Defining LoT based on observed EHR profiles introduces a degree of uncertainty into classifications. When seen in the context of inter-rater agreement between clinicians, algorithmic performance is comparable to clinician defined lines, while allowing the processing of LoT classifications at scale.