An Examination of Preference-based Reinforcement Learning for Treatment Recommendation