Recently, investigations on extracellular RNA (exRNA) as the new possible biomarker of complex diseases were put under great attention. Plasma is one of the easiest biopsies to obtain among all the liquid samples from which exRNA can be sequenced. Previous studies by Zhong lab have brought up the potential of increasing plasma exRNA PHGDH level as a new biomarker during noninvasive medical screening tests for early identification of Alzheimer’s Disease (AD). This paper aims to evaluate the viability of exRNA PHGDH level in predicting the cognitive status of sporadic AD patients using the Dementia Rating Scale-2 (DRS) score and longitudinal exRNA data sequenced by SILVER-seq (Small Input Liquid Volume Extracellular RNA Sequencing). To this end, we designed a predictive-model-based screening test. We started with formulating the definitions of the increase in exRNA PHGDH level and the decrease in the DRS score for data labeling, considering the variations’ time interval and fluctuation rate. Then, we designed two algorithms, called the simple test and the full test, considering the different depths at which molecular-level data are analyzed. We trained the model with labeled data and two algorithms and evaluated the performance using accuracy, positive predictive value (PPV), and negative predictive value (NPV). Models trained by both algorithms achieved great predictive power with 68.2% accuracy per simple test and 95.0% accuracy per full test. This study proves the feasibility of predicting human cognitive status using plasma exRNA level and opens up new possibilities for building complex predictive models using exRNA AD-associated biomarkers.