Prevalence of unvoiced desires. The top data series shows the percentage with an unvoiced desire for each of 8 types of resources for all patients. The bottom data series shows the percentage for just those patients who desired each of these resources but who did not request it. Entries have been sorted by prevalence.
Bell RA, Kravitz RL, Thom D, Krupat E, Azari R. Unsaid but Not ForgottenPatients' Unvoiced Desires in Office Visits. Arch Intern Med. 2001;161(16):1977-1984. doi:10.1001/archinte.161.16.1977
Copyright 2001 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2001
To examine patient, physician, and health care system characteristics associated with unvoiced desires for action, as well as the consequences of these unspoken requests.
Patients and Methods
Patient surveys were administered before, immediately after, and 2 weeks after outpatient visits in the practices of 45 family practice, internal medicine, and cardiology physicians working in a multispecialty group practice or group model health maintenance organization. Data were collected at the index visit from 909 patients, of whom 97.6% were surveyed 2 weeks after the outpatient visit. Before the visit, patients rated their trust in the physician, health concerns, and health status. After the visit, patients reported on various types of unexpressed desires and rated their visit satisfaction. At follow-up, patients rated their satisfaction, health concerns, and health status, and also described their postvisit health care use. Evaluations of the visit were also obtained from physicians.
Approximately 9% of the patients had 1 or more unvoiced desire(s). Desires for referrals (16.5% of desiring patients) and physical therapy (8.2%) were least likely to be communicated. Patients with unexpressed desires tended to be young, undereducated, and unmarried and were less likely to trust their physician. Patients with unvoiced desires evaluated the physician and visit less positively; these encounters were evaluated by physicians as requiring more effort. Holding an unvoiced desire was associated with less symptom improvement, but did not affect postvisit health care use.
Patients' unvoiced needs affect patients' and physicians' visit evaluations and patients' subjective perceptions of improvement. Implications of these findings for clinical practice are examined.
PHYSICIANS and their patients are gradually moving away from a paternalistic model of medicine1 to embrace a shared decision-making approach.2 Collaborative decision making requires openness, mutual information sharing, and clinical negotiation.3 Understanding patients' expectations and responding appropriately are fundamental goals of clinical practice4 that can promote patient satisfaction, encourage adherence to treatment plans, and deter malpractice suits.5 Research shows that physicians' actions are influenced considerably by their perceptions of patient preferences.6
Given that physicians' wield almost exclusive control over the order sheet and prescription pad, the primary means by which patients exert influence in the medical partnership is through requests for information and action.7 The questions patients ask and the resources they solicit can reveal much about their psychological states, health-related perceptions, treatment expectations, and life circumstances.7,8 Request fulfillment can increase patient satisfaction,7,9 foster perceptions of physician attentiveness and communication competence,9 and may even enhance adherence and outcomes (R.L.K., R.A.B., R.A., E.K., D.T., and Steven Kelly-Reif, MD, unpublished data, 1999).9 In particular, patient requests are an integral part of clinical negotiations, which can resolve differing views of the patient's situation,10 lead to a mutually acceptable treatment plan, and promote patient satisfaction.11
To date, far more attention has been given to what patients say during clinical encounters (communication behaviors) than to what they do not say (unexpressed desires).12,13 A patient desiring action from the physician may remain silent for many reasons. Desires might not be articulated owing to the sensitivity of the topic, beliefs about what constitutes appropriate patient behavior, a poor relationship with the physician, cultural gaps,14 time constraints in the visit, or the physician's practice style. Whatever the reason, the unvoiced need is not likely to be met, potentially resulting in poorer health outcomes.12 Soliciting the patient's goals and desires is deservedly recognized as a fundamental physician skill15 that can be taught,16 but it is a skill that is poorly demonstrated in many medical interviews.17,18
The concepts "agenda," "desire," and "request" are not synonymous. The patient's agenda is his or her goal (eg, "I want to determine if I have a ruptured lumbar disk"), whereas a desire is a means for satisfying that goal (eg, "I want my doctor to order a magnetic resonance imaging"). Agendas are typically formulated prior to a visit, whereas a desire can be held before or emerge over the course of the visit. A desire becomes a request when it is verbalized to the physician.
Elsewhere we have examined requests and their fulfillment (R.L.K., R.A.B., R.A., E.K., D.T., and Steven Kelly-Reif, MD, unpublished data, 1999).7,8 We extend this line of work by focusing on unspoken desires–requests that are never verbalized. Four issues are addressed. First, we quantify the extent to which patients leave their outpatient visits having remained silent about their wishes for medical information, tests, referrals, and other kinds of physician action. Second, we identify patient, physician, and practice factors associated with unvoiced desires for care. Third, we investigate the effects of unexpressed desires on patients' and physicians' evaluations of the office visit. Fourth, we examine the association of patients' unvoiced desires with health outcomes and subsequent health care utilization. To assess the generalizeability of our conclusions, we address these issues in both primary care and specialty settings.
This study is a component of the Physician Patient Communication Project, a large-scale investigation carried out in Sacramento (California) County's 2 largest health care systems, the University of California, Davis, Medical Group (UCDMG) and Kaiser-Permanente. The University of California, Davis, Medical Group is a multispecialty group practice with explicit utilization review where physicians are compensated according to a Resource-Based Relative Value Scale (discounted fee-for-service) and specialists are salaried. Physicians at Kaiser-Permanente, a group model health maintenance organization, are salaried, with limited bonuses earned by those who meet utilization, quality, and patient satisfaction targets. The study was reviewed and approved by the University of California, Davis, Human Subjects Review Committee.
Physicians were invited to participate in the study if they were practicing family medicine, internal medicine, or cardiology and if they were providing direct patient care at least 20 hours per week. At UCDMG, the group's associate medical director identified suitable practice sites and helped to recruit several physicians from each site. At Kaiser-Permanente, potentially eligible physicians were recruited by mail and through interpersonal contact. Efforts were made to balance the cohort for sex, ethnicity, and physician experience.
English-speaking patients 18 years or older were sampled from among patients scheduled to see a participating physician during screening periods held from February 1, 1999, through November 22, 1999. A sampling frame of potentially eligible patients was created from appointment lists obtained 1 to 2 days prior to the visit and then randomly sampled until daily quotas were met. Patients were eligible for study if they could complete a written questionnaire with minimal assistance and if they were willing to provide written informed consent. In an effort to focus on problem-based visits, we also imposed the eligibility requirement that a patient needed to report a new or worsening problem or be concerned about having a serious disease (79% were enrolled in the study because of a new or worsening health condition).
Of 2606 patients contacted by telephone during the enrollment period, 737 were ineligible, usually because they neither had a new or worsening condition nor were worried about having a serious undiagnosed condition. An additional 677 declined to participate, most often before their eligibility could be ascertained. In total, 80.4% of patients known to be eligible agreed by telephone to participate and 84.5% (n = 909) of these completed patient questionnaires at the index visit. We were able to administer by telephone 2-week follow-up questionnaires to 97.6% (n = 887) of the patients who completed index visit questionnaires.
Data were collected from patients at 4 time points. Participating patients completed a brief telephone screening questionnaire administered 1 to 2 days prior to the index visit, a self-administered questionnaire immediately before the index visit, a self-administered questionnaire immediately after the visit, and a telephone follow-up questionnaire approximately 2 weeks after the index visit. For 99.3% of the patients, physicians completed postvisit encounter forms that included questions on visit type, chronic conditions, treatments, patient behavior, and visit demandingness and satisfaction.
Patients indicated in the immediate postvisit questionnaire if they had asked their physician for medical information, a physical examination, a diagnostic test or procedure, new medications, a specialist referral, physical therapy or medical equipment, assistance with paperwork, or any other form of resource or help. Patients who did not voice a particular type of request were asked if such a request was one that they wanted to make, but did not. Their responses to these questions were analyzed separately, but also summed to create an unexpressed desires index (potential range, 0-8).
Immediately after the index visit and at the 2-week follow-up, patients rated their satisfaction with the care they received on a widely used instrument composed of five 5-point agreement scales.19 These items were averaged to create a satisfaction with care score (α = .8820 on both occasions; range, 1-5 [least to greatest satisfaction, respectively]). Endorsement of the physician was measured after the index visit and at the follow-up visit by averaging responses on three 5-point agreement scales that addressed the patients' willingness to make a special effort to see the index physician again, their eagerness to recommend the physician to a friend, and their intention to follow the physician's advice (α = .90 and .89 after the index visit and at the follow-up visit, respectively; range, 1-5 [least to strongest endorsement, respectively]). At the follow-up visit, patients also reported via single 5-item agreement scales the extent to which their physician made them feel well taken care of and did everything possible on their behalf (range for both items, 1-5 [least to most positive evaluation, respectively]).
At the 2-week follow-up, postvisit use of health care was assessed in 2 ways. First, patients indicated (yes/no-question format) if they had gone to an emergency department, had been an overnight hospital patient, made contact by telephone or in person with the index physician, or had any contacts with another physician concerning the index visit problem since the visit. Their responses were used to create a health system contacts index that reflected the number of affirmative responses to these questions (range, 0-4 [4 = high usage]). Second, patients who reported that they had postvisit contact with the index physician were asked in a subsequent series of questions if they had requested further medical information, tests, or procedures; specialist referrals; physical therapy or medical equipment; new medications; or assistance with paperwork. Responses to these questions were used to create a postvisit requests index, which was the number of yes responses to these queries (range, 0-6 [6 = high usage]).
Patients' health concerns were measured by averaging responses on three 5-point items that asked about symptom bothersomeness, health worries, and concerns about having a serious condition not yet diagnosed. These items were administered immediately after the index visit (α = .68) and at the 2-week follow-up visit (α = .72) (range, 1-5 [5 = greatest concern]). At the 2-week follow-up visit, patients also made a direct rating of symptom improvement on a single 5-point scale (1 = much worse; 5 = much better). Physical functioning, role–physical functioning, and general health perceptions were assessed with instruments taken from the Medical Outcomes Study Short Form-36 survey21; the first 2 of these were also assessed at the 2-week follow-up visit. These measures were scored such that higher scores indicate better functioning and more positive self-evaluations of general health (range, 0-100). The α reliabilities exceeded .93 for the physical functioning measure, .86 for the role–physical functioning instrument, and .81 for the general health perception instrument.
Following each visit, the physician rated on 2 single-item 5-point scales their impressions of how demanding the visit was for the amount of effort required and how satisfying it was in comparison with the typical visit (range, 1-5 [5 = far more demanding or satisfying than typical]).
We also assessed 3 other variables for statistical control. Previsit trust in the physician seen during the index visit, used as a proxy for the quality of the prior relationship, was assessed by averaging responses on 9 items developed for this study. This instrument was developed based on results from previous patient focus groups and piloted for clarity and acceptability prior to being used in the current study. The wording of some of the items was patterned after items appearing in 2 published instruments.22,23 These items asked patients to judge their level of trust in their physician's honesty, competency, and agency (the extent to which the physician acts in their interests at all times) on 5-point response scales, where 5 indicates complete trust (α reliability = .90).24 Because trust could not be assessed for those 195 patients (21.5% of the sample) who had no prior experience with the index physician, we created a 4-category trust variable that classified patients into low, moderate, or high (relative) trust groups of nearly equal size or into a fourth category labeled "no prior relationship." These 4 categories were represented as 3 dummy-coded covariates in our analyses. Visit length was estimated in minutes after the outpatient visit by physicians; such estimates have been found in our data to be highly correlated with actual visit length.8 Number of requests made by patients was also controlled for on the assumption that the patient with a long list of expectations might have less of an opportunity to verbalize 1 or more of those expectations. This variable was assessed by asking patients after the outpatient visit to report within 8 categories of resources those for which a request had been made (eg, medical information, referrals, tests or procedures, and others).
Basic descriptive statistics were used to describe the sample. Primary analyses were corrected for the clustering of patients within physicians using the Stata 6.0 svytab, svylogit, svymean, and svyreg (Stata Statistical Software: Release 6.0; Stata Corp, College Station, Tex) procedures for complex surveys. In these analyses, the physician was identified as the cluster (primary sampling unit) and a 6-level stratification variable was created by crossing specialization (internal medicine, family practice, and cardiology) with site (UCDMG or Kaiser-Permanente). Probability weights were assigned to the patients in a cluster to account for differences among physicians in the number of patients enrolled in the study from their practice and the number of patients they see regularly. Specifically, a weight was assigned to each patient within a cluster (ie, physician practice) by (1) multiplying for each physician the number of patients seen weekly on an outpatient basis by sample size and (2) dividing this value by the product of the number of patients seen weekly by all physicians in the study and the number of patients enrolled in the study by the index physician. By so doing, observations obtained from very busy practices were given greater influence. Weights produced by this method ranged from 0.18 to 1.90 (mean weight, 1.0).
A total of 45 physicians were enrolled in the study, 22 from UCDMG (29% of eligible physicians) and 23 from Kaiser-Permanente (15% of eligible physicians). Sixteen practiced general or family medicine, 18 general internal medicine, and 11 cardiology. The mean (SD) age was 44 (8.3) years, 14 (31%) were female, and 13 (29%) were nonwhite. Participants were affiliated with their current institution for a mean (SD) of 8.3 (6.3) years, had held their medical degrees an average (SD) of 17 (9.1) years, and reported spending an average of 39 hours per week in direct patient care; 43 (96%) were board certified. Data were collected at 6 UCDMG locations and 5 Kaiser-Permanente sites.
Among 909 patients completing baseline questionnaires, the mean (SD) age was 57 (15.3) years, 56.1% were female, and 81.0% were white. Seventy-seven percent reported completing at least some college, and 30.2% had at least a bachelor's degree. About half (45.3%) were employed at least part-time, 60.4% were married, 19.1% reported annual household incomes of less than $20 000, and 96.0% had health insurance. Slightly more patients came from Kaiser-Permanente (51.3%) than from UCDMG (48.7%). Physicians reported that 13.0% of these visits were for comprehensive evaluation, 71.8% were follow-up visits for a patient known to the physician, 5.3% were urgent care appointments with a colleague's patient, and the remaining 9.9% were for other purposes.
Not surprisingly, cardiology and primary care patients differed significantly in many ways. Cardiology patients tended to be older, on average (64 vs 55 years, P<.001), and were more likely to be male (57% vs 40%, P<.001), married (71% vs 57%, P<.001), and a member of the lowest household annual income group of under $20 000 (28% vs 17%, P = .003). Physicians reported significantly more chronic medical conditions for cardiology patients than for primary care patients (1.9 vs 1.3 conditions, P<.001) and cardiology patients rated themselves to be less healthy on all of the Medical Outcomes Study Short Form-36 survey21 measures (all P values <.001). Cardiology visits were longer than primary care visits by about 3 minutes (20 vs 17 minutes, P<.001) and were rated by physicians as being more effortful (P = .02). These 2 groups did not differ on the patient and physician visit satisfaction measures, but cardiology patients did rate their physicians more positively on the endorsement and trust measures (both P values <.001).
We initially examined the prevalence with which patients reported unspoken desires for each of the 8 prespecified categories. The top data series in Figure 1 shows the percentage of all patients who reported at least 1 unvoiced desire for that category. Patients were most likely to report an unvoiced desire for a specialist referral (3.4% of patients) and least likely to report an unvoiced desire for paperwork assistance (0.2%). The bottom data series in Figure 1 shows the percentage of patients who desired a particular resource but did not request it. (For example, among just those 317 patients who said they were interested in getting a test or procedure, 95.6% mentioned it to their physicians and 4.4% remained silent.) Desires for referrals and for physical therapy or medical equipment were least likely to be conveyed to the physician (16.5% and 8.2% of desiring patients, respectively). In contrast, a desire for medical information was seldom left unsaid (2.7%).
A total of 91.1% of the patients reported no unvoiced desire, 6.8% reported 1 unvoiced desire, 0.7% reported 2 unvoiced desires, and 1.4% reported 3 or more unvoiced desires. Given the skewed nature of this distribution, we created a dichotomous variable for subsequent analyses by classifying patients into 1 of 2 groups: those who reported no unvoiced desires (91.1%, coded as 0) and those who reported at least 1 unvoiced desire (8.9%, coded as 1).
Patient, physician, and practice behaviors associated with unvoiced desires were initially examined in a series of univariate analyses using the Stata svytab (cross-tabulation) procedure. After weighting and correction for design effects, 4 patient characteristics were found to be associated with the tendency to leave wishes unvoiced: younger age, female sex, nonwhite ethnicity, and being unmarried. In addition, patients with lower trust in the treating physician were more likely to remain silent (Table 1). Unvoiced desires were more common in visits of intermediate length than in shorter and longer visits, but this effect failed to reach significance (P<.06). We anticipated that patients with long agendas would be more likely to leave a request unvoiced owing to the time constraints of the visit. In fact, those patients with 1 or more unvoiced requests did not differ significantly from patients with no unvoiced requests on the number of requests made of their physicians (1.56 vs 1.50 requests, P = .18).
The svylogit multivariate procedure was used to assess the independent effects of these variables. Unspoken desires were more common among patients who were younger, less educated (high school education or less), unmarried, and less trusting of the treating physician (all P values <.04). The difference in prevalence found in the univariate analyses for patient sex and race were not identified when covariates were considered. Other analyses, not reported, found no effect on unvoiced desires of patient-physician congruence on age, sex, or ethnicity.
Patients' postvisit evaluations of their encounters were examined by comparing the group of patients with no unexpressed desires to those patients with at least 1 unexpressed desire, after correcting for design effects and adjusting means for patient age, sex, race, previsit general health perceptions, prior relationship quality, and the number of self-reported requests (Table 2). After adjusting for these variables, we found that patients with unvoiced desires were less satisfied with the care they received when assessed immediately after the index visit and less likely to endorse their physicians; these effects persisted for the duration of the 2-week follow-up period. In addition, 2 weeks after the index visit, those with an unvoiced desire felt less well taken care of and were less likely to feel that their physician did everything possible for them. With regard to physicians' ratings, visits in which patients reported 1 or more unvoiced desires were judged to be significantly more demanding (Table 2). No significant effect was found for physicians' visit satisfaction.
The possibility that these results might vary as a function of physician specialty was examined. For the endorsement measure at follow-up and physicians' ratings of visit demandingness measure, significant effects (P<.05 criterion) were found for at least 1 of the 2 dummy variables created to represent the interaction between the unvoiced desires grouping variable and the 3-level specialty categorical variable. Subsequent analyses revealed that the difference between patients with voiced and unvoiced desires on the endorsement measure at follow-up was substantially greater in family practice settings (r = −0.64, P<.001) than in internal medicine (r = −0.20, P = .25) and cardiology settings (r = 0.01, P = .97). As for the demandingness measure, the effect of unvoiced desires on physicians' ratings of the effort required of them was greater for cardiologists (r = 0.60, P = .02) than for internal medicine (r = 0.31, P = .06) and family practice physicians (r = 0.20, P = .18).
The effect of having unvoiced desires on postvisit health use and clinical outcomes was examined, again correcting for design effects. In these analyses, utilization and outcome measures were adjusted for patient age, sex, race, previsit general health perceptions, prior relationship quality, visit length, and the number of requests. Analyses of health concerns, physical functioning, and role–physical functioning measures were further adjusted for their baseline values. The only significant effect was obtained for patients' direct rating of symptom improvement at follow-up. Patients with an unvoiced desire reported the least amount of symptom improvement (3.4 vs 3.8 on a 5-point scale, P = .02). They also reported more health concerns, lower physical functioning, and more physical limitations (ie, worse role–physical functioning measure), but these effects failed to reach significance. There was no evidence that these results varied as a function of physician specialty.
In this sample, fewer than 1 in 10 patients reported that there was something they wanted to ask of their physicians but did not. It would seems as if unvoiced requests are less common than hidden agendas, which have been estimated to be present in perhaps 30% to 50% of the visits.13 Our finding is not necessarily at odds with this estimate because patients can ask for all that they desire while remaining mute on their goals for the visit. The prevalence of unvoiced desires differed dramatically as a function of the nature of the resource requested. For instance, patients almost always asked for the medical information they wanted, but were more hesitant to ask the physician for referrals and for physical therapy or medical equipment. Patients may appropriately assume that it is the physician's responsibility to answer their questions. A desire for a referral, however, might be kept private owing to a fear that it would be construed by the physician as a challenge to his or her medical competence. Alternatively, the high incidence of unvoiced desires for specialty care may reflect patients' awareness of the role of managed care in the referral process. For whatever reasons, research has shown that desires for specialty care often remain unexpressed by patients and unrecognized by physicians.25
The multivariate analyses suggest that trust, a key component of which is clear and complete communication,24 may encourage patients to present their desires to physicians. Conversely, individuals with a high school education or less were more likely to report unspoken desires. This effect may be owing to less awareness among these patients about alternative treatment options. Desires were also much more likely to remain unspoken by unmarried individuals and by the youngest patient group (aged, 18-29 years). We speculate that unmarried patients may be less likely to be accompanied by family or friends who could serve as intermediaries, encouraging the patient to voice concerns and requests. The finding for age could possibly reflect greater intimidation among younger patients. Unfortunately, we do not have the data to test these possibilities. The univariate analyses suggested that female and minority patients were more likely to remain silent about their desires, but this effect did not hold up when covariates were considered.
This study evidences the importance of encouraging patients to communicate fully to physicians the information and interventions they hope to receive. Unvoiced desires represent just one aspect of patients' evaluations of their medical encounters, but an important one. Patients who withheld requests were less satisfied with their care and their physician. These negative feelings were persistent enough to be detected 2 weeks after the visit. While it is true that some patients with unvoiced requests had less positive and possibly less established relationships with their physicians to begin with, the effects on patient evaluations of these unspoken requests held even when controlling for baseline relationship quality and other factors.
The visits in which patients failed to convey their wishes were judged by physicians to require more effort than the typical visit. This result may seem ironic since the unspoken request is presumably unknown to the physician. It is possible that the physician in such visits sensed that these patients were "holding back" and felt frustrated as a result. Alternatively, unspoken requests may be an indicator of unresolved tensions that make for a more challenging consultation.
It makes intuitive sense that unvoiced desires compromise health care quality by forcing the physician to treat a patient whose issues have not been fully communicated,13 but this study provides no compelling support for this hypothesis. For instance, we found no evidence that unvoiced desires led to poorer treatment, resulting in greater health care use after the index visit. Perhaps one should not be surprised to find that patients who felt unable to express their desires at the index visit did not seek care later from these same physicians. However, there was no indication of postvisit contacts with sources of care other than the index physician. There was also no evidence that unvoiced requests were simply expressed later in postvisit contacts with the index physicians and their staffs.
In contrast, patients' reports of their health concerns and health status were consistently more positive at follow-up for those patients who let the treating physician know of their expectations for his or her behavior. The only significant effect, however, was for patients' perceptions of symptom improvement. It would appear that when patients keep their desires to themselves, perceptions of quality of care are affected more than the care itself. One possibility is that patients prioritize their requests, leaving to themselves those that are less critical and less likely to be informative for the physician.
This study is not without limitations. It is possible that our sample statistics do not represent the true incidence of unvoiced desires in the patient population. First, by drawing patients from a health maintenance organization and medical group, we recruited a sample of largely insured patients. These individuals may feel comfortable asking for anything they think they might need because they would not have to shoulder much of the financial burden if these requests were fulfilled. Second, our eligibility requirement specified that only those patients with a medical condition or health-related anxiety could be enrolled in the study. These patients may have been especially motivated to get to the bottom of their problem, leading to fewer unarticulated desires for information and treatment. Conversely, by requiring that a patient have a new or worsening condition to be eligible, our sample may overrepresent patients with chronic conditions. It is unknown if such patients are more or less likely to have unvoiced desires.
An additional limitation is that the prevalence of unvoiced desires we report may reflect biases in our methods and sampling. For instance, some patients may be hesitant to report that they did not "stand up for themselves" by communicating their expectations to the physician owing to impression management concerns. However, it is possible that patients "invent" unvoiced desires in response to the questionnaire. Finally, our sample comes from a single geographic region and a managed care market that may not represent the nation as a whole.
Research on physician-patient interaction has focused extensively on patterns of verbal and nonverbal behaviors exhibited in clinical encounters. This study highlights the need for investigators of physician-patient communication to give attention to both what is said by patients and what remains unspoken. For practitioners, these findings suggest that a consideration of their relationship with a particular patient is key in promoting patient openness about their perceived needs. Silence is not always golden; what is left unsaid is not necessarily forgotten by the patient.
Accepted for publication January 9, 2001.
This investigation was supported in part by grant 03484 from the Robert Wood Johnson Foundation, East Princeton, NJ (Dr Kravitz).
We gratefully acknowledge the assistance of the 45 participating physicians and their patients. Thanks are also owed to Sara Lu Vorhes, Steven Kelly-Reif, MD, and David Ormerod, MD, for assistance with physician recruitment and data collection; to Christine Harlan for budgetary management; and to the staff of the Patient-Provider Relationship Initiative (Bernard Lo, MD, director) for technical assistance. Appreciation is also expressed to 3 anonymous reviewers of this journal for their outstanding critiques.
Reprints: Robert A. Bell, PhD, Department of Communication, University of California, Davis, Davis, CA 95616 (e-mail: firstname.lastname@example.org).