Clinical versus neuropathological diagnosis in an ADRC sample: Factors impacting diagnostic concordance rates

Abstract

Background Improving clinical diagnostic accuracy and refining criteria is challenging in dementia research. To explore our ADRC’s diagnostic accuracy, a QI project was initiated to identify potential systemic areas affecting our center’s performance. Method Clinical consensus diagnoses were compared to the neuropathological findings of 68 ADRC brain bank participants between 2008‐2017. Concordance level (full, partial, discordant) was assigned by three authors (MS, JN, and HG) with input from our neuropathologist (DP). Fisher's exact test was used to examine concordance and Clinical Dementia Rating (CDR), race, age at death, education, number of visits, time between last visit and death, onset age, and depression. Results Participants were 56% female, 88% non‐Hispanic white, education 14.6+/3.9 (range 2‐22), age of onset 71+/‐11 (range 47‐93), age at death 82.5+/‐10.4 (range 52‐102), CDR sum of boxes 12.8+/‐6.5, CDR Global 2.2+/‐1.03, number of follow‐up visits 2.5+/‐2.8, one visit only 40%, 56% had a visit within 18 months before death ( range 0‐132 months). Depression was present in 28%. Primary clinical diagnosis at death was 82.4% AD, 7.4% normal, 4.4% MCI, 2.9% DLB and 3% other. 88% were fully or partially concordant (clinical diagnosis matched at least one neuropathological diagnosis). Of the 8 discordant cases, 2 had no identifiable neuropathology despite dementia; 2 had tauopathies; 1 was an AD/FTD mismatch. Three who were >92yo had no clinical diagnosis but exhibited AD pathology. For partially concordant cases where the second diagnosis was mismatched, it was most frequently DLB (80%). Concordance rate was not associated with any of the above clinical/demographic variables except for CDR and race (p < 0.05). Regarding CDR, 5/8 (63%) discordant cases had a CDR of 0.5 (mild symptoms); the other three had a CDR of 3 (severe dementia). 3/8 minority cases were discordant but two had diagnoses currently only available at autopsy. Conclusion Consistent with the literature, the most common error was either missing DLB when present or diagnosing it when absent. No systemic issues were identified that contributed to inaccuracies.