Both prosodic and acoustic speech features have been shown to reflect changes in speakers' cognitive abilities across diverse clinical contexts. Prosodic features (e.g., speech rate) have particularly been found useful for Cancer-Related Cognitive Impairment (CRCI). CRCI refers to cognitive difficulties experienced by cancer survivors, which often go undetected by neuropsychological tests. While extracting prosodic features remains time-consuming, incorporating acoustic features (e.g., MFCCs) in automated tools may be more applicable in healthcare settings. However, little is known about their discrimination power relative to cognitive impairment. This study aims to compare the performance of prosodic features and MFCCs for identifying cancer survivors with CRCI. Nine breast cancer survivors with CRCI and thirteen healthy controls completed a narrative task in a soundproof booth. Five prosodic features and twelve MFCCs were extracted and used to train classifiers with either prosodic features or MFCCs. Results showed that both feature types were effective in discriminating between groups, achieving 86% accuracy with speech-to-silence ratio and speech rate (ROC = .872), and 82% accuracy with 11 out of 12 MFCCs (ROC = .829). While these findings suggest that MFCCs can be used for identifying cancer survivors, the link between MFCCs and CRCI is discussed in the paper.