Estimation of unascertained diabetes prevalence: different effects on calculation of complication rates and resource utilization.

The incidence and prevalence of insulin-dependent (Type 1) diabetes mellitus (IDDM) in populations are both well defined. In the more prevalent non-insulin-dependent (Type 2) diabetes mellitus (NIDDM), which is responsible for the bulk of diabetes-related morbidity, true prevalence is uncertain because of delayed diagnosis and problems of definition, particularly with increasing age. Estimates therefore vary widely. We have previously presented evidence of increased relative probability of hospital admission for people with diabetes. These absolute and relative rates of admission were based on a large scale community-derived prevalence for diabetes of 1.36%. Assuming that the true prevalence of diabetes is higher, recalculation of activity data in a sensitivity analysis suggests a theoretical maximum prevalence of diabetes of 5% in our population, since a higher value would imply less morbidity associated with diabetes than 'non-diabetes'. This approach identifies the possible range of unascertained diabetes in a population and defines it in functional terms as that state carrying any excess risk of admission for complications when compared to non-diabetes. Higher estimates of prevalence have little impact on the calculation of overall resource use for diabetes, since the great majority of costs are related to fixed hospital activity for people with identified diabetes. The unascertained diabetes sub-group will cost little by comparison. Paradoxically, the tendency to use higher estimates of unascertained diabetes increases the denominator for calculation of complication rates and reduces both the absolute and relative risk of complications. This dilutes the epidemiological significance of diabetes in the aetiology of its related complications.