doctor-patient interactions in oncology

Pergamon o277-gs360'J)ooT.t~-o

Soc. Sci. Med. Vol. 42, No. II, pp. 1511-1519, 1996 Copyright © 1996 Elsevier Science Ltd

Printed in Great Britain. All rights reserved 0277-9536/96 $15.00 + 0.00

DOCTOR-PATIENT INTERACTIONS IN ONCOLOGY

SARAH FORD, a* LESLEY F A L L O W F I E L D l and SHON LEWIS 2

~CRC Communication and Counselling Research Centre, Department of Oncology, UCL Medical College, 48 Riding House Street, London WIP 7PL, England and 2School of Psychiatry and Behavioural Sciences, Withington Hospital, West Didsbury, Manchester M20 8LR, England

Abstract--Studies which apply content analysis techniques to the cancer consultation are few. This descriptive study examines the structure and content of the bad news cancer consultations of ! 17 out- patients newly referred to the Medical Oncology Department of a large London teaching hospital. From previous communication research three main hypotheses are formed: (i) the cancer consultation is clinician-dominated rather than patient-centred; (ii) the level of psychosocial discussion between clinicians and patients is low and (iii) patient characteristics such as sex, age and prognostic category influence clinician behaviours. Each patient had two consultations with one of 5 oncologists. Both these were audiotaped with the patients' consent. The tapes were content coded using the Roter Interaction Analysis System. Results showed that clinicians tended to use closed rather than open questions. Patients asked few questions and were seldom given space to initiate discussion. Thus, the level of patient-centredness was low. Despite the fact that consultations concerned life threatening disease and often contained information regarding toxic treatment which is known to provoke psychological dys- function, the number of questions relating to patients' psychological health were few. The amount of discussion concerning medical topics from both parties was 2.5 times greater than the amount of psychosocial discussion. Although there was a suggestion in the data that 3 clinicians showed variations in behaviour according to patient age and prognostic group, the number of patients for each doctor was small. Patients were well informed about their diagnosis, prognosis and treatment options, but their emotional well-being was rarely probed. Copyright © 1996 Elsevier Science Ltd

Key words--cancer consultation, doctor-patient interactions, psychosocial exchange, patient-centredness

INTRODUCTION

Research into the interactive behaviours which occur between doctors and patients during medical encounters is essential in order to improve the outcome of patient care. Assessing strengths and weak- nesses in doctors' interaction behaviours in relation to the effects these have on patients is important for devising models of communicat ion and interviewing skills for use in general medical education. Effective communicat ion skills have been associated with beneficial outcomes for patients including, enhanced information recall [1], satisfaction [2] and improved psychological adjustment to cancer [3]. Clearly the ability to communicate effectively is fundamental to the practice of clinical medicine, but it is a skill which many doctors lack. Studies have shown that many clinicians fail to meet the informational [4, 5] and emotional needs [6] of patients with cancer.

A significant minority of patients seen by oncologists have clinically significant levels of psychological distress [7] but, as much of this goes unrecognized, it is estimated that only 15% of those

*Author for correspondence.

whose condition merits psychiatric intervention are referred [8]. One reason suggested for this lack of recognition concerns the evidence that specific aspects of clinicians' basic interviewing techniques discourage patients disclosing details concerning their psychological state. For example, avoiding eye contact and asking many closed questions about physical symptoms [9]. Behaviours found to be con- ducive to patients' disclosure of sensitive emotional information include the adoption of a reassuring and empathetic interviewing style [10] and listening to patients attentively and sympathetically [11]. Employing a patient-centred approach can help in this respect, and facilitates patient participation in the consultation. The patient-centred encounter is one in which the clinician behaves in a manner that facilitates patient expression so that he or she feels free to speak openly and ask questions [12].

Doctor-pat ient interaction and behaviour in can- eer care is an under researched area. The few existing studies offer little or no evidence to suggest that a patient-centred interview style is practised by the majority of oncologists. For example, using the Observer Checklist for coding discussions of specific treatment related topics, Siminoff et al. [13, 14] found that it was the oncologist who initiated the

1511

1512 Sarah Ford et al.

Table 1. Demographic and prognostic characteristics of patients

(1) (2) (3)

Sex Female Male

Age group* 21-39 40-59 60-75

Cancer diagnoses GTDt Testicular Breast Bowel Ovary Other

% N Partner status % N 59 (69) Married/cohabiting 72 (84) 41 (48) No partner 28 (33)

% N Socioeconomic status % N 49 (57) Professional/intermediate 37 (44) 22 (26) Skilled non-manual/manual 38 (45) 29 (34) Partly skilled/unskilled 24 (28)

% N Bad news category % N 27 (32) Primary bad news 77 (90) 17 (20) Secondary bad news 23 (27) 13 05) 06 (07) 05 (06) Prognosis group % N 32 (37) Good prognosis 49 (57)

Poor prognosis 51 (60)

*Mean age 44.6 (SD = 16.5). "t'Gestational trophoblastic disease.

discussion of most topics and patients' emotional issues were least often discussed.

The development of interaction process systems has made it possible to code and classify doctor- patient encounters in order to study their structure and content. Apart from one recent study which used a computerized coding system [15], these methods have not so far been applied to the exam- ination of consultations in oncology. Most studies of doctor-patient interaction have concentrated on general practice or primary care situations [16]. How these might apply to doctor-patient interactions in oncology is unclear as the latter may be more emotionally laden than those in general practice as the information discussed is often complex and anxiety provoking for the patient.

In this study we examine the interactions between oncologists and patients during bad news consultations. We used the Roter Interaction Analysis System (RIAS) a modification of Bale's Interaction Process System. This is a well documented and widely used system in the U.S.A. [17] and Holland [18] and was judged favourably in a comparison study [19].

Our study is mainly descriptive and exploratory being part of a larger project investigating the efficacy of audiotapes in reducing psychological distress in patients with cancer [20]. The main objective was to conduct an in-depth content analysis of the cancer consultation to find out how it was structured in terms of the type of information exchanged between patient and oncologist (e.g. biomedical vs psychosocial), the degree of physician dominance vs patient dominance and shifts in consultation style according to patient differences in terms of prognosis, sex and age. Previous primary care studies have reported that female patients tend to elicit more interpersonal involvement from doctors compared to men [21]. Higher ratings have also been given to physicians on information giving and empathic skills, in the presence of female

patients [22], and more counselling (both biomedical and psychosocial) offered to male patients [23].

On the basis of the existing research we hypoth- esized that:

(1) Rather than being patient-centred, the cancer consultation would be largely dominated by the clinician.

(2) Despite the distressing nature of the encounter the level of psychosocial discussion between clinician and patient would be low.

(3) Patient characteristics such as sex, age and prognostic category would influence specific clinician behaviours. For example, emotional responsiveness and information giving.

SUBJECTS AND METHODS

The subjects were 117 out-patients with cancer, recruited prospectively from a consecutive series of patients newly referred to the Medical Oncology Department at a large London teaching hospital. Demographic and clinical characteristics of these patients are shown in Table 1. Patients were eligible for the study if they were to be given potentially distressing information, either: (i) newly-diagnosed patients receiving 'primary bad news' of the diagnosis itself or (ii) patients with an established diagnosis in whom initial treatment had so far been unsuccessful ('secondary bad news'). Inclusion cri- teria also required a patient to be aged between 21 and 75, to be able to speak and write in English, to be free of primary or secondary brain disease and to have given written informed consent to having their consultation audiotaped.

Each patient had 2 linked consultations with one of five clinicians, (the second approximately 4 weeks after the first), during which information concerning diagnosis, treatment and prognosis was given. Both these consultations were audiotaped and clinicians instructed not to change their normal interview styles. The clinicians involved included 3 male

Doctor-patient interactions in oncology

Table 2. Interaction process categories

1513

Socioemotional/affective exchange

Social talk Social, non-medical exchange (Patient and clinician) Positive talk Laughter, agreements, approval, compliments (Patient and clinician) Negative talk Disagreements and criticisms (Patient and clinician) Emotional responsiveness Concern and reassurance/optimism, (Patient and clinician)

empathy, legitimization (Clinician only)

Task-focused/instrumental exchange

Partnership building Asking for opinion, understanding, paraphrasing, interpretation (Clinician only) Orientation Direct instructuions, setting the agenda of the visit (Clinician only) Open questions Across all topics (i.e. medical history, treatment, psychosocial and lifestyle issues (*Patient and clinician) Closed questions Same as above (Clinician only) Biomedical info Information re medical condition, treatment and side effects (Patient and clinician) Psychosocial info Information re emotional issues and lifestyle (Patient and clinician) Biomedical counselling Persuasive attempts re medical condition and treatment (Clinician only) Psychosocial counselling Persuasive attempts re emotional issues and lifestyle (Clinician only)

*All patient questions are treated as open.

consultant oncologists, a male senior registrar, and a female senior registrar.*

Content analysis

Analysis of the audiotaped consultations was conducted using the RIAS. Each discernable seg- ment of speech or verbal utterance is the unit of analysis. A unit may vary in length from a single word to a lengthy sentence. A sentence is considered one unit if it conveys only one thought or is related to one item of interest. Every utterance was assigned to one of 34 mutually exclusive content categories. These were condensed into fewer and larger groups resulting in 12 clinician and 8 patient categories shown in Table 2. The main content categories shown in this table are emotional responsiveness, partnership building, open and closed questions, biomedical and psychosocial information giving and biomedical and psychosocial counselling. A closed question is one that asks for specific information and results in either a short response or a 'yes' or 'no ' answer, for example, "Have the tablets helped your pain?". An open question is distinguished from a closed question by its non-specificity or probing nature, for instance, "Tell me how you've been feeling since I last saw you". These questions allow the responder to give an expanded answer and often begin with 'what ' , 'why' and 'how'. Information giving categories are differen- tiated from counselling categories in that they do not explicitly direct behaviour. Statements of information are delivered in a neutral manner whereas counselling statements suggest or imply some kind of action to be taken by the other person (usually the patient).

Coding was carried out directly from audiotapes rather than transcripts which allows affective utter- antes (those that transmit emotion) to be assessed such as concern and empathy. The RIAS categories are sub-divided into two main types of behaviours, namely task-focused/instrumental (e.g. giving infor-

*A U.K. senior registrar is equivalent to a U.S. senior resident.

marion, asking questions) and socio-emotional/ affective exchange (e.g. showing concern, giving reassurance). Both these types of behaviour are recognized as important in medical communication and correspond to the two main purposes of a medical consultation: information exchange, necessary for arriving at a medical diagnosis and the cre- ation of a good interpersonal relationship [24].

The main coder was trained by the creator of the system in the United States over a two week period using a coding manual and training tapes. This coder was responsible for coding and analyzing all the tapes collected. A random sample of 10% of the tapes (N=20) were double coded by a second coder who was given intensive training by the main coder to assess inter-coder reliability. Pearson correlation coefficients between the two coders were calculated for all categories with a mean frequency greater than 2. The overall average correlation was 0.78. Clinician categories had a mean correlation of 0.77 (range 0.60-0.92) and the average correlation for patient categories was 0.80 (range 0.464).92). The lowest patient category correlation was for 'giving treatment/test information' which was one of the categories amalgamated to form the 'biomedical information giving category'. The reasons for this particular category having a modest coefficient the authors find difficult to explain as all other patient categories had high levels of reliability (0.82 and above). Other researchers using the RIAS have reported similar levels of inter-coder reliability. For example Bensing and Dronkers [18] reported Pearson correlations ranging from 0.76 to 0.99 for the clinician categories and between 0.67 and 0.99 for patient categories. Similarly Roter et al. [23] achieved overall levels of 0.76 (range 0.58--0.90) for clinician categories and 0.81 (range 0.71-99) for patient categories occurring more frequently than 2%. As Roter and other investigators have noted inter-coder reliability could not be calculated confi- dently for under-utilized categories due to the effect of skewed data on correlation coefficients. Those


Table 3. Under-utilized RIAS categories

Interview 1 Interview 2 Patient Mean/Range Mean/Range

Affective categories Compliments 0.99 (0-7) 0.22 (0-11) Shows concern 1.50 (0-14) 0.96 (0-13) Shows optimism 0.46 (0-7) 0.71 (0-6) Shows disapproval 0.19 (0-5) 0.18 (0-8) Exhibits criticism 1.08 (0-12) 0.71 (0-1 I) Asks for reassurance 1.74 (0-14) 1.62 (0-18)

Tasked-focused categories Seeks understanding 0.55 (0-1 I) 0.17 (0-4) Questions--lifestyle 0.42 (0-7) 0.16 (0-3) Questions--psychosocial 0.13 (0-2) 0.02 (0-2)

Clinician Affective categories Compliments 0.39 (0-15) 0.06 (0-5) Empathy 1.00 (0-14) 0.28 (0-7) Legitimises 0.13 (0-8) 0.15 (0-5) Shows disapproval 0.19 (0-2) 0.13 (0-2) Exhibits criticism 0.18 (0-3) 0.13 (0-5)

Tasked-focused categories Asks for opinion 0.49 (0-2) 0.17 (0-7) Open questions treatment 0.46 (0--6) 0.19 (0-3) Open questions lifestyle 0.29 (0-7) 0.04 (0-2) Open questions psychosocial 0.49 (0-4) 0.44 (0-5) Closed questions psychosocial 0.37 (0-5) 0.24 (0-3)

categories with mean frequencies of less than 2 are listed in Table 3.

RESULTS

A total of 113 (out of a possible 117) tapes from consultation 1 and 95 from consultation 2 were available for analysis, During consultation 1 there were 4 tape failures and prior to the second 18 patients dropped out of the study: 4 died, 4 refused to remain in the study, 5 were not re-interviewed as they required no further treatment, 1 developed

brain disease and a further 4 tape failures occurred. The mean duration of patients' first consultation was 30 minutes (range 10-78) and 14 minutes (range 4-34) for the second. There were significant differences between the average consultation times for the 5 clinicians during consultation 1 only. A Seheffe range test found clinician B's time to be significantly longer than clinicians' A, D and E and clinician C's consultation time to be significantly longer than clinician D's. Clinician B who had with the longest mean consultation time spent on average 11 minutes longer with each patient (F=8.4, P = < 0.0000). Seventy-three percent of patients saw the same doctor for both consultations.

The structure of the cancer consultation

The percentaged breakdown of the structure of consultations 1 and 2 is shown in Table 4. During consultation 1 the most dominant category for both doctor and patient was 'biomedical information giving'. Although over 40% of patient utterances were concerned with giving biomedical information, clinicians still had a higher number of total biomedical utterances. The next major category for patients was agreements (27%) and for clinicians biomedical counselling (14%), followed by closed questions relating to biomedical issues (11%). The amount of psychosocial/lifestyle information given by patients (0.3%) and the percentage of psychosocial counselling (3%) and questions from the clinician (3%) was low compared to questions (14%), information (31%) and counselling (14%) relating to biomedical topics. Patient biomedical question asking amounted to 5% of their total utterances. Clinicians tended to use closed rather than open

Table 4. The structure of the cancer interview

Breakdown of Total Utterances--Interview 1 Patient categories % N Clinician categories % N

Biomedical information 41 7362 Biomedical information 31 8281 Agreements 27 4848 Counselling biomedical 14 3739 Gives psychosocial/lifestyle 14 2 5 1 3 Biomedical closed Qs 11 2938 Biomedical questions 5 897 Checking/reassurance 11 2938 Psyehosocial/lifestyle Qs 0.3 53 Agreements 9 2404 Other 13 2334 Instructions/orientation 5 1335

Biomedical open Qs 3 801 Psychosocial/lifestyle O/C Qs 3 801 Counselling psychosocial/lifestyle 3 801 Other 10 2671

Total patient utterances (40%) 17,956 Total clinician utterances (60%) 26,712

Breakdown of Total Utterances--Interview 2 Patient categories % N Clinician categories % N

Agreements 32 2328 Biomedical information 30 2985 Biomedical information 30 2183 Counselling biomedical 21 2089 Gives psychosocial/lifestyle 9 655 Checking/reassurance 12 I 194 Biomedical questions 9 655 Biomedical closed Qs 7 696 Psychosocial/lifestyle Qs 0.2 14 Agreements 5 497 Other 20 1 4 5 5 Instructions orientation 5 497

Biomedical open Qs 2 199 Counselling psychosocial/lifestyle 2 199 Psychosocial/lifcstyle O/C Qs 2 199 Other 14 1393

Total patient utterances (42%) 7278 Total clinician utterancas (58%) 9951

Doctor-patient interactions in oncology

Table 5. Patterns of interaction identified in the cancer consultation (numbers are m e a n utterances)

1515

Biomedical restricted Interview 1 (n = 38) Patient Clinician Interview 2 (n = 65) Patient Clinician

Questions--all topics 5.00 18 .68 Questions---all topics 3.45 5.8,1 Biomed information 34,79 34.92 Biomed information 14.60 13.38 Psychosocial discussion 10.42 3. I 1 Psychosocial discussion 4.25 I. 18

Biomedical expanded Interview 1 (n = 52) Patient Clinician Interview 2 (n = 10) Patient Clinician

Questions--all topics 7.48 36 .21 Questions--all topics 8.56 11.00 Biomed information 81.38 44.08 Biomed information 57.22 26.67 Psychosocial discussion 23.85 2.90 Psychosocial discussion 15.33 1.56

Consumerist Interview I (n = 22) Patient Clinician Interview 2 (n = 13) Patient Clinician

Questions--all topics 13.73 28 .09 Questions--all topics 9.50 8.14 Biomed information 54.64 148.82 Biomed information 18.50 39.55 Psychosocial discussion 18.50 13.50 Psychosocial discussion 7.64 2.23

questions for all topics and their low level of agreements, i.e. attentive responses (9%), indicates that patients were seldom given space to initiate discussions. The ratio of all clinicians' open questions vs closed was 0.21. For every open question there were 4 closed questions. During consultation 2 the ratio increased, rising by 28% to 0.27 (1 open for every 3 closed questions). The ratio of total patient vs total clinician utterances was 0.67 during consultation 1, rising by 9% to 0.73 during the second consultation 2.

During consultation 2, biomedical information formed 30% of all utterances for both doctor and patient, but the total number of biomedical utterances from doctors was still larger than those for patients. Once again the number of patients' agreements was high (32%) and those for clinicians' were low (5%). Patients asked 4% more biomedical questions than in their first consultation. However, there still remained a large gap between the amount of clinician biomedical questions (9%) and counselling (21%) compared to psychosocial based probing (2%) and counselling (2%).

Three other patient-clinician ratios were calculated to examine the following: (1) patient direction vs clinician direction of the consultation, (2) patient-centredness and (3) clinician-patient psychosocial exchange vs biomedical exchange. The first ratio of effective communication control was based on patient's biomedical questions and biomedical information from the clinician divided by clinician's closed biomedical questions, directions and biomedical information from the patient. For consultation 1 this ratio was 0.85 which increased by 42% to 1,21 during the second consultation as patients gained more control. Ratio 2 comprised all patients' questions, psychosocial/lifestyle discussion, clinicians' partnership building statements and psychosocial/lifestyle discussion divided by clinicians' closed questions, biomedical information giving and patients biomedical information giving. This" ratio was 0.33 at the first consultation, rising by 24% to 0.41 at consultation 2 indicating a low level of patient centredness during both consultations. The

ratio of psychosocial to biomedical exchange for clinician and patient was 0.20 at consultation 1 fall- ing by 25% to 0.15 at consultation 2.

Patterns of interaction

Hierarchical cluster analysis was carried out as another measure of patient centredness. The vari- ables of interest entered into the analysis for both doctor and patient were: (1) closed questions (all topics), (2) biomedical information giving and (3) psychosocial discussion. These groupings represent distinct functions in the medical dialogue and are those most often considered in the patient-centred vs physician-centred consultation debate [25]. On the basis of the mean scores of the 6 key vari- ables entered into the analysis, each cluster was assigned a name reflecting its distinguishing characteristics. Three main clusters were identified for both consultations 1 and 2. These were biomedical restricted, biomedical expanded and consumerist (shown in Table 5). The biomedical restricted category is distinguished by the equal levels of biomedical information exchange, high levels of clinician question asking (e.g. qs there a history of cancer in your family?', 'When did you first notice that something was wrong?') and low levels of psychosocial discussion. The biomedical expanded pattern is discernable by the large amount of patient biomedical information giving vs the clinician. Finally, the consumerist group is characterized by high levels of clinician information giving, more patient questions (e.g. 'how long will the treatment last?', 'Will I lose my hair?') and a greater amount and a more equal division of psychosocial discussion with the patient contributing slightly more (e.g. Patient: 'I 'm so frightened'. Doctor: 'Is there anything in particular that makes you feel frightened?' Patient: 'Everything about cancer is frightening'). There was no evidence of an independent psychosocial cluster. These three main types of interaction patterns accounted for 96% of the 113

! 5 ! 6 Sarah Ford et al.

consultations analyzed from stage 1 and 93% of the 95 consultations from stage 2.

Differences in level of specific interview behaviours for each clinician

Analyses of covariance (controlling for consultation length) were carried out to identify differences in the frequencies of four main interviewing behaviours for each of the 5 clinicians during consultation 1. The behaviours were emotional responsiveness, partnership building, psychosocial/lifestyle discussion and biomedical information giving. There were significant differences in mean frequencies for all these behaviours. Clinician C was more emotionally responsive and provided more biomedical information and psychosocial discussion than the other clinicians. Clinician E offered a higher level of partnership building behaviour compared with the other clinicians. Oneway analyses of var- iance were carried out to test for the same effects during consultation 2. Significant differences emerged for emotional responsiveness and biomedical information giving for clinician C and partnership building behaviours from clinician E. The results of these tests for both consultations are shown in Table 6.

Clinician behaviours and patient characteristics

Non-parametric Mann-Whitney U tests were car- fled out to examine the same four interviewing behaviours for each clinician according to patients' sex, age and prognostic group. Unfortunately, it was not possible to investigate bad news category, i.e. primary bad news vs secondary bad news for each clinician as numbers were very small. From the outset the ratio of primary bad news patients to those receiving secondary bad news was 3:1 and 4 of the latter group died prior to the second consultation. Furthermore, it was not possible to analyze the mean behaviours for all clinicians and bad news group collectively due to the significant differences (reported above) between specific interview behaviours and individual clinicians.

Due to the small numbers of patients for each doctor, age was split above and below the median to form two groups viz: 21-40 and 41-74. For consultation 1, significant differences were found for Clinicians A, B and C, but not for D and E. Clinician A gave more biomedical information to poor prognosis patients than good prognosis patients (median=37 vs 18, Z=-2.2, P=0.02). All but one of these patients were aged between 21 and 40 years. Clinician B offered more biomedical information (median=83 vs 30, Z=-2.3, P--0.02) and was more emotionally responsive (median 26 vs 9, Z=-2.3, P=0.02) to good prognosis compared to poor prognosis patients although numbers were very small (N=4 and 12 respectively). Clinician B also exhibited more emotional responsiveness (median 15 vs 8, Z=-2.7, P=0.006), psychosocial/ lifestyle discussion (median 6 vs 1, Z=-2.2, P = 0.02) and biomedical information (median 71 vs 26, Z=-2.0, P=0.04) to patients aged between 21 and 40. Clinician C offered more partnership building to poor prognosis patients (median 18 vs 11, Z=-2.0, P=0.03) and those aged 41-74 (median 17 vs 11, Z= - l . 9 , P=0.04). Once again numbers of patients were very small (4 poor prognosis vs 20 good prognosis). The 4 patients in the poor prognosis group were the same individuals as those aged between 41 and 74. It is therefore impossible to say if these patients received more partnership building statements because they had a poor prognosis or were aged between 41 and 74. Clinician C also gave more biomedical information to patients with a good prognosis (median 152 vs 98, Z=-2.0, P=O.04) and those aged between 21 and 40 (median 152 vs 74, Z=-2.3, P=0.02). As before, the patients in these two groups are the same indicating the presence of an interaction effect between prognosis and age group. Due to the diminutive cell sizes it is not possible to explore this issue further. Only one significant difference was found for consultation 2. This was for clinician A who continued to offer more biomedical information to poor prog-

Table 6. Differences in frequencies of clinician's interviewing behaviours

Clinician Consultation l--Bchaviours, mean frequencies and ranges

*N Emotional responsiveness Partnership building Psychosocial discussion Biomedical information

A (24) 06 (2-19) 10 (3-27) B (16) 13 (1-30) 20 (3-35) C (24) 43 (12-125) 13 (3-29) D (20) 13 (0-28) 14 (5-28) E (29) 10 (0-30) 21 (4-48)

(F=25.3; P= > 0.000) (F=7.1; P= >0.000)

04 (0--19) 32 (9-74) 06 (0-25) 52 (7-143) 19 (0--128) 145 (36-321) 03 (0-15) 64 (23-143) 05 (0-24) 29 (5-70)

(F=6.7; P= >0.000) (F=45.7; P= > 0.000)

Clinician Consultation 2--Behaviours, mean frequencies and ranges

*N Emotional responsiveness Partnership building Psychosocial discussion Biomedical information

A (16) 03 (0-10) 02 (0-8) 0.6 (0-4) B (10) 08 (0-18) 06 (0-23) 03 (0-6) C (31) 15 (1-51) 06 (0-23) 04 (0-21) D (18) 08 (0-20) 08 (2-19) 02 (0-12) E (20) 09 (1-27) 12 (3-28) 02 (0-10)

(F=6.7; P=0.0001) (F=9.5; P= >0.0000) (F= 1.6; P=0.16)

19 (0-37) 24 (0-66) 37 (4-149) 24 (.5-80) 20 (3-38)

(F= 2.3; P=0.05)

*N= Number of patients per doctor.

Doctor-patient interactions in oncology 1517

nosis (N=3 , median 31) than good prognosis patients (N= 13, median 16; Z=-2.0, P=0.03).

DISCUSSION

The main aim of this study was to examine the structure of the 'bad news' cancer consultation. In addition to identifying the types and frequency of information exchanged between patient and clinician, three main areas were examined. These were the degree of patient centredness in the consultation, level of psychosocial exchange and variations in clinicians' interviewing style in relation to specific patient characteristics, namely sex, age and prognostic group.

The most frequent behaviour that dominated consultations was, not surprisingly, biomedical information-giving from both clinicians and patients. Patients, on the whole, received a large amount of information from their clinicians about their diagnosis, prognosis and treatment options. They also received much reassurance and counselling on biomedical topics, but although frequently queried on whether or not they understood what was being conveyed to them, patients were rarely given 'space' to respond fully to these queries. The few numbers of patient biomedical and psychosocial questions and the high percentage of agreements (verbal head nods), indicates that patients were attentively listening to the large volume of information being conveyed to them with little opportunity for asking questions. Furthermore, levels of psychological probing and counselling from clinicians were very low, especially considering these were predominantly 'bad news' consultations. Some patients did disclose their psychosocial concerns, but in the majority of cases these were not pursued by clinicians who kept to a rigid biomedical agenda. Clinicians tended to rely on the use of closed questions across all topics especially during history taking. This tech- nique limits the range of patients' responses and discourages the disclosure of information. Open questions give patients more space to respond and can often cut down the total number of questions required to elicit specific information. The over use of closed questions can make patients defensive [26] actively deterring them from becoming involved in the interaction. Another discouraging factor, not apparent from the content analysis, but very audi- ble, was the use of a computer by the clinician during the consultation. This particular teaching hospital has recently introduced a data management policy, whereby oncologists are responsible for up- dating patients' records and making their own appointments. This is carried out using mobile computer terminals which the clinicians bring with them to each clinic. Consequently, data are often being entered by clinicians whilst patients are speaking. The lack of eye contact resulting from this practice can communicate a lack of interest to the patient.

The ratio of patient--clinician communication control was high at consultation 1. Clinicians asked 4 times as many biomedical questions than patients. This ratio improved slightly during the second consultation due to a decrease in the number of clinician biomedical questions and hence a reduction in the level of biomedical information-giving from patients. The level of patient-centredness was very low during consultation 1. This was mainly due to clinicians asking 4 times as many questions (all topics) than patients and the amount of biomedical discussion from both parties was 2.5 times more than the amount of psychosocial discussion and partnership building behaviour. This ratio improved during,the second consultation not because of an increase in psychosocial discussion, but due to decreases in the number of closed questions from clinicians and an increase in patients' biomedical questions. The ratio of psychosocial vs biomedical exchange was the smallest of all with 4.5 times more biomedical than psychosocial information being exchanged during consultation 1 and 6 times as much during the second consultation. Empathic responses towards all patients were very few (as listed in the under-utilized RIAS categories) having a mean frequency of 1.

The three patterns of interaction identified in the cluster analysis, namely, biomedical restricted, biomedical expanded and consumerist reflect clinician- dominated interviewing styles. These interaction groups have previously been reported in a primary care study [27]. In that study two other distinct patterns were noted these were biopsychosocial (containing moderate levels of all categories) and psychosocial (dominated by psychosocial discussion). In the present study there was no evidence of either a biopsycho or psychosocial cluster. This is due to the strong medical bias in interviewing styles of the five clinicians and emphasizes their general failure to explore psychosocial issues.

The differences in frequencies of the four main interviewing behaviours displayed by clinicians E and C draw attention to the fact that it can not be assumed that patients with cancer will experience the same consultation style no matter which clinician they are referred to. For example, clinician C was more emotionally responsive than the other clinicians, whilst clinician E exhibited more partnership building behaviours than the other clinicians. Unfortunately due to the extremely small cell sizes and clinicians having unequivalent case-mixes we are unable to draw any firm conclusions concerning the effect of patient characteristics on clinician behaviours. For example, it seems likely that those patients who were receiving an initial diagnosis of cancer would receive more information than those with recurrence of disease, the latter group having acquired the necessary biomedical information from previous visits. Unlike previous studies we found no evidence of any behavioural differences according


to patient sex, although this may be due to sample deficiencies per doctor. Three of the clinicians displayed differences in at least one category of behaviour according to age and/or prognostic group. It seems likely that there is an interaction effect between prognostic group and age group. However, the size of our data set prevents us from itavesti- gating this possibility further and is, therefore, a topic for future research.

In summary, patients in this study were given a large volume of biomedical information and received much counselling concerning their medical condition. However, there were few open questions across all topics and the lack of psychosocial questions from clinicians demonstrates how emotional issues remained mostly uninvestigated. The type of information exchanged between patient and oncologist was predominantly biomedical rather than psychosocial. Patients were rarely given 'space' to express their feelings and initiate discussions, hence the level of patient-centredness was low. We h.ave no reason to suppose that the clinicians in this study behaved any differently from oncologists in general in the U.K. Recent communications skills workshops held for senior oncologists have shown that discussing psychosocial concerns was acknowl- edged as a major problem area. Few medical schools teach clinicians how to incorporate such issues into an ordinary clinic situation [28].

In a recent study Wissow et al. [29] found that three simple communication skills exhibited by pae- diatricians were associated with the increased disclosure by mothers' of sensitive information concerning the mental health of their children. These were asking questions about psychosocial issues, making supportive statements and listening attentively. An appropriate avenue for future communication research in oncology would be to conduct a similar study to ascertain if behaviours such as these could he employed by oncologists to facili- tate patient disclosure of emotional problems. Another vital yet neglected area is the influence of specific communicative behaviours on patients' quality of life and coping strategies. Of prime im- portance is the implementation of the findings of doctor-patient interaction studies into communication and interviewing skills training programmes for use in general medical education and practice.

Acknowledgements--The authors are grateful to the clinicians and patients from the Department of Medical Oneology, Charing Cross Hospital. Special thanks go to Dcbra Roter for her general advice and instruction concerning the RIAS, and to Angela Hall for her help in double-coding the tapes. The original data were collected by Patrieia McHugh to whom we also owe our thanks. This study was funded by a Cancer Research Campaign grant awarded to Dr Fallowtield and Professor Lewis.

REFERENCES

1. Roter D. L., Hall J. A. and Katz N. R. Relations between physicians' behaviours and analogue patients' satisfaction, recall, and impressions. Med. Care 25, 437, 1987.

2. Bertakis K. D., Roter D. and Putnam S. M. The relationship of physician medical interview style to patient satisfaction. J. Family Practice 32, 175, 1991.

3. Fallowfield L. J., Hall A., Maguire G. P. and Baum M. Psychological outcomes of different treatment policies in women with early breast cancer outside a clinical trial. Br. med. J. 301, 575, 1990.

4. Sutherland H. J., Llewellyn-Thomas J. A., Lockwood G. A., Tritchler D. L. and Till J. E. Cancer patients: their desire for information and participation in treatment decisions. J. R. Soc. Med. 82, 260, 1989.

5. Fallowfield L., Ford S. and Lewis S. Information pre- ferences of patients with cancer. Lancet, 344, 1576, 1994.

6. Maguire P. and Faulkner A. How to improve the counselling skills of doctors and nurses in cancer care. Br. meal. J. 297, 847, 1988.

7. Ford S., Fallowfield L. and Lewis S. Can oncologists detect distress in their out-patients and how satisfied are they with their performance during bad news consultations? Br. J. Cancer 70, 767, 1994.

8. Magnire P., Tait A., Brooke M., Thomas C. and S¢llwood R. Effect of counselling on the psychiatric morbidity associated with mastectomy. Br. med. J. 2~1, 1454, 1980.

9. Davenport S., Goldberg D. and Millar T. How psychiatric disorders are missed during medical consultations. Lancet 2, 439, 1987.

10. Cox A., Holbrok D. and Rutter M. Psychiatric interviewing techniques, VI: experimental study: eliciting feelings. Br. J. Psychiat. 139, 144, 1981.

11. Putnam S., Stiles W., Jacob W. and James S. Teaching the medical interview: an intervention study. J. Gen. Intern. Med. 3, 38, 1988.

12. Stewart M. What is a successful doctor-patient interview? A study of interactions and outcomes. Soc. Sci. Med. 19, 167, 1989.

13. Siminoff L, A., Fetting J. H. and Abeloff M. D. Doctor-patient communication about breast cancer adjuvant therapy. J. clin. Oncol. 7, 1192, 1989.

14. Siminoff L. A. and Fetting J. H. Factors affecting treatment decisions for a life-threatening illness: the case of medical treatment of breast cancer. Soc. Sci. Med. 32, 813, 1991.

15. Butow P. N., Dunn S. M., Tattershall M. H. N. and Jones Q. J. Computer based interaction analysis of the cancer consultation. Br, J. Cancer 71, 1115, 1995.

16. Roter D. U and Hall J. A. Studies of doctor-patient interation. Ann. Rev. Publ. Hlth 10, 163, 1989.

17. Roter D. U The Roter Method of Interaction Process Analysis. RIAS Manual, Johns Hopkins University, Baltimore, 1991.

18. Bensing J. M. and Dronkers J. Instrumental and affective aspects of physician behavior. Med. Care. 30, 283, 1992.

19. Inui T. S., Carter W. B., Kukull W. A. and Haigh V. H. Outcome-based doctor-patient interaction analysis, 1: Comparison of techniques. Med. Care 20, 535, 1982.

20. MeHugh P., Lewis S. W., Ford S. T., Newlands E. S., Rustin G. J., Coombes R. C., Smith D. B., O'R¢illy S. M. and Fallowfieid L. J. The efficacy of audiotapes in promoting psychological wellbeing in cancer

Doctor-patient interactions in oncology 1519

patients: a randomised, controlled trial. Br. J. Cancer 71, 338, 1995.

21. Waitzkin H. Doctor-patient communication: clinician implications of social scientific research. J. Am. reed. Ass. 252, 2441, 1984.

22. Hooper E. M., Loretto M., Comstock M. S. Goodwin J. M. and Goodwin J. S. Patient characteristics that influence physician behaviour. Med. Care 20, 630, 1982.

23. Roter D., Lipkin M. and Korsgaard. Sex differences in patients' and physicians' communication during primary care medical visits. Med. Care 29, 1083, 1991.

24. Bensing J. M. Doctor-patient communication and the quality of care. Soc. Sci. Med. 32, 1301, 1991.

25. Stewart M. A. and Roter D. L. Introduction. In Communicating with Medical Patients (Edited by Stewart M. A. and Roter D. L.). Sage Publications, Newbury Park, CA, 1989.

26. Clin¢ R. J. Interpersonal communication skills for enhancing physician-patient relationships. Md State Med. J. 32, 273, 1983.

27. Roter D., Lipkin M., Stewart M. and Putnam S. Physicians' communication patterns in primary care. Unpublished manuscript. 1991.

28. Fallowfield L. J. Can we enhance the personal and professional fulfillment of doctors in cancer medicine7 Br. J. Cancer71, 1132, 1995.

29. Wissow L. S., Roter D. L. and Wilson M. E. H. Pediatrician interview style and mothers' disclosure of psychosocial issues. Pediatrics 93, 289, 1994.

doctor-patient interactions in oncology

Documents