Concerns voiced by patients and GPs’ responses during psychosocial visits in primary care: a historical cross-sectional study

We aim to study the role of patients during psychosocial consultations prior to and after the introduction of guidelines on psychological disorders in Dutch general practice in the 1990s. We decided to further investigate previously examined consultations that were considered psychosocial by GPs [1], and focus on whether patients share their worries and how GPs respond to these worries.

First, given the previously found decline in empathy by GPs over time, we expect to find a decline in expression of concerns or cues to underlying concerns by patients during recent consultations. Second, we aim to explore the timing of expressed worries of patients during consultations from the two time periods. Greater reluctance among patients to share their worries may also imply postponement of expressing concerns. On the other hand, active symptom exploration may motivate GPs to invite patients to share concerns during diagnostic phases of consultations. Last, we expect that GPs have become less responsive and inviting to patients’ worries.

Dutch general practice consultations were videotaped in the period from 1977 to 2008 as part of previous studies on doctor-patient communication [8]-[13]. Participating GPs were followed for at least a full day (or a series of consecutive days). An unmanned camera was placed in the consultation room to record consultations. A research assistant was present in the waiting room to ask patients’ informed consent for participating in the study and answer any additional questions. All participating GPs and patients provided some general background information in a basic questionnaire (e.g. age, gender, assessment of psychosocial aspects), which enabled the selection of consultations for additional observational research in the present study. We included consultations from six previous study samples (1977-1980, 1982-1984, 1989, 1995, 2000-2001, 2007-2008) and assigned these samples to the two periods of interest (prior to clinical guidelines: 1977-1980, 1982-1984, 1989, versus after the introduction of guidelines: 1995, 2000-2001, 2007-2008). The total database of available videotaped consultations consisted of 5,184 consultations (1,895 in the first period, 3,289 in the second period). First, the GP from each videotaped consultation assessed the degree to which psychosocial aspects determined the consultation (1 = completely somatic; 2 = mainly somatic; 3 = both somatic and psychosocial; 4 = mainly psychosocial; 5 = completely psychosocial). Second, we selected consultations assessed by the GP as being `completely psychosocial’. GPs assessed 150 consultations in the first period and 394 consultations in the second period as psychosocial. Owing to deterioration in the technical quality of some videotaped consultations, we excluded 31 consultations (28 in the first period, 3 in the second period) and one consultation was excluded because patient characteristics (e.g. age) were not available. Our analyses were conducted on 512 consultations (121 in the first period, 391 in the second period) and we specified whether psychological (N = 185), social (N = 62) or physical (N = 265) symptoms were presented by patients during these consultations by using the International Classification for Primary Care (ICPC) codes that were available for all consultations in the database.

The studies were carried out in accordance with Dutch privacy legislation. The privacy regulations were approved by the Dutch Data Protection Authority. According to Dutch legislation, approval by a medical ethics committee was not required for these observational studies. All participating physicians and patients gave their informed consent.

Patient cues and concerns were coded using the Verona coding definitions of emotional sequences, VR-CoDES-CC [14]. A cue is defined as `a verbal or non-verbal hint which suggests an underlying unpleasant emotion but lacks clarity’ (“I cannot stand it anymore”), while a concern is `a clear and unambiguous expression of an unpleasant current or recent emotion where the emotion is explicitly verbalized’ (“I feel very anxious”). GPs’ immediate responses (lag 1) to cues and concerns were coded using VR-CoDES-P [15]. We only coded lag 1 responses, which refers to the first utterance after a voiced cue or concern, while excluding delayed responses (e.g. lag 2 or 3). Responses were coded according to two major conceptual factors: explicitness (explicit versus non-explicit responses) and space provision for further disclosure of the cue of concern (space-providing versus space-reducing responses). See Table 2 for definitions and examples of the response categories.

Approximately 10% of the observed consultations were coded by the two coders involved in this study. The interrater reliability of the VR-CoDES-CC and VR-CoDES-P was found to be satisfactory to good. We calculated an intraclass correlation (ICC) of 0.56 for concerns and 0.89 for cues. The mean intraclass correlation of the GPs’ response categories was 0.71 (range 0.43 - 0.86).

Total visit duration was timed in seconds for all videotaped consultations in the database. We timed the initial opening statements by patients, which starts after GPs solicitations (What can I do for you today?) and ends when GPs initiate the next phase of exploration by asking either open-ended (“Tell me more about the head aches”) or closed-ended questions (“Is it worse in the morning or in the evening?”). The last phase of the consultation can be considered the therapeutic phase of the consultation and is characterized by GPs giving information and advice. The interrater reliability of the durations of the different phases during consultations (initial opening statements, exploration including physical examination, and giving information and advice) was found to be good. The mean intraclass correlation of the duration of the phases was 0.93 (range 0.90 - 0.99).

To account for the variation in communication skills between GPs, we used multilevel models with random intercepts (multilevel Poisson regression analysis for count variables). The multilevel models consisted of consultations (level 1) nested within GPs (level 2). The number of consultations per GP in the sample varied between 1 and 15. However, since 80% of the GPs had five or less consultations included in the present study, we could not calculate or report on intraclass correlations. We included dummy variables for both periods (1977-1989 versus 1995-2008) and examined estimated frequencies for the three types of symptoms (psychological, social and physical symptoms). First, we used multilevel Poisson regression models to estimate frequencies of communication categories by GPs per consultation. In these analyses, we included duration of consultation, patient characteristics and GP characteristics as centred covariates. Second, based on these estimates we tested whether there were differences in communication categories between the two periods.

Mean duration of consultations in the second period (1995-2008) was significantly longer than during than in the first period (see Table 3). When comparing the patient and GP characteristics between the two time periods, we found that patients and GPs were significantly older in 1995-2008 compared to 1977-1989. The gender ratios of patients did not differ significantly between the two periods, but the percentage of female GPs was higher in the second period (31% versus 7%).

In 92% of the consultations in this study, patients presented at least one cue or concern. The proportion of consultations containing at least one implicit cue did not differ between the two periods. However, the proportion of consultations containing at least one explicit concern was larger in the second period (see Table 4).

When looking at the timing of the cues and concerns expressed by patients, we see that the percentage of cues and concerns voiced during initial statements and exploration was higher in the first period compared to the second period. Patients more often expressed cues and concerns during the last phase of information and advice in more recent consultations (see Table 2).

In both periods, GPs’ responses to cues and concerns were mostly characterized by giving space for patients to talk about their concerns in a non-explicit way (see Table 5): being silent or giving minimal responses (back channels) were the most frequent ways for GPs to respond to cues and concerns. We did not find changes in response categories by GPs when comparing the two periods.

In contrast to our expectations, we did not find a decrease in cues and concerns expressed by patients over time. Patients have become more familiar with psychosocial problems such as anxiety and depression and therefore may be more willing to open up about their concerns. Information on mental health has become readily available, for example through the Internet. Internet users seem to have high levels of mental health literacy and are well able to answer knowledge-based questions regarding psychological disorders [16]. Interestingly, although the proportion of explicitly expressed concerns increased over time, the number of expressions during consultations did not change. patients’ expressiveness has become somewhat more explicit, but not more extensive over time.

Patients tend to share their worries most often during the therapeutic phase as opposed to the initial or exploration phase of consultations. Late arising cues and concerns were more likely in recent consultations, compared to older consultations. With the introduction of clinical guidelines in the 1990s, more emphasis is placed on giving information and advice [1],[2]. It is important that GPs are aware that providing new information to patients may also elicit new questions and concerns.

We found that the number of expressions of cues and concerns was partly dependent on GP characteristics. GPs in the second sample (1995-2008) were older, had more professional experience, were more often female, and were more likely to evoke more expressions of cues and concerns by patients. This is in line with previous research that shows that patients are more likely to share psychosocial issues and feel more empowered when talking to female GPs [17]. Accordingly, feminization of Dutch general practice care may partly account for changes in the way GPs facilitate communication about emotional issues.

Moreover, the GPs in our study were not likely to respond in an explicit way to patients’ concerns, either by showing empathy or by further exploration of the worries voiced. This indicates that GPs remain relatively passive listeners, rather than actively providing emotional support for their patients. Studies in psychiatric settings show similar findings; physicians often hesitated to respond and were likely to respond with passive listening skills as opposed to engaging in active emotion-focused skills [18]. In some situations indirect responses to patients’ emotions may be less intrusive and more adequate than explicit responses. However, active solicitation of concerns by GPs can prevent late arising, or even unvoiced, concerns [6]. Recent versions of clinical guidelines for depression and anxiety published in 2012 refer to the importance of empathy during consultations involving these psychological problems [19]. We consider this allusion to the importance of empathy in clinical guidelines as a positive development.

One of the strengths of this study is that we used the concerns and cues of patients as the starting point for our observations and analyses and focused on the patient’s role. Furthermore, we examined consultations using videotaped real-life general practice consultations over a thirty-year period, enabling a comparison of doctor-patient communication prior to and after the introduction of national clinical guidelines in the 1990s. Video recording is a valid method of examining doctor-patient communication: the influence of the video recorder on participants is marginal [20] and the participants were unaware that the analyses would focus on psychosocial problems.

A possible weakness of the study is that we did not include consultations in which GPs were unable to recognize or identify psychosocial issues during consultations. We decided to only include consultations assessed by GPs as completely psychosocial, since we believe that cues and concerns by patients are most likely to be clearly expressed in these types of consultations. However, to be able to generalize the results to other health problems there is a need for replication studies within other consultation types, which may include more hidden psychosocial problems. Another limitation of our study is that we did not specify the nature and content of expressed cues and concerns by patients. Therefore, we do not know for each specific expressed cue or concern whether the patient was referring to psychosocial issues. However, due to the low frequency of the cues and concerns voiced, we decided to not further categorize these as this would complicate the interpretation of quantitative analyses and we could not guarantee interrater reliability. Furthermore, we only looked at immediate responses (lag 1) of GPs after patients’ expressed cues and concerns; we did not include GPs’ delayed or random empathic responses. This may explain why the previously found decline in empathy - assessed over the entire consultation - was not found in the present study. We decided to look at lag 1 responses following expressed worries because we were interested in GPs immediate responsiveness; it can be argued that empathic responses are most appropriate and supportive when following directly after patients’ expressed worries.