Search | Search by Center | Search by Source | Keywords in Title
Abrams T, Richardson KK, Lund BC. Variations in Pharmacotherapy Associated with a Diagnosis of PTSD. Poster session presented at: VA HSR&D / QUERI National Meeting; 2012 Jul 16; National Harbor, MD.
Objectives: Most of the larger, observational research studies used to estimate rates of evidenced-based pharmacotherapy among Veterans with PTSD have relied on administrative data derived algorithms. Unfortunately, such studies lack uniformity on defining PTSD. Thus, we sought to examine the variation in three commonly prescribed drugs as identified by three PTSD coding algorithms. Methods: A cross-sectional study design was used to identify all Veterans with a visit for PTSD (n = 371,467) in fiscal year 2009 using three PTSD coding algorithm methods: 1) one or two outpatient visits, 2) three or more outpatient visits, or 3) at least one inpatient visit. Three classes of medication use were identified, selective serotonin reuptake inhibitors (SSRI), second generation atypical (SGA), and benzodiazepines (BZD). Each medication was identified using electronic pharmacy data and was defined by receipt of at least one prescription fill for at least 30 days. Unadjusted analyses used analysis of variance to compare rates of medication use and adjusted analyses controlled for demographical and medical comorbidity. Results: Average age was 55.8 (SD 14.6) and 92% were male. Variations in BZD fills were substantial by coding algorithm, for example SSRI fills varied from 47.6% (n = 53,903) to 70.2% (n = 80,511); SGA fills varied from in 13.8% (n = 31,473) to 44.1% (n = 100,577); and BZD fills varied from 22.4% (n = 6,431)to 36.2% (n = 10,393). Adjusted analyses found that a PTSD diagnosis as identified by method 3 relative to method 1 was substantially more likely to be associated with higher rates of SSRI, SGA, and BZD use; OR 2.3 (2.22 - 2.35) for SSRI, 4.0 (3.84 - 4.09) for SGA, and 2.39 (2.32 - 2.46) for BZD, respectively. Implications: Estimating the use of evidenced (SSRI) and non-evidenced based (SGA and BZD) medications remains quite variable depending on the coding algorithm employed with an observed prevalence of varying from 48-70% for SSRI use, 14-44% for SGA use, and 22-40% for BZD use. Impacts: As comparative effectiveness research moves forward using administrative databases, selecting algorithms used to identify PTSD remains paramount in deriving interpretable estimates for pharmacotherapy rates. This study demonstrates that rates vary widely depending on how PTSD is identified