Introduction: Most patients who suffer a minimal trauma fracture (MTF) remain undiagnosed and untreated1. Osteoporosis refracture prevention (ORP) services are implementing different models of care for early identification of patients with MTFs. Case finding may be assisted by electronic search tools to automatically screen medical records for fracture. Our study assessed the efficacy of two new tools in identifying MTFs at two tertiary hospitals in Sydney.
Methods: ‘XRAIT’ uses natural language processing of imaging reports to detect fractures while the ‘AES’ tool identifies fractures through hierarchic disease code and text-based search of the electronic Medical Record (eMR) and radiology reports, respectively. Data were collected from 1/7/2018 to 31/12/2018. A sample of the extracted reports was then manually reviewed to determine the specificity and sensitivity of each search tool in detecting MTFs. The eMR was accessed to determine the fracture mechanism and treatment status.
Results: The true positive rate was similar for both tools (76.6-88.4%). However, in terms of detecting MTFs (rather than just fractures or a code-based diagnosis of osteoporosis) the tools performed differently at different sites, with 55.3-87.7% correct identification. Each tool identified separate subsets of patients. When all patients detected by both tools were combined, the AES tool identified 52.3-55.3% of patients with a MTF, while the XRAIT tool identified 87.7-93.2% of these patients (similar in both centres). Less than half (43-45.4%) of patients with a MTF were detected by both tools.
Conclusion: Both tools had a high true positive rate of identifying fractures, however one tool missed almost half of all MTFs. A hybrid tool that combines the methodology of both XRAIT and AES could improve patient identification. In the meantime, ORP service providers should be cognisant that subsets of patients continue to be missed with both tools.