The technical capacity to monitor patients with a mobile device has drastically expanded, but data produced from this approach are often difficult to interpret. We present a solution to produce a meaningful representation of patient status from large, complex data streams, leveraging both a data-driven approach, and use clinical knowledge to validate results. Data were collected from a clinical trial enrolling chronic pain patients, and included questionnaires, voice recordings, actigraphy, and standard health assessments. The data were reduced using a clustering analysis. In an initial exploratory analysis with only questionnaire data, we found up to 3 stable cluster solutions that grouped symptoms on a positive to negative spectrum. Objective features (actigraphy, speech) expanded the cluster solution granularity. Using a 5 state solution with questionnaire and actigraphy data, we found significant correlations between cluster properties and assessments of disability and quality- of-life. The correlation coefficient values showed an ordinal distinction, confirming the cluster ranking on a negative to positive spectrum. This suggests we captured novel, distinct Pain Patient States with this approach, even when multiple clusters were equated on pain magnitude. Relative to using complex time courses of many variables, Pain Patient States holds promise as an interpretable, useful, and actionable metric for a clinician or caregiver to simplify and provide timely delivery of care.