Benzodiazepine Prescribing Variation and Clinical Uncertainty in Treating Posttraumatic Stress Disorder

Posttraumatic stress disorder (PTSD) has affected the lives of hundreds of thousands of U.S. veterans and continues to afflict thousands of veterans returning from combat. Fortunately, several internationally recognized, evidenced-based clinical practice guidelines are available to aid clinicians in caring for veterans with PTSD (1–3). Among the more controversial positions expressed in these guidelines is that benzodiazepine use is inappropriate in the management of PTSD. Despite this recommendation, benzodiazepines are among the most common medications prescribed in this population. In 2009, 30.6% of veterans with PTSD who sought care through the Veterans Health Administration (VHA) received a benzodiazepine, which was a higher proportion than those who received second-generation antipsychotics (24.3%), trazodone (23.0%), nonbenzodiazepine hypnotics (12.8%), and prazosin (9.1%) (4). Benzodiazepine prescribing was second only to selective serotonin reuptake inhibitors (52.5%), the only therapeutic class containing medications with U.S. Food and Drug Administration approval for PTSD. These findings suggest considerable uncertainty across VHA prescribers concerning the implementation of guideline recommendations against benzodiazepine use. One common strategy to examine clinical uncertainty is to study practice variation, typically across geographic regions or health service catchment areas. As the nation’s largest centralized health system, caring for nearly 500,000 individuals with PTSD, the VHA is uniquely positioned to examine variation in its prescribing practices.

One specific area of concern related to practice variation is the quality of care delivered by VHA community-based outpatient clinics (5). The goal with the establishment of these clinics is to offer access nearer to the veteran’s home, and over 800 clinics were in operation as of 2010. The challenge is to ensure high-quality care equivalent to that provided by Veterans Affairs (VA) medical centers, despite differences in the breadth of services available and many other important resources. One particular concern has been mental health monitoring and a perceived lack of access to mental health specialists (6).

Access to high-quality care at community-based outpatient clinics is particularly important for rural veterans, who account for approximately 41% of VHA enrollees (7). An important driving force behind clinic expansion has been the increasing number of veterans serving in combat roles who came from and are returning to rural areas (7). Rural residence has been associated with problems accessing health care, worse health status, and higher prevalence of chronic diseases (8–12). Differences in physician practice patterns between rural and urban settings have been observed, although much of this work has been done outside the VHA (13–16). Although there do not appear to be important disparities in access to prescription medications (17), some studies suggest that rural residents may be at increased risk for certain types of inappropriate prescribing practices (18,19). Ensuring access for rural veterans has been a focal point for the VHA, but the potential impact on prescribing quality has neither been sufficiently studied nor well characterized among veterans with PTSD.

To address these important issues, this study included two primary objectives related to benzodiazepine prescribing variation among veterans with PTSD. Our first objective was to characterize the variation in prescribing frequency across multiple levels of aggregation, including census region, Veterans Integrated Service Network (VISN), and individual medical center. We hypothesized that clinically significant variation in benzodiazepine prescribing would be observed at all levels but that the extent of variation would decrease over time. Our second objective was to evaluate differences in benzodiazepine prescribing frequency between rural and urban residents and between community-based outpatient clinics and medical centers. We hypothesized that benzodiazepine prescribing would be more common among rural residents and for patients receiving care at community-based clinics.

Methods

Data sources

National administrative VHA data were obtained for the 11-year period spanning fiscal years (FYs) 1999 through 2009 (October 1, 1998, to September 30, 2009). Prescription drug records were obtained from the VHA Pharmacy Benefits Management Services. Inpatient discharge and outpatient encounter data sets were obtained from the Austin Information Technology Center. This study was approved by the University of Iowa Institutional Review Board and the Iowa City VHA Research and Development Committee.

Patients

Eligible veterans for this study included all VHA enrollees in FY 1999 through FY 2009 who had an inpatient or outpatient encounter with an ICD-9 code of 309.81. Veterans were considered to have PTSD during a given year if they had at least one encounter coded for PTSD as either a primary or secondary diagnosis. This PTSD case definition has been used in several prior studies examining psychiatric medication use among veterans with PTSD (20–22). The estimated rate of false-positive cases resulting from administrative miscoding is infrequent (<4%) with this methodology (23,24). As previously reported, the number of veterans treated for PTSD in the VHA increased nearly threefold during the study time frame, from 170,685 in FY 1999 to 498,081 in FY 2009 (22). In FY 2009, 7.5% of veterans with PTSD were women, and the mean±SD age was 53.8±14.6 years.

Benzodiazepine use

Benzodiazepine use was defined as any outpatient prescription fill for the following medications: alprazolam, chlordiazepoxide, clonazepam, clorazepate, diazepam, estazolam, flurazepam, halazepam, lorazepam, oxazepam, prazepam, quazepam, temazepam, and triazolam. This definition did not include any requirement for a minimum quantity, days’ supply, or specific dosage form. A majority (94%) of veterans with any benzodiazepine use received ≥30 days’ supply, and approximately two-thirds received more than 90 days of continuous benzodiazepine treatment (22).

Site of care

The primary site of PTSD care was assigned on the basis of the station where the veteran had the most PTSD-coded encounters during a given fiscal year. Each site of care was classified as a community-based outpatient clinic or medical center and aggregated at multiple organizational levels for different analyses. Each community-based outpatient clinic in the VHA is assigned to a parent medical center, and medical centers are assigned to a VISN. We further grouped VISNs into regions based on overlapping boundaries with U.S. census regions (Northeast, South, Midwest, and West). Stations were considered a medical center according to the classification used by the 2009 VHA Facility Quality and Safety Report (25). Medical centers and community-based outpatient clinics are predominantly located in urban areas (86% and 58%, respectively).

Rural or urban residence

Rural or urban residence was determined by using the Rural-Urban Commuting Areas (RUCA) system, which was mapped with the zip code of the veteran’s residence (26). Beginning with the RUCA four-category classification system (urban, large rural towns, small rural towns, and isolated rural towns), we further collapsed all nonurban categories (large, small, and isolated rural towns) into one rural category.

Analyses

Benzodiazepine frequencies were reported at three hierarchical levels: medical center (N=137), VISN (N=21), and region (N=4). Variation in prescribing frequency was described at the medical center and VISN levels by using range and interquartile range. Inferential statistics were not used to make comparisons across these levels because our data included the entire population of veterans receiving care within the VHA. However, we discussed the clinical significance of the observed frequencies and variability. Comparisons in benzodiazepine prescribing frequencies by rural and urban residence were expressed as odds ratios (ORs) with 95% confidence intervals (CIs). Statistical comparisons between community-based outpatient clinics and medical centers used a chi square test. All statistical analyses were conducted with SAS version 9.2.

Results

Region, network, and medical center

Benzodiazepine prescribing frequencies among veterans with PTSD, stratified by geographic region, are shown in Table 1. Substantial variation was observed early in the study period and ranged from 31.1% in the Northeast to 43.3% in the South in FY 1999. Benzodiazepine use declined over the study period in all regions but declined most notably in the South. The range in regional prescribing frequencies decreased from 12.2% in FY 1999 to 4.0% in FY 2009.

Decreased variation was also observed at the VISN level and across medical centers (Table 2). Variation at the network level declined by just over 50% from FY 1999 to FY 2009—a finding supported by the reduction in the interquartile range. Relative to the network level, the magnitude of variation was greater at the medical center level (FY 1999 minimum 14.0% to maximum of 75.9%), although variation declined at the facility level over time (from 14.9% in FY 1999 to 9.1% in FY 2009).

Whereas the national trend toward declining benzodiazepine use was clear, several VISNs (N=3) and medical facilities (N=33) experienced increases in benzodiazepine prescribing. At the network level, two VISNs had absolute frequency increases <1%, but one VISN had an increase of 5.9%. Increases at the medical center level ranged from .3% to 19.2% and tended to occur among medical centers with low initial rates, which likely reflected regression to the mean.

Rural residence

Nationally, 26.9% of veterans with PTSD resided in rural areas in FY 2009, and this frequency varied by region: Northeast, 20.1%; West, 21.9%; South, 28.8%; Midwest, 36.0%. Benzodiazepine prescribing frequencies for rural and urban veterans with PTSD can be compared by national region in Table 3. At the national level, rural veterans were more likely than their urban counterparts to receive benzodiazepines (40.4% versus 35.2%) in FY 1999 (OR=1.24). By FY 2009, prescribing frequencies decreased for both rural (33.2%) and urban (29.4%) veterans, but the magnitude of the rural-urban difference remained consistent (OR=1.19).

However, aggregate national findings were not representative of rural-urban differences across individual geographic regions. Rural-urban differences in benzodiazepine prescribing in the Northeast and West regions were statistically significant in some years, but the relationship was inconsistent and effect sizes were small. In contrast, rural-urban differences were substantial in the Midwest and South. In FY 1999, the benzodiazepine frequency for rural veterans in the Midwest was 40.2%, compared with 32.4% for urban veterans (OR=1.41, CI=1.34–1.48). Benzodiazepine prescribing declined substantially by FY 2009 for both rural (35.6%) and urban (29.8%) Midwest veterans, but the rural-urban difference persisted (OR=1.30, CI=1.27–1.34). A similar pattern was seen in the South, where benzodiazepine prescribing frequencies decreased among both groups from FY 1999 to FY 2009 (rural, from 46.5% to 34.0%; urban, from 41.9% to 29.8%), although rural-urban differences remained consistent.

Community-based outpatient clinics

Rates of care at community-based outpatient clinics increased during the study period, particularly among veterans in rural areas. In FY 1999, 31.7% of rural veterans had the majority of their PTSD encounters at a community-based outpatient clinic, which increased to 50.9% in FY 2009. Community-based outpatient clinic use also increased for urban veterans, with PTSD encounters ranging from 35.0% in FY 1999 to 45.3% in FY 2009. Benzodiazepine prescribing frequencies at community-based outpatient clinics and medical centers are shown in Table 4. Among all veterans with PTSD, benzodiazepine prescribing frequencies were generally similar for patients receiving care at community-based outpatient clinics and medical centers across all years during the study period. However, stratification by rural and urban residence revealed some differences. Benzodiazepine prescribing frequencies were generally similar between community-based outpatient clinics and medical centers for urban veterans. In contrast, significant differences between care at community-based outpatient clinics and medical centers were apparent for rural veterans beginning in FY 2003, where benzodiazepine prescribing frequencies were lower for veterans treated at community-based outpatient clinics.

Discussion

This study examined variation in benzodiazepine prescribing over an 11-year period using several aggregation approaches, based on both geographic and health system boundaries. Our primary finding was that the level of practice variation was extensive, with benzodiazepine prescribing frequencies ranging from 14.0% to 75.9% across VA medical centers in FY 1999. This finding suggests the existence of clinically relevant uncertainty among VHA prescribers concerning the role of benzodiazepines for veterans with PTSD. Building on prior work demonstrating a general declining trend in the absolute frequency of benzodiazepine prescribing (22), this study documented important decreases in variation over the same period. This observation was consistent across multiple levels of aggregation, including medical center, VISN, and regional levels. We further observed that benzodiazepine prescribing was elevated among rural veterans at a national level but that this association appeared to be limited to the South and Midwest regions of the United States. Finally, benzodiazepine prescribing at community-based outpatient clinics was similar to that at medical centers for veterans with PTSD overall and was somewhat lower for the veterans who were rural residents.

The driving forces behind declining variation in benzodiazepine prescribing are unclear. Potential regional influences present in FY 1999 (range 31.1%–43.3%) were essentially eliminated by FY 2009 (range 27.9%–31.9%), driven principally by a decrease of 12.1 percentage points in prescribing in the South. VISN-level differences were also substantially reduced, from an interquartile range of 10.8% to 5.3%. In FY 2009, the 25th and 75th percentiles were 27.9% and 33.2%, respectively, suggesting minimal influence of VISN-level policies or organizational characteristics on current rates of benzodiazepine prescribing. However, these factors may have played an important role in the decline in prescribing variation over time.

One potential driver with national scope may be the 2004 clinical practice guideline from the U.S. Department of Veterans Affairs and the Department of Defense. Anecdotally, the largest biannual decrease in benzodiazepine prescribing variation, at both the VISN and the medical center levels, occurred between 2003 and 2005. However, our study was not designed to measure the impact of the guidelines on prescribing practices, and drawing a causal connection remains speculative.

With benzodiazepine prescribing variation significantly diminished at the regional and VISN levels, most of the remaining variation seems to be attributable to local differences among individual medical centers. Further research is needed to determine the extent to which prescribing variation may be accounted for by differences in patient-level characteristics across facilities and in organizational characteristics of individual facilities. Facility characteristics could include specific local policies as well as differences across facilities in the prescribing culture regarding benzodiazepine use for PTSD. However, facility-level prescribing variation is ultimately the aggregate behavior of individual prescribers. Each prescriber caring for veterans with PTSD has an individual propensity to prescribe benzodiazepines. Although administrative data are useful in characterizing patterns of prescribing, they offer little toward explaining why these variations occur. Future qualitative work will be needed to gain a richer understanding of what factors drive facility-level variation, including local policies, prescribing culture, and individual prescriber decision making. An additional opportunity is to examine the relative contribution of prescribers in primary care versus mental health care toward the frequency and variability of benzodiazepine prescribing. Primary care physicians’ awareness of PTSD guidelines could be evaluated along with likelihood of prescribing benzodiazepines and understanding of the risk for potential harms among veterans.

Our second objective was to determine whether benzodiazepine prescribing was more common for rural veterans and for patients receiving care at community-based outpatient clinics. On a national basis, rural veterans were more likely to receive benzodiazepines, but this difference was limited to the Midwest and South. This finding has important implications for studies of urban-rural discrepancies, which often look only at national estimates and may miss relevant regional differences. Urban-rural differences remained consistent throughout the study period, which is noteworthy given the overall backdrop of declining variation. However, we again face the limitations of administrative data to identify the underlying reasons for these associations. One important consideration is that rural veterans are disproportionately represented among combat troops, which may predispose these individuals to developing more severe or treatment-refractory PTSD and lead prescribers to resort to treatment options with less supporting evidence (27–29). It is also possible that urban veterans had better access compared with rural veterans to evidenced-based psychotherapy treatment, which could have decreased benzodiazepine prescribing. A further consideration is whether higher benzodiazepine prescribing among rural veterans was explained by a greater reliance on community-based outpatient clinics. However, we found that veterans with PTSD receiving care at community-based outpatient clinics were actually less likely to receive benzodiazepines than individuals treated at medical centers. This finding is important given the VHA’s ongoing efforts to ensure access to high-quality mental health services.

It is important to outline the limitations of this study. First, the study population was limited to veterans who received care within the VHA, and our findings may not be generalizable to nonveterans or to veterans receiving PTSD care outside the VHA. The relative quality and impact of PTSD treatment outside the VHA will be of increasing importance because VHA enrollment is below 50% for veterans returning from recent conflicts abroad (30). Second, we identified veterans with PTSD by using administrative diagnostic codes, which are susceptible to sensitivity and specificity problems. Diagnostic misclassification could have multiple competing effects on the absolute frequency of benzodiazepine prescribing but are unlikely to explain the wide variation across the VHA or the declining variation over time. Third, we do not know the clinical indication for which the benzodiazepine was prescribed. There are certainly case examples where benzodiazepine use could be clinically appropriate, such as with severe comorbid panic disorder. We do not advocate a target rate of zero for benzodiazepine use, because this practice would disallow individual circumstances to drive patient-centered care. However, it seems unreasonable that the one medication class that clinical practice guidelines recommend avoiding in this population would be the second most commonly prescribed (4).

Regardless of how absolute benzodiazepine prescribing frequencies are interpreted, it seems clear that substantial facility-level variation is indicative of uncertainty regarding best practices. Further research is necessary to identify potential subgroups of veterans with PTSD for whom the benefit of benzodiazepines may outweigh risks. A final limitation was that we did not include case-mix adjustment to account for potential differences across facilities in patient-level characteristics related to benzodiazepine prescribing. For example, one prior cross-sectional study found that a comorbid anxiety disorder increased the likelihood of benzodiazepine use among veterans with PTSD (20). Although case-mix differences seem unlikely to account for the extent of prescribing variation observed in our study, future research examining patient- and facility-level correlates of benzodiazepine use would be informative.

Conclusions

Benzodiazepine prescribing among veterans with PTSD remains common despite clinical practice guideline recommendations. Wide variation across the VHA health care system reflects uncertainty among providers regarding best practices and is ultimately due to the limited number of effective PTSD treatments supported by a strong evidence base. Our findings also serve to highlight the challenges faced by veterans with PTSD in working toward recovery, as well as the providers involved in their care. As the number of veterans affected by PTSD continues to climb, expanding our understanding of evidence-based treatments becomes ever more important. Benzodiazepine use among veterans with PTSD remains controversial, and further research is crucial to maximize guideline-concordant prescribing practices in the VHA.