Previous research indicates that patient, provider, and health system characteristics are important factors in the adoption of healthcare innovations, including new prescribing patterns and/or medication utilization. However, efforts to increase evidence-based prescribing have been only modestly successful. These efforts have been of two main types: administratively based formulary manipulations such as drug restriction or preauthorization, and educational efforts based on diffusion theory such as academic detailing of opinion leaders or high-volume prescribers. We propose that evidence-based prescribing can be enhanced by focusing such interventions for maximal effectiveness and efficiency, and that this depends on knowledge of the flow of information and influence among providers working within a specific organizational context with specific types of patients. Such knowledge requires, specifically: (a) early identification of change in prescribing behavior (surveillance), (b) identification of characteristics of prescribers with a propensity to early adoption while controlling for confounding factors (including patient characteristics), and (c) identification of specific points of intervention by determining the relative importance of social vs. administrative factors in a provider¿s decision to prescribe. Geographic methods provide an innovative, large-scale approach to understanding the process of diffusion of innovation because provider, practice environment (location), and patient characteristics can be integrated into one empirical model that can be used to design and target practice-change interventions. In particular, space-time cluster analysis (STC) allows us to identify groups of providers and/or locations that change their behavior first (¿early adopters¿). These geographic methods first construct STCs of prescription events. Specifically, STCs are defined as geographical areas characterized by relatively higher rates of prescribing in a given time interval. Hierarchical General Linear Modeling then allows us to explain the development of these clusters using patient, provider and facility characteristics. We therefore propose to identify STCs that describe the spread of prescribing of second generation antipsychotics (SGAs) for two serious mental illnesses of high cost and priority to the VHA: bipolar disorder and PTSD. Focusing on SGAs in bipolar disorder takes advantage of several discrete events in 2004¿new FDA indications for bipolar disorder¿that anchor the investigation of diffusion. PTSD provides an important complementary disorder by which to study innovation spread since SGAs have become widely used for PTSD¿though without FDA indications. We will utilize national VHA data from Decision Support System (DSS), the Personnel and Accounting Integrated Dataset (PAID) and related datasets to identify STCs of early adopter providers prescribing SGAs for bipolar disorder, and within these STCs to evaluate: (a) the demographic characteristics of prescribers; (b) the demographic characteristics of the patients who receive prescriptions, and (c) the structural and cultural organizational characteristics at the VISN, VAMC, and CBOC levels within which the prescribing occurs. We will then characterize the robustness of early adopter prescriber profiles by determining the consistency of early adopter characteristics across SGAs, and determining whether the same characteristics that identify early adopters for bipolar disorder also identify early adopters for PTSD. Finally, we will develop an integrated model that characterizes the relative strength of diffusion-based versus organizational factors in prescribers¿ likelihood of adopting SGAs for bipolar disorder. We hypothesize that both geographic factors, consistent with classic diffusion theory, and organizational factors, as articulated in more recent applications of diffusion theory to dissemination within healthcare organizations, will shape SGA spread and, therefore, identify opportunities for intervention.

The use of newer, "second generation" antipsychotics (SGAs) is of substantial importance to the VHA and the Veterans we serve. SGAs now represent 4 of the top 10 most costly drugs in the VHA, accounting for expenditures of over $350M/year. Much of this use is not evidence-based, and identifying the factors that contribute to the spread of SGA prescribing will support the development of maximally effective and efficient interventions to guide prescriber behavior. Geographic methods provide an innovative, large-scale approach to understanding the spread of prescribing, integrating provider, practice, and patient characteristics into one empirical model that can be used to design and target provider behavior change interventions. This study utilizes VHA national datasets to model the spread of SGA prescribing for two serious mental illnesses of high cost and priority to the VHA: bipolar disorder and PTSD. This study also tests and refines diffusion theory, which underlies the majority of provider behavior change strategies within and beyond the VHA.