Predictors of Treatment Adequacy During Evidence-Based Psychotherapy for PTSD

To ensure high quality of care for veterans with posttraumatic stress disorder (PTSD), the U.S. Department of Veterans Affairs (VA) health system mandates that all VA medical centers provide access to frontline evidence-based psychotherapies (EBPs) for PTSD (1). These psychotherapies include prolonged exposure (PE) and cognitive processing therapy (CPT), and their recommendation as frontline treatments are a result of an accumulation of strong evidence for their efficacy (2, 3). Despite this evidence, many individuals who initiate PE or CPT do not complete a full course of treatment (4). This is concerning given findings of a dose-response effect for such EBPs (5, 6) and evidence that attrition during the acute phase of trauma-focused treatment is related to worse outcomes at long-term follow-up (7).

Low rates of treatment completion among individuals receiving trauma-focused psychotherapy have been documented in the PTSD literature. For example, literature reviews and meta-analyses of randomized controlled trials (RCTs) and open trials have reported rates of dropout ranging from 12% to 39% during EBPs for PTSD (8, 9). By comparison, a meta-analysis of adult psychotherapy reported a dropout rate of 19.7% across psychological conditions (10). However, there are concerns that results produced from tightly controlled studies may not generalize to larger clinical settings that serve the broader patient population with greater complexity (11, 12). Indeed, a recent systematic review of psychotherapy for PTSD with Operation Enduring Freedom and Operation Iraqi Freedom veterans observed higher rates of dropout in routine clinical care settings (43%) relative to RCTs (28%) (13). Thus, the cumulative evidence from meta-analyses, RCTs, and clinical data suggests that many individuals who begin trauma-focused treatment do not receive a full course of psychotherapy. However, a large-scale study using clinical data from a national health system sample is needed to extend these prior explorations, which have relied upon specialized patients from clinical trials and smaller clinical samples.

Predictors of treatment adequacy have been explored in prior research with equivocal results (14–20). The exception, although still not consistent across all studies, is that older age has been found to be associated with treatment completion during CPT and PE (21–25). Data obtained in routine care provide an opportunity to examine treatment outcomes among a broader PTSD patient population, including those with more complex clinical presentations.

This study had three aims related to the evidence-based treatment of PTSD. First, we are unaware of any prior study that has examined completion of an adequate course of treatment during EBPs for PTSD across clinical practice in a national health system. Therefore, we sought to examine EBP treatment adequacy in the nation’s largest integrated health system to determine whether the low rates of treatment completion reported in controlled research trials generalized to broader clinical contexts. Second, as EBPs for PTSD become more accessible, it is important to characterize the individuals who initiate but do not complete a course of treatment. Identifying factors related to treatment adequacy may allow clinicians to apply adjunctive strategies to increase the likelihood of completing therapy. Third, we explored treatment adequacy at the facility level to identify characteristics of facilities that have larger percentages of veterans completing EBP treatment. The VA health care system provided an excellent context for this work, given the prevalence of PTSD in its patient population (26–28) and the systemwide efforts that have been made in the training, dissemination, and use of EBPs (29).

Methods

This study used a national sample of 16,559 veterans in the VA health system who had at least one psychotherapy visit in any VA facility, had a diagnosis of PTSD coded as the primary diagnosis in the electronic medical record (EMR), used an EBP note template for PTSD treatment (CPT or PE), and began a course of treatment during fiscal year (FY) 2015. CPT and PE are recommended as first-line treatments for PTSD by the VA/Department of Defense clinical practice guidelines (30) and are currently the only PTSD treatments with associated EMR note templates in the VA system.Other psychotherapies that are indicated for PTSD, such as eye movement desensitization and reprocessing, do not have associated note templates and thus cannot be assessed through administrative data sources. All study data were obtained from the VA Corporate Data Warehouse (CDW), which provides national-level reporting capabilities.

The primary outcome of interest was whether an individual received a minimum of eight sessions of evidence-based treatment within 14 weeks of treatment initiation. This metric was based in previous explorations of psychotherapy treatment adequacy, which consistently defined adequate treatment as treatment consisting of between seven and nine sessions (17, 29, 31–38). Further, eight or more sessions was used as a current mental health performance measure by the VA (39), and clinical practice recommendations by the American Psychological Association are based on clinical trials that have generally included eight or more sessions of psychotherapy (40). Sensitivity analyses were conducted by using the outcome of 12 or more sessions within 6 months as criteria for adequate treatment. Participants were excluded from the sample if their first EBP visit was within the last 13 weeks of FY 2015 (or within the last 6 months for the sensitivity analysis), given that their course of treatment may have extended beyond the range of the study.

The number of sessions was quantified by the number of EBP note templates in the VA EMR. EBP templates are tools for writing clinical notes when administering an EBP. The templates list the essential components of the treatment, which vary by EBP type, on a session-by-session basis. Templates allow the provider to click on checkboxes to insert standardized descriptions of EBP components and session information into the note text. Usage of EBP templates generates data tags (i.e., searchable records) of therapy visit information, which become accessible in the CDW for research and program evaluation purposes on a national level.

Predictor variables were obtained through EMR data. Demographic variables of interest included age, gender, race, and ethnicity. Comorbid depression, bipolar or psychotic disorders, substance use disorders, and anxiety disorders other than PTSD were included as dichotomous variables. Service connection for PTSD was also included as a dichotomous variable indicating absence or presence of a current service-connected disability. Record of a symptom measure for PTSD (PTSD Checklist) within 30 days before or after the first session was coded dichotomously. Additionally, the total number of past-year mental health visits (not including the current course of EBP treatment) was included as a predictor of treatment adequacy. Number of past-year visits represents a behavioral indicator of treatment attendance and is readily accessible to VA providers.

Facility-level analyses were also conducted by using variables obtained though EMR and CDW data to predict the percentage of patients with PTSD who completed eight or more sessions within 14 weeks of initiating EBP treatment. All clinics and hospitals were grouped by parent facility. Variables of interest for each facility included geographic region, percentage of patients with a primary PTSD diagnosis who received a PCL during FY 2015, percentage of patients with a primary PTSD diagnosis who received an EBP, and number of certified EBP providers. The list of certified providers and their affiliated sites was obtained through the VA intranet training registry.

Individual-level analyses were conducted using generalized estimating equations (GEE) with the logit link function. GEE models accounted for correlated observations within each VA medical center. Using GEE, the binary variable indicating eight or more sessions of EBP within 14 weeks of initiating treatment was regressed on the predictors of interest in unadjusted and adjusted models. Primary analyses used the total sample, and secondary analyses examined the relationship between predictors of interest and treatment adequacy for CPT and PE, separately. Facility-level analyses used generalized linear models with the logit link function and binomial distribution to predict the overall percentage of EBP treatment adequacy among all patients who initiated an EBP at any of the VA facilities in the sample.

Results

The sample characteristics for the 16,559 veterans are presented in Table 1. The majority of the sample was male (N=13,594, 82.1%), white (N=11,035, 66.6%), and identified as not Hispanic or Latino (N=15,025, 90.7%).

Individual-Level Models

Overall, 5,142 (31.1%) of the veterans who initiated an EBP for PTSD received eight or more sessions of psychotherapy within 14 weeks, and 1,561 of 10,914 (14.3%) veterans received 12 or more sessions within 6 months in sensitivity analyses. CPT (N=12,267, 74.2%) was used more commonly compared with PE (N=4,273, 25.8%). The mean±SD number of EBP visits attended within 14 weeks of the first visit was 5.53±3.58, with noncompleters averaging 3.51±2.02 sessions and completers averaging 10.04±1.69 sessions. The proportion of veterans who received eight or more sessions was greater for CPT (N=3,988, 32.5%) than for PE (N=1,150, 26.9%), and this difference was statistically significant (χ²=46.36, df=1, p<0.001).

Results of the unadjusted and adjusted models for the overall sample are presented in Table 2. In the adjusted model, older age was associated with greater odds of receipt of eight or more sessions of psychotherapy, and comorbid bipolar or psychotic disorders were associated with lower odds. Separate models were fit for PE and CPT (Table 3). For the sensitivity analysis, which examined completion of 12 or more sessions of psychotherapy within 6 months, older age was associated with greater odds of completion, and compared with females, males had lower odds of completion.

Facility-Level Models

There were 129 parent facilities identified in FY 2015. Among all patients with a PTSD diagnosis across sites, 29.3%±10.2% (range 0.0%−65.8%) received eight or more sessions within 14 weeks, 16.3%±9.6% received a PCL during FY 2015, and 3.5%±2.3% received an EBP for PTSD. The mean number of EBP-certified providers per facility was 37.32±22.33. After adjustment, the facility-level model showed that receipt of the PCL by a greater percentage of PTSD patients was associated with lower rates of treatment completion (≥8 sessions within 14 weeks). Receipt of EBPs by a greater percentage of patients and more EBP-certified providers were associated with higher rates of treatment completion (Table 4).

Discussion

Overall, our results suggest that less than one third (31.1%) of veterans with a primary diagnosis of PTSD who initiate an EBP with a VA mental health provider receive an adequate dose of psychotherapy within 14 weeks, and only 14.3% complete 12 or more sessions within 6 months. In general, our findings using systemwide clinical data are consistent with previous reports of low rates of treatment completion during clinical trials of trauma-focused treatment and are concerning given evidence that attrition during the acute treatment phase is related to reduced effect sizes (7, 8, 13).

The results regarding the greater rate of completion during CPT are notable and necessitate further study. Several prior studies have compared dropout rates between CPT and PE, but they have produced equivocal results. For example, one study using retrospective chart review reported higher dropout rates in CPT (17) and another found higher attrition during PE (23). A recent study using secondary data found no difference in rates of dropout between PE and various forms of CPT (41). Last, a study using administrative data from one VA PTSD clinic reported greater use of CPT (N=231) relative to PE (N=188), with more attrition occurring during PE (44.9%) than during CPT (33.3%) (24). Additional data are needed from both randomized trials and routine care to clarify factors related to dropout and improve treatment completion for these frontline treatments.

These analyses identified several demographic characteristics associated with treatment adequacy. Our findings add support to the existing literature, which has found that older age is associated with increased odds of completing an EBP for PTSD (17, 22, 23). Our results are also consistent with findings from a review of psychotherapy, which reported that younger age is a significant predictor of dropout. Relative to men, women had greater odds of completing at least 12 sessions within 6 months (i.e., in the sensitivity analysis). This finding is consistent with recent reports that women are more likely than men to complete treatment for PTSD (31, 32, 36, 42), although this relationship was not observed in explorations of discontinuation from psychotherapy in general (10, 18). Differences in rates of treatment adequacy by race and ethnicity were also observed in several models. Consistent with previous reports of PTSD treatment adequacy (36), identifying as Hispanic or Latino was associated with lesser odds of receiving eight or more sessions within 14 weeks in the combined model; whereas identifying as Asian or American Indian was associated with greater odds of PE and CPT completion, respectively. Given that these associations have also been observed during non-EBP treatment of PTSD, these data may reflect general trends rather than EBP-specific patterns. Additional data are needed to clarify relationships between these demographic characteristics and EBP treatment completion among patients with PTSD.

The presence of a comorbid bipolar or psychotic disorder was related to reduced odds of completing an adequate course of treatment in the overall sample and among those who received CPT. This is somewhat inconsistent with prior studies showing that patients with and without comorbid serious mental illness may discontinue psychotherapy for PTSD at similarly high rates (43–45). However, previous research has suggested that, among those with comorbid serious mental illness, greater PTSD symptoms at baseline were related to increased risk of dropout (44). It is possible that treatment adequacy was negatively related to greater PTSD symptoms in our subsample with comorbid serious mental illness, although data regarding symptom severity were not available in the current study. These findings may also reflect differences between samples of patients receiving routine clinical care relative to those studied in trial and other research contexts. Nonetheless, our findings suggest that many individuals with comorbid PTSD and serious mental illness complete fewer than the recommended number of sessions of EBP for PTSD in routine care settings and highlight the need for more research with this population.

At the facility level, receipt of EBPs by a higher percentage of patients with PTSD and greater number of EBP-certified providers were related to a greater percentage of veterans receiving eight or more sessions of psychotherapy within 14 weeks. These preliminary findings suggest that the percentage of veterans completing adequate treatment during EBPs for PTSD may be higher at facilities that more frequently offer and have provider expertise in delivering EBPs. Nonetheless, further research is needed to explore reasons for this finding. The result that completing the PCL within the FY was associated with lower rates of treatment completion was unexpected. We have previously reported that greater rates of facility-level PCL administration are related to greater frequency of documented EBP use (46). Thus, facility-level PCL administration may be related to whether patients receive documented EBP but not to whether the treatment is adequate.

The finding that only 31.1% of veterans who begin EBPs for PTSD completed an adequate course of treatment suggests additional efforts are warranted at the individual and facility levels. At the individual level, patients at higher risk for receiving less than adequate treatment may benefit from adjunctive interventions, such as addressing comorbidity concurrently with or prior to initiating EBP for PTSD, or carefully assessing treatment barriers and motivations. At the facility level, future work may seek to examine whether promotion of EBPs through clinic policies and provider training may increase completion of adequate treatment. Further, Sayer and colleagues (47) identified qualitative themes that were associated with high use of EBPs for PTSD in VA clinics. These themes included factors such as leadership engagement and promotion of EBPs, effective screening and monitoring of patient progress, clinician beliefs about the effectiveness of EBPs, and scheduling flexibility. In light of our findings that the rate of EBP delivery and number of EBP providers were associated with EBP adequacy at the facility level, it is possible that efforts in these areas could help increase use and completion of EBPs for PTSD in the VA. Although overall rates of treatment completion were low, efforts have been made to increase treatment adequacy during EBPs. For example, some evidence suggests that shared decision making during treatment planning and ongoing feedback to patients and providers may reduce dropout (33, 48) and increase preference for EBPs (49). Additionally, it is important to obtain further data on those individuals who do not complete treatment, given that some may discontinue because of early treatment response (50).

This study had several limitations. First, treatment adequacy was calculated based on template data in the EMR and may underestimate the true rate of EBP usage. Some providers delivering EBPs for PTSD may not use templates or symptom measures, which introduces the potential for sample bias at the provider and facility levels. However, template usage is mandated by the VA for those delivering EBPs; thus, any sample bias is likely toward providers who are more adherent to EBP protocols (29, 51). It is also unlikely that providers who use EBP templates do so only with some patients and not others. Facility-level bias was likely mitigated through use of GEEs, which accounted for within-facility clustering of observations. Administrative efforts are underway in the VA to increase use of templates when delivering EBPs and to implement measurement-based care (2, 29). These efforts will likely lead to improvements in use of administrative data for systemwide estimates of EBP use, completion, and outcomes. Second, retrospective study data are based on standard clinical care, which has several implications. Although the clinical nature of the data represents a strength in that the results reflect rates of treatment adequacy in VA clinics and hospitals, lack of randomization precludes causal inference. Additionally, psychiatric diagnoses in the EMR data were based on providers’ clinical decisions during care and may not correspond with diagnostic interviews and self-reported sources (52). Unfortunately, at the time of EBP initiation, self-reported PTSD symptom measures were documented in the EMR for less than 30% of patients in the sample, precluding the analysis of symptom severity as a potential marker of treatment adequacy. Thus, these data are subject to biases related to misdiagnosis or underdiagnosis during routine care. Third, other aspects of care at the individual (e.g., treatment fidelity) and facility (e.g., promotion of EBPs by leadership and clinical operations designed to deliver and sustain EBP treatment) levels may be important for treatment completion; however, we were unable to address them with these administrative data. Such factors represent important areas for further exploration.

Conclusions

The rate of treatment completion during EBPs for PTSD was low in a national sample of veterans, according to national administrative data obtained during the usual course of care. Younger patients and those with comorbid bipolar or psychotic disorders were at greater odds of not receiving adequate psychotherapy. Facilities with a higher percentage of EBP use and more EBP trained providers had higher rates of treatment completion. Additional research is needed to clarify these factors and to develop strategies to increase the likelihood of PTSD treatment adequacy.