Treatment Receipt Patterns Among Individuals with Co-Occurring PTSD and Substance Use Disorders

Abstract

Objective:

To determine latent classes of treatment-receipt among people with comorbid Posttraumatic Stress Disorder (PTSD) and Substance Use Disorder (SUD) and describe each class by demographics, disease characteristics, and psychiatric diagnoses.

Method:

Participants were NESARC-III respondents with lifetime PTSD and SUD (n = 1,349; mean age 40.3; 62.5% female; 30.9% non-White or Hispanic-White). Cross-sectional data were collected using the DSM-5 Alcohol Use Disorder and Associated Disabilities Interview Schedule. Latent class analysis was used to identify subgroups of participants with different patterns of treatment receipt.

Results:

36% of patients received at least one SUD treatment while 79% received at least one mental health (MH) treatment. Six latent classes were identified: No Treatment (17.3%); Outpatient MH (34.0%); Outpatient + Inpatient MH (17.9%); SUD (7.3%); SUD + Outpatient MH (15.7%), and SUD + Outpatient MH + Inpatient MH (7.7%). The SUD treatment classes evidenced greater social instability, higher alcohol use disorder symptom severity, and used more drug types than the non-SUD classes. Classes receiving inpatient MH treatment had greater incidence of additional comorbid conditions and suicidal behaviors. Across all six classes, most respondents met diagnostic criteria for chronic PTSD (overall: 68.9%) while fewer met diagnostic criteria for chronic SUD (overall: 38.7%).

Conclusions:

Most people with lifetime PTSD and SUD have sought either SUD or MH treatment or both, with substantially greater receipt of MH treatment. This comorbid group has complex clinical presentations that differ depending upon treatment subgroup, and for most, their comorbid disorders persisted despite high rates of treatment engagement.

Posttraumatic stress disorder (PTSD) frequently co-occurs with a host of other psychiatric disorders (Goldstein et al., 2016), including substance use disorders (SUD; Blanco, Xu, Brady, Perez-Fuentes, Okuda, & Wang, 2011, Mills, Teeson, Ross, Peters, 2006; Smith, Goldstein, & Grant, 2016). A recent study using data from the National Epidemiologic Survey on Alcohol and Related Conditions-III (NESARC-III) found that 58% of those with lifetime PTSD had a co-occurring lifetime alcohol use disorder (AUD) and/or a drug use disorder (DUD; Simpson, Rise, Browne, Lehavot, & Kaysen, 2019). Of those with lifetime PTSD/AUD and PTSD/DUD, 80% and 87%, respectively, reported having received some form of SUD and/or mental health (MH) treatment, rates that were significantly and substantially higher than those reported by the three single disorder groups (AUD: 32%; DUD: 57%; PTSD: 70%; see also Blanco et al., 2011; van den Berk-Clark & Silver Wolf, 2017). Studies using other epidemiologic data sources have also found that individuals with co-occurring mental health conditions and SUD are more apt to seek treatment than those with SUD only (Edlund, Booth, & Hahn, 2012; Harris & Edlund, 2005; Watkins, Burnam, Kung, & Paddock, 2001).

While it appears that individuals with SUD and psychiatric comorbidities, including PTSD, engage with treatment at high rates, little is known about where they receive treatment and the type(s) of treatment they receive. Further, the demographic and clinical correlates of various treatment seeking patterns are poorly understood. Learning which individuals seek what types of treatments is critical because such information would enable policy makers and healthcare systems to deploy empirically supported treatments (Back et al., 2019; Roberts et al., 2015; Ruglass et al., 2018; Simpson, Lehavot, & Petrakis, 2017) where individuals with PTSD/SUD are most apt to seek care.

Those with co-occurring PTSD and SUD could pursue care in either MH or SUD treatment settings, in both, or neither. Whether treatment is received likely depends on the severity of the two disorders, the availability of care, and individuals’ available resources for accessing care (Dhingra, Zack, Strine, Pearson, & Balluz, 2010; Walker, Cummings, Hockenberry, & Druss, 2015; see also Ali, Teich, & Mutter, 2016). What type of treatment is received likely depends on people’s ability to access various types of care combined with which aspect of their comorbidity they find more distressing or perceive to be causal (Back et al., 2014; Gielen et al., 2016). There is evidence that more individuals with various mental health conditions co-occurring with SUD pursue MH-oriented treatment than SUD-oriented treatment (Edlund et al., 2012; Han, Compton, Blanco, & Colpe, 2017; Harris & Edlund, 2005; Manuel, Stebbins, & Wu, 2016; Watkins et al., 2001), and this was found to be true for NESARC-III participants with co-occurring PTSD and SUD (Simpson et al., 2019).

Despite the consistency of these findings across the literature, there is little specific information about the types of treatment people receive in part because there is marked heterogeneity in both SUD and MH treatment that heretofore has been collapsed into those two general categories. Both SUD and MH treatment vary with regard to setting and intensity (e.g., outpatient visits vs. inpatient admissions), formality (e.g., contact with clinical providers vs. self-help), and use of medication. Thus, what constituted SUD treatment in previous studies may have encapsulated both best practices (i.e., individual or group psychotherapy with medication management) and less effective approaches (e.g., a detox visit and no structured aftercare), and such differences may be patterned across demographically meaningful characteristics, such as economic disadvantage (Ilgen, et al., 2011). Additionally, research in this area has not evaluated whether different treatment receipt patterns are associated with increased or decreased likelihood of remission from these disorders, an issue of critical concern given the chronicity often associated with both PTSD (see Cougle, Resnick, & Kilpatrick, 2013; Magruder et al., 2016; North & Oliver, 2013; Pietrzak, Feder, Singh, & Schechter, 2014) and SUD (McCabe, West, Strobbe, & Boyd, 2018).

In order to address these gaps in the literature, the current exploratory study used Latent Class Analysis (LCA) to characterize classes of NESARC-III participants with lifetime PTSD/SUD across different patterns of lifetime treatment receipt (e.g., SUD: 12-step, inpatient SUD treatment, social services, outpatient SUD treatment, private provider/Employee Assistance Program [EAP] providers, clergy, emergency room [ER]; MH: self-help, outpatient psychotherapy, inpatient psychiatric treatment, ER, medication). Based on prior literature, it was anticipated that there would be a small group of respondents that had received no treatment, a somewhat larger group that received some form of SUD treatment with or without additional MH treatment, and a substantial group that received MH treatment only (Blanco et al., 2011; Edlund et al., 2012; van den Berk-Clark & Silver Wolf, 2017). While it seemed likely the LCA would discriminate treatment receipt based on care intensity (i.e., outpatient vs. inpatient) and formality (i.e., self-help vs. provider driven), there was not a specific a priori sense of how these factors would influence the results.

Once latent classes were identified, our study aimed to identify between-group comparisons on demographics, functional stability, chronicity and past-year diagnostic status, additional psychiatric comorbidities, and risk indicators (e.g., suicidal ideation and past suicide attempts). While formal a priori hypotheses were not specified given the exploratory nature of the LCA and the subsequent comparisons between all of the to-be determined classes, some individual characteristics were anticipated to differentiate between latent classes. For example, it was expected that severity of the respective disorders would be associated with receipt of SUD or MH care (i.e., those with more severe SUD would be especially likely to receive SUD care and those with more severe PTSD would be especially likely to receive MH care). Additionally, based on prior findings it was expected that women would likely be disproportionately represented in the MH care class(es) and men in the SUD care class(es) (Edlund et al., 2012; Gilbert et al., 2019; Manuel et al., 2016).

Methods

Study Sample and Procedures

The NESARC-III includes 36,309 community-dwelling US residents aged 18 and older not currently on active-duty in the military (Grant et al., 2014). Data collection took place from 2012 to 2013 and used multi-stage probability sampling to include a representative sample. Counties or groups of contiguous counties were the primary sampling units; census blocks the secondary sampling units, and households within census blocks the tertiary sampling units. Eligible adults were randomly selected within households (Grant et al., 2015). Several racial/ethnic minority groups (Hispanic, Black, Asian) were oversampled. The data were weighted to represent the US population and adjusted for non-response. The overall response rate was 60.1%, similar to other US national surveys (American Community Survey, 2012). The current study involves respondents who met criteria for both lifetime PTSD and a lifetime SUD (n = 1,349) with the following SUDs included: alcohol, sedative/tranquilizers, cannabis, stimulants, cocaine, non-heroin opioid, heroin, hallucinogens, club drugs, and solvent/inhalants (note: tobacco use disorder was not included). NESARC-III data were collected via in-person structured interviews by non-clinicians.

The VA Puget Sound Human Subjects IRB determined the study exempt from oversight.

Measures

The Alcohol Use Disorder and Associated Disabilities Interview Schedule (AUDADIS) for DSM-5 was used to collect all data (Grant et al., 2014). The primary variables included in the current study are lifetime and past-year diagnostic status for PTSD, AUD, and DUD and receipt of specific forms of MH and SUD treatment (yes/no; treatment types are delineated below).

We used the lifetime and past-year indicators of PTSD, AUD, and DUD status provided by NESARC-III. Both the AUD and DUD diagnostic indicators conform to the eleven DSM-5 symptoms for SUD disorders. The NESARC-III PTSD diagnostic indicator used slightly different criteria from the final DSM-5 criteria; it required one additional symptom for criteria D (negative alterations in mood and cognitions) and criteria E (alterations in hyperarousal and reactivity). Sensitivity checks (Lehavot, Katon, Chen, Fortney, & Simpson, 2018; Simpson et al., 2019) have found no differences in the individuals identified with either definition.

Utilization of the same types of treatment was assessed via the AUDADIS for both AUD and DUD (with the exception of methadone maintenance, which was assessed for DUD only but was omitted from the LCA models due to minimal endorsement). Endorsement of treatment types was combined such that a “yes” response on either or both the AUD or the DUD indicator was counted as “yes.” Thirteen different types of SUD treatment were queried and these were consolidated into eight subcategories to facilitate an interpretable LCA result and to manage low counts in some treatment types: 12-step/self-help, inpatient, social services, outpatient, private provider/EAP providers, clergy, ER, or other (see Table S1 for details).

Similarly, utilization of the same treatment variables for the various mental health conditions was assessed, and these were counted as affirmative if the respondent said “yes” regarding the treatment type for any of the mental health conditions. The mental health conditions assessed by NESARC-III for which respondents may have received treatment include depression, dysthymia, bipolar disorder, agoraphobia, social anxiety disorder, specific phobia, panic disorder, generalized anxiety disorder, and PTSD. We opted to include treatment received for any mental health condition rather than solely for PTSD because PTSD is associated with multiple comorbidities (Goldstein et al., 2016) and to focus only on PTSD-related treatment would underestimate actual care received. Information regarding which condition(s) respondents reported seeking any MH care for was also described. The AUDADIS included five types of MH treatment (see Table S1) and each was used as originally specified: self-help, outpatient psychotherapy, inpatient, ER, or medication.

Information regarding the timing of receipt of SUD care and receipt of MH care was extracted from the AUDADIS to form the following four indicators: no treatment, past year only, prior to past year only, and both past year and prior to past year. This information was tallied for any SUD care, any PTSD care, and any non-PTSD mental health care.

In order to describe the latent treatment classes, demographic information, trauma type indicators, diagnostic status on the 10 additional lifetime mental health comorbidities assessed (i.e., depression, dysthymia, bipolar disorder, agoraphobia, social anxiety disorder, specific phobia, panic disorder, generalized anxiety disorder, and schizotypal, borderline, and antisocial personality disorders), past-year suicidal ideation, and lifetime suicide attempt status were utilized. Following Amato (2014), an instability index was created by adding the number of affirmative responses regarding no college, lifetime legal problems, no employment, income below poverty line, public assistance, food stamps, and no private insurance (range: 0–7), with the last five anchored to the past year. Past-year homelessness was examined separately because it represents an appreciably more severe degree of social instability than the other indicators.

In addition, information from the SUD and PTSD modules was used to characterize the order of onset of the two disorders for each person (i.e., PTSD before SUD, SUD before PTSD, and same-age onset) and severity of the disorders. Each of the three order of onset options was coded yes/no for each individual. Severity was gauged by adding the number of DSM-5 symptoms endorsed for PTSD and AUD. Degree of drug use involvement was approximated by the number of DUD diagnoses, including sedative/tranquilizers, cannabis, stimulants, cocaine, non-heroin opioid, heroin, hallucinogens, club drugs, and solvent/inhalants. Although both PTSD and SUD diagnoses can remit and recur over time and therefore may not be continuous, they may still be considered chronic and thus we defined chronic PTSD and SUD as the presence of positive diagnostic status for both past year and prior to past year. Chronic SUD could refer to the same SUD being present in both timeframes or to two different SUDs (e.g., prior to past year cocaine use disorder and past year AUD), as we deemed the presence of persistent SUD-related signs and symptoms to be the clinically relevant issue rather than substance consistency. New onset of PTSD or SUD was defined as presence of only past-year diagnostic status in the absence of prior to past-year diagnostic status.

Data Analysis

Latent Class Analysis

LCA (McCutcheon, 1987) was used to identify classes of people with different patterns of treatment receipt among individuals with comorbid PTSD and SUD. LCA is a latent variable model that probabilistically categorizes people into latent classes defined by internally similar responses to a set of indicator variables (e.g., types of treatment).

Our analysis had two phases. In the class enumeration phase, we determined the optimal number of classes to characterize different patterns of treatment-seeking. We chose the number by considering fit to the data, parsimony, and substantive interpretability (Nylund, Asparouhov, & Muthen, 2007; Masyn, 2013). In the participant characteristics phase, we described each latent class by its demographics, disease characteristics, and psychiatric diagnoses. We incorporated these additional characteristics into the model using the 2-step method (Bakk & Kuha, 2017) to account for classification uncertainty. In this second phase of analysis, parameters of the latent class measurement model were held fixed at the values obtained in the class enumeration phase while the additional participant characteristics were freely estimated. A Wald chi-square test was used to block-test whether each characteristic was significantly different across classes. When the block test was significant at p < .05 after adjusting for multiple comparisons, between-class differences were probed by comparing the characteristics across all possible pairs of classes.

Multiple Comparisons

Due to the large number of comparisons in the participant characteristics phase of analysis, we managed Type I error using the Benjamini-Hochberg correction (Benjamini & Hochberg, 1995; Cribbie, 2007) with the false discovery rate set to 5%. We adjusted both block tests and pairwise comparisons.

Survey Weights

To account for multistage probability sampling and reweighting of the model to reflect the US civilian population, we included sampling weights and stratification variables in the final model (Grant et al., 2015).

Missing Data

Because we used a maximum likelihood estimator, estimates were unbiased to the extent reasons for missingness were included in the model (i.e., missing at random; Enders, 2010). For 54 of the 57 variables considered in this study, the rate of missingness was 2% or less. The exceptions were the three variables assessing order of onset among the PTSD and SUD diagnoses, which were missing 220/1349 observations (15%). Muthen and colleagues suggested that bias levels less than 10% to 15% are unlikely to be problematic in latent variable models (Muthen, Kaplan, & Hollis, 1987), and simulation studies have found that even under extreme conditions, bias is well below 10% for levels of missingness up to 15% (Enders, 2010). Thus, bias due to nonrandom missingness is expected to be minimal in the present study.

Analyses were conducted in Mplus version 8.2 (Muthen & Muthen, 1998–2017).

Results

Overall Treatment-Seeking

Sample-wide rates of treatment-seeking are presented in Table 1. Fewer participants received at least one type of SUD treatment (36%) than at least one type of mental health treatment (84%). Among SUD treatments, 12-step programs were the most common, followed by detox or inpatient stay, then outpatient treatment. Most participants in this sample attended psychotherapy (78%), and more than half used a medication to treat mental health symptoms. Additionally, across the entire sample, more individuals reported receiving MH care for a depressive disorder (66%) than for PTSD (61%) and more than half received care for an anxiety disorder (56%), while relatively few reported receiving care for either bipolar disorder or any type of eating disorder.

Table 1.

Overall and class-specific proportions for each type of treatment

Type of Treatment	Overall	No Treatment (n ~ 234; 17.3%)	OP MH Only (n ~ 459; 34.0%)	OP + IP MH (n ~ 242; 17.9%)	SUD Only (n ~ 99; 7.3%)	SUD + OP MH (n ~ 212; 15.7%)	SUD + OP + IP MH (n ~ 104; 7.7%)
SUD Treatments
12-Step	.27	.02	.04	.04	.84	.75	.91
Detox/Inpatient/Rehab	.23	.00	.00	.00	.62	.69	.98
EAP/Private Provider	.22	.00	.04	.03	.36	.64	.93
Outpatient	.16	.01	.00	.01	.33	.43	.80
ER-SUD	.15	.00	.00	.02	.31	.39	.84
Social or Family Services	.11	.00	.00	.00	.19	.31	.56
Clergy	.08	.00	.00	.00	.26	.18	.37
Other/Crisis/Half-way House	.11	.01	.01	.01	.26	.17	.68
Forms of Mental Health Treatments
Therapy	.78	.06	1.00	.97	.45	.97	1.00
Medication	.67	.11	.74	.90	.07	.99	.98
Self-Help	.32	.04	.27	.47	.20	.40	.76
ER-Mental Health	.32	.07	.09	.77	.01	.46	.83
Inpatient-Mental Health	.28	.02	.00	.78	.03	.41	.88
Mental Health Treatment By Disorder (Any Form)
PTSD	.61	.04	.70	.83	.33	.73	.86
Depression or dysthmia	.66	.10	.75	.85	.25	.84	.96
Bipolar	.18	.01	.13	.25	.00	.27	.53
Any anxiety disorder	.56	.09	.57	.80	.12	.73	.93
Any eating disorder	.07	.00	.07	.14	.01	.07	.12
Any SUD*	.36	.00	.03	.01	1.00	1.00	1.00
Any Mental Health*	.84	.20	1.00	1.00	.50	1.00	1.00

Note. The Mental Health Treatment by Disorder (Any Form) variables and the Any SUD and Any Mental Health variables were not included as indicators in the latent class model but rather calculated post-hoc.

OP = Outpatient; IP = Inpatient; MH = Mental Health; SUD = Substance Use Disorder

Latent Class Enumeration

We chose the six-class solution as the best blend of fit, parsimony, and interpretability. Fit statistics are presented in supplementary Table S2. Notably, the six-class solution was preferred by the three fit criteria that perform best in mixture models, the Bayesian Information Criterion, Consistent Akaike Information Criterion, and adjusted Lo-Mendell-Rubin Likelihood Ratio Test (Nylund et al., 2007). People were classified with high precision (entropy = .87) and latent classes had clear qualitative separation, with large differences in the type of services used by each class. The following six classes were identified: No Treatment (17.3%); Outpatient MH (34.0%); Outpatient + Inpatient MH (17.9%); SUD (7.3%); SUD + Outpatient MH (15.7%), and SUD + Outpatient MH + Inpatient MH (7.7%). The LCA solution did not make qualitative distinctions among types of SUD treatment and so SUD refers to receipt of any of the 8 types of SUD treatment included in the model.

The six latent classes can be further categorized based on receipt of SUD treatment: three that did not receive SUD treatment and three that did receive SUD treatment (see Figure 1). Additionally, each class was characterized by receipt of outpatient and/or inpatient MH treatment utilization. Within the classes that did not receive SUD treatment, the No Treatment class reported receiving no to minimal MH or SUD services (i.e., in this class up to 11% received some form of MH care and 2% received some form of SUD care), while almost all people in the Outpatient MH class received outpatient therapy and 74% were prescribed a medication for a mental health condition; very few used self-help resources (see Table 1). Of those in the Outpatient MH + Inpatient MH class, nearly 8 in 10 visited an ER and/or an inpatient unit for mental health reasons.

Figure 1.

Patterns of treatment receipt by latent class

Note. The three mental health classes are shown on the left. The three mental health + SUD classes are shown on the right.

Within the three classes that received SUD treatment, people used a variety of SUD-oriented services (see Table 1). Specifically, the SUD Treatment class was comprised of people who had primarily attended 12-step programs, SUD inpatient care, and who may have participated in an additional one or two other SUD services. Forty-five percent of those in this class also received some outpatient MH counseling. Nearly all members of the SUD + Outpatient MH class used both outpatient psychotherapy and medication. The SUD + Outpatient MH + Inpatient MH class endorsed the highest rates of treatment exposure across all services.

Participant Characteristics

Omnibus tests adjusting for a false discovery rate of 5% indicated statistically significant differences across classes in 33 of the 45 characteristics examined, meaning that we examined 495 possible pairwise comparisons across classes. Overviews of between-class differences in demographics and disease characteristics may be seen in Figures 2a and 2b (corresponding to Supplementary Tables 3a and 3b). Overviews of between-class differences regarding PTSD/SUD lifetime, prior to past year and past year (i.e., chronic), and new onset diagnostic statuses can be seen in Figure 3a, and additional psychiatric comorbidities and risk indicators in Figure 3b (see corresponding Supplementary Tables 4a and 4b).

Figure 2.

Forest plots depicting differences in demographics (Panel A) and disease characteristics (Panel B) across latent classes

Note. Standardized means were used for continuous variables (top axis; above the horizontal line) and percentages were used for categorical variables (bottom axis; below the horizontal lines)

Figure 3.

Diagnostic status for PTSD, SUD (Panel A) and additional psychiatric comorbidities (Panel B) across latent classes

Note. IPV = Intimate Partner Violence. SI = Suicidal Ideation. GAD = Generalized Anxiety Disorder.

Demographics and Social Instability (Figure 2a; Supplementary Table 3a)

Broadly, differences in demographic patterning were most evident in latent classes that received SUD care versus those that did not. The three SUD classes tended to be comprised of people who were older, had higher levels of social instability, and were men, whereas the MH-only classes tended to have more women. Although the three SUD classes had the highest levels of social instability, it is notable that the No Treatment class had higher levels of social instability than those in the Outpatient MH class. The No Treatment and SUD Only classes had the largest proportion of racial and ethnic minority participants, followed fairly closely by the SUD + Outpatient MH + Inpatient MH class. Those in SUD + Outpatient MH + Inpatient MH class were significantly more likely to have been homeless in the past year than those in the No Treatment and non-SUD treatment classes.

PTSD and SUD Disease Characteristics (Figure 2b; Supplementary Table 3b)

The strongest patterning evident in these results was that SUD and non-SUD treatment classes were distinguished by AUD severity and number of DUDs, with SUD classes reporting higher AUD severity and a greater number of DUDs than non-SUD classes. A second pattern showed graded increases in treatment receipt by severity of PTSD characteristics, irrespective of SUD treatment status. For example, within both the SUD and non-SUD classes, participants tended to receive more services—and particularly more inpatient services—with higher PTSD severity, more traumas, and a history of child abuse. That same graded response pattern was not seen across levels of AUD severity or number of DUDs. Order of diagnosis onset (e.g., PTSD before SUD, etc.) did not differ across classes.

Diagnostic Status for PTSD and SUD (Figure 3a, Supplementary Table 4a)

While nearly 69% of people had a chronic PTSD diagnosis, only 39% had a chronic SUD diagnosis. The largest differences across classes appeared to be related to the presence of a lifetime DUD, which was associated with more SUD-specific treatment (although not necessarily more intensive treatment within SUD classes). Rates of chronic PTSD conjoint with chronic AUD did not significantly differ across latent classes; however, when PTSD alone was the chronic diagnosis, individuals were particularly likely to have received MH treatment (i.e., Outpatient MH Only, Outpatient + Inpatient MH, SUD + Outpatient MH classes) relative to the No Treatment or SUD Only classes. Additionally, those in the SUD and the SUD + Outpatient + Inpatient MH classes were more likely to have a chronic SUD than those in the Outpatient MH Only class. Participants with new onset diagnoses were most likely to belong to the No Treatment class, followed by classes characterized by MH treatments; participants with new onset diagnoses were among the least likely people in the sample to receive SUD treatments.

Psychiatric Comorbidities (Figure 3b, Supplementary Table 4b)

Generally, individuals with more comorbidities received more intensive levels of care within both the MH and SUD classes. This pattern was strongest for those participants with suicidal ideation, suicide attempts, and Borderline Personality Disorder. The two classes receiving inpatient treatment (Outpatient + Inpatient MH and SUD Treatment or SUD + Outpatient MH + Inpatient MH) —regardless of SUD treatment status—clustered together with similar rates for any particular comorbidity except for Bipolar and Antisocial Personality Disorder. People in the No Treatment class had the fewest comorbidities, although their levels were still moderately high for depression (37.7%), Borderline (56.4%), and Schizotypal Personality Disorders (40.2%).

Who Went Where? (Table 2)

Table 2.

Who went where? Patient characteristics (%) by latent class treatment groupings (N = 1,349).

	No Treatment (n ~ 234; 17.3%)	OP MH Only (n ~ 459; 34.0%)	OP + IP MH (n ~ 242; 17.9%)	SUD Only (n ~ 99; 7.3%)	SUD + OP MH (n ~ 212; 15.7%)	SUD + OP + IP MH (n ~ 104; 7.7%)	Class characterized by SUD Treatment (n ~ 415; 30.7%)	Class characterized by MH Treatment (n ~ 1017; 75.3%)
Sex
Female	14.3	38.0	23.0	4.8	14.6	5.4	24.8	81.0
Male	22.4	27.4	9.5	11.5	17.6	11.6	40.6	66.0
Sexual Orientation
Sexual Minority	10.3	41.4	19.2	7.8	11.3	10.4	29.5	82.2
Heterosexual	18.2	33.1	17.8	7.3	16.2	7.4	30.9	74.5
Race and Ethnicity
Non-White/Hispanic	22.4	26.6	19.0	9.1	14.4	8.3	31.9	68.3
White	15.0	37.3	17.4	6.5	16.3	7.5	30.2	78.5
Veteran Status
Veteran	16.8	25.7	15.9	8.3	20.8	12.8	41.8	75.1
Non-Veteran	17.4	35.1	18.2	7.2	15.1	7.1	29.4	75.4
Homelessness
Homeless past-year	7.7	26.0	12.9	6.5	23.8	22.6	52.9	85.4
Stable Housing	18.1	34.7	18.3	7.4	15.0	6.5	28.9	74.5
Suicide Attempt
Suicide Attempt	7.9	19.8	32.8	5.4	16.4	17.5	39.2	86.5
No Suicide Attempt	22.0	41.0	10.6	8.3	15.4	3.0	26.6	69.9
Suicidal Ideation
Suicidal Ideation	8.5	29.6	25.9	4.4	18.4	13.2	35.9	87.1
No Suicidal Ideation	26.4	38.5	9.7	10.4	13.0	2.2	25.5	63.3
Lifetime Diagnostic Status
PTSD/DUD	16.6	25.1	15.9	9.7	20.9	11.7	42.3	73.5
PTSD/AUD only	18.1	43.5	20.1	4.8	10.1	3.5	18.4	77.3

OP = Outpatient; IP = Inpatient; MH = Mental Health; SUD = Substance Use Disorder

In addition to detailing the constitution of each latent class (e.g., 51.5% of participants in the No Treatment class were women), we selected eight participant characteristics to show the type of treatment received by specific groups of people (e.g., 14.3% of women and 22.4% of men were in the No Treatment class). (Note: this information is presented descriptively and was not included in the class enumeration analysis.) Approximately 42% of people with a DUD received SUD-specific treatment, compared to only 18.0% of people with AUD-only. More men than women attended SUD treatment, whereas more women attended MH treatment. Although White and non-White/Hispanic people were roughly equally likely to attend SUD treatment, more White people received some form of MH treatment. People experiencing homelessness in the past year received more treatment in general, and this was largely driven by receipt of SUD treatment. Veterans and non-veterans were equally likely to have received MH treatment but veterans were more likely to have received SUD treatment than non-veterans, a difference likely driven by gender differences between the two groups.

Sensitivity Analysis

We were concerned about the possibility of overestimating the chronicity of PTSD and SUD and underestimating the possible influence of treatment due to either (1) ongoing but incomplete courses of treatment or (2) differing rates of PTSD-specific treatment across classes. To further investigate this, we first examined the timing of treatment across latent classes (Supplementary Table 5) and found that very few participants (≤ 4% for each treatment type) had treatment in the past year alone, suggesting ample opportunity to complete at least one course of treatment. We then examined receipt of PTSD treatment compared to receipt of other MH treatment and found they were similar across latent classes (Supplementary Figure 1), suggesting that differential types of MH treatment did not drive differing chronicity rates across classes.

Discussion

In this large, nationally representative sample, most people with lifetime PTSD/SUD received either SUD or MH treatment or both, with approximately 83% having been categorized to one of the treatment classes. This high rate of treatment engagement is somewhat larger than was found for NESARC-III respondents with PTSD and no SUD (70%) and markedly greater than what was found for those with SUD and no PTSD (AUD: 32%; DUD: 57%; Simpson et al. 2019; see also Spoont, Murdoch, Hodges, & Nugent, 2010), suggesting that the comorbid group appears to have a higher rate of lifetime receipt of care than single disorder groups. In the current sample, MH treatment generally entailed therapy and medications with little use of self-help options, while SUD treatment was more variable; most people attended a 12-step program and some form of inpatient SUD care (i.e., detoxification, inpatient for SUD, and/or residential rehabilitation), along with a mix of private or EAP therapy, outpatient programming, family and social services, and ER visits. Similar to other studies involving comorbid samples (Edlund et al., 2012; Harris & Edlund, 2005; Watkins et al., 2001), across the entire sample, receipt of MH treatment was more common than SUD treatment (84% and 36%, respectively). Our review of disorder-specific treatment types suggests that receipt of MH treatment beyond that for only PTSD was the norm in this sample and further, we found that very few participants received treatment in the past year alone.

Beyond these global treatment patterns, we found that six latent classes best characterized participants’ treatment receipt and that the classes varied more by treatment intensity than formality (i.e., the LCA did not distinguish self-help options from those provided in formal clinical or healthcare settings). Latent classes were differentiated by receipt/non-receipt of SUD treatment (three classes totaling 30.7% of participants were characterized by receipt of SUD treatment; three classes totaling 69.3% of participants had negligible amounts of SUD treatment) with each treatment receipt class further differentiated by whether they received outpatient and/or inpatient MH care.

We found substantial differences across latent classes with regards to demographics, social stability, symptom severity, chronicity of one or more of the diagnoses comprising the PTSD/SUD comorbidity, and the presence/absence of additional psychiatric comorbidities. Not surprisingly, those who received no treatment generally had the fewest number of comorbidities, but relative to those receiving only outpatient MH treatment, they evidenced greater social instability and were more likely to be members of a racial or ethnic minority group. These findings suggest that while some in the no treatment class may have actively chosen not to seek care, others may have faced financial, practical, and/or cultural barriers preventing them from accessing needed care (Lee-Tauler, Eun, Corbett, & Collins, 2018). We also found that people who received any MH treatment were more likely to be White/non-Hispanic than non-White/Hispanic, which is consistent with results from the National Survey on Drug Use and Health indicating that Black respondents with SUD and psychiatric comorbidities were at least as likely to receive SUD treatment as White respondents, but far less likely to receive MH treatment (Nam, Matejkowski, & Lee, 2017; see also Alvanzo, et al., 2014; Creedon & Cook, 2016).

There were also noteworthy findings with respect to other key individual characteristics. People who received SUD treatment were older on average than those who did not, suggesting that perhaps as people with this comorbidity age and continue to have difficulties with substances, they become more apt to seek out SUD treatment (Alvanzo et al., 2014). This treatment-seeking pattern contrasts with research involving general SUD samples, which has found that those in older age cohorts are less likely to seek SUD treatment than those in younger age cohorts (Blanco et al., 2015; Ilgen et al., 2011). Perhaps the presence of PTSD symptoms suggests a path to treatment that is viewed as less stigmatizing, with SUD treatment only pursued later (i.e., when people are older) in the face of SUD issues that persist in spite of MH treatment receipt. In addition, SUD and MH treatment receipt patterns for women and men were consistent with other literature; MH treatment was more likely to have been accessed by women and SUD treatment was more likely to have been accessed by men (Edlund et al., 2012; Gilbert et al., 2019; Manuel et al., 2016). The present study also found that those receiving SUD treatment showed greater social instability than those receiving no treatment or MH treatment only, and the two classes receiving both SUD and MH treatment were particularly likely to have experienced homelessness in the past year.

As expected, people receiving SUD treatment had greater AUD severity and had a larger number of DUDs than those receiving MH treatment only. Additionally, presence of a DUD was particularly associated with receipt of SUD treatment (42.3%), while presence of AUD without DUD was not (18.4%). These results suggest that people with PTSD/SUD may appropriately self-select for SUD treatment based on their SUD severity (Tuithof, ten Have, van den Brink, Vollebergh, & Graaf, 2016). Along these same lines, those receiving inpatient MH treatment had more severe PTSD, higher rates of additional disorders, and more suicide attempts than those receiving only outpatient MH treatment and/or SUD treatment, again suggesting that reasonable triage is taking place such that those with more serious impairment and acute safety issues are receiving higher levels of care.

The current results also indicate that not only is SUD severity relatively low among those in the MH treatment classes, but there are not appreciable differences along this parameter among treatment classes receiving only outpatient MH treatment versus both inpatient and outpatient MH treatment. Additionally, we found that the predominately MH treatment classes had lower chronic SUD rates than the SUD treatment classes. The latter results may suggest that some individuals with PTSD/SUD, particularly those with lower SUD severity, may be able to successfully address their SUD issues in MH treatment contexts. For example, it is possible that having the opportunity to address one’s PTSD mitigates reliance on substance use to cope (Lopez-Castro, Hu, Papini, Ruglass, & Hien, 2015) or that the acquisition of new coping skills and strategies in MH treatment helps ameliorate substance-related problems. However, because we are not able to ascertain the temporal sequence of receipt of MH care vis a vie onset (or recovery from) of SUD, this supposition needs to be tested in future research.

Although findings suggest that people largely self-selected or were referred into appropriate levels of treatment, the actual treatment received may have been inadequate. As can be seen in Supplemental Table 4b, 69% of the current sample had chronic PTSD and 39% had a chronic SUD diagnosis, indicating large proportions of people who maintained their diagnoses even after receiving fairly intensive services. Of note, very few people reported receipt of SUD, PTSD, or other MH treatment only in the past year (≤ 4% for each type of treatment; see Supplemental Table 5), suggesting that incomplete courses of treatment were unlikely to play a large role in mitigating the high rates of chronic diagnoses. We also found that 70–86% of those in one of the mental health treatment classes received PTSD-specific care and 75–96% reported receipt of depression-related care, proportions that are both high and similar. Thus, it appears unlikely that the apparent PTSD chronicity is associated with lack of PTSD care, although it is quite possible that that care was not in the form of an empirically supported intervention, an issue we cannot resolve with the NESARC-III data set as this level of precision is lacking. These findings suggest that people with both PTSD and SUD are perhaps not obtaining empirically supported treatments for this comorbidity (Roberts et al., 2015; Simpson et al., 2017), may not be responding well to such treatments, and/or may have difficulty completing recommended courses of treatment. The available NESARC-III data do not allow us to address either quality of treatment received nor treatment dropout, but with regards to the latter, the PTSD/SUD treatment outcome literature has repeatedly noted that treatment dropout is a significant issue for these individuals (Roberts et al., 2015; Simpson et al., 2017; see also Krawczyk, Feder, Saloner, Crum, Kealhofer, & Mojtabai, 2017). Because the present study found that across all latent classes there is a stronger likelihood of retaining one’s PTSD diagnosis than retaining a SUD diagnosis (see Hien et al., 2010), provision of effective and acceptable PTSD interventions is particularly important (see Jonas et al., 2013). Thus, future research should address the degree of access that patients with comorbid PTSD/SUD have to effective PTSD or integrated PTSD/SUD treatment, the barriers to engagement and retention in such care, and how best to overcome those barriers.

The current study has a number of strengths in that it used a large, nationally representative sample of individuals with co-occurring PTSD/SUD, took a novel approach in delineating classes of treatment seekers, and thoroughly described key demographic and clinical characteristics differentiating the latent classes while controlling Type I error. The study is limited in that it is unknown whether lifetime PTSD and SUD occurred simultaneously, and we cannot make causal inferences regarding the role of treatment receipt and either PTSD or SUD recovery (or lack thereof) with these cross-sectional, non-experimental data. Although some information is available regarding the timing of treatment, the available information is not specific enough to determine whether non-PTSD MH treatments were received only before the onset of PTSD, though this limitation is somewhat offset by the large proportion of respondents in the MH treatment classes reporting having received care specifically for PTSD. Information is also lacking regarding treatment dose or the specific types of treatment participants received (e.g., empirically supported treatments versus supportive counseling). Additionally, because PTSD-related MH treatments and those for other non-SUD psychiatric conditions were combined, the current results do not address whether those who specifically received PTSD-related MH care fared better with regard to chronicity or differed from those receiving only other types of MH care, though the finding that between 70–86% of those who received any MH care reported receipt of care specifically for PTSD suggests that there was likely little impact of such care on chronicity. Finally, despite the relatively large sample size, some cells were too small to conduct more fine-grained analyses (e.g., racial and ethnic minority groups).

In sum, findings demonstrate that most people with both lifetime PTSD and SUD have sought either SUD or MH treatment or both, with an apparent preference for MH treatment over SUD treatment. This comorbid group also has complex clinical presentations that differ depending upon treatment subgroup, and for most of these individuals one or the other or both of their comorbid disorders persisted despite high rates of treatment engagement. These findings suggest that interventions with demonstrated efficacy and acceptability need to be widely disseminated and implemented across both SUD and MH treatment settings.

Supplementary Material

Supplemental Information

Table 1. Substance Use Disorder and Mental Health Treatment Mapping

Table 2. Latent Class Analysis Fit Statistics

Table 3a. Demographic characteristics by latent class

Table 3b. Disease characteristics and order of onset by latent class

Table 4a. Lifetime and past-year PTSD and SUD diagnostic status by latent class

Table 4b. Additional lifetime psychiatric comorbidity diagnostic status and suicide behaviors by latent class

Table 5. Overall and class-specific timing of each global type of treatment List of Supplemental Figures

Figure 1. Differences in PTSD versus other MH treatments Across latent classes

Data Transparency Statement:

We have published six other manuscripts from the NESARC-III dataset and have two additional papers under review (see below). The current paper is an extension of the paper that was published in Addiction in 2019. In that paper we described the clinical characteristics of NESARC-III participants with co-occurring PTSD and alcohol use disorder (AUD) only, PTSD and drug use disorder (DUD) with or without AUD, as well as those with AUD, DUD, and PTSD without the other co-occurring condition present. Among other things, we observed in that paper that those with PTSD and a co-occurring SUD had high rates of mental health and/or SUD treatment engagement relative to the non-comorbid groups.

Public Health Statement.

The results of this epidemiologic study broadly demonstrate that most individuals with co-occurring Posttraumatic Stress Disorder (PTSD) and Substance Use Disorders (SUD) have sought mental health treatment (79%) while fewer have sought SUD treatment (36%). Six different treatment receipt patterns were identified that varied in terms of type, amount, and intensity of care, and the people in different treatment classes differed from one another with regards to functional stability, disease severity, disease chronicity, and the presence of additional comorbidities. Nearly 69% of the sample had chronic PTSD while approximately 39% had a chronic SUD, suggesting that health care systems serving individuals with this comorbidity are not successfully connecting them with treatments to effectively address PTSD in the context of SUD.