Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2022 Mar 1.
Published in final edited form as: Cogn Behav Ther. 2020 Oct 6;50(2):89–103. doi: 10.1080/16506073.2020.1819865

Objective Analysis of Language Use in Cognitive-Behavioral Therapy: Associations with Symptom Change in Adults with Co-Occurring Substance Use Disorders and Posttraumatic Stress

Anthony N Jennings a, Heather E Soder a, Margaret C Wardle a,b, Joy M Schmitz a, Anka A Vujanovic c
PMCID: PMC7897212  NIHMSID: NIHMS1626063  PMID: 33021143

Abstract

Substance use disorders (SUD) commonly co-occur with posttraumatic stress disorder (PTSD) symptoms; and the comorbidity is prevalent and difficult-to-treat. Few studies have objectively analyzed language use in psychotherapy as a predictor of treatment outcomes. We conducted a secondary analysis of patient language use during cognitive-behavioral therapy (CBT) in a randomized clinical trial, comparing a novel, integrated CBT for PTSD/SUD with standard CBT for SUD. Participants included 37 treatment-seeking, predominantly African-American adults with SUD and at least four symptoms of PTSD. We analyzed transcripts of a single, matched session across both treatment conditions, using the Linguistic Inquiry and Word Count (LIWC) program. The program measures language use across multiple categories. Compared to standard CBT for SUD, patients in the novel, integrated CBT for PTSD/SUD used more negative emotion words, partially consistent with our hypothesis, but less positive emotion words. Further, exploratory analyses indicated an association between usage of cognitive processing words and clinician-observed reduction in PTSD symptoms, regardless of treatment condition. Our results suggest that language use during therapy may provide a window into mechanisms active in therapy.

Keywords: trauma, PTSD, substance use, cognitive-behavioral, cognitive processing therapy, treatment, communication and language measures

Introduction

Substance use disorder (SUD) is associated with high rates of posttraumatic stress disorder (PTSD) symptoms (McCauley et al., 2012), and both diagnostic and subthreshold PTSD symptoms are associated with more severe SUD (e.g., Coffey et al., 2002; Vujanovic & Back, 2019) and less improvement in SUD treatment (e.g., Ouimette, Moos, & Finney, 2015; Vujanovic & Back, 2019). The PTSD/SUD comorbidity is prevalent, complex, and difficult-to-treat (e.g., Vujanovic & Back, 2019). Enhancing our understanding of treatment mechanism has significant potential to improve or refine extant treatments or to develop novel intervention approaches.

Few tools exist for the objective analysis of cognitive, affective, or behavioral mechanisms that may be active during psychotherapy sessions and may impact treatment outcomes for patients with SUD and PTSD. One way to study these mechanisms objectively may be analysis of language. Several studies suggest that language use in multiple contexts predicts real-world outcomes in PTSD and SUD. For example, perceptual language (e.g., “saw”, “feels”) used in narratives about traumatic experiences predicted increased PTSD symptoms measured six years after the narrative was collected (Ng et al., 2015). Furthermore, increased positive emotion words and decreased “I” word use in an expressive writing treatment predicted better psychological outcomes among sexual abuse victims (Pulverman et al., 2015). Similarly, past work has shown that strength of commitment language used during therapy predicted attempted-abstinence outcomes in SUD patients (Campbell et al., 2010). Notably, personal pronoun use has been associated with depression and depression-vulnerability in college students (Rude et al., 2004). Previous studies examining language use in trauma narratives have found that higher levels of both positive and negative emotion words are associated with decreased symptom manifestation (Jaeger et al., 2014). Additionally, increased use of cognitive processing words (“cause”, “know”, “ought”) in written trauma narratives has been associated with greater mental health improvements (Pennebaker et al., 1997; Tausczik & Pennebaker, 2010), potentially because use of these words suggests an internal active reappraisal process (Tausczik & Pennebaker, 2010). This is potentially important because emotional experiences are a necessary mechanism for the successful processing of trauma (Wardecker et al., 2017; Zoellner et al., 2011) and abstinence or reduction of substance use (Kang et al., 2019; Sloan et al., 2017). Emotion words used aloud during psychotherapy may be an indication of this emotional experience mechanism being active internally (Ng et al., 2015). Notably, no studies to date have compared language use across different in-person psychotherapies in an attempt to examine potential treatment mechanisms.

The Linguistic Inquiry and Word Count (LIWC) is an automated text-analysis program that measures use of different, theoretically-defined categories of words in text or speech, such as “negative emotion” and “past focus” (Tausczik & Pennebaker, 2010). Applying the LIWC to examine the content of therapy sessions may allow quantification of putative treatment mechanisms, such as emotional engagement (McCarthy et al., 2017; Zoellner et al., 2011) or analytical thinking (McCarthy et al., 2017), via objective measurement of language used in session. The present study examined the language use of patients in a pilot randomized clinical trial comparing a novel, integrated cognitive-behavioral treatment (CBT) for patients with PTSD/SUD (i.e., Treatment of Integrated Posttraumatic Stress and Substance Use [TIPSS]) with standard CBT for SUD (Vujanovic et al., 2018). While the most effective treatments for PTSD/SUD have been demonstrated as those that integrate trauma-focused psychotherapy with CBT for SUD, such treatments have generally combined elements of prolonged exposure therapy for PTSD with standard CBT for SUD (e.g., Roberts et al., 2015; Simpson, Lehavot, & Petrakis, 2017; Vujanovic & Back, 2019). The TIPSS approach presented here offers a novel treatment program (see Vujanovic et al., 2018) that instead integrates cognitive processing therapy (CPT), another gold standard, evidence-based treatment for PTSD (Resick et al., 2002), with standard CBT for SUD. The TIPSS program is in the development stage and represents an integration of two well-established evidence-based treatments for SUD and PTSD, respectively, although its efficacy as a stand-alone intervention has not been documented. Most importantly for the current study, TIPSS incorporates the original CPT program, which includes a written trauma account component, or patient-written account of the most distressing traumatic event experience, with emphasis on cognitive restructuring of maladaptive trauma-related cognitions relevant to the self, others, and the world across five themes: safety, trust, power/control, intimacy, and esteem (Resick & Schnicke, 1992).

In this secondary analysis, we examined how language use differs between TIPSS and CBT for SUD, with the intent of testing the ability of language use to detect differences in the mechanisms active during critical sessions of these therapies. We hypothesized that patients would use more positive and negative emotion words in the TIPSS trauma narrative review session (session 7) than in the active cognitive restructuring session (session 7) of the CBT for SUD condition, indicating greater emotional engagement and ultimately better PTSD and SUD treatment outcomes. This hypothesis was based upon our expectation that greater emotional awareness and expression would be fostered during the TIPSS trauma account review session, as compared to the CBT for SUD session, and that this would be predictive of better treatment outcomes (i.e., lower PTSD symptoms, lower substance use) given the pertinent role of emotion regulatory processes in PTSD and SUD treatment outcomes (e.g., Vujanovic & Back, 2019). We also conducted two post hoc exploratory analyses. First, we examined differences in use of cognitive processing and personal pronoun words between treatment conditions. We further explored the predictive utility, collapsed across treatment conditions, of language categories (i.e., use of emotion, cognitive processing, and personal pronoun words) with regard to PTSD symptoms and substance use post-treatment to preliminarily evaluate associations of these mechanisms with treatment outcomes, generally.

Methods

Participants

Participants were included if they met criteria for substance dependence per DSM-IV-TR (APA, 2000), reported a history of trauma exposure per DSM-5 PTSD Criterion A (APA, 2013) with at least four past-month DSM-5 PTSD symptoms, and were seeking treatment for both substance dependence and trauma-related symptoms. Individuals with and without full-criteria past-month PTSD were included based on previous literature indicating PTSD symptoms have a meaningful impact on outcomes even at sub-threshold levels (Ruglass et al., 2018; Vujanovic et al., 2018); however, we also repeated all analyses only in the sub-set of individuals meeting full PTSD criteria. Recruitment took place shortly after the change to DSM-5, therefore SUD diagnoses continued to be based upon DSM-IV criteria to maintain consistency with ongoing studies of substance use in the research center, while PTSD diagnoses utilized DSM-5 criteria. Other inclusionary criteria were: being 18–65 years old and proficient in English. Exclusionary criteria were: dependence exclusively on nicotine, alcohol or opioid dependence requiring medical detoxification, past-month suicidal or homicidal ideation with intent or plan, pregnancy, or inability to provide verbal and written consent. Participants were recruited through advertisements.

A total of 41 participants were eligible to be included in this secondary analysis, as they attended a specified “critical therapy session” (see Treatment section). However, three of those participants dropped out of the study prior to completing the final therapy session, and one was determined to have invalid responding during the final assessment (i.e., answering “no” to all questions and refusing to elaborate). The final sample was thus comprised of 37 participants.

During an extended piloting period prior to the start of the randomized controlled trial, eight participants from this sample were assigned to a therapy condition based on therapist availability and not by randomization (n = 6 nonrandomly assigned to CBT for SUD; n = 2 nonrandomly assigned to TIPSS). All analyses were performed both with and without the nonrandomized participants to examine the impact of their inclusion upon results (see Data Analytic Strategy).

Treatment

Each participant attended two therapy sessions per week over the course of 6 weeks (12 total sessions). Therapists for both conditions were M.A.-level psychological counselors and M.A.-level doctoral students in counseling psychology. Therapists had training in CBT and drug counseling for SUD prior to the study. At the outset of the trial, therapists were assigned to either the TIPSS or CBT for SUD conditions; therapists were not permitted to cross-over between treatments.

Across both treatment conditions, the “critical” session was the seventh session. In the seventh session of the CBT for SUD condition (N = 23), participants were tasked, for the first time in treatment, with active cognitive restructuring, a key element of CBT for SUD (McHugh et al., 2010). In the seventh session of the TIPSS condition (N = 14), participants were asked to process their trauma narrative with the therapist, a key element of CPT for PTSD (Resick et al., 2002). The entire 60-minute session was transcribed for the seventh session of both treatment conditions.

Standard CBT for SUD.

The SUD treatment components are based upon cognitive-behavioral relapse prevention principles (Carroll, 1998; Marlatt & Donovan, 2007) that are intended to facilitate awareness and management of cravings, review coping skills for high-risk substance-related cognitions and situations, and provide a greater understanding of the associations between thoughts, feelings, and substance use behaviors, including cravings, via functional analyses.

TIPSS.

The TIPSS integrates CBT for SUD with CPT for PTSD (Resick et al., 2017), which is designed to target trauma-related cognitions in order to facilitate cognitive-emotional processing of the trauma and thus alleviate PTSD symptomatology. Connections between PTSD and SUD are drawn through each session (see Vujanovic et al., 2018; Vujanovic, Smith, Tipton, & Schmitz, 2019). PTSD/SUD comorbidity is approached as an integrated syndrome rather than as two separate disorders.

Measures

Language use.

Audio recordings of the entire critical session were transcribed by a HIPAA-compliant service. Each was then compared to the corresponding audio by the first author, with any transcriptional errors logged and corrected. Additionally, bracketed notes (“[inaudible]”, “[laughter]”) were removed, and spoken numbers and symbols were written out (“5” to “five”, “&” to “and”). Comments made by the therapist were removed, so only client speech was analyzed. These edited transcripts were processed by the LIWC program (2015 version) to produce the reported category statistics, which are the percentage of words from the transcript in each category (Pennebaker et al., 1983).

PTSD symptoms.

PTSD symptom severity was assessed using the (1) Clinician-Administered PTSD Scale for DSM-5 (CAPS-5; (F.W. Weathers et al., 2013)) and (2) PTSD Checklist for DSM-5 (PCL-5; (Blevins et al., 2015). The CAPS-5 is a clinician-administered and rated interview, characterized as the “gold standard” for measuring PTSD symptom severity (Griffin et al., 2004; Grubaugh et al., 2007; Prins et al., 2004). The PCL-5 is a 20-item self-report measure for assessing PTSD symptom severity (Blanchard et al., 1996; Wortmann et al., 2016).

Substance use.

Substance use was measured using the Timeline Follow Back (TLFB), a self-report measure of days of substance use (Sobell & Sobell, 1996)(Hjorthøj et al., 2012). Participants completed the TLFB at baseline and each time they came to the clinic, reporting on use since the last visit. TLFB scores were determined by dividing number of reported days of use by the number of days in the reporting period for both baseline and from visit 11 to visit 12. TLFB scores were computed for each participant’s self-designated primary drug of concern, as the sample included participants using various substances. Self-report was used rather than biochemical verification because this was a mixed sample of substance users, and different drugs have widely different windows for biochemical detection (Moeller et al., 2019), which could confound results.

Data Analytic Strategy

All analyses were performed in IBM SPSS 26. Normality was confirmed via histograms and non-significant (p’s > .05) Shapiro-Wilk tests for all language use variables. For our first analysis, positive and negative emotion words, “I” pronouns, and cognitive processing words were compared between treatment conditions via independent samples t-tests using a Holm-Bonferroni correction (Aickin & Gensler, 1996). For our second analysis, three separate multiple regressions were performed to test the association between language use in therapy and treatment outcomes (CAPS-5, PCL-5, TLFB), while controlling for effects of therapy condition. Prior to the main analyses, the following potential covariates were evaluated: age, gender, race, education, addiction severity, and therapist experience level. Per guidelines on inclusion of covariates in clinical trials (Assmann et al., 2000; Pocock et al., 2002), we only included variables that were related to both a predictor (therapy words, treatment condition) and an outcome (CAPS-5, PCL, TLFB). None of the potential covariates met this criterion; therefore, we did not include any covariates in the final analyses. Baseline scores for each dependent variable (CAPS-5, PCL-5 and TLFB) were entered into the analysis of that dependent variable as covariates predicting post-treatment scores (at week 12) to account for baseline distributions (van Breukelen, 2013). This method is preferred to change scores, as initial starting points can bias estimates of treatment effects (Senn, 2006; Vickers, 2001). We used Holm-Bonferroni corrected thresholds that adjusted for the four main predictors of interest, within each regression. In addition, we covaried for baseline substance use (TLFB) in the analyses of PTSD outcomes (CAPS-5 and PCL-5), and for baseline PTSD symptoms (CAPS-5) in the analysis of substance use outcomes (TLFB). Note that CAPS-5 baseline scores were used to represent baseline PTSD symptoms in the TLFB analysis, but use of baseline PCL-5 scores as a covariate resulted in the same pattern of findings. Exclusion of potential outliers in a separate analysis revealed no major differences in the outcome of the tests. Normality of residuals was confirmed via visual inspection of Q-Q plots and non-significant Shapiro-Wilk tests (Shapiro & Wilk, 2006). Standardized residuals, Cook’s Distance, and leverage tests revealed no influential observations. Finally, we also examined correlations between the language use categories and CAPS-5 symptom clusters. All analyses were repeated in two subgroups to address randomization and diagnosis concerns: 1) only those participants who were randomized (n = 29) and 2) only those participants who met full PTSD criteria (n = 29). These subgroup analyses revealed similar patterns of results (although some were no longer statistically significant, as might be expected with reduced sample sizes) and will not be discussed further.

Results

Language Use by Condition

Sample characteristics are displayed in Table 1. Overall word count did not differ between the treatment conditions, [t(35) = −1.11, p =.276]. T-tests indicated that several language use categories differed significantly between the treatment conditions (Figure 1). Specifically, use of negative emotion words was higher in the TIPSS condition compared to the CBT for SUD condition. Cognitive processing and positive emotion words were higher in the CBT for SUD condition compared to the TIPSS condition. Please see Table 2 for t-test statistics and effect sizes.

Table 1.

Sample Characteristics

Variable CBT for SUD TIPSS
Age M(SD) 46.13(9.70) 49.29(10.19)
Sex % Male 56.5% 50.0%
Race % African American 73.9% 71.4%
Education M(SD) 13.41(1.74) 13.23(3.32)
Monthly Income M(SD) 323.91(470.40) 470.36(1188.18)
Addiction Severity Alcohol 0.23(0.29) 0.40(0.32)
Addiction Severity Drug .22(.09) .23(.11)
Number of Diagnoses M(SD) 1.70(0.88) 1.79(0.70)
Number of Substance Diagnoses M(SD) 2.87(0.97) 3.29(1.44)
Therapist % Experienced1 56.5% 100%*
CAPS-5 baseline M(SD) 34.30(11.81) 40.43(13.77)
CAPS-5 final M(SD) 25.96(14.28) 27.29(18.35)
PCL-5 baseline M(SD) 49.39(11.23) 47.14(15.61)
PCL-5 final M(SD) 27.87(19.03) 33.36(25.28)
TLFB baseline M(SD) 55.51(39.13) 62.26(38.09)
TLFB final M(SD) 37.58(47.09) 38.64(41.81)
Open in a new tab

Note.

1

Refers to whether the therapist was a doctoral student trainee or a licensed Master’s level clinician. The TIPSS condition was conducted by licensed counselors, while the other was comprised of both licensed counselors and doctoral student trainees.

*

denotes a significant difference between the groups.

CAPS-5 = Clinical Administered PTSD Scale for DSM-5, PCL-5 = PTSD Checklist for DSM-5, TLFB = Timeline Follow Back, CBT = Cognitive Behavioral Therapy, TIPSS = Treatment of Integrated Posttraumatic Stress and Substance Use

Figure 1.

Figure 1.

Open in a new tab

Language use categories by treatment condition. Top left: Use of positive emotion words by treatment condition. Top right: Use of I pronouns by treatment condition. Bottom left: Use of negative emotion words by treatment condition. Bottom Right: Use of cognitive processing words by treatment condition.

Table 2.

Language Use by Treatment Condition

LIWC category
(n=37) 		CBT for SUD M(SD)
(n=23) 		TIPSS
M(SD)
(n=14) 		t p d 95%CI
Lower		 95%CI
Upper
Total word count 3950.09 (1777.68) 4681.14 (2206.36) −1.12 .276 .36 −2071.57 609.45
“I” pronouns 15.63(1.60) 17.11(2.33) −2.30 .244 .74 −1.69 0.44
Positive emotion 2.90(0.89) 2.11(0.73) 2.78 .009 .97 0.21 1.36
Negative emotion 1.82(0.50) 2.23(0.53) −2.38 .023 .80 −0.76 −0.06
Cognitive processing 16.26(2.09) 14.43(2.06) 2.60 .014 .88 0.40 3.26
Open in a new tab

Note. Holm-Bonferroni corrected p value thresholds were used to determine significance. Tests that met this threshold for significance are bolded. LIWC = Language Inquiry and Word Count; CBT for SUD = Standard Cognitive Behavioral Therapy for SUD, TIPSS = Treatment of Integrated Posttraumatic Stress and Substance Use

Language Use and Treatment Outcomes

In the model predicting CAPS-5 PTSD symptom severity final scores, greater use of cognitive processing words predicted lower CAPS-5 scores at treatment end. No other language use category emerged as a significant predictor of CAPS-5 outcomes. Full model results are presented in Table 3. No language use categories emerged as significant predictors of PCL-5 (Table 3) or TLFB outcomes (Table 3).

Table 3.

Linear Regression Results: PTSD and Substance Use Outcomes

PTSD Symptoms: CAPS-5 Final Scores
B SE B β p CI(95) L CI(95) U F df p R2
Outcome: Final CAPS-5 Scores 8.45 7, 29 < .001 .82
 Intercept 13.79 21.26 .52 −29.70 57.27
 “I” Pronouns −0.05 1.27 −0.01 .97 −2.64 2.54
 Positive Emotion 1.08 2.21 0.06 .63 −3.43 5.59
 Negative Emotion 1.60 3.44 0.06 .65 −5.44 8.63
 Cognitive Processing −2.30 0.93 −0.33 .02 −4.20 −0.39
 TLFB baseline 0.12 0.05 0.28 .03 0.01 0.22
 Treatment −9.30 4.16 −0.29 .03 −17.82 −0.79
 CAPS-5 baseline 0.96 0.15 0.78 <.001 0.66 1.26
Outcome: Final PCL-5 Scores 2.55 7, 29 0.08 .58
 Intercept 19.60 42.21 .65 −66.72 105.92
 “I” Pronouns −1.93 2.40 −0.14 .43 −6.83 2.97
 Positive Emotion −8.47 4.35 −0.36 .06 −17.36 0.42
 Negative Emotion 9.87 6.60 0.25 .15 −3.62 23.37
 Cognitive Processing −1.11 1.82 −0.12 .55 −4.82 2.61
 Treatment −5.17 8.02 −0.12 .52 −21.58 11.24
 TLFB baseline 0.15 0.11 0.27 .17 −0.07 0.37
 PCL baseline 0.86 0.29 0.52 <.001 0.26 1.45
Outcome: Final TLFB Scores 1.26 7, 29 .30 .48
 Intercept −133.57 92.12 .16 −321.97 54.83
 “I” Pronouns 7.81 5.49 0.27 .17 −3.42 19.04
 Positive Emotion −10.89 9.55 −0.22 .26 −30.43 8.65
 Negative Emotion 15.65 14.91 0.19 .30 −14.85 46.14
 Cognitive Processing 5.10 4.04 0.26 .22 −3.16 13.37
 Treatment −9.82 18.03 −0.11 .59 −46.70 27.06
 CAPS baseline −0.10 0.63 −0.03 .88 −1.39 1.19
 TLFB baseline 0.14 0.22 0.12 .53 −0.30 0.58
Open in a new tab

Note. Holm-Bonferroni corrected p value thresholds were used to determine significance. Tests that met this threshold for significance are bolded. CAPS-5 = Clinician Administered PTSD Scale for DSM-5; PCL = Posttraumatic Stress Disorder Checklist for DSM-5; TLFB = Timeline Follow Back

Correlations between language use categories and CAPS-5 scores at baseline and at treatment end are presented in Table 4. Greater use of “I” pronouns was associated with PTSD intrusion symptom severity at baseline and PTSD arousal and reactivity symptom severity at treatment end. Greater use of negative emotion words was associated with higher PTSD intrusion symptom severity at baseline and treatment end.

Table 4.

Language use and PTSD CAPS-5 Symptom Cluster Correlations

Baseline Final
CAPS-5
B		 CAPS-5
C		 CAPS-5
D		 CAPS-5
E		 CAPS-5
B		 CAPS-5
C		 CAPS-5
D		 CAPS-5
E
“I” Pronouns .068 .400 .239 .268 .135 .226 .263 .484
Positive Emotion Words −.128 −.289 −.037 −.122 −.154 −.131 −.059 −.096
Negative Emotion Words .331 .291 .121 .246 .382 −.050 .135 .195
Cognitive Processing Words .023 −.288 .012 −.234 −.241 −.219 −.093 −.301
Open in a new tab

Note. Bolded correlations are significant at the p < .05 level. CAPS-5 = Clinician Administered PTSD Scale. CAPS-5 B: PTSD Intrusion Symptoms; CAPS-5 C: PTSD Avoidance Symptoms; CAPS-5 D: PTSD Negative Alterations in Cognitions and Mood Symptoms; CAPS-5 E: PTSD Arousal and Reactivity Symptoms

Given the discrepancy between CAPS-5 and PCL-5 outcomes, bivariate associations between CAPS-5 and PCL-5 scores were examined. At the bivariate level, CAPS-5 baseline scores were correlated to PCL-5 baseline scores (r = .49, p < .01), PCL-5 final scores (r = .53, p < .01), and CAPS-5 final scores (r = .72, p < .01). CAPS-5 final scores were correlated with PCL-5 baseline scores (r = .37, p < .05) and PCL-5 final scores (r = .65, p < .01). PCL-5 baseline scores were not significantly associated with PCL-5 final scores (r = .29, p > .05).

Discussion

Comparing language use between a novel, integrated CBT for PTSD/SUD (TIPSS) and standard CBT for SUD partially confirmed our hypotheses. Participants in the TIPSS condition, as compared to participants in the standard CBT for SUD, used a significantly greater percentage of negative emotion words, possibly indicating greater emotional engagement (Ng et al., 2015). However, in contrast to our hypothesis, patients in the TIPSS condition used fewer positive emotion words compared to patients in the CBT for SUD condition. Finally, exploratory analysis of treatment outcomes found that increased patient use of positive emotion language predicted lower self-reported PTSD symptom severity at the conclusion of treatment. As this study examined language use during one psychotherapy session only, findings may or may not be consistent should language use throughout the course of the treatment programs be analyzed and compared.

In addition to our primary findings regarding emotional language, use of cognitive processing words was significantly higher in the CBT for SUD condition, while the use of personal pronouns did not differ between conditions. These results suggest that language may measure other mechanisms potentially important to psychotherapy, such as the analytical engagement necessary for cognitive restructuring, or greater self-focus when emotionally recounting a trauma narrative and re-balancing maladaptive thoughts related to the self. However, it is also important to consider that findings may be an artifact of the sessions chosen for evaluation. In the TIPSS condition, session 7 focused on the trauma account review, a highly personalized emotional exploration of the index trauma, while session 7 of the CBT for SUD condition focused on active cognitive restructuring of substance use-related beliefs. It will be important to replicate and extend this line of inquiry with larger sample sizes and across more diverse types of treatment sessions in order to draw more definitive conclusions.

Regarding treatment outcomes, only increased use of cognitive processing words during psychotherapy was related to greater reduction in PTSD symptoms. In addition, this finding was exclusively evident with regard to the clinician–administered (CAPS-5) PTSD outcomes but not self-report (PCL-5) PTSD outcomes. This discrepancy between self-report and clinician-administered interviews of PTSD symptoms may be due to moderate convergence between these measures (r’s = .37 to .65) and underscores the need for replication. Furthermore, while the PCL-5 was administered at each study session (twice-weekly), the CAPS-5 was administered at only two time-points (i.e., baseline, end-of-treatment); thus, issues of (a) method variance and (b) sensitivity of the PCL-5 to day-to-day change cannot be ruled out. Relatedly, although the CAPS-5 is considered a “gold standard” measure of PTSD, the sample is comprised of an understudied population, mostly low-income African-American adults with SUD and PTSD symptoms. Another extension of this work might be the psychometric evaluation of the CAPS-5 and PCL-5 in low-income, racially diverse populations with PTSD/SUD. However, findings are consistent with previous literature demonstrating an association between increased use of cognitive processing words and greater mental health improvements (Pennebaker, Mayne, & Francis, 1997). In contrast, we did not replicate previous findings regarding a correlation between higher negative emotion word use and lower PTSD symptomatology (Wardecker et al., 2017). These data also did not replicate prior work suggesting that increased use of positive emotion words in trauma narratives is associated with decreased PTSD symptomatology post-treatment (Pulverman et al., 2015), or the association between increased use of personal pronouns and worse treatment outcomes (Rude et al., 2004). One possible explanation for these differences is that our research measured word use specifically during therapy sessions rather than in self-written narratives. Use of spoken versus written language may indeed have distinct correlates with treatment outcomes and warrants further exploration. Another caveat to our findings is that the language analysis was conducted during only one session. It is possible that evaluation of language use throughout the course of treatment would demonstrate more consistent findings. Further, the sample was comprised of predominantly low-income, African-American adults presenting with multiple SUD diagnoses and PTSD symptoms, thus representing a severe psychiatric population that has historically been understudied and underserved. It is possible that language use among racial/ethnic minority populations and/or low-income populations is distinct from majority populations (Hoff & Tian, 2005).

Findings should be considered in light of study limitations. First, the current study had a relatively small sample size, which both reduces our chances of detecting associations with smaller effect sizes and increases the chance of artifactual findings. The small sample size also did not allow us to examine potential moderators of associations. For instance, the small sample size precluded examinations of within-session changes in language use, which is an interesting avenue for future empirical exploration. Relatedly, tests of mediation or moderation to examine mechanisms of change, such as positive and negative emotion word use, also would require larger samples. Second, this study was a secondary analysis of data from a single pilot clinical trial, and therefore, implications for other types of psychotherapy are highly preliminary. Third, we analyzed only a single critical session chosen a priori, so the active mechanisms of either therapy may lie in an unexamined session. Fourth, results relied on two measures of PTSD (self-report and interview) and only one measure of substance use (self-report). Future work should use multimodal indices of PTSD and substance use to affirm findings. Notably, results were not consistent across the self-report and interview indices of PTSD, possibly due to lower convergence between measures (Weathers et al., 2018) underscoring the need for replication. Fifth, the parent trial employed an extended pilot program during which participants were not randomized, decreasing the ability of our analysis to account for unknown confounding variables through randomization. Finally, the use of commercially available software with generic dictionaries offers both advantages, in that this makes language analysis of therapy widely accessible and replicable, but also disadvantages, in that generic categories such as “positive emotions” include words referencing positive judgments (e.g. “sweet”, “nice”), that may not be relevant to hypothesized mechanisms of change, such as emotional engagement. Future studies could consider examining additional language categories, which might include a more fine-grained analysis of which emotions are most important (e.g. anger, sadness, anxiety), or constructing custom dictionaries based on hypothesized mechanisms of change (as has been done in studies of motivational interviewing; Campbell et al., 2010). Language analysis also offers the opportunity to test changes in patient language across therapy sessions, correlations between language use and baseline values, therapist language use and outcomes, and correlations between therapist and patient language use.

Despite the limitations of this study, our analyses indicate an association between language use and psychotherapy among adults with PTSD/SUD that is worth investigating on a larger scale. This investigation offers an important, initial step to understanding what is happening “in the room” during psychotherapy that can be used to improve outcomes of patients with comorbid SUD and PTSD, a historically difficult-to-treat population.

Figure 2.

Figure 2.

Open in a new tab

Partial regression plot of cognitive processing words predicting CAPS-5 treatment outcomes.

Funding

This study was funded by the National Institutes of Health KL2 Career Development Award (KL2TR000370-07: PI: Vujanovic). The work was also supported by the National Institute on Drug Abuse (P50 DA009262; PIs: Schmitz, Lane, Green; K08 DA040006, PI: Wardle)

References

  1. Aickin M, & Gensler H (1996). Adjusting for multiple testing when reporting research results: The Bonferroni vs Holm methods. American Journal of Public Health, 86(5), 726–728. 10.2105/AJPH.86.5.726 [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Assmann SF, Pocock SJ, Enos LE, & Kasten LE (2000). Subgroup analysis and other (mis)uses of baseline data in clinical trials. The Lancet, 355(9209), 1064–1069. 10.1016/S0140-6736(00)02039-0 [DOI] [PubMed] [Google Scholar]
  3. Blanchard EB, Jones-Alexander J, Buckley TC, & Forneris CA (1996). Psychometric properties of the PTSD checklist (PCL). Behaviour Research and Therapy, 34(8), 669–673. 10.1016/0005-7967(96)00033-2 [DOI] [PubMed] [Google Scholar]
  4. Blevins CA, Weathers FW, Davis MT, Witte TK, & Domino JL (2015). The Posttraumatic Stress Disorder Checklist for DSM-5 (PCL-5): Development and Initial Psychometric Evaluation. Journal of Traumatic Stress, 28(6), 489–498. 10.1002/jts.22059 [DOI] [PubMed] [Google Scholar]
  5. Campbell SD, Adamson SJ, & Carter JD (2010). Client language during motivational enhancement therapy and alcohol use outcome. Behavioural and Cognitive Psychotherapy, 38(4), 399–415. 10.1017/S1352465810000263 [DOI] [PubMed] [Google Scholar]
  6. Coffey SF, Saladin ME, Drobes DJ, Brady KT, Dansky BS, & Kilpatrick DG (2002). Trauma and substance cue reactivity in individuals with comorbid posttraumatic stress disorder and cocaine or alcohol dependence. Drug and Alcohol Dependence, 65(2), 115–127. [DOI] [PubMed] [Google Scholar]
  7. Coffey SF, Stasiewicz PR, Hughes PM, & Brimo ML (2006). Trauma-focused imaginal exposure for individuals with comorbid posttraumatic stress disorder and alcohol dependence: Revealing mechanisms of alcohol craving in a cue reactivity paradigm. Psychology of Addictive Behaviors, 20(4), 425–435. 10.1037/0893-164X.20.4.425 [DOI] [PubMed] [Google Scholar]
  8. Dixon LJ, Leen-feldner EW, Ham LS, Feldner MT, & Lewis SF (2010). Alcohol use motives among traumatic event-exposed treatemnt-seeking adolecents. Addictive Behaviors, 34(12), 1065–1068. 10.1016/j.addbeh.2009.06.008.Alcohol [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Griffin MG, Uhlmansiek MH, Resick PA, & Mechanic MB (2004). Comparison of the posttraumatic stress disorder scale versus the clinician-administered posttraumatic stress disorder scale in domestic violence survivors. Journal of Traumatic Stress, 17(6), 497–503. 10.1007/s10960-004-5798-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Grubaugh AL, Elhai JD, Cusack KJ, Wells C, & Frueh BC (2007). Screening for PTSD in public-sector mental health settings: the diagnostic utility of the PTSD checklist. Depression and Anxiety, 24(2), 124–129. 10.1002/da.20226 [DOI] [PubMed] [Google Scholar]
  11. Hjorthøj CR, Hjorthøj AR, & Nordentoft M (2012). Validity of Timeline Follow-Back for self-reported use of cannabis and other illicit substances — Systematic review and meta-analysis. Addictive Behaviors, 37(3), 225–233. 10.1016/J.ADDBEH.2011.11.025 [DOI] [PubMed] [Google Scholar]
  12. Hoff E, & Tian C (2005). Socioeconomic status and cultural influences on language. 38, 271–278. 10.1016/j.jcomdis.2005.02.003 [DOI] [PubMed] [Google Scholar]
  13. Jaeger J, Lindblom KM, Parker-Guilbert K, & Zoellner LA (2014). Trauma Narratives: It’s What You Say, Not How You Say It. Psychological Trauma : Theory, Research, Practice and Policy, 6(5), 473–481. 10.1037/a0035239 [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Kang D, Fairbairn CE, & Ariss TA (2019). A meta-analysis of the effect of substance use interventions on emotion outcomes. Journal of Consulting and Clinical Psychology, 87(12), 1106–1123. 10.1037/ccp0000450 [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. McCarthy KL, Caputi P, & Grenyer BFS (2017). Significant change events in psychodynamic psychotherapy: Is cognition or emotion more important? Psychology and Psychotherapy. 10.1111/papt.12116 [DOI] [PubMed] [Google Scholar]
  16. McCauley JL, Killeen T, Gros DF, Brady KT, & Back SE (2012). Posttraumatic Stress Disorder and Co-Occurring Substance Use Disorders: Advances in Assessment and Treatment. Clinical Psychology: Science and Practice, 19(3), 283–304. 10.1111/cpsp.12006 [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. McHugh RK, Hearon BA, & Otto MW (2010). Cognitive behavioral therapy for substance use disorders. The Psychiatric Clinics of North America, 33(3), 511–525. 10.1016/j.psc.2010.04.012 [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Moeller KE, Lee KC, & Kissack JC (2019). Urine Drug Screening: Practical Guide for Clinicians. Mayo Clinic Proceedings, 83(January 2008), 66–76. 10.4065/83.1.66 [DOI] [PubMed] [Google Scholar]
  19. Ng LC, Ahishakiye N, Miller DE, & Meyerowitz BE (2015). Narrative characteristics of genocide testimonies predict posttraumatic stress disorder symptoms years later. Psychological Trauma : Theory, Research, Practice and Policy, 7(3), 303–311. 10.1037/tra0000024 [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Ouimette PC, Moos RH, & Finney JW (2015). Two-year mental health service use and course of remission in patients with substance use and posttraumatic stress disorders. Journal of Studies on Alcohol, 61(2), 247–253. 10.15288/jsa.2000.61.247 [DOI] [PubMed] [Google Scholar]
  21. Ouimette P, Finney J, & Moos R (1999). Two-year posttreatment functioning and coping of substance abuse patients with posttraumatic stress disorder. Psychology of Addictive Behaviors, 13(2), 105–114. 10.1037/0893-164X.13.2.105 [DOI] [Google Scholar]
  22. Ouimette P, Moos RH, & Finney JW (2003). PTSD treatment and 5-year remission among patients with substance use and posttraumatic stress disorders. Journal of Consulting and Clinical Psychology, 71(2), 410–414. 10.1037/0022-006X.71.2.410 [DOI] [PubMed] [Google Scholar]
  23. Pennebaker JW, Boyd RL, Jordan K, & Blackburn K (1983). The Development and Psychometric Properties of LIWC2015. Environment and Planning D: Society and Space, 1(2), 163–180. 10.1068/d010163 [DOI] [Google Scholar]
  24. Pennebaker JW, Mayne TJ, & Francis ME (1997). Linguistic predictors of adaptive bereavement. Journal of Personality and Social Psychology, 72(4), 863–871. 10.1037/0022-3514.72.4.863 [DOI] [PubMed] [Google Scholar]
  25. Pocock SJ, Assmann SE, Enos LE, & Kasten LE (2002). Subgroup analysis, covariate adjustment and baseline comparisons in clinical trial reporting: Current practice and problems. Statistics in Medicine, 21(19), 2917–2930. 10.1002/sim.1296 [DOI] [PubMed] [Google Scholar]
  26. Prins A, Ouimette P, Kimerling R, Camerond RP, Hugelshofer DS, Shaw-Hegwer J, Thrailkill A, Gusman FD, & Sheikh JI (2004). The primary care PTSD screen (PC–PTSD): development and operating characteristics. Primary Care Psychiatry, 9(1), 9–14. 10.1185/135525703125002360 [DOI] [Google Scholar]
  27. Pulverman CS, Lorenz TA, & Meston CM (2015). Linguistic changes in expressive writing predict psychological outcomes in women with history of childhood sexual abuse and adult sexual dysfunction. Psychological Trauma: Theory, Research, Practice, and Policy, 7(1), 50–57. 10.1037/a0036462 [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Resick PA, Nishith P, Weaver TL, Astin MC, & Feuer CA (2002). A comparison of cognitive-processing therapy with prolonged exposure and a waiting condition for the treatment of chronic posttraumatic stress disorder in female rape victims. Journal of Consulting and Clinical Psychology, 70(4), 867–879. http://www.ncbi.nlm.nih.gov/pubmed/12182270 [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Resick PA, & Schnicke MK (1992). Resick Cognitive Processing Therapy for Sexual Assault Victims. Journal of Consulting and Clinical Psychology, 60(5), 748–756. [DOI] [PubMed] [Google Scholar]
  30. Rude SS, Gortner EM, & Pennebaker JW (2004). Language use of depressed and depression-vulnerable college students. Cognition and Emotion, 18(8), 1121–1133. 10.1080/02699930441000030 [DOI] [Google Scholar]
  31. Ruglass LM, Lopez-castro T, Papini S, Killeen T, Back SE, & Hien DA (2018). HHS Public Access. 86(3), 150–161. 10.1159/000462977.Concurrent [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Saladin ME, Drobes DJ, Coffey SF, Dansky BS, Brady KT, & Kilpatrick DG (2003). PTSD symptom severity as a predictor of cue-elicited drug craving in victims of violent crime. Addictive Behaviors, 28(9), 1611–1629. 10.1016/j.addbeh.2003.08.037 [DOI] [PubMed] [Google Scholar]
  33. Senn S (2006). Change from baseline and analysis of covariance revisited. Statistics in Medicine, 25(24), 4334–4344. 10.1002/sim.2682 [DOI] [PubMed] [Google Scholar]
  34. Shapiro SS, & Wilk MB (2006). An Analysis of Variance Test for Normality (Complete Samples). Biometrika, 52(3/4), 591 10.2307/2333709 [DOI] [Google Scholar]
  35. Sloan E, Hall K, Moulding R, Bryce S, Mildred H, & Staiger PK (2017). Emotion regulation as a transdiagnostic treatment construct across anxiety, depression, substance, eating and borderline personality disorders: A systematic review In Clinical Psychology Review (Vol. 57, pp. 141–163). Elsevier Inc; 10.1016/j.cpr.2017.09.002 [DOI] [PubMed] [Google Scholar]
  36. Sobell LC, & Sobell MB (1996). Timeline Followback In Measuring Alcohol Consumption (pp. 4–5). Humana Press; 10.1007/978-1-4612-0357-5_3 [DOI] [Google Scholar]
  37. Stewart SH, Mitchell TL, Wright KD, & Loba P (2004). The relations of PTSD symptoms to alcohol use and coping drinking in volunteers who responded to the Swissair Flight 111 airline disaster. Journal of Anxiety Disorders, 18(1), 51–68. 10.1016/j.janxdis.2003.07.006 [DOI] [PubMed] [Google Scholar]
  38. Tausczik YR, & Pennebaker JW (2010). The psychological meaning of words: LIWC and computerized text analysis methods. Journal of Language and Social Psychology, 29(1), 24–54. 10.1177/0261927X09351676 [DOI] [Google Scholar]
  39. van Breukelen GJP (2013). ANCOVA versus CHANGE From baseline in nonrandomized studies: The difference. Multivariate Behavioral Research, 48(6), 895–922. 10.1080/00273171.2013.831743 [DOI] [PubMed] [Google Scholar]
  40. Vickers AJ (2001). The use of percentage change from baseline as an outcome in a controlled trial is statistically inefficient: a simulation study. BMC Medical Research Methodology, 1(1), 6 10.1186/1471-2288-1-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Vujanovic AA, Smith LJ, Green CE, Lane SD, & Schmitz JM (2018). Development of a novel, integrated cognitive-behavioral therapy for co-occurring posttraumatic stress and substance use disorders: A pilot randomized clinical trial. Contemporary Clinical Trials, 65, 123–129. 10.1016/j.cct.2017.12.013 [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Wardecker BM, Edelstein RS, Quas JA, Cordon IM, & Goodman GS (2017). Emotion Language in Trauma Narratives is Associated with Better Psychological Adjustment among Survivors of Childhood Sexual Abuse. Journal of Language and Social Psychology, 36(6), 628–653. 10.1177/0261927X17706940 [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Weathers FW, Blake D, Schnurr P, Kaloupek D, Marx B, & Keane T (2013). The clinician-administered PTSD scale for DSM-5 (CAPS-5). Interview Available from the National Center for PTSD at Www.Ptsd.va.Gov.
  44. Weathers Frank W., Bovin MJ, Lee DJ, Sloan DM, Schnurr PP, Kaloupek DG, Keane TM, & Marx BP (2018). The clinician-administered ptsd scale for DSM-5 (CAPS-5): Development and initial psychometric evaluation in military veterans. Psychological Assessment, 30(3), 383–395. 10.1037/pas0000486 [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Wortmann JH, Jordan AH, Weathers FW, Resick PA, Dondanville KA, Hall-Clark B, Foa EB, Young-McCaughan S, Yarvis JS, Hembree EA, Mintz J, Peterson AL, & Litz BT (2016). Psychometric analysis of the PTSD checklist-5 (PCL-5) among treatment-seeking military service members. Psychological Assessment, 28(11), 1392–1403. 10.1037/pas0000260 [DOI] [PubMed] [Google Scholar]
  46. Zoellner LA, Feeny NC, Bittinger JN, Bedard-Gilligan MA, Slagle DM, Post LM, & Chen JA (2011). Teaching trauma-focused exposure therapy for PTSD: Critical clinical lessons for novice exposure therapists. Psychological Trauma: Theory, Research, Practice, and Policy, 3(3), 300–308. 10.1037/a0024642 [DOI] [PMC free article] [PubMed] [Google Scholar]

RESOURCES