(P-095) Fracture Outcome Coding in Observational Osteoporosis Drug Effects Studies: A Scoping Review

Background: Algorithms that define fractures using healthcare administrative data (hospitalization, outpatient, and emergency) vary and can include diagnostic or procedural codes, as well as washout windows between eligible codes. There is currently no standard for fracture outcome coding, resulting in inconsistencies in fracture definitions.

Objectives: Describe fracture outcome coding definitions in observational studies that consider osteoporosis drug effects and use healthcare administrative data.

Methods: We searched MEDLINE, Embase, and CINAHL for studies published in English from January 1, 2000 to June 29, 2020. We included North American observational studies that leverage healthcare administrative data to code fractures and estimate the association between osteoporosis medications and fracture risk. Screening and data extraction were completed independently by two reviewers.

Results: We identified 56 eligible papers, most evaluated effectiveness (n=49, 88%), 2 (4%) evaluated safety, and 5 (9%) explored both. Over half (52%) did not provide a citation for their algorithm, and the same amount did not indicate specific data sources for one or more of their outcome definitions. Hip fracture was the most common fracture site studied (n=51, 91%), followed by humerus (68%), vertebral (63%), radius/ulna (59%), and atypical femoral fractures (9%). There was heterogeneity in the algorithms used to define fractures. Among studies that indicated their data sources, 13 distinct algorithms were used to define hip fracture, 9 for humerus fractures, 8 for radius/ulna fractures, and 10 for vertebral fractures. The most common algorithm within each fracture site was used by less than 40% of studies that examined effectiveness. The most common algorithms included 1 diagnosis from hospitalization data (hip, 39%), 1 diagnosis from hospitalization data or 2 diagnoses from outpatient data (vertebral: 20%; radius/ulna: 25%), 1 diagnosis from hospitalization or outpatient data (humerus, 29%). Two algorithms were used to define the safety outcome, atypical femoral fracture, with 75% using 1 diagnosis from hospitalization data. Washout periods were most common for humerus and vertebral fractures (29%), followed by radius/ulna (24%), atypical femoral (20%) and hip fracture (14%).

Conclusions: There is ambiguity and heterogeneity in fracture outcome definitions in studies that leverage healthcare administrative data. Better transparency in reporting is needed. Future exploration on how fracture definitions can impact study results is warranted.