Our study provides data from a small sample of physicians about the complex relationship between patient requests and these physicians’ subsequent behavior. Using data from 6 focus groups with 22 participants, we have described some of the affective and cognitive mechanisms and some of the attitudinal and contextual factors which influenced participants’ clinical judgment and decision to prescribe. We found that some were initially unaware of the degree to which their reasoning was swayed by patient requests until prompted in a setting that facilitated reflection.
Other studies that trained patients to ask questions and be assertive have also noted some patient–physician friction as a result of the intervention,20–22
and that the benefits of assertive patient behavior (e.g., greater patient involvement in decision making) are observed only when physicians adopt a “patient-centered” attitude.23
Consistent with these studies, we found that the most commonly observed affective response to patient requests was annoyance because of the disruption of usual routines or encroachment on physician authority. DTCA-driven patient self-diagnosis was considered especially noxious because participants perceived encroachment by patients and also by pharmaceutical companies who were able to sway patient self-assessments. Reactions to that annoyance determined whether the request led to further inquiry to establish a more conclusive diagnosis or to acquiesce to the patient’s request to move on to the next patient.
Participants commonly used 3 cognitive strategies to justify complying with patient requests that they viewed as inappropriate. First, some prescribed to establish rapport. There may, however, be more effective ways of developing rapport without the dangers of inappropriate prescribing.24,25
Relationship-centered methods of developing rapport include eliciting patients’ ideas and expectations, using empathy, presenting choices, and supporting patient autonomy in decision making. Second, participants assumed that complying with the request would save time, avoid conflict, or achieve an earlier closure of the visit—assumptions which were not supported by a retrospective analysis of data from the SIP study. The data showed that visits in which requests were granted were no shorter than those in which requests were denied. The third involved the misapplication of a “representativeness heuristic”—drawing generalizations from limited prior experience.26
Participants expected to encounter patient resistance to the use of antidepressants. When there was an explicit request for antidepressants and a lack of expected resistance, it was inappropriately assumed that the patient must need the medication.
Existing theories and models of decision making can help elucidate our findings. Sound clinical judgment depends on both affective and cognitive processes, which are inextricably linked in the formation of memory, heuristics, and reason.27–29
These have been conceptualized using different terminologies; however, theories converge over the use of 2 primary modes of affective and cognitive decision processes that are consistent with our observations. The first mode is script-driven, using intuitive heuristics in a fast and effortless way; the second mode is deliberative, using slower, rational processes in a more careful and effortful way.30,31
Script-driven processes consist of elaborated compilations of knowledge and experience characteristic of expert practitioners.32–34
Deliberative processes, in contrast, might use pretest probabilities and decision thresholds.35
In our study, there was little evidence of use of these formal quantitative probability approaches. Our observations suggest that when choosing to “err on the side of overtreating”, focus group participants implicitly considered disease likelihoods and side effect risks, although they did not explicitly estimate probabilities or calculate thresholds. Employing only script-driven processes can lead to errors involving misapplication of heuristics, whereas deliberative processes are too slow for routine use and sometimes ignore the “gist information” often used by experts.36
We observed that patient requests may sometimes disrupt the balanced application of these 2 modes. Treatment decisions are further influenced by 3 domains: patient characteristics and values, experience and knowledge, and external clinical evidence.37
Whereas in acute conditions, clinical evidence and physician knowledge have the most influence, in chronic conditions, such as depression, patient characteristics and attitudes are much more important in the decision process, which is consistent with our focus group findings.
The lack of conscious awareness of the influence of DTCA-driven patient requests may increase vulnerability to pharmaceutical companies’ efforts to alter physician behavior without decreasing perceived autonomy. As we have noted, 1 characteristic of expertise is having the ability to make rapid intuitive script-driven decisions based on limited information.36
When unexamined, this process appears to be problematic in the face of unclear or ambiguous requests for medications.
During the focus groups, most participants were able to identify extrinsic influences on their thought processes and describe actions they might take to improve their clinical judgment. They were often able to articulate the strategies they were using by hypothetically considering differences between their own clinical behavior and what they would suggest if they were in supervisory roles. “Metacognition” is a term used to describe the ability to observe one’s own thinking as if an external observer were present.38
In principle, if patients’ requests trigger cognitive scripts that tend to favor inappropriate prescribing, physicians who develop a greater metacognitive self-awareness of these influences can promote more appropriate responses to these requests. The focus group discussions provided hope that this level of self-awareness is possible. It remains to be seen to what degree physicians, in general, can achieve sufficient self-awareness in real clinical settings and recalibrate their impressions to improve quality of care.39
Synthesis of Themes and Proposed Framework
We present a framework for understanding influences on physician decision-making in response to patients’ requests (Fig. ). It represents a hypothesis that requires further empirical study, but is consistent with the data from the present analysis. Direct responses to requests are made through script-driven processes, conditioned by affective factors such as annoyance or empathy, and through more deliberative, cognitive processes such as choosing to “err on the side of overtreating.” Patient characteristics, such as attitude and perceived severity of illness, have a greater than usual influence on physicians’ prescribing decisions for chronic conditions like depression. Assumptions based on prior experience with inaccurate self-diagnoses may make it more likely that a patient’s self-diagnosis is rejected. Conversely, assumptions based on prior experience of encountering resistance to taking antidepressants may make it more likely that a prescription is made. In addition, contextual factors (such as time pressure) and metacognitive self-awareness influence whether a physician would use a script-driven approach or a deliberative approach to clinical reasoning. All these factors played a role in whether the physician explored the patient’s concern further or a made a premature move toward closure.
Influence of patients’ requests on physicians’ prescribing.
The physicians in our study may not have been representative of the larger primary care physician population for several reasons. Participants were a subsample of those who participated in the SIP study who constituted only a subsample of the primary care physician population. SIP study participants may have had greater than average confidence in their clinical skills and might have been more willing to reveal aspects of their thought processes. Physicians from other cities, practice settings, and health systems with different patient populations might have responded differently. Despite these limitations, we used SIP participants because they all had a common experience of an SP antidepressant medication request.
Two of the focus groups were conducted by telephone, so that we were unable to see the nonverbal aspects of communication related to the discussion of prescribing behaviors. In spite of the different methods of focus group moderation, our findings across the 6 focus groups were consistent. The most significant limitation of our study is our 22% recruitment rate for from the SIP study; however, because this is one of the first studies of its kind to examine reflection on decision-making behaviors related to prescribing, we believe the data it generated is of interest. Larger, more representative samples would have to be drawn and analyzed to reach more generalizable conclusions about physician prescribing behaviors.
Several other limitations are noteworthy. The extent to which discussions reflected actual cognitive processes is uncertain, and more than a year had passed since the most recent SIP visit. Most participants did, however, view verbatim video reenactments and felt adequately familiarized with the study protocol. Some participants may have felt uncomfortable disclosing responses among colleagues, but a review of transcripts suggests that all participants were actively engaged.