This economic evaluation was based on a Markov model with three health states: healthy, dead, and post hip fracture. A diagram of the model was given. The analysis was subcategorised by gender and two age groups (65 to 79 years and 80 or older). The time horizon was from the study start until the age of 100 years or death. The authors stated that the societal perspective was adopted.
The effectiveness of the programme versus no intervention was derived from a quasi-experimental time series analysis, with six control areas, which was carried out by the authors of this paper (Johansson. 2008, see 'Other Publications of Related Interest' below for bibliographic details). The analysis used data from 1990 to 1995 for pre-intervention and 1996 to 2001 for post-intervention. The hip fracture rates were obtained from the national Swedish Discharge Register. The key clinical input was the number of hip fractures avoided with the programme.
Monetary benefit and utility valuations:
The reduction in utility estimate as result of hip fracture was derived from a Swedish published study. Population-based, age-specific utility values were taken from another Swedish general-population study and the utility decrements, due to fracture, were applied. The instruments used to derive these utilities were not reported.
Measure of benefit:
Quality-adjusted life-years (QALYs) and hip fractures avoided were the benefit measures. A 3% annual discount rate was applied to the QALYs.
The cost categories included programme costs and costs related to hip fracture. Programme costs included wages and running costs. The programme resource use data were collected prospectively using document analysis or self reports. The medical, pharmaceutical, community, and informal care resource use data for hip fracture were estimated from Swedish databases. All costs were expressed in Swedish kronor (SEK). The price year was 2004 and all costs were discounted by 3%.
Analysis of uncertainty:
The model parameters of costs, effectiveness, and the discount rate were altered in univariate, multivariate and threshold analyses. The overall model uncertainty was investigated in a bootstrap analysis and the 95% confidence intervals were calculated by the percentile method. The results were reported in a table and bootstrap analysis plots.