Reporting delay and corrected incidence of multiple sclerosis.

The incidence of multiple sclerosis can be adjusted for reporting delay in a way similar to that for AIDS incidence as described in this paper. Two models are applied to data from the Danish Multiple Sclerosis Registry. The first simply describes the delay distribution, the second is a regression model. For the latter it is shown how goodness-of-fit assessment for both time-dependent and time-independent covariates can be handled using ideas from the Cox proportional hazards model. The goodness-of-fit analysis shows that people with onset age below 40 years and people with onset age from 40 years and above should be analysed separately and this is done throughout the paper. The incidences are adjusted with the two estimated delays and an estimate of the variance for the adjusted incidence is given. Finally the crude incidences are compared with the adjusted ones. We find that the incidence has been increasing since 1965 but mostly for people with onset age below 40 years. Copyright 1999 John Wiley & Sons, Ltd.