Ascertainment bias in dementias: a secondary to tertiary centre analysis in Central Italy and conceptual review

Background and aims Ascertainment bias (AB) indicates a bias of an evaluation centre in estimating the prevalence/incidence of a disease due to the specific expertise of the centre. The aim of our study was to evaluate classification of different types of dementia in new cases appearing in secondary and tertiary centres, in order to evidence possible occurrence of AB in the various (secondary to tertiary) dementia centres.Methods To assess the mechanism of AB, the rates of new cases of the different forms of dementia reported by different centres were compared. The centres involved in the study were 11 hospital-based centres including a tertiary centre, located in the University Department of Clinical Neurology. The tertiary centre is endowed with state-of-the-art diagnostic facilities and its scientific production is prominently focused on dementia with Lewy bodies (DLB) thus suggesting the possible occurrence of a bias. Four main categories of dementia were identified: Alzheimer's disease (AD), DLB, fronto-temporal dementia (FTD), vascular dementia (VaD), with other forms in a category apart. The classification rate of new cases of dementia in the tertiary centre was compared with rates reported by secondary centres and rates of recoding were calculated during a follow-up of 2 years.Results The study classified 2,042 newly diagnosed cases of dementia in a population of 1,370,000 inhabitants of which 315,000 were older than 65. AD was categorized in 48-52 % of cases, DLB in 25-28 %, FTD in 2-4 % and VaD in 17-28 %. During the 2-year follow-up the diagnosis was re-classified in 40 patients (3 %). The rate of recoding was 5 % in the tertiary centre, 2-8 % in referrals from secondary to tertiary centre, 2-10 % in recodings performed in secondary centres and addressed to tertiary centre. Recoding or percentages of new cases of AD or DLB were not different in the comparison between secondary or between secondary and tertiary centres. FTD and VaD were instead significantly recoded.Conclusion The results of the study suggest that in a homogeneous area, AB is not interfering with diagnosis of AD or DLB.