IdentIfIcatIon of subjects wIth alzheImer’s dIsease and other dementIa from the permanent benefIcIarIes sample

Abstract

The number of patients with Alzheimer’s disease or other dementia (ADD) in France has been estimated at 856 000 in 2004 and the incidental cases (subjects > 65 years) between 186 000 and 225 000. The use of the medical-administrative databases could be proposed to identify these patients with epidemiological purposes. We used the EGB (Echantillon Generaliste des Beneficiaires), a 1/97 sample of the beneficiaries of the usual French health insurance plan, (concerning 77% of the French population in 2007). Three criteria of identification of the patients > 60 years having an ADD in 2007 were used: the “long-term disease n°15” (ALD15) -meaning 100% reimbursement of health costs for the patient”, the consumption of the specific drugs of ADD (MED) and hospitalisation with an ADD code as the main diagnosis in the French Diagnosis Related system called Programme de Medicalisation du Systeme d’Information (PMSI). 3 802 beneficiaries distributed in 7 separate populations were identified: patients in ALD15 (17.3 %), the MED group (21.4 %), the PMSI group (14.4 %) and the various combinations of these 3 criteria (ALD15+MED (27.1 %), ALD15+PMSI (2.5 %), PMSI+MED (6.8 %) and ALD15+MED+PMSI (9.7 %)). Patients declared in ALD15 were older (85.1 ± 9.1 years) than those of the MED group (81.9 ± 7.4 years) and of the PMSI group (83.4±9.5 years). The distribution of genders was uneven (sexratio F/H: 3.0, 2.0 and 1.5, respectively). The rate of a general practitioner consultation was 57.6 % for the patients ALD15 against 89.8 % and 84.8 % in 2 other groups, respectively. The 2-year death rate ranged from 8.6% (MED) to 34.4% (ALD15+PMSI). The use of the various criteria of identification of the ADD identifies groups of different sociodemographic characteristics and gravity. The institutionalization of the patients or the expectation of declaration in ALD15 could partially explain these dissimilarities. It seems difficult to use administrative data for identifying AD patients with the absence of bias and the precision required for epidemiological studies.