Wang, Tiansheng

Abstract There is an urgent need to improve dementia ascertainment robustness in real-world studies assessing drug effects on dementia risk. We developed algorithms to dementia identification algorithms using Medicare claims (inpatient/outpatient/prescription) from 3318 Visit 5 (2011-2013) and 1828 Visit 6 (2016-2017) participants of the Atherosclerosis Risk in Communities (ARIC) Study, validated against ARIC's rigorous syndromic dementia classification. Algorithm performance was compared to existing algorithms (Jain, Bynum, Lee). We further evaluated algorithms effectiveness in a 20% random Medicare sample aged ≥70 years who initiating liraglutide or dipeptidyl peptidase 4 inhibitors (DPP4i) to assess 3-year adjusted risk difference (aRD) for dementia. Our incident dementia algorithm required two dementia diagnostic codes within 1-year, or one dementia code plus a new dementia prescription within 90 days. It achieved a positive predictive value (PPV) of 69.2%, specificity of 99.0%, and sensitivity of 34.6% (population prevalence: 8.8%), comparable to extant algorithms (PPV, 58.7–68.6%; sensitivity 25.5–40.4%). Prevalent dementia algorithm (without requiring incident diagnoses/prescriptions) demonstrated similar performance. In the Medicare sample, dementia risk ranged from 3.0% to 12.5%, aRD comparing liraglutide to DPP4i varied −1.2% to −3.6%, with our algorithm closely matching the Bynum algorithm. Algorithm selection significantly impacts treatment effect estimates, highlighting its importance in in pharmacoepidemiologic research.

AbstractBackgroundPharmacoepidemiologic studies assessing drug effectiveness for Alzheimer’s disease and related dementias (ADRD) are increasingly popular given the critical need for effective therapies for ADRD. To meet the urgent need for robust dementia ascertainment from real‐world data, we aimed to develop a novel algorithm for identifying incident and prevalent dementia in claims.MethodWe developed algorithm candidates by different timing/frequency of dementia diagnosis/treatment to identify dementia from inpatient/outpatient/prescription claims for 6,515 and 3,997 participants from Visits 5 (2011‐2013; mean age 75.8 (SD 5.3) years) and 6 (2016‐2017; mean age 79.8 (SD 4.8) years) of the ARIC cohort with ≥ 1‐year continuous Medicare Part A & B enrollment prior to each visit. We evaluated algorithm performance (sensitivity/specificity/positive predictive value (PPV)/negative predictive value (NPV)) against the gold standard of the ARIC dementia classification based on an extensive battery of neurocognitive tests administered at each visit. We compared performance of the new algorithm with that of three existing algorithms (Jain, Bynum, Lee) and their relaxed forms (4‐year window, dementia diagnosis plus a depression diagnosis or dementia evaluation procedure, ignoring code position) to identify incident (n = 90 at V6) and prevalent dementia (n = 190 at V5; n = 160 at V6).ResultThe newly developed primary algorithm requires presence of a dementia diagnosis code followed by another diagnosis code within 1 year (under condition that the following non‐specific ICD‐9 diagnostic codes, 290.40, 290.41, 290.11, 290.3, should appear no more than once), or a subsequent dementia treatment prescription within 90 days (requiring Medicare Part D enrollment ≥ 6 months). The PPV of the new algorithm in the identification of incident dementia was PPV of 72.7%, with a specificity of 99.1%, outperforming existing algorithms. Its sensitivity of 38.7% and NPV of 96.2% were comparable to those achieved by existing algorithms. Sensitivity analysis restricting to ICD‐10 diagnostic codes increased the PPV and attenuated its sensitivity by 6%. Similar comparative performance was observed in the identification of prevalent dementia.ConclusionA novel algorithm for identifying dementia from Medicare claims, developed using linked ARIC‐Medicare data, outperforms previous algorithms and stands as a valuable tool in the assessment of ADRD drugs from real‐world data.