Background/Purpose: Self-reported fracture (fx) history data is frequently used in epidemiological studies of osteoporosis. Self-reported fx data may differ from fx history coded in electronic health records (EHR) due to imperfect patient recall, incomplete communication with clinicians, or lack of a universal EHR. Because both self-reported fx history and EHR data can define phenotypes for clinical research studies, it is important to understand how these 2 data sources compare. Our objective was to compare self-reported fx history using survey data with fx codes from an available EHR dataset.

Methods: Self-reported fx data was derived from the Activating Patients at Risk for OsteoPOroSis (APROPOS) trial, which recruited participants from the Global Longitudinal study of Osteoporosis in Women (GLOW) cohort. Prior fx data was collected using a survey deployed June – August 2015. Women were asked if they ever had a fx and for each fx type the date of the most recent one. Data on fx recorded in the EHR September 2011 – June 2015 was obtained from Kaiser Permanente Washington Health Research Institute. We excluded skull, toes and fingers fxs. We defined concordance between the EHR and self-reported data if the location of fx was reported to be the same and if the reported dates were within 1 year of each other. Kappa (κ) statistic described the concordance between the 2 sources of fx history. Descriptive statistics evaluated potential factors associated with discordance between the self-reported and EHR-coded fx history.

Results: A total of 133 fxs from 360 women (91% white, mean[SD)] age 74.5(7.5) years, 82% had some college education) were included. There were 35 fxs reported on the survey but not in the EHR and 39 fxs coded in the EHR but not in the survey. Agreement between self-reported and EHR fxs was κ 0.48. Of the discordant fxs, we were more likely to find claims for fxs in EHR referent to self-report among whites (OR=5.5, 95%CI 1.1-27.9), for major osteoporotic fxs (OR=2.8, 95%CI 1.1-7.1), and for fragility fxs that typically require hospitalization (vertebral, hip, femur, pelvis) (OR=3.8, 95%CI 1.3-10.7). Discordance between EHR codes and self-reported fxs did not vary by age, formal education, or health literacy.

Conclusion: There was only modest correlation between self-reported fx history and EHR fx codes. This discrepancy may have implications for clinical and epidemiological studies of fxs suggesting that combining both types of data may be optimal.

« Back to 2018 ACR/ARHP Annual Meeting

ACR Meeting Abstracts - https://acrabstracts.org/abstract/self-report-of-fracture-history-compared-to-fracture-codes-from-an-electronic-health-record-dataset/