| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
ORIGINAL ARTICLES |
From the Divisions of Clinical Psychology (P.S., A.R.) and Primary Care (A.R., C.F.D.) and Computing Services Department (J.C.D.), University of Liverpool, Liverpool, England; Bute Medical School, University of St. Andrews, Fife, England (G.M.H.).
Address correspondence and reprint requests to Peter Salmon, Department of Clinical Psychology, University of Liverpool, Whelan Building, Brownlow Hill, Liverpool, L69 3GB England. E-mail psalmon{at}liv.ac.uk
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODRESULTSDISCUSSIONNOTESREFERENCES |
|---|
Methods: Primary care physicians identified consecutive patients who consulted with symptoms that the physician considered unexplained by physical disease. Four hundred twenty consultations with 36 physicians were audio recorded and transcribed, and physician and patient speech was coded turn by turn. Hierarchical logistic regression analysis modeled the probability of the physician proposing medical care as a function of the quantity of patients’ speech of specific kinds that preceded it.
Results: Whether physicians proposed medical care was unrelated to patients’ attributions to disease or demands for treatment. Proposals of explicitly somatic responses (drugs, investigation or specialist referral) became more likely after patients had elaborated their symptoms and less likely after patients indicated psychosocial difficulties. Proposals of a further primary care consultation were responses simply to lengthening consultation.
Conclusions: The findings are incompatible with the influential assumption that physicians offer medical care to patients with unexplained symptoms because the patients demand treatment for a physical disease. Instead, the reason why many of these patients receive high levels of medical care should be sought by investigating the motivations behind physicians’ responses to patients’ symptom presentation.
Key Words: somatization • communication • physician-patient relations • symptoms
Abbreviations: MUS = medically unexplained symptoms.
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODRESULTSDISCUSSIONNOTESREFERENCES |
|---|
Contrasting explanations for why physicians offer medical care for medically unexplained symptoms (MUS) have been proposed, with correspondingly divergent implications for how they might manage these patients better. Emotional disorders have often been found to be more prevalent in such patients (11–13), and a longstanding and influential view is that unexplained physical symptoms can reflect the "somatization" of psychological distress that patients otherwise deny (14). Interviews with physicians have been consistent with this view in supporting the assumptions that patients believe that physical disease is present and that they press physicians to treat it (15–18). On this analysis, physicians should help patients to think psychologically about their problems (19). An alternative view arose from observational studies of consultations about MUS. Patients’ overt pressure for treatment was rare, and their suggestions of physical disease were tentative. Instead, presentations that had the potential to provoke physicians to offer medical care were characterized by graphic, elaborate, and extended accounts of symptoms and their effects (20). It was suggested that physicians offered medical care because they misperceived patients as seeking it or because it provided a tangible response to the sense of pressure that this presentation caused (21). According to this view, physicians should engage better with these patients’ need for explanation and support (20,22).
The present study tested these contrasting explanations. We tested whether the probability that the physician would propose medical care at any point in the consultation was related to how much patients had previously a) extended and elaborated their symptom presentation or b) linked their symptoms to physical disease and requested physical intervention. We predicted that physicians’ proposals of care would depend on a) but not on b). Most patients presenting MUS indicate psychosocial needs (20) and we also examined the influence of these cues. In principle, they should help divert physicians from offering medical care, but we anticipated that their effect would be small or absent because physicians routinely disregard them (20). As tangible offers of care, we included a further consultation with the primary care physician, pharmacotherapy, investigation, or specialist referral. We went on to explore whether physicians proposed further consultation or explicitly somatic interventions (drugs, investigations, and referral) in response to different aspects of patients’ presentation.
|
METHOD |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODRESULTSDISCUSSIONNOTESREFERENCES |
|---|
After approval of local research ethics committees, 50 primary care physicians from eleven practices with research or educational links to the University of Liverpool were asked to take part, of whom 42 (84%) agreed (22 males, 20 females) with 5 to 42 years’ medical experience. Practice size ranged from 1 to 10 physicians (mean, 4.5) and 2087 to 13,116 patients (mean, 7564). Six practices were urban, four were suburban, and one was rural. Jarman deprivation scores ranged from –11 to 56 (mean, 21.3), indicating that this sample was drawn from a relatively deprived section of the UK population. Consecutive patients attending participating physicians were approached by a researcher before consultation and asked for written consent to audio record their consultation. Of 5083 patients attending surgeries, 1086 (21%) were excluded (under 16 years or unable to consent because of visual impairment, learning disability, or extreme distress); 75 (2%) were missed; therefore, 3922 (77%) were asked for consent. Data were collected in 2000 to 2002.
There are no agreed research diagnostic criteria for primary-care patients with unexplained symptoms. We used criteria based on those proposed by Peveler et al. (1) to identify patients who, in the physician’s opinion, have unexplained symptoms. Immediately after each consultation, the physician completed a checklist to indicate whether or not the consultation involved a) presentation of a physical symptom b) that could not entirely be explained by a recognizable physical disease. Consultations satisfying these two criteria were retained for analysis. These criteria have face validity in that physicians can readily understand and apply them, and they allow for the clinical reality that physicians differ in the symptoms that they regard as unexplained by physical disease. Two additional questions, concerning duration of symptoms and whether they were causing the patient clinically significant distress or impairment, were used initially (n = 36) but discarded, as explained previously (4). Although it is possible that some symptoms identified as "unexplained" might prove to have a pathological cause, our selection procedure ensures, crucially, a patient group that is defined by their clinicians’ belief that such a cause is absent.
Procedure
Each physician operated a Sony MZ-R55 minidisk Walkman and Sony ECM-F8 Electret condenser desktop microphone to record consultations with consenting patients and completed the checklist (see above) after each consultation. Audio recordings of consultations meeting the research criteria were identified and anonymously transcribed, including all speech and noting silences exceeding 10 seconds and simultaneous speech. Patients’ symptoms were noted from transcripts and categorized by the research team using the British National Formulary categorization. Treatment decisions were also noted from transcripts and were checked by the research team against medical records for those patients who consented to this.
Interaction Analysis
Consultations were coded, turn by turn, using the Liverpool Clinical Interaction Analysis Scheme (4), which was developed specifically for primary care consultations about MUS. It contains 25 and 30 substantive codes for the patient and physician, respectively, and has good reliability. The unit of coding is an "utterance," defined pragmatically as a piece of speech that has sufficient meaning to be coded. So that sequences of communication could be examined, codes are linked to the speech turn (numbered sequentially from the start of consultation, separately for the physician and patient). A single turn could include more than one utterance. Occasionally, an utterance extended across more than one turn. Coding was by A.R.; 16 transcripts were also coded by C.D., G.H., and P.S. to confirm reliability of coding, reported previously (4). Speech categories relevant to the hypotheses were selected for use in the present study (Table 1). Categories were, where appropriate, formed by aggregating specific codes from the coding scheme, which were functionally equivalent in relation to the present research questions. Patients extended their symptom presentation either by providing further information about symptoms or reporting new ones; they extended psychosocial presentation by offering psychosocial factors as explanations for symptoms or independently of them. As proposals of somatic interventions by patients or physicians, we included references to the value or potential value of drugs, investigations, or specialist referral, as well as references in which the precise nature of the intervention was not specified. Speech codes not relevant to the present research questions were disregarded. The application of these codes is illustrated in Table 2.
|
|
Background to Sequence Analysis
The coding of physician and patient speech contains information not just on content and frequency but also on interdependence and serial dependence. A range of statistical approaches has been used to analyze such information (24), and each necessarily makes assumptions about the nature of the sequential processes. Lag sequential analysis, log-linear analysis, and Markov chain modeling test the influence of a given type of utterance on the types of utterance that occur a specific number of speech turns later. Pattern recognition techniques are less restrictive about the precise location of utterances but assume that specific types of utterance are linked repeatedly in a sequence. Neither is appropriate for our predictions, according to which the probability of a physician response depends on the cumulative number of specific patient strategies that precede it; we make no assumptions about the precise location of those preceding strategies, and we assume that their influence on physicians’ speech lies in their cumulative effect rather than in a repeated sequence. We used hierarchical generalized linear modeling to model the probability of a specific type of physician utterance as a function of the numbers of patient utterances of specific types that had preceded it. Important advantages of this new approach to sequence analysis are that it is based on clearly defined statistical assumptions and distinguishes variability between patients and between physicians, as well as between speech turns.
Statistical Procedures
Groups were compared on age by t test and on gender ratio by 
2.
There are three levels of variability in the speech data. Turns are clustered within consultations, and there is variability between consultations and between groups of consultations with the same physician. Therefore, in multilevel analyses, we distinguished variability at the turn, patient (consultation) and physician levels (25). In the first analysis, the response variable was whether, on any turn, the physician proposed medical care, including a further consultation with the primary care physician, a prescription for somatically oriented drugs, investigation, or referral to a specialist. Predictor variables were patients’ symptom and psychosocial presentations, disease explanations, and somatic proposals (Table 1). By also including turn number as a factor, we controlled for lengthening consultation.
The response variable was modeled with a logit link function and a Bernoulli sampling distribution. Mixed-effects models were fitted with random intercepts and, to avoid the restrictive assumption that independent variables should affect all patients and physicians similarly, random slopes. Preliminary analyses indicated that patient and physician gender did not influence the physicians’ proposals of medical care, and these variables were therefore not included in subsequent analyses. Independent variables were first examined singly to identify those with heterogeneity of slope (26). In preliminary analyses, not shown here, each variable was, in turn, given a random slope at the patient or physician level, whereas all other slopes were fixed. Significant variance components indicated heterogeneity of slope. Hence, corresponding slopes were set to random in the final analysis; others were fixed. Coefficients from the unit-specific model for the physician level were used to test the significance of the independent variables. Variance components were examined to indicate the relative magnitude and the significance of variability among patients and physicians in the mean (intercept) and in the influence of dependent variables (slope). The first analysis confirmed significant influences on physicians’ proposals of medical care. Therefore, to find out whether these influences applied similarly to proposals of care from the physician (further consultation) and of immediate somatic intervention (drugs, investigation, and referral), two further analyses, conducted as above, examined these as dependent variables.
SPSS (v12 for windows) and HLM 6.0 were used for analysis. The significance criterion was p < .01.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODRESULTSDISCUSSIONNOTESREFERENCES |
|---|
Sequence Analysis
The mean number of physician speech turns per consultation was 78.01 (median, 69; maximum, 195). Table 1 summarizes descriptive statistics for the speech categories. Results of the regression analyses are shown in Table 3. For each component of physicians’ proposals of care, Figure 1 describes the effects as odds ratios associated with the proposal, and Figure 2 depicts how the significant variables cumulatively changed the probability of such a proposal.
|
|
|
Physicians’ proposals were unrelated to patients’ references to physical disease or requests for somatic treatment, but depended on turn number and on both symptom and psychosocial presentation (Table 3). Whereas proposals of medical care increased with turn number and after symptom presentation, they became less likely after psychosocial elaboration.
Subsequent analyses examined how each component of medical care that physicians proposed was linked to patients’ presentation. The probability of explicitly somatic proposals grew as patients added to their symptom presentation and decreased after they added to their psychosocial presentation (Table 3, Figures 1 and 2). For neither of these effects was the variance component significant at the physician level, indicating consistent effects across physicians. However, although on average patients’ proposals for physical intervention did not influence physicians’ proposals, the significant variance component at the physician level indicates that physicians varied in their response to this presentation (Table 3, Figure 1).
The only influence on the probability that physicians would ask patients to return for a further consultation was the turn number, i.e., how long the consultation had lasted (Table 3, Figures 1 and 2). Controlling for this, physicians’ offers were unrelated to what the patients had said. The variance components indicate that, although physicians’ offers of further consultation were consistently unaffected by patients’ speech of the types that we examined, the average effect of turn number masked significant variability in its influence.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODRESULTSDISCUSSIONNOTESREFERENCES |
|---|
There is no support for the influential and longstanding assumption that this is because patients insist on physical disease explanations and demand intervention. Physicians instigated further consultations in response simply to lengthening consultation. Whether they proposed immediate somatic intervention depended, however, on what patients said. As predicted from qualitative evidence that patients’ symptom elaboration is the main potential source of subjective pressure on physicians in these consultations (21), they proposed somatic intervention in response to patients’ complaints about and elaboration of their symptoms.
We do not know the specific properties of symptom presentation that led physicians to respond in this way. Increasing complexity of presentation might be important, as symptoms defy physicians’ attempts to explain them (21). Reports of the emotional and social consequences of the symptoms, whereby patients emphasize their own privileged knowledge (27), might also be important. Alternatively, simple repetition of suffering might suffice. Neither do we know why such presentation provokes physicians to propose somatic intervention for symptoms that they believe are not caused by disease. Physicians might misperceive patients as seeking such responses, or they might consider that they lack alternative responses to patients’ suffering. Many primary care physicians feel negative toward MUS patients, consider their symptoms to be invalid (18), do not want to provide support (15), and consult in a less patient-centered way with these patients than with others (23), so somatic responses might be a way to avoid engagement. Finally, interviews with physicians have indicated feelings of powerlessness in consultations about unexplained symptoms (18), and observations of consultations have suggested that patients can emphasize areas in which they, rather than the physician, have authority, including psychosocial complaints and subjective details of symptoms (21,27). One way in which physicians might seek to reestablish their authority could be to emphasize the area in which they retain exclusive authority: the administration of physical intervention.
We have previously shown that patients with MUS request medical care less than their physicians propose it (4) and that, although they want more support from their physicians than do other patients, they do not seek more medical care (28). Therefore, whatever the explanation for physicians’ responses to patients’ accounts of their symptoms, offering medical care that the patient was not requesting could itself "somatize" patients’ beliefs and increase their dependence on medical care. Our findings have implications for how primary care physicians can better manage MUS. The objective of reattribution therapy is that they should encourage patients to think psychologically in order to divert consultation from somatic intervention (19,29). Consistent with this, the present study confirmed that physicians became less likely to propose somatic intervention after patients’ psychosocial presentation. Most patients with MUS already provide psychosocial cues (4,20). Therefore, reattribution training might help physicians to think more psychologically and to attend to the psychological cues that patients with MUS already provide. On this analysis, physicians need not risk the negative reactions that many believe are associated with introducing psychological explanations to patients but can, where appropriate, facilitate the patients’ existing psychological thinking. The effect of psychosocial presentations was modest in the present study, perhaps because physicians routinely disregard such cues (20), and future research should test whether reattribution training for physicians can enhance it.
Our interpretation of our findings requires several caveats. Methods for identifying cases of unexplained symptoms have varied greatly. Criteria derived from psychiatric diagnoses of somatization disorder are problematic because of poor agreement among them or poor discriminating capacity compared with psychiatric interview (30,31). A common procedure in UK studies has been to select from patients with symptoms designated by the physician as unexplained those who are psychologically disturbed and who respond to a single question by attributing their symptoms to a physical rather than psychological cause (32). This procedure restricts concern to those who manifest psychological disturbance on a screening questionnaire and assumes that patients readily distinguish physical from psychological causes. Reliance on negative investigations restricts consideration to only the minority of primary care patients in which such tests would be carried out (3). It is likely that some of the symptoms that we studied would become explained after future consultations, and only a proportion of our sample would be identified by criteria based on psychiatric categories. Moreover, it is likely that physicians differed in their readiness to regard specific symptoms as unexplained. Therefore, although our definition ensured consultations defined by each physicians’ perception that symptoms were unexplained, comparison with studies using different methods of case definition should be regarded as tentative until the relationships between these different approaches are clearer.
Generalizability is inevitably restricted. The patients were from a single region of the UK and predominantly white European. Moreover, physicians who agree to participate in such research are likely to be more psychologically minded than others. Any speech category in a coding scheme contains a degree of heterogeneity, and our findings might mask relationships with specific components of different codes. The effects that we found could be due to other variables, correlated with both patient presentation and physician decisions, that we did not measure, including nonverbal communication and patients’ history. A strength of our statistical approach is that it detects not just mean effects but heterogeneity. The effects of turn number and patients’ presentation belied significant variability among both patients and physicians. In particular, patients had varying degrees of influence on physicians’ proposals of somatic intervention, and, despite no average effect, physicians had differing responses to patients’ own somatic proposals. An aim for future work is to identify the patient and physician characteristics that underlie these sources of heterogeneity. Finally, the findings do not preclude the possibility that patients who, exceptionally, propose disease explanations or request somatic intervention beyond the levels seen in this study would thereby prompt their physicians to propose medical care.
However, it is no longer plausible that primary care physicians’ proposals of medical care for patients who they think have no physical disease should generally be attributed to overt patient pressure. Physicians’ own attributions of somatic intervention to patient pressure (18) are therefore better seen as justifying their interventions than as explaining them. The present findings therefore add to the growing view that to attribute the problems associated with MUS to patients is too limited a view. They need to be understood also as products of their context. This includes the cultural context (33,34) but also the health care systems that shape or compound patients’ problems (3,6–10,35). Research should therefore focus on physicians, as well as patients. Ultimately, improved clinical management of patients with MUS will require interventions directed at physicians (36) and particularly at the processes that lead them to provide medical care to patients who they think do not need it and who have not sought it.
We are grateful for the enthusiastic cooperation of the participating primary care physicians.
|
NOTES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODRESULTSDISCUSSIONNOTESREFERENCES |
|---|
The study was funded by the UK Medical Research Council Grant G9900294.
DOI:10.1097/01.psy.0000227690.95757.64
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODRESULTSDISCUSSIONNOTESREFERENCES |
|---|
This article has been cited by other articles:
 |
|
 |
|
  T. olde Hartman, L. Hassink-Franke, C Dowrick, S Fortes, C Lam, H. van der Horst, P. Lucassen, and E. van Weel-Baumgarten Medically unexplained symptoms in family medicine: defining a research agenda. Proceedings from WONCA 2007 Fam. Pract., July 1, 2008; (2008) cmn041v1. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 P. Salmon, G. M. Humphris, A. Ring, J. C. Davies, and C. F. Dowrick Primary Care Consultations About Medically Unexplained Symptoms: Patient Presentations and Doctor Responses That Influence the Probability of Somatic Intervention Psychosom Med, July 1, 2007; 69(6): 571 - 577. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. Persaud PATIENTS DO NOT DEMAND AND DOCTORS DO NOT MISUNDERSTAND--HOW MEDICALLY UNEXPLAINED SYMPTOMS BECOME MEDICALIZED Psychosom Med, November 1, 2006; 68(6): 993 - 993. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
