This study collects the views of 38 health policy leaders, answering one open-ended question in a 1-hour interview: What state-level reforms do you believe are necessary to implement a feasible model of universal health insurance in Oregon? Interviewees represented seven groups: state officials, insurers, purchasers, hospitals, physicians, public interest, and experts. About 370 coded arguments in the interview transcripts were condensed into 95 categorical topics. A code outline was constructed to present a dialogue among stakeholders in one comprehensive narrative. Topical sections include the cost imperative, politics, model systems, insurance, purchasing, delivery system, practice management, and finance. Summary results show the prevalence of group attention to each topic, group affinities, and proximity correlations of different arguments mentioned by individuals. The most common arguments related to problems of low-value care and delivery system reform. There was a generally felt imperative to control costs. Regarding universal health insurance, stakeholders were split between two main alternatives. One model, favored mostly by insurer and purchaser groups, supported the state-sponsored individual mandate. This plan, embodied in the current Oregon Action Plan to implement universal health insurance, involved managed competition for insurers and clinical governance over professional practice. A separate set of arguments, favored mostly by expert and physician groups, emphasized the need for a unified public system, or utility model, possibly with centralized funds and regional global budgets. The ability of the individual mandate plan to control costs or manage quality appears doubtful, which strengthens opposition. The utility model is more likely to work at cost control and governance, but it disrupts the status quo and its details are vague, which strengthens opposition. Neither model is endorsed by a majority of the stakeholders, and political success for either one alone is not promising. Possibly, a close analysis of the two models could find a way to combine them and generate unified support.