Outcome-Based Doctor-Patient Interaction Analysis

Interactional analysis (IA) systems have been devised and applied to doctor-patient dialogues to describe encounters and to relate process to outcomes. Prior work in this area has been typified by the use of a single taxonomy for classifying verbal behaviors and limited outcomes (compliance and/or satisfaction). We applied three different IA systems (Bales, Roter's modified Bales with affective ratings, and Stiles' "Verbal Response Modes") to 101 new-patient visits to a general medical clinic for which multiple outcomes had been determined: several measures of patient knowledge of problems at conclusion of visit; patient compliance with drugs (over the ensuing three months); and patient satisfaction with the visit (perceived technical, interpersonal and communication quality). Within IA systems, cross tabulations and multiple regressions were performed to relate encounter events to outcomes. Across IA systems, multiple regression R2 and R2 adjusted (R2a) for the number of independent variables entering were used to characterize strength of relationships. Roter's IA system showed stronger relationships to outcomes of knowledge (41% R2, 27% R2a) and compliance (44% R2, 28% R2a) than did Bales' or Stiles' systems. R2 for patient satisfaction was identical for Bales and Roter (35%), and greater than R2 for Stiles (14%). We conclude that choice of IA system for research or teaching purposes should be based on behaviors and outcomes of particular interest and importance to the user. Based on audioreview of tapes, Roter's approach is less time-consuming and may perform as well as more complex systems requiring transcript analysis.