Three types of ambiguity in coding empathic interactions in primary care visits: implications for research and practice.