Analytical approach:
This economic evaluation was performed using a decision analytic model, the Diabetes Cost-Benefit Model, which projected the incidence and progression of diabetes and its complications in the whole Australian population from 2000 to 2010. The time horizon of the analysis was 10 years. The perspective of the study was not explicitly stated.
Effectiveness data:
The approach used to identify clinical estimates from among the available literature was not described. However, most of the primary sources were described. The epidemiological data came from Australian sources. For example, population projections came from the Australian Bureau of Statistics, while prevalence data were derived from the national Health Survey. Data on the effect of screening in detecting undiagnosed patients and in reducing diabetes cases were derived from two trials conducted in Finland and the USA. The reduction in death and complications due to early detection of diabetes was taken from the UK Prospective Diabetes Study. The key clinical end point was the reduction in diabetes prevalence in previously undiagnosed patients.
Monetary benefit and utility valuations:
The sources of the disability values were not reported.
Measure of benefit:
The summary benefit measures used were disability-adjusted life-years (DALYs).
Cost data:
The analysis included the costs of screening, lifestyle programme, visits to doctors (general practitioners and specialists), visits to other health care professionals (educator, dietician, podiatrist), in-hospital services, prescription medications and diabetes-related supplies (i.e. those related to self-monitoring of blood glucose levels). The sources of the costs and resource use data were unclear. The authors made assumptions about the costs of lifestyle programme costs. The costs were in Australian dollars (AUD). The price year was 2000 and the costs were discounted at an annual rate of 3%.
Analysis of uncertainty:
A deterministic sensitivity analysis was carried out to consider nine alternative scenarios where key model inputs were varied. An exhaustive list of these items was provided. The sources of the alternative values were not stated clearly.