A comparison of variance estimators for logistic regression models estimated using generalized estimating equations (GEE) in the context of observational health services research
Austin PC. Stat Med. 2024; Oct 31 [Epub ahead of print].
We identified inconsistency in fracture definitions in a prior review of studies that utilized claims data. Here, we aimed to compare fracture rates estimated using thirteen hip and seven radius/ulna fracture definitions. Our primary analysis compared results in a cohort of 120,363 older adults treated with oral bisphosphonates for ≥3 years. The most inclusive definition (hip: inpatient or emergency diagnosis; radius/ulna: inpatient, emergency, or outpatient diagnosis) served as a referent to compare the number and proportion of fractures captured. In sensitivity analyses, we considered a 180-day washout, excluded fractures associated with trauma; and hip only, excluded: (1) subtrochanteric fractures, and (2) hip replacement procedures. Hip fractures varied by definition in number (52–8058) and incidence (0.7–111.8/10,000 person-years). The second most inclusive definition required one inpatient diagnosis and identified 8% fewer hip fractures than the referent. Excluding hip replacements missed 33% of hip fractures relative to the primary analysis. Radius/ulna fractures also ranged in number (1589–6797) and incidence (22.0–94.3/10,000 person-years). Outpatient data were important, when restricted to inpatient or emergency data, only 78% of radius/ulna fractures were identified. Other than hip replacement procedures, sensitivity analyses had minimal impact on fracture identification. Analyses were replicated in a cohort of patients treated with long-term glucocorticoids. This study highlights the importance and impact of coding decisions on fracture outcome definitions. Further research is warranted to inform best practice in fracture outcome identification.
Konstantelos N, Burden AM, Cheung AM, Kim S, Grootendorst P, Cadarette SM. Pharmacy. 2023; 11(2):53. Epub 2023 Mar 14.
The ICES website uses cookies. If that’s okay with you, keep on browsing, or learn more about our Privacy Policy.