Methodology for identifying patients at high risk for osteoporotic fracture

Osteoporotic fractures are associated with significant morbidity, mortality, and health care costs.

The purpose of this paper is to present and validate a mathematical model that managed care organizations can apply to administrative claims data to help locate members at risk for osteoporotic fracture and estimate future fracture rates.

Using known risk factors from previous clinical studies, 92,000 members of a large Midwest health plan were placed in 1 of 4 risk categories based on historical claims markers: demographic/lifestyle (age, sex, smoking, alcoholism); steroid use; medical history (previous osteoporotic fracture, ordinary bone fracture, osteoporosis diagnosis, bone mineral density test); or steroid use with medical history. Logistic regression was used to assign a probability of fracture for the 4 groups over the next 2 years. These predictions were compared with actual fracture rates, and refined models were produced. The models were then validated by applying them to current data and comparing the predicted fracture rate for each group to known results.

The model predicted that 1.26% of the study members would experience osteoporotic fracture over the next 2 years; the actual result was 1.27%. Within the 4 risk groups, the predicted fracture rates were lower than the actual rates for the demographic risk group (0.87% predicted vs 0.97% actual) and higher than the actual rates for the steroid use (1.78% predicted vs 1.58% actual), medical history (5.90% predicted vs 4.94% actual), and the steroid use with medical history groups (7.80% predicted vs 6.42% actual).

The application of this risk model to an administrative claims database successfully identified plan members at risk for osteoporotic fracture.