitely true' = 4).
Higher scores indicate greater perceived social support. Responses were summed to create a composite PSS score ranging from 12-48. A unidimensional model has previously been found to be both valid and reliable (Merz et al., 2014). The internal consistency among the current sample was satisfactory (Cronbach's alpha = .85).

Sociodemographic variables:
Several sociodemographic variables were assessed including age, sex, marital status, residency (urban/rural), education, and past-year household income. Education was measured using 14 categories ranging from "no formal schooling" to "completed master's degree or equivalent or higher graduate degree".
Household income was measured using 21 categories that ranged from "less than $5,000" to "$200,000 or more".

Analytical Plan
The analytical plan was three-fold. First, prevalence rates of various psychiatric disorders were estimated among the older adult sample that met the symptom criteria for ICD-11 PTSD. Second, latent classes of PTSD comorbidity were examined through LCA (conducted using Mplus 7.4; Muthén and Muthén, 2012), using robust maximum likelihood classes were examined. 500 random sets of starting values were used followed by 100 final stage optimizations in order to avoid solutions based on local maxima. The fit of each latent class model was determined using several fit indices: the Akaike information criterion (AIC; Akaike, 1987), the Bayesian information criterion (BIC; Schwarz, 1978), the sample sizeadjusted BIC (ssaBIC; Sclove, 1987), entropy values, and the Lo-Mendell-Rubin adjusted likelihood ratio test (LMR-A; Lo, Mendell, and Rubin, 2001). Lower AIC, BIC, and ssaBIC values, and higher entropy values, indicate better model fit. A non-significant LMR-A value indicates that the model with one less class should be accepted. Previous Monte Carlo simulation studies indicated that the BIC is the best indicator for class enumeration (Nylund et al., 2007). Regarding the individual psychiatric disorders, and akin to studies of a similar nature (Burstein et al., 2012;Galatzer-Levy et al., 2013), a probability ≥ .15 indicated a class characteristic; a probability ≥ .15 and ≤ .59 was indicative of a moderate probability of comorbid diagnosis; and a probability ≥ .60 suggested that the disorder was highly probable within the respective class.
Third, a multinomial logistic regression was performed by regressing the latent classes (identified during the class enumeration process) onto several covariates (age, sex, marital status, residency, education, household income, social support, attempted suicide, and multiple forms of traumatic exposure). This was conducted using the R3STEP function in Mplus (Muthén and Muthén, 2012;Vermunt, 2010). This three-strep procedure involves first identifying the most appropriate latent classes; then obtaining the most likely class memberships based on the posterior probabilities of the LCA, while accounting for the classification uncertainty rate (i.e. measurement error); and finally the most likely class memberships are analysed with the covariates, thus accounting for at least some of the misclassification error (Asparouhov, and Muthén, 2014;Vermunt, 2010). This method is also preferable as it does not result in a shift in latent classes when the covariates are included.

ICD-11 PTSD Comorbidity
The sample characteristics and descriptive statistics of the variables used in the current study are reported in supplementary Table 1. Among the older adults who met symptom criteria for ICD-11 PTSD, the most common co-occurring disorder was MDD (40.5%), followed by AUD (30.1%), BPD (28.0%), and GAD (21.9%) (  Table 2 presents the LCA results for the different class solutions. Two diagnostics indicate that a two-class solution fit the data best: the BIC value and the LMR-A test finding.

Latent Class Analysis
However, a single further diagnostic favoured the three-class solution (the ssa-BIC). The twoclass solution was determined to be the most appropriate model on the grounds of statistical fit and parsimony. However, for the purposes of transparency, the three-class solution is included as supplementary material (see Supplementary Figure 1).

ICD-11 PTSD Patterns of Comorbidity Covariates
The results of the multinomial logistic regression are reported in Table 3

Discussion
There is limited research regarding ICD-11 PTSD comorbidity, especially with older adults. In line with emerging evidence from general population and clinical samples Karatzias et al., 2019;, we found that ICD-11 PTSD was highly comorbid with a range of lifetime psychiatric disorders. Moreover, despite an overall decline in psychiatric morbidity among older adults (Gum et al., 2009;Kessler et al., 2005;Thomas et al., 2016) and comorbidity with PTSD (Reynolds et al., 2016), we found that nearly four fifths of the sample met the diagnostic criteria for at least one other psychiatric disorder. This suggests that although there is a general downward trend in the prevalence of psychiatric disorders among older adults, psychiatric comorbidity remains high for those with PTSD.
Consistent with our second hypothesis, the highest rates of comorbidity were identified for other 'distress' disorders including MDD, BPD, and GAD. The notable exception to this pattern was the association with AUD which, according to the HiTOP model, sits within the 'externalising' dimension of psychopathology. A putative explanation for the increased probability of comorbid AUD is the self-medication hypothesis, which proposes that PTSD symptom manifestation precedes the development of drug/alcohol addiction and that individuals misuse drugs/alcohol in an attempt to assuage the distressing nature of their traumatic symptoms (Khantzian, 1997;Leeies et al., 2010). However, given the cross-sectional nature of the data, and that a lifetime traumatic exposure assessment was used, it is not possible to garner the precise temporal ordering to test the self-medication hypothesis. In other words, it is unclear whether trauma exposure and/or PTSD symptoms preceded the onset of substance abuse.
Among the different psychiatric disorders that were comorbid with PTSD, similar prevalence rates were found compared to the general population .
However, among the current sample of older adults, PTSD comorbidity was substantially lower among several of the internalising 'fear' disorders (consisting of social phobia, specific phobia, agoraphobia, and panic disorder), compared to the general population. Most notably, panic disorder and specific phobia. It is also important to note that a subset of the current sample is likely to have also met the requirements of CPTSD, the sibling disorder to PTSD in the ICD-11. However, given the available items in the AUDADIS-5 we were not able to extract sufficient items to represent this sibling disorder. A diagnosis of CPTSD may partially explain the observed comorbidity rates, most notably among MDD, GAD, and BPD (Frost et al., 2020;Karatzias et al., 2019). However, previous research has found similar comorbidity rates between PTSD and CPTSD regarding AUD and suicidality (Karatzias et al., 2019).
It is also possible that the PTSD symptoms are covered by diagnostic overshadowing, whereby symptoms are falsely attributed to PTSD but are in reality manifestations of another disorder (Hryvniak and Rosse, 1989). However, a strength of the ICD-11 model of PTSD focusing on the 'core PTSD' symptoms should, in theory, reduce the likelihood of diagnostic overshadowing. For example, the ICD-11 model does not include symptoms pertaining to difficulty concentrating, or the inability to recall certain aspects of a traumatic event.
Therefore, diagnostic overshadowing with other psychiatric disorder, or age-related conditions (e.g. dementia) are unlikely to be an issue within the context of the current study.
Given the heterogenous nature of trauma response, we applied a person-centred approach to examining patterns of PTSD comorbidity. Somewhat inconsistent with previous findings from the general population (Galatzer-Levy et al., 2013;Müller et al., 2014) and our third hypothesis, we found that a two-class solution was the most appropriate representation of latent patterns of PTSD comorbidity. This suggests that psychiatric comorbidity among older adults with PTSD may manifest in somewhat different ways than in the general population. These previous studies reported a moderate probability of MDD within the lower comorbidity class, however, we found that this class ('PTSD with moderate probabilities of depressive/alcohol use disorders') was characterised by an increased probability of AUD and MDD. This suggests that PTSD may carry an inherent risk for comorbid depressive/alcohol use disorders among older adults, even in the more resilient class.
The second class ('PTSD with general psychopathology') was characterised by increased likelihood of nearly all other disorders with the exception of ASPD and EDs. Individuals in this class had moderate or high probabilities of meeting requirements for diagnoses that cut across the internalising, externalising, and thought-disorder spectra of the HiTOP model. Thus, diagnostic comorbidity in this sample did not follow a simple dimensional-specific pattern; this finding is consistent with the HiTOP's description of correlated superordinate dimensions of psychopathology which ultimately reflect a general vulnerability to all forms of psychopathology (Kotov et al., 2017). This description could be evaluated in future analyses taking advantage of superordinate analytic approaches within LCA.
Although there is no consensus regarding the precise reason for the observed psychiatric morbidity and comorbidity differences among older adults compared to their younger and middle-aged counterparts, possible explanations include: (1) generational differences in one's willingness to disclose psychiatric symptoms due to fear of stigma (Cook and Simiola, 2018;Pless Kaiser et al., 2019); (2) that older adults are more likely to report psychological symptoms as somatic complaints (Cook and Simiola, 2018;Palmer et al., 1997;Pless Kaiser et al., 2019); (3) the inability to accurately report psychological symptoms due to cognitive impairments (Thomas et al., 2016); and (4) due to the distressing nature and concomitant health-risks of PTSD, and most notably PTSD and comorbid psychiatric diagnoses, it may possible that age-related differences reflect a form of survivor bias whereby individuals with PTSD are less likely to survive into older adulthood (Cook and Simiola, 2018;Thomas et al., 2016). These explanations may, in-part, account for the differences in comorbidity patterns among the current sample compared to the general population studies.
The higher rates of SPD in the second class may be explained through a number of different reasons. First, this high comorbidity may be due to the deleterious effects of trauma exposure as a shared risk factor of psychosis and PTSD (Bendall et al., 2008;Varese et al., 2012). Second, there may be an overlap in the clinical manifestation of PTSD symptoms and psychotic indicators (Seow et al., 2016). For example, re-experiencing the traumatic event (e.g. flashbacks) may be misconstrued as hallucinations, and hypervigilance may be misconstrued as paranoia (O'Conghaile, and DeLisi, 2015). Third, high comorbidity rates between PTSD and psychosis may be mediated through numerous other psychiatric comorbidities such as substance abuse and depressive symptoms (Sareen et al., 2005).
Consistent with our fourth hypothesis, those with a history of suicidal attempts were nearly three times more likely to belong to the class characterised by high rates of comorbidity. This is consistent with previous findings (Galatzer-Levy et al., 2013;Müller et al., 2014), and provides additional evidence that an increased burden of psychopathology is a risk factor of suicidality. Membership of the highly comorbid class was also associated with lower levels of social support. This predictor being associated with high comorbidity levels aligns with results found among general (non-PTSD) populations. For instance, there is a wealth of existing data demonstrating the important role that social support and social connection plays in maintaining good mental health (Wang et al., 2018). Moreover, we found that spousal/partner physical abuse was the only type of traumatic experience to predict membership of the higher comorbidity class. This finding is similar to Galatzer-Levy and colleagues (2013) who found that partner physical abuse was the only trauma-specific variable that predicted class membership for the two higher-comorbidity classes, compared to the low comorbidity class.
The findings of this study have important clinical and research implications. First, it was found that ICD-11 PTSD was a highly comorbid disorder among older adults. Clinicians working with geriatric populations should be aware of this high comorbidity among patients exhibiting symptoms of PTSD, as psychiatric diagnostic assessments in later life can carry difficulties such as older adults being less likely to endorse social or occupational impairment that may lead to psychiatric disorders being under-, or mis-diagnosed (see Bodner et al., 2018). Second, results provide evidence that there are distinct patterns of PTSD comorbidity among older adults. Clinicians working with older adults should be cognizant that increased rates of disorder comorbidity are associated with less social support, spousal/partner physical abuse, and history of suicide attempts. Older adults with multiple mental health problems require urgent and substantial clinical interventions. Third, the current findings provide additional evidence in support of the HiTOP model, and demonstrate that this framework may be useful for clinicians in understanding which disorders are most likely to exist comorbidly.
The findings of this study should be interpreted in light of several limitations. First, the current study used a nationally representative household sample of U.S. older adults therefore the generalisability to older adults in other nations, or in in-patient clinical settings, cannot be assumed. Second, the measure of ICD-11 PTSD did not consider the functional impairment criterion, and as such comorbidity rates may be overestimated. However, it has been argued that a limitation of current psychiatric diagnostic assessments, specifically related to the functional impairment criteria, may be inaccurate among older adults (Bodner et al., 2018). For example, older adults may be less likely to attribute social impairment to PTSD symptomatology if they are physically impaired, or occupational impairment if they are retired. Third, it will also be important to replicate this study using measures specifically designed to capture the ICD-11 PTSD diagnostic criteria (such as the ITQ; Cloitre et al., 2018), including items that explicitly refer to the "here-and-now" quality of the reexperiencing symptoms. Moreover, future research should aim to include an assessment of ICD-11 CPTSD. Fourth, as LCA is an exploratory and data-driven approach, future studies should aim to examine latent classes of PTSD within varying contexts to ensure that these findings are robust. Fifth, it should be noted that the measure of suicidal attempt history does not take into account factors such as the intensity or outcome of the attempted suicide (e.g. serious life-threatening injuries/hospitalisation). Moreover, due to the sensitive nature of this question, it is possible that the prevalence rate was under-estimated. Sixth, the measure of comorbidity used was a count variable denoting the presence/absence of a disorder and was not weighted to reflect differences in the functional impairment of the disorder. The limitation of such a counting approach is that all comorbid psychiatric disorders are incorrectly assumed to be equal in terms of distress/functional impairment. Seventh, as the measures of psychiatric comorbidity used were lifetime assessments, it was not possible to precisely infer temporal ordering among the disorders. For example, whether PTSD symptoms had abated before the onset of the other forms of psychopathology. As such, future research should aim to address this limitation by employing longitudinal methods of psychiatric assessment.
In summary, previous research has found that diagnostic psychiatric comorbidity reduces among older adults. Despite this reduction among older adults, it was found that ICD-11 PTSD appears to remain a highly comorbid disorder in later life. Diagnostic psychiatric comorbidity was most notably observed among other internalising 'distress' disorders (MDD, BPD, and GAD) and AUD. Similar to studies among the general population, the latent class with higher diagnostic psychiatric comorbidity was associated with a greater likelihood of having a history of suicidal attempts. Moreover, these findings provide a useful addition to the literature pertaining to both the structure of general psychopathology and the structure of ICD-11 PTSD comorbidity. These findings demonstrate the importance of identifying early signs of maladaptive posttraumatic responses and can inform clinicians and researchers of potential comorbid subtypes that may manifest as a response to trauma. Note: n = 530; a = Percentages are adjusted for the complex survey design of the NESARC-III, based on the stratification, clustering, and weighting of the study population; b = sample size is based on the unweighted data.  Note: Class 1 = PTSD with moderate probabilities of depressive/alcohol use disorders; Class 2 = PTSD with general psychopathology; OR = odds ratio; 95% CI = 95% Confidence Intervals; a = sex coded as 0 = male, 1 = female; b = marital status coded as 0 = not married/cohabiting, 1 = married/cohabiting; c = residency coded as 0 = urban, 1 = rural. Statistical significance: *p < .05.