Analytical approach:
An individual-level state transition model of osteoporosis screening and treatment was used to assess the costs and outcomes of each of the different strategies under study. The model was validated by comparing the model’s predictions about life expectancy and fractures with actual outcomes reported in different USA sources. A lifetime time horizon was used. The authors reported that the perspective was that of the payer.
Effectiveness data:
Clinical and effectiveness data were derived from previously published studies and national sources. The main measures of effectiveness were sensitivity and specificity of QUS and SCORE as prescreening tests. These estimates of effectiveness were derived from two published studies. Baseline DXA T-score values were derived from a national survey (femoral neck) and from a DXA manufacturer (lumber spine).
Monetary benefit and utility valuations:
The authors reported that health state utility values were derived from a nationally representative non-institutionalised sample of elderly women. Disutilities associated with fractures, nursing home residence and adverse events were derived from published studies.
Measure of benefit:
Quality-adjusted life years (QALYs) gained were used as the summary measure. As benefits could be generated over the lifetime of the patient, future benefits were discounted using an annual rate of 3%.
Cost data:
Direct costs included in the analysis were for screening tests, oral bisphosphonate treatment, physician visits, fracture-related treatment, nursing home stays and adverse events. Nursing home rates were derived from a published study. Costs for fracture-related treatment and other medical services were from Medicare diagnosis-related group information, reimbursement rates and published studies. The price year was 2010. All costs were reported in USA dollars ($). Costs were discounted using an annual rate of 3%.
Analysis of uncertainty:
The authors undertook a probabilistic sensitivity analysis by fitting probability distributions alongside all model parameters.