Prediction and Utility of a clinical fracture risk score in administrative claims Public Deposited

Downloadable Content

Download PDF
Last Modified
  • March 20, 2019
  • Overman, Robert
    • Affiliation: Eshelman School of Pharmacy, Division of Pharmaceutical Outcomes and Policy
  • The clinical manifestation of osteoporosis is osteoporotic fracture, which has been estimated to cost $25.3 billion by 2025 within the US healthcare system. Osteoporotic fracture risk has been measured using various risk scores with the most prevalent being FRAX® from the World Health Organization. FRAX® scores are used clinically to guide treatment, but these scores and key inputs (such as bone mineral density and body mass index) cannot be measured in administrative claims. The objectives of this dissertation are 1) to create a claims-based fracture risk score to determine if administrative claims data can be used to predict FRAX® (interval validation); to evaluate how the risk score performs in a different population (external validity); and 3) to determine the best way to utilize the fracture risk score in a research study. For this project, we linked registry data including clinical fracture risk factors from a multispecialty academic hospital with Medicare administrative claims for individuals receiving a dual energy x-ray absorptiometry scan (DXA) between 2009 and 2013. FRAX® has 4 different scores for 10-year fracture risk of hip and major osteoporotic fracture (MOF) with and without bone mineral density. We created the Calculated Fracture Risk Index (CFRI) to estimate these 4 scores. We found that we were able to predict a continuous FRAX® score with an adjusted R2 that accounted for between 21 to 43% of variation in the estimates. We found these estimates to be internally valid. Subsequently we used the linked dataset and a 20% random selection of fee-for-service Medicare beneficiaries to evaluate the external validity of our CFRI scores. We found no significant differences in CFRI and FRAX® ability to predict 1 year fractures. Additionally, we found CFRI and FRAX® to be similarly calibrated. Lastly, we found that we were not able to sufficiently reduce confounding in a non-experimental comparative effectiveness study of alendronate users versus non-users to that of a randomized clinical trial using CFRI as a regression component or a restriction device. Although estimates including CFRI reduced confounding, residual confounding remained and estimates differed from those in the Fracture Intervention Trial (FIT); the gold standard in for our comparisons. Overall CFRI appears to be internally and externally valid and a useful tool in reducing confounding compared to its non-use in osteoporosis research, though not to the level of an RCT. It also appears to be a reasonable proxy score for FRAX® when only administrative claims data are available. Therefore, CFRI when calculated in administrative claims should be useful for both researchers and policy makers to determine who is at risk for osteoporotic fracture.
Date of publication
Resource type
Rights statement
  • In Copyright
  • Dusetzina, Stacie
  • Farley, Joel
  • Sleath, Betsy
  • Gourlay, Margaret
  • Brookhart, M. Alan
  • Doctor of Philosophy
Degree granting institution
  • University of North Carolina at Chapel Hill Graduate School
Graduation year
  • 2017

This work has no parents.