Difficult Patient Encounters in the Ambulatory Clinic:  Clinical Predictors and Outcomes FREE

NEARLY 1 of 6 outpatient visits is considered difficult by clinicians,^1- 2 who have labeled the patients seen in such encounters as "heartsink"^3- 5 or "black holes"⁶ in the United Kingdom, and as "difficult,"^1,7- 15 "problem,"^16- 17 "disliked,"¹⁸ "frustrating,"¹⁹ "troublesome,"²⁰ or even "hateful"²¹ in the United States. Most previous studies of such encounters have been anecdotal,^{7- 10,12- 15,21} focused on highly selected patient groups,^17- 19 or used unproved methods for assessing difficulty.^{1,5,9- 10,13,19} Scanty data on outcomes of care among "difficult" patients have been collected,^2,19 and none have measured provider characteristics associated with labeling encounters as difficult,²² an important omission given the fact the physician-patient relationship is dyadic and that both parties (or their interaction) may contribute to difficulty in the encounter.²³

In a prospective study of 500 patients presenting to a general medicine walk-in clinic with physical symptoms, we sought to determine what patient or physician characteristics predict difficult physician-patient interactions and how these interactions related to outcomes.

Adults presenting to the general medicine walk-in clinic at Walter Reed Army Medical Center, Washington, DC, with a chief complaint of a physical symptom were eligible. Exclusion criteria included an upper respiratory tract infection as the primary symptom or dementia. Since one aim of the study was to assess correlates with 3-month outcomes and 90% of upper respiratory tract infections resolve within 2 weeks, such patients were excluded. Immediately before seeing the physician, all patients completed surveys on symptom severity (0-10 scale) and duration (days), whether they had previously seen a clinician for this problem (yes/no), whether they were worried that the symptom could represent a serious illness (yes/no), and whether they had experienced stress in the previous week (yes/no). There were also questions on the 5 most common symptom-related expectations²⁴ (causal explanation, expected duration, prescription, diagnostic test, referral, other). In addition, patients completed the Medical Outcomes Study Short-Form Health Survey (SF-6), a 6-item measure of functional status,²⁵ and were evaluated for Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition depressive and anxiety disorders with the PRIME-MD (an acronym for Primary Care Evaluation of Mental Disorders).^26- 27 The SF-6 measures functional status in 6 domains: general health, role functioning, physical functioning, social functioning, emotional health, and physical pain. The PRIME-MD is a validated instrument, with 2 parts, a patient questionnaire followed by a semistructured interview among patients endorsing certain questions. Immediately after the visit, patients completed the RAND 9-item satisfaction survey²⁸ and a questionnaire that asked about postvisit serious illness worry and the presence of any unmet expectations.²⁴ Using the automated clinic database, the total number of patient visits with a physician within the Walter Reed system, which includes a number of outlying clinics, were determined for 3 months before and after the index visit.

Two weeks and 3 months after the visit, patients were mailed a questionnaire on symptom outcome, residual serious illness worry, unmet expectations, functional status, and satisfaction with the index encounter. Telephone contact of nonrespondents was attempted.

All eligible patients presenting to the clinic on selected half-days were invited to participate. While selected half-days were selected for convenience, the specific half-day sampled was rotated to ensure balanced sampling among all possible clinic half-days during the study period (October 1994 to December 1995). This clinic primarily provides continuity-of-care appointments, but has walk-in appointments available for patients desiring to be seen that day. Such walk-in patients are assigned to physicians on a queue system: first come, first assigned to the next available clinician. The clinician is assigned from a roster of available appointments, with clinicians listed alphabetically within each time slot. Nearly all study patients were seen by physicians seeing them for the first time. The demographics and case mix of patients seen in a military outpatient clinic are comparable to civilian settings.²⁹ This protocol was approved by our institution's human use committee. Informed consent was obtained prior to the visit by 1 of 2 clinical researchers.

Prior to study participation, physicians completed the Physician's Belief Scale, a 32-item questionnaire that measures attitudes regarding psychosocial aspects of patient care.³⁰ Each question is followed by a 5-point Likert-type response scale ranging from 1 (disagree) to 5 (agree). Scores range from 32 to 160, with higher scores reflecting poorer psychosocial attitudes. This instrument has been found to be a reliable and valid brief measure of physicians' psychosocial beliefs.³⁰

After each patient visit, physicians completed the 10-item Difficult Doctor-Patient Relationship Questionnaire (DDPRQ)^2,31 to assess clinician-perceived difficulty of the encounter, and indicated whether a subspecialty referral was provided. The DDPRQ has been previously shown to be a reliable instrument with an internal consistency of 0.88 to 0.96.

From the clinical database, prescription and diagnostic test orders were obtained for each patient. These were tabulated using Health Care Financing Administration relative value units and converted to dollars using the 1996 Medicare schedule. Prescription costs were based on either Health Care Financing Administration rates or generics (when available).

Analyses were performed using STATA.³² Continuous variables were analyzed using analysis of variance or covariance, standard linear regression techniques, or the Kuskal-Wallis test as appropriate. Categorical variables were analyzed using Mantel-Haenszel χ² or logistic regression. The relationship between difficulty and presenting symptoms was analyzed using the Fisher exact test for overall significance and χ² techniques by collapsing the tables into specific symptom vs all other symptoms. Predictors of difficulty were evaluated using logistic regression. Potential predictive variables were selected by univariate screening using P<.25 (Table 1). In addition, potential confounding or interactive effects of other variables were explored. For example, functional status was adjusted for well-described confounders such as mental disorders and age. Because of the clustered sampling technique, the Huber formula was used to produce robust confidence intervals.³³ Model fitting and variable selection, as well as interaction and confounding assessment followed the methods of Hosmer and Lemeshow, with the goal of producing the best-fitting, most parsimonious model.³⁴ Internal consistency of the DDPRQ and Physician's Belief Scale were assessed using Cronbach α.

Difficulty was analyzed as a dichotomous variable using a score of 30 or greater as indicative of difficulty. Data exploration using various cut points for difficulty confirmed the previously held cut point of 30 as functionally identifying a unique subgroup of patients. An attempt to use DDPRQ scores to categorize patients into 3 groups, "difficult," "average," and "satisfying," after the suggested hierarchy of Lin et al¹⁹ was not successful.

A cut point of 70 (on a scale of 32-160) was used as indicative of "poor" physician psychosocial attitude score for several reasons. First, dichotomization was necessary because the Physician's Belief Scale score was not linear over the logit function. Second, preliminary evidence suggests that a cut point of 70 may discriminate among physician specialties that differ in the degree of their biopsychosocial orientation.³⁰ Finally, a score of 70 clearly demarcated a group of clinicians with higher rates of experiencing encounters as difficult.

These data were collected as part of a clinical trial of providing previsit information to clinicians on patient symptom-related expectations and mental disorders.³⁵ Clinicians received this information on the second group of 250 patients enrolled. Since one outcome of the intervention³⁶ was a reduction in the frequency that patient encounters were labeled as difficult, adjustment for study group was made in all analyses.

Five hundred twenty-eight adults were invited to participate. The 500 participants were similar to nonparticipants in terms of age, race, sex, and type of symptom. Study patients were seen by 38 physicians with no physician contributing more than 34 patients. Patients had a mean age of 54.7 years, half were women, 49% were white, and 45% were African American. Seventy-six patient encounters (15%) were rated as difficult by the provider. In our sample of 500 subjects, the internal consistency of the DDPRQ was 0.89.

A depressive or anxiety disorder was present in 29.0% of patients, with 11.0% experiencing 2 or more disorders. Major depression was present in 8.4%, other depressive disorders in 17.4%, panic disorder in 1.4%, and other anxiety disorders in 14.2% of patients.

Among 38 participating physicians, 35 were general internists and 3 were family practitioners. Sixteen were faculty members and 22 were house staff. The 16 staff physicians saw 76% of the study patients (n=382). The mean age of participating clinicians was 35.6 years (median, 32.5 years), with staff physicians averaging 13 years of practice (range, 4-25 years). Thirty-nine percent of clinicians were women, 11% were African American, 5% Hispanic, and 5% Asian. The Physician's Belief Scale had a Cronbach α of 0.91.

Encounter difficulty was not associated with patient sex, age, ethnicity, educational level, or marital status (Table 1). Previsit expectations of care, including desiring an explanation of the symptom's cause, a prescription, an estimate of likely symptom duration, and subspecialty referral, did not differ between difficult and not-difficult patient encounters. There was no relationship between the duration of the presenting symptom or patient report of recent stress and difficulty.

Patients in difficult encounters were more likely to have an underlying mood or anxiety disorder, were worried that their symptom might represent a serious illness, reported greater symptom severity, and had higher rates of somatization (Table 1). Encounters with patients endorsing 0 to 1, 2 to 5, 6 to 9, or more than 10 other common physical symptoms on the PRIME-MD were found to be difficult 6%, 13%, 23%, and 36% of the time, respectively. Patients from difficult encounters had lower functioning on all domains of function assessed (overall health, physical, pain, role, social, and emotion), although the difference in emotional functioning was eliminated after adjusting for age and the presence of mood or anxiety disorders (Figure 1). There was no relationship between the type or number of mental disorders and physician-perceived difficulty.

No specific type of presenting physical symptom was associated with difficulty. However, when symptom groups were dichotomized into those with a specific symptom vs all other symptoms, dermatological (P=.01), otolaryngologic (P=.003), and genitourologic (mostly urinary tract infection) (P=.007) symptoms were less likely to be considered difficult than those presenting with any of the remaining physical symptoms.

Physician age, sex, ethnicity, house staff vs staff status, and the number of years of practice were not associated with the likelihood of rating patient encounters as difficult. There was no interaction between patient or physician age, sex, or ethnicity with regard to difficulty ratings. Clinicians whose Physician's Belief Scale score exceeded 70 (n=17) found 23% of patient encounters difficult, compared with only 8% of encounters involving clinicians (n=21) whose scores were 70 or less (P<.001).

Individual questions from the DDPRQ were also analyzed, using responses of 4 to 6 as indicative of difficulty.² Physicians reported that they were significantly less eager to see patients from difficult encounters again (82% vs 15%; P<.001), were more likely to be frustrated by such patients (66% vs 5%; P<.001), found communication difficult (37% vs 0.9%; P<.001), felt uneasy (51% vs 9%; P<.001), and secretly hoped these patients would not return for follow-up (65% vs 3%; P<.001).

The presence of a depressive or anxiety disorder, multiple physical symptoms (≥5 on the PRIME-MD checklist of 15 physical symptoms), symptom severity greater than 6 (on 10-point scale), and poorer physician attitude scores toward psychosocial problems independently increased the likelihood of difficulty (Table 2). The proportion of encounters rated as difficult when 0, 1, 2, 3, and 4 risk factors were present were 3%, 8%, 18%, 23%, and 47%, respectively. When difficulty was analyzed as a continuous variable, these 4 variables were found to account for 34% of the variance in difficulty, with physician psychosocial attitudes accounting for 5% of total variance.

Follow-up data were available on all patients immediately after the visit, on 463 (92.6%) at 2 weeks, and 413 (82.6%) at 3 months. There were no differences between respondents and nonrespondents at any time, including the proportion of encounters considered difficult.

There were no differences in visit costs or in the likelihood that a patient would receive a specific intervention from the physician, including a diagnostic test, prescription, or subspecialty referral. Patients from difficult encounters had no increase in the number of previsit expectations for care and were neither more nor less likely to receive desired interventions. Despite this, patients from difficult encounters were more likely to have unmet expectations for care both immediately after the visit and at 2 weeks.

Patients from difficult encounters were less likely to be satisfied overall with the care they had received immediately after the visit and at 2 weeks, and were more critical of all aspects of the physician-patient encounter measured. They reported themselves to be "somewhat" or "very dissatisfied" with the physicians's technical competence (9% vs 1%; P<.001), bedside manner (7% vs 0.7%; P<.001), explanation of what was done for them (12% vs 3%; P<.001), and the time spent with the clinician (13% vs 3%; P=.002), despite the uniformity of visit duration.

During the 3 months of follow-up, the difficult group had a higher number of visits (median, 4 vs 2; P=.004) and higher use of health services for a 6-month period, including 3 months before and after the index visit (5.5 vs 4.0 visits; P=.002).

There were no differences between difficult and not-difficult groups at any follow-up points with regard to the persistence of serious illness worry or improvement of the presenting symptom (70% at 2 weeks, 83% at 3 months). Patients also experienced significantly improved functioning in all 6 domains by 2 weeks after which functional status did not change appreciably over the ensuing several months. Initial functional status differences between difficult and not-difficult groups (Figure 1) were no longer present at either 2-week or 3-month follow-up. Also, the greater dissatisfaction with the index visit that the difficult group expressed immediately after the index visit and persisting out to 2 weeks was no longer apparent at 3 months.

Fifteen percent of encounters involving walk-in patients presenting with physical symptoms to a walk-in clinic were experienced as difficult by the clinician. Patient characteristics associated with difficult encounters included the presence of depressive or anxiety disorders, more somatic symptoms, and greater symptom severity. Poor physician psychosocial attitude was strongly predictive of experiencing more encounters as difficult. Adverse outcomes associated with difficult encounters included more unmet expectations, higher utilization rates, and greater dissatisfaction with the overall care received as well as with all aspects of the physician-patient relationship.

Several of our findings are similar to those seen previously. A relationship between difficult encounters and the presence of mental disorders,^{1- 2,16- 19,37} greater somatization,^2,19,37 and higher health care utilization^2,10,19 have been reported. Similar to previous reports, the relationship between patient functional status and difficulty disappeared when adjusted for the other patient characteristics.² We are the first to report on the importance of physician variables in the labeling process, to find higher rates of unmet expectations after the encounter (despite receiving a similar number of diagnostic tests, prescriptions, and subspecialty referrals), and to measure patient dissatisfaction with specific aspects of the encounters (Table 1). We are also the first to report on the lack of impact of "difficulty" on symptom or functional status improvement or on visit-specific costs.

METHODS FOR assessing the physicians' perception of a difficult encounter can vary. For example, Lin et al¹⁹ used a single question asking physicians to rate their experience with 228 high users of ambulatory services as "satisfying," "average," or "frustrating." A substantial proportion (37%) were labeled as "frustrating," probably due to the highly selected patient sample. Walker et al³⁷ used the DDPRQ as a continuous variable in asking rheumatologists to rate the difficulty of 68 patient encounters with fibromyalgia or rheumatoid arthritis. Hahn et al² used the DDPRQ as both a continuous and categorical measure in a study of 627 patients presenting for primary care. All 3 studies corroborated our own findings of more psychopathology among patient encounters rated as difficult or frustrating.

We found no interaction between physician and patient sex, age, or ethnicity and the likelihood difficult ratings, similar to a previous report.² We also found no relationship between the years of clinician experience and the likelihood of encountering patients as difficult, similar to those of one,² but conflicting with other reports.^13,33 Hahn and colleagues² found that clinicians with a higher than usual interest in psychosocial disorders were more likely to rate patients as difficult, conflicting with our results. Their finding was limited by clinician interest measurement based on responses to a single, nonvalidated question. Additionally, their physicians were a select group, volunteering to participate in the PRIME-MD Study and seeing their own clinic patients. Physicians with an interest in managing patients with psychosocial disorders tend to accumulate such patients in their practice. A higher interest in psychosocial disorders may result in a clinic population with a higher proportion of patients with such disorders. In our study of physicians seeing new, arbitrarily assigned walk-in patients, a 32-item validated measure of psychosocial interest found that clinicians with better psychosocial attitudes experienced significantly fewer patient encounters as difficult.

The Physician's Belief Scale used in our study has been found by other investigators to correlate with better physician communication skills and with a higher proportion of time spent discussing psychosocial issues.³⁸ One recent study of physician's communication styles found that a "narrowly biomedical" or "expanded biomedical" pattern was seen in 65% of patient encounters.³⁹ These 2 patterns were associated with the lowest levels of patient and physician satisfaction and had the least participatory interactions.⁴⁰ Only 8% of patient encounters were "psychosocial," the pattern most likely to reveal patient emotional distress.⁴⁰

It is possible that difficulty could be reduced by recognizing and treating mental disorders and by improving physician skills or attitudes toward addressing psychosocial problems or patient's serious illness concerns. Specific training in caring for "difficult" patients has also been found to help trainees gain understanding and empathy for such patients, rendering them less difficult.³⁹ Strategies for managing somatization^41- 43 and improving physician communication skills^44- 45 might be other measures to consider. Most reports suggesting approaches to managing difficult patients have focused on patient-physician communication.^{4- 5,7- 8,12,16,19} Physician training in communication skills has been previously associated with more accurate recognition of emotional distress and a more participatory medical decision-making style.⁴⁶ Despite the call for increased emphasis on improved interviewing skills and psychosocial aspects of care,^47- 50 training programs often do not offer high-quality teaching of interpersonal skills^51- 52 or adequate teaching of psychiatry.^53- 54

By 3 months, the relationship between encounter difficulty and patient dissatisfaction had faded. However, after 3 months, the intensity of patient dissatisfaction is likely to shift to more recent clinical encounters. Most patients had at least 1 interim follow-up visit, usually with a different clinician than seen in the index visit, and patients from difficult encounters averaged more than 7 follow-up visits. The patient's recall for the index encounter may be limited. Reports of dissatisfaction may be more likely to reflect attitudes about recent visits or symptom outcome. Indeed, in this study, symptom resolution by 3 months was the strongest correlate with 3-month satisfaction.

Our study has several limitations. First, because the sample consisted of walk-in patients seeing new physicians, one should be cautious in generalizing these findings to established clinician-patient relationships. One study found that new patients were less likely to be considered difficult than those that were "somewhat known" or "well known."² While this would suggest that we could be underestimating the prevalence of difficult encounters, it is interesting that the other study had a nearly identical rate of difficulty and that similar factors were associated with difficult encounters.² In addition, patients may be more willing to report themselves dissatisfied or to report unmet expectations in walk-in settings than those seen in the context of an established patient-physician relationship.

A second limitation was the inclusion of patients with a variety of physical complaints rather than 1 specific symptom. Although this may limit conclusions regarding individual symptoms, it does more broadly reflect the range of symptoms presenting in primary care.

Third, we used a brief checklist to detect symptom-related expectations. Kravitz et al⁵⁵ as well as other investigators^56- 57 using much longer instruments or interviews have identified other common expectations (eg, physician preparation for the visit, history taking, physical examination) that we did not measure. Although we invited patients to write in other expectations, they did so infrequently.

Finally, studies have found that patients with somatoform disorders are particularly likely to be considered difficult.^2,37 Our study focused on depressive and anxiety disorders and did not specifically diagnose somatoform disorders. However, the fact that multiple, severe somatic complaints were strongly associated with being considered difficult corroborates the impact of such disorders on difficulty.

The strengths of this study include its prospective design, large sample size, measurement of multiple symptom-relevant outcomes at different time points in the patient's care, inclusion of physician psychosocial belief measurements, and the use of a validated instrument for determining patient difficulty. The arbitrary method of assigning patients to clinicians prevented biased sampling from potential concentration of patients with known psychosocial problems into particular clinician's practices. Response rates were also high, with 93% two-week and 83% three-month follow-up.

The portrait of patients likely to be considered difficult is gradually being clarified as a consequence of a number of studies. Difficult patients are more likely to have multiple somatic complaints, to be seriously worried about their symptom, to report greater symptom severity, have underlying mental disorders, and to report poorer self-reported functional status. Adverse patient outcomes of difficult encounters can include more unmet expectations, less satisfaction with care, and higher utilization rates. Finally, clinicians with poorer psychosocial attitudes are more likely to experience patient encounters as difficult. The magnitude of the impaired relationship as viewed by both participants is troublesome. As continued emphasis is placed on patient-centered medicine⁵⁸ and on the patient-physician relationship,⁵⁹ research that focuses on interventions to improve clinician skills in managing such patients is needed.⁶⁰

Accepted for publication August 15, 1998.

This work was supported by intramural grant 32001 from the Uniformed Services University of the Health Sciences, Bethesda, Md.

The views expressed in this article are those of the authors and should not be construed to represent in any way those of the Department of Defense or the Department of the Army.

Reprints: Jeffrey L. Jackson, MD MPH, Department of Medicine-EDP, Uniformed Services University of the Health Sciences, 4301 Jones Bridge Rd, Bethesda, MD 20814.