Identification of fractures from computerized Medicare files

Area of Science:

• Gerontology
• Public Health
• Health Informatics

Background:

• Non-hip fractures pose a significant public health challenge for individuals aged 65 and older.
• Economical and efficient methods for identifying fracture cases are crucial for research and intervention.

Purpose of the Study:

• To evaluate the effectiveness of using computerized Medicare claims data for identifying non-hip fractures in an elderly population.
• To determine the accuracy and efficiency of claims data compared to medical record review for fracture ascertainment.

Main Methods:

• Utilized computerized Medicare inpatient, emergency room, hospital outpatient, and physician claims data from 1987 for an elderly Tennessee Medicaid population.
• Identified 3086 potential fractures and conducted a detailed medical record review for a sample of 1440 cases.
• Developed a definition of probable fractures by excluding claims unlikely to represent new diagnoses.

Main Results:

• The developed definition for probable fractures achieved a positive predictive value of 94% for all fractures, with site-specific values ranging from 79% (tibia/fibula) to 98% (hip).
• Approximately 91% of fractures in the reviewed sample were identified as probable fractures, demonstrating high sensitivity.
• Upper bounds for sensitivity varied by fracture site, from 75% (femoral shaft) to 100% (patella).

Conclusions:

• Computerized Medicare claims data provide a rapid and economical method for identifying fractures in individuals aged 65 and older.
• This approach can significantly aid epidemiological studies and public health surveillance of fractures in the elderly.
• Further research is recommended to refine sensitivity estimates for specific fracture types.