Although several investigators have described the physician [6, 23, 24] and patient [24–26] communication behaviors that are desirable elements of patient-centered communication, the literature lacks a generally-accepted model for how both sets of behaviors might unfold and influence the course and outcome of a clinical encounter. An overarching model of patient-physician communication might prompt researchers to examine the degree to which interventions directed toward one party (i.e., either physicians or patients) influence the communication behaviors of the interaction partner. Moreover, a framework that describes the sequence of communication behaviors could help guide interventions directed toward specific segments of the interaction. The Four Habits Model has been particularly useful in describing the appropriate physician communication behaviors during the course of a clinical interaction and serving as a research framework for interventions promoting patient-centered communication skills among physicians . We believed that this Model could serve as a useful starting point for developing a framework that includes both physician and patient communication behaviors. As a first step in this direction, we conducted a consensus process involving a group of communication experts to identify an initial set of patient communication behaviors that could affect the course, direction and outcomes of the medical encounter.
Seventeen international experts in communication research, medical education and health care delivery participated in the consensus process. We found the Delphi method useful in soliciting input and building consensus within this group. The experts were quite engaged throughout the entire process, as evidenced by the numerous suggestions (25 modifications, 20 additions) they made in round 1 and the greater than 90% response rate to all three rounds of ratings. Most of their suggestions pertained to Habit 2 ("Provide your perspective") with Habit 3 ("Communicate your feelings and concerns") receiving the least number of suggestions.
After three rounds of ratings, the experts agreed to retain 22 behaviors which were distributed throughout the Model. Interestingly, 6 behaviors that the experts agreed to retain were among the 14 originally-proposed by the project team. The final list addresses key aspects of patient communication, such as sharing and prioritizing concerns, expressing feelings, and summarizing information and recommendations provided by the physician. While most behaviors can be assessed using existing coding schemes, the list also includes a few behaviors that may require methods that examine the content of statements made by the patient. Examples of such behaviors include "describe how the illness affects (one's) daily life" and "share your story".
Of the behaviors that the experts agreed to drop or which ultimately did not achieve consensus (in either direction), most were similar to other behaviors that were retained. For example, "give opinions about possible causes of (my) illness," a suggestion that received 0 votes, is comparable to "state own understanding of the problem," a retained behavior. Similarly, "respond to the clinician's welcome," which did not achieve consensus is comparable to the retained behavior, "greet or verbally welcome the physician." Several experts also noted that certain behaviors were not practical for the typical patient (e.g., "solicit the physician's goals for the visit"), which may explain why they were dropped from the Model.
The experts were unable to reach consensus on whether or not to retain "participate in decision-making" or any of its suggested modifications. Interestingly, the literature lacks a detailed definition of what constitutes patient participation or involvement in decision making [27, 28]. In a recent study of physician-patient communication related to breast cancer decisions, Brown and colleagues  assessed whether the patient: presented her agenda; declared her preferences (for information and involvement); declared her perspectives (costs and benefits of treatment); and portrayed herself in an active role. To a certain extent, these elements were retained, even though they are not labeled as participation in decision-making. For example, Habit 2 includes: "express preferences...". Similarly, Habit 4 includes: "use own words to summarize information...," "give frank opinions about the choices offered," "ask questions about explanations and choices," "describe any barriers to following recommendations," and "ask about options interested in but not mentioned by the health care provider".
The following limitations should be considered when reviewing our findings. First, our project focused on verbal communication behaviors only because we believe that these behaviors might serve as a logical starting point for developing educational initiatives for patients. Second, our results should not be interpreted as representing all the views of experts in the field of communication, particularly patient communication. Our panel included those with expertise in developing interventions to enhance physician communication as well as those who focus specifically on improving patient communication skills. We chose experts with diverse interests to acknowledge the interactive nature of physician-patient communication (i.e., to recognize that the physician's communication behaviors influences the patients' communication behaviors).
Third, it is important to note that the Four Habits Model describes a set of basic physician communication behaviors for the clinical encounter. Studies of the medical interview which were conducted in primary care with adult patients, other conceptual models of communication and consensus statements (i.e., Kalamazoo consensus statement) informed the development of the Four Habits Model [9, 30]. Thus, the original model and patient behaviors suggested through our consensus process may not apply to certain situations, such as communication between physicians and caregivers, pediatric patients, or psychiatric patients. Likewise, the Four Habits Model and our consensus behaviors do not reflect the potential differences in communication styles due to the gender or ethnic background of the physician and/or patient. Additional research is necessary to examine (and adapt) the Four Habits Model and the consensus behaviors to these other circumstances.
Finally, our consensus process involved professionals, and we chose this approach for several reasons. Because of our interest in identifying measureable communication behaviors for an overarching conceptual framework, we believed that experts who were familiar with the communication research and education literature would be helpful in generating an initial list in an efficient manner. We also believed that developing the list of patient behaviors was a necessary step for guiding our subsequent work with patients. All 17 experts who participated in the consensus process have years of experience working to enhance the patient's experience of health care, and it is noteworthy that 7 members of the panel were behavioral scientists and not physicians. The fact that the group eliminated behaviors that are not practical for the average patient is a demonstration of their commitment to the patient's perspective.
We recognize that starting with expert opinion could potentially bias the subsequent modification of the communication skills identified through this process and is an important limitation of our study. We also acknowledge that our approach may appear to "privilege" the experts' voice in determining what communication skills and behaviors patients might find useful in communicating with their physicians. By the same token, however, one could argue that a similar bias would exist had we conducted a Delphi process with patients only, not to mention the difficulty that would be entailed in selecting different "types" of patients to participate in such a process. We firmly believe that patients should be involved in the validation process, and this process should involve patient populations: with different demographic characteristics (i.e., gender, race/ethnicity); who are vulnerable (e.g., limited health literacy, older adults, lower socioeconomic status); or with varying degrees of medical co-morbidity (i.e., relatively healthy versus relatively complex). Potential approaches could include focus groups or semi-structured interviews with patients, analogous to previous studies which examined patients' perspectives regarding informed decision making [31–33]. In addition, the patient behaviors should be validated using audiotapes of actual encounters, an approach that was used to validate the physician behaviors within the Four Habits Model .