(P-075) Visualization Methods to Examine Treatment Patterns in Pharmacoepidemiology: An Example Using Osteoporosis Medications

Background: Treatment uptake in real-world populations is constantly changing in response to new guidelines, approvals and emerging evidence. It can be challenging to display prescription treatment patterns that span multiple years and include many different therapies in a changing landscape.

Objectives: Using Sankey and sunburst diagrams, we depicted treatment patterns among users of osteoporosis therapies.

Methods: We used claims data from Optum Clinformatics to identify a cohort of women aged >=55 years who filled a prescription for therapy between February 2012-March 2019. We excluded women with history of Paget’s disease and cancer and women without continuous enrollment >=455 days. We developed Sankey diagrams, which depict the flow of patients between cross-sectional treatment states (ie, individual therapies, discontinuation, or death) at treatment initiation and during follow-up at 6, 12, 18 and 24 months post-initiation. Treatment state transition probabilities were calculated between successive time points by estimating the conditional probability that a patient transitioned to a particular treatment state, given her prior state and continued enrollment. Sunburst diagrams display hierarchical structures of treatment patterns, with the first treatment state represented by the origin circle and each subsequent treatment state shown as successive outer circles. Sunbursts present the frequency of treatment progression patterns, while Sankey diagrams provide a complementary view of transitions at fixed timepoints. Separate diagrams were created for clinically important subgroups (age <65 vs >=65, treatment-experienced vs new users, history of fracture vs no fracture).

Results: We identified 264,787 women initiating an osteoporosis therapy. Using a Sankey diagram, we identified increasing numbers of patients who had discontinued therapy after initiation (27.6% at 6 months; 37.8% at 12 months; 46.9% at 18 months; 49.6% at 24 months). Similarly, the sunburst diagram shows a common pattern of alternating periods of treatment and discontinuation. For example, 12% of all identified patients initiated, discontinued and then resumed alendronate within two years of initiation. Switching medications was less common across therapies (eg, <4% of alendronate initiators switched to another therapy).

Conclusions: Sankey and sunburst diagrams convey a wealth of information concisely and are effective tools for conveying complex information about treatment patterns over time to a broad audience of stakeholders, including clinicians and patient advocates.