As the elderly demographic grows larger, the need for efficient healthcare becomes pressing. One solution is to introduce Artificial Intelligence in the healthcare domain, which implies relevant dataset exploitation. In this paper, we address the issue of imbalance prediction using open-access sensor-based datasets recorded on both young adults and senior citizens. We highlight the need to tailor feature choice to the target age-range: strong signals for one population may be classified as weak for another. Although the obtained results are encouraging, there remains a need for adaptive models developed using AI and under the guidance of medical experts.