Nursing home providers are moving towards a model of care that is person centered in order to improve the quality of care and quality of life for individuals residing in their communities. The State of Ohio has mandated that providers use the Preferences for Everyday Living Inventory (PELI) to assess resident preferences. This paper and pencil assessment adds to an increasing data management barrier to efficiently incorporate preferences into care. We are in the process of developing the Care Preference Assessment of Satisfaction or ComPASS system which supports data collection and reporting in order to better integrate preferences into the everyday care of residents. With this platform we are exploring how machine learning can be used to provide more personalized care in nursing homes by providing insights and recommendations based on resident preferences while lessening the data collection and management burden. In this paper, we describe ComPASS, discuss our initial investigations into using machine learning for long-term care, present initial findings, and suggest future directions for this research.