Incidence analysis in ADRS group defined by the first recording of one of the following criteria: (i) LTD registration for ADRS (ICD-10 codes: "F00-F03", "G30", or "G31"), (ii) hospital stay reporting a diagnosis code of ADRS (similar ICD-10 codes) or (iii) reimbursement for at least one acetylcholinesterase inhibitor (rivastigmine, galantamine or donepezil) or memantine.

Incidence analysis in non ADRS group. Each incident ADRS case was paired (1:1) to a beneficiary without any ADRS criteria, matched on age (same birth year), sex, residence area (based on of the 100 administrative 'départements') and insurance scheme.

Incidence analysis between the 2 groups (ADRS/non ADRS). For each pair, an index date was defined as the ADRS identification date. In both groups, a 5-year period free of these three ADRS criteria was required before the index date, to ensure incident ADRS cases.

Given the absence of consensual guidelines describing the frequency of diabetes monitoring among elderly subjects, we defined a conservative minimal threshold as follows: ambulatory biological monitoring :

• ≥ 1 annual HbA1c determination (primary endpoint)
• ≥ 2 annual HbA1c determination
• ≥ 1 annual lipid profile (≥ 1 annual LDL cholesterol, ≥ 1 annual triglyceride)

≥ 1 annual eye examination, defined by a visit to an ophthalmologist or a dilated fundus examination, in or out of the hospitals

≥ 1 annual hospitalization for hypoglycemia