OBJECTIVES: Newer treatments for Alzheimer’s disease target patient subgroups by severity or biomarkers. This study examines real-world evidence to understand utilization of cognitive and biomarker screening tools around Alzheimer’s diagnosis.

METHODS: Patients with mild cognitive impairment and/or Alzheimer's disease consented to contribute their records to medical research via PicnicHealth, a patient-centric virtual platform. This study included the first 300 patients enrolled in 2022 with a provider-confirmed Alzheimer’s diagnosis who were at least 50 years old at onboarding. Medical records across US facilities (including physician notes, laboratory results, imaging reports) were retrospectively and prospectively collected. Measurements were abstracted from both structured and unstructured record data. Cognitive tests, neuroimaging, and cerebral spinal fluid (CSF) measurements within 12 months of diagnosis were examined. Impairment was considered mild if Mini-Mental State Examination (MMSE)≥20 or Montreal Cognitive Assessment (MoCA)≥18 and moderate/severe otherwise.

RESULTS: The study population was 57% female, 85% White and 8% Black. 5% identified as Hispanic/Latinx. Longitudinal data was available for a median of 5 years. Median age at diagnosis was 71. Cognitive scores were available for 121 (40%) patients; of these 72% exhibited mild impairment and 28% were moderate/severe. 58 patients were screened with MMSE alone, 35 MoCA alone, and 28 with both. Neuroimaging was available in 157 (52%) patients; 118 had at least one MRI, 71 at least one CT, and 22 at least one PET-CT. 22 (7%) had at least one CSF measurement.

CONCLUSIONS: Many Alzheimer’s patients did not have evidence of cognitive or biomarker testing, indicating these screens are limited in clinical practice. Previous studies identified cognitive scores in smaller percentages of patients than the 40% in this sample, suggesting unstructured physician notes may identify more measurements than structured data sources. Further research is needed for stratifying severity in the real-world setting, potentially using patient histories from unstructured clinical notes or patient-reported outcomes.