Opening the Black Box: How Do Physicians Communicate about Advance Directives?

Professional societies, citizens' groups, legislators, and the courts all advocate the use of advance directives to ensure that physicians respect the wishes of patients with regard to treatment at the end of life [1-5]. Physicians are encouraged to discuss these issues in the outpatient setting while patients are healthy and competent. Despite enthusiasm for the use of advance directives, research raises doubts about the ability of advance directives to influence care [6-12]. To understand why these documents do not work as intended, investigators have examined many facets of the advance directive process, such as the form of the document used, the durability of directives, and the ways in which directives are interpreted [6, 8, 13, 14]. However, the quality of communication between physicians and patients that leads to the creation of a written directive remains largely an unexplored black box. Studies of advance directive discussions suggest that physicians do not provide patients with adequate information [15, 16]. However, these studies used role playing or standardized patients rather than actual patientphysician encounters or lacked generalizability, typically studying housestaff conversations with inpatients about do-not-resuscitate orders [16-20]. We sought to learn how attending physicians in outpatient practices discuss advance directives with their patients and to observe how closely these discussions adhere to the informed consent model described in the literature. Methods Participants All primary care internists at five practice sites in Durham, North Carolina, and Pittsburgh, Pennsylvania, were eligible for the study; only the study investigators were excluded. The sites were two university-based general medicine practices, two Veterans Affairs general medicine practices, and one university-based geriatrics practice. Patients were eligible for the study if they were at least 65 years of age or had a serious medical illness (including cancer; previous cardiac arrest; HIV infection; renal insufficiency [creatinine concentration >3 mg/dL (250 mol/L) or long-term dialysis]; and chronic obstructive pulmonary disease, congestive heart failure, or cirrhosis severe enough to cause two hospitalizations in the past year) that made discussion of advance directives relevant. They had to speak English, had to be judged competent by their physician to make medical decisions, and had to have not previously discussed advance directives with their physician. We studied one patient per physician. After physicians indicated a convenient clinic session, we randomly selected an eligible patient with whom the physician would discuss advance directives in whatever way you think is appropriate for this patient. If physicians felt that such a discussion was inappropriate for that patient, they were asked, in order, about the next eligible patient until a patient was chosen. If no eligible patients were available, we selected another clinic session and repeated the process. We called eligible patients before their visit and requested their consent to participate in a study to learn how doctors communicate with their patients when making decisions about future medical treatments. Data Collection We audiotaped the selected encounters from 1 April 1994 and 30 October 1994. Physicians completed a self-administered survey that asked about their background and attitudes toward advance directives. We obtained demographic and attitudinal data from patients in face-to-face interviews. The protocol was approved by institutional review boards at the Duke University, Durham Veterans Affairs, and University of Pittsburgh medical centers. Data Analysis Code Book Development All audiotapes were transcribed and coded. We developed the code book through an iterative process [21]. First, we created general coding categories for all pertinent topics identified in the literature [4, 18, 22-25]. Although no clear consensus exists about the ideal content of these discussions, experts generally recommend a model that reflects the established standards of informed consent [26, 27]. After explaining the rationale for advance directives, physicians should describe the nature of potential procedures and the risks, benefits, and likely outcomes of and alternatives to these treatments. They should learn whether patients desire treatments under specific scenarios [23, 28]. Others also advise physicians to attend to patients' values and to identify surrogate decision makers [16, 22, 29]. All agree that these discussions should be conducted in an empathic manner, with sensitivity to patients' emotional needs [16-18, 25]. Because this model is not empirically derived, it is only a best approximation for judging discussion quality; it is not a gold standard. On the basis of the model, we coded the ways in which physicians accomplished the following tasks: introducing the topic, describing advance directives, giving rationales for advance directives, describing pertinent forms, discussing scenarios and treatments, providing medical information, eliciting patient values, and identifying surrogate decision makers. We also developed codes to evaluate the quality of the overall communication process. For each general content area, we formulated a list of specific codes (Table 1). Table 1. Major Coding Categories for Discussions about Advance Directives and Samples of Specific Codes Next, the research team read the study transcripts to ensure that the coding system captured all relevant issues. Coding categories were adjusted as necessary. After developing a draft of the code book, the research team coded a subset of transcripts to identify problems and formulate coding rules. The final code book contained 70 items spanning 11 topic areas (Appendix). A copy of the complete coding instrument is available from the authors on request. Reliability of the Code Book Two raters independently coded all transcripts. The raters applied as many codes as necessary to the content of each speaker's turn, which was defined as a segment of uninterrupted speech. scores, used to measure interrater reliability, were greater than 0.40 for 58 of the 70 codes, indicating moderate or better agreement. No codes with scores less than 0.40 were used, with one exception. The code for dire scenario had a score of 0.39, was present in 91% of cases, and represented a central concept that emerged repeatedly in qualitative analysis. Disagreements in coding were resolved by consensus between the two coders. Disagreements remained in 11% of cases and were resolved by the entire research team. The following statement illustrates a case in which the team initially disagreed about whether the scenario was truly dire but, after discussion, decided to apply the code. OK, what if you became ill, for example with something like a stroke. You were in the hospital and you were in a coma, all right. There was very little chance of you ever coming out of that coma. In other words, you could breathe and your heart was still beating, but you couldn't really think about things. Statistical Analysis We entered the coded transcripts into the Unix Text Analyzer database (University of Pittsburgh Center for Medical Ethics), which facilitates indexing and retrieval and is capable of quantitative assessments of qualitative text (such as frequency counts on codes) [30]. We calculated frequencies for each of the codes and used the chi-square and t-tests to identify variables associated with presence or absence of specific codes. We used standard qualitative methods to analyze communication approaches and relevant sections of coded text [21]. Results Participants Sixty physicians in five practice sites were eligible, and 56 agreed to participate (93%) (Table 2). The median age of the physicians was 37 years (range, 28 to 63 years); 56% were men, and 93% were white. They spent a median of 18 hours per week in direct patient care (range, 2 to 50 hours per week) and had practiced medicine for a median of 10 years (range, 3 to 38 years). Ninety-five percent of physicians stated that they felt comfortable talking to patients about advance directives, but 61% rarely discussed advance directives in the outpatient setting. Table 2. Characteristics of Physicians and Patients Eighty-nine patients were eligible for the study. In 8 cases, physicians refused the discussion because they felt that the patient was emotionally unstable or had had too few previous visits. Fifty-six of the 81 recruited patients (69%) agreed to participate. Their median age was 72 years (range, 58 to 88 years); 68% were men, 84% were white, and 66% had less than a 12th-grade education. Twenty-five percent of patients rated their overall health as very good or excellent, 27% rated it as good, and 48% rated it as fair or poor. The patients had known their physicians for a median of 1.2 years (range, 1 month to 12 years). The median chance that patients would survive for 5 years, according to their physicians, was 72% (range, 9% to 97%). Participants and persons who refused did not differ significantly in any measured variables, except that more participants were Catholic. Communication Process The median advance directive discussion lasted 5.6 minutes (range, 0.9 to 15.0 minutes). Physicians spoke for a median of 3.9 minutes (range, 0.6 to 10.9 minutes), and patients spoke for the remaining 1.7 minutes (range, 0.3 to 9.6 minutes). When introducing the topic, 93% of physicians stated why they were discussing advance directives, although 20% attributed the discussion only to a research project. Usually, the conversation ended without any specific follow-up plan. Forty-three percent of physicians mentioned the possibility of future conversations, 55% discussed advance directive forms, and 25% asked patients whether they had questions. Only 16% of physicians told patients that they could change their mind. In general, physicians were unlikely to attend to t