Title: Leveraging Edge-Aware Interaction Risk Modeling to Detect Financial Exploitation in Alzheimer’s Patients Through Medication Context

Abstract:

As cognitive stability fluctuates, individuals with Alzheimer’s disease face an escalating threat of financial exploitation. Traditional fraud detection mechanisms typically depend exclusively on financial behavior patterns, overlooking critical clinical variables that influence vulnerability. To address this gap, this study introduces a medication-aware framework that integrates transaction-level monitoring with medication adherence data to better identify financially risky events linked to cognitive decline.

The research utilized a hybrid simulation dataset comprising 180 patients over a 45-day period, generating a total of 8,100 medication records and 30,855 financial transactions. The proposed framework assesses five key indicators: amount anomaly, vendor novelty, transaction frequency, time deviation, and medication adherence. These metrics were analyzed using three distinct logistic models: a financial-only baseline, an additive medication-aware model, and an interaction-aware model.

The results indicate that while the financial-only baseline achieved the highest global F1-score of 0.5000, the interaction-aware model demonstrated superior performance in specific contexts. Specifically, it increased recall during windows of medication-induced vulnerability from 0.7442 to 0.9070 and secured the highest average precision for ranking high-risk cases. These findings suggest that medication adherence serves best as a contextual modifier for financial risk assessment rather than functioning as a standalone predictor.

Source: arXiv Generated at: 2026-06-02 00:00:00 UTC