Identifying proximal humerus fractures: an algorithmic approach using registers and radiological visit data

Area of Science:

• Medical informatics
• Orthopedic research
• Public health surveillance

Background:

• Accurate identification of proximal humerus fractures is crucial for managing osteoporosis and fragility fractures.
• Traditional administrative register analysis has limitations in capturing all fracture cases.
• Automated methods are needed for efficient large-scale fracture tracking.

Purpose of the Study:

• To evaluate the reliability of identifying proximal humerus fractures using administrative datasets without manual review.
• To compare the accuracy of traditional register analysis with a combined register and radiological data approach.
• To develop and validate algorithms for automated fracture identification.

Main Methods:

• Utilized national healthcare registers and a regional radiological image archive (PACS).
• Developed algorithms for automated proximal humerus fracture identification.
• Established a gold standard using patient records and self-reports from the Kuopio Osteoporosis Risk Factor and Prevention Study (OSTPRE) for validation.
• Analyzed data from 11,863 post-menopausal women (2004-2022).

Main Results:

• Traditional register analysis achieved 75% coverage of proximal humerus fractures over 19 years.
• Combining register and radiography visit data improved fracture identification coverage from 74% to 81%.
• The combined approach reduced the false discovery rate from 8% to 7% compared to register analysis alone.

Conclusions:

• The proposed method offers a more reliable automated approach for identifying proximal humerus fractures from administrative data.
• This advancement supports the automated tracking of fragility fractures in extensive datasets.
• Improved accuracy in fracture identification can enhance patient care and epidemiological studies.