Effectiveness of the ACA (Availability, Current issues and Anticipation) training programme on GP-patient communication in palliative care; a controlled trial

This controlled trial was conducted in the context of an existing post-graduate two-year Palliative Care Peer Group Training Course (PCPT), consisting of four two-day residential courses, followed by two-hour peer group sessions with five GPs in each group, facilitated by a palliative care consultant, every six to eight weeks. All GPs enrolled in the four PCPT courses in 2006 and 2007 were invited to take part in the study. As our intervention was added to an existing training course, we had to assign whole training groups to either the intervention or the control condition. Because we wanted to start with an intervention group in 2006, and to prevent contamination between groups, GPs enrolled in the PCPT courses conducted in Eindhoven (2006) and Rotterdam (2007) were assigned to the intervention condition in which the ACA training programme was integrated into the PCPT course. GPs who enrolled in the PCPT courses in Amsterdam (both 2007) were assigned to the control condition in which the ACA training programme component was not included.

The development of the ACA training programme has been reported elsewhere [14]. The programme consists of eight steps (see Table 1) and is supported by the ACA checklist (see Table 2). Steps 1 and 2 took place on the first day of the training programme. Within two months all participants received individual feedback on their videotaped simulated consultation (step 3). During the following months they had to complete step 4 in order to formulate their personal learning goals (step 5). Six months after the start of the programme, the GPs participated in role-play exercises that were tailored to their learning goals (step 6). Finally, a second consultation with an actor simulating a patient was videotaped (step 7) to allow participants to assess their communication skills against the ACA checklist (step 8).

For calculating sample size, we used the outcome measure ‘number of issues discussed by the GP’ and considered a difference of 0.5 standard deviation (which corresponded with one extra issue discussed by the GP) between intervention and control conditions as a clinically relevant difference. Such a difference can be detected with 64 GPs in each group (power 0.80, two-sided alpha 0.05).

Outcome measures of this study were determined in discussion with a panel of experts in palliative care research. We decided to measure both how the GP communicated with the patient and what he discussed with him. These outcomes fit in well with the content of the ACA training programme on how to communicate with the patient (availability items) and what to discuss (the current and anticipated issues). Both ‘how’ and ‘what’ were measured quantitatively.

The number of issues discussed (‘what’) was defined as the summed number of 13 current and anticipated issues about which the GP made at least one utterance concerning that issue, during the simulated consultation. Additionally, we calculated for each issue the percentage of consultations in which the GPs discussed that issue.

The quality of a GP’s communicative behaviour ('how’) was defined as their scores on the six availability items. Because this complex outcome consisted of several numbers and percentages its sub-scores could not be summed up and were reported separately. Additionally, verbal dominance was calculated to evaluate whether the training influenced the GP’s dominance during the consultation.

To measure both outcomes (‘how’ and ‘what’) from the videotaped consultations we used the Roter Interaction Analysis System (RIAS) [15, 16]. The RIAS, which was developed in the United States, has been used successfully in previous studies in Dutch general practice settings [17]. It distinguishes mutually exclusive and exhaustive categories into which verbal utterances that convey a complete thought can be classified. A distinction is made between instrumental or task-oriented categories, and affective or socio-emotional categories. Task-oriented categories refer to utterances that address a patient’s physical or psychosocial problems. Affective categories carry explicit emotional content and refer to aspects of communication that are needed to establish a therapeutically effective relationship. The RIAS also rates ‘global affects’ on 6-point scales (e.g. friendliness/warmth)

For the outcome ‘number of issues discussed’ we added the current and anticipated issues to the task-oriented categories of the original RIAS. For the outcome quality of GP’s communicative behaviour we added several study-specific 6-point scales to the RIAS (e.g. the extent to which the GP took time with the simulated patient). Four of the six availability items could be scored positively (e.g. ‘taking time’) as well as negatively (e.g. ‘not taking time’). As we were especially interested in the communication by the GPs, we only calculated scores for the GPs (and not for the simulated patient).

For each GP participating in the study, we videotaped a 15-minute consultation with a simulated palliative care patient at baseline and at follow-up. The baseline assessment took place on the first day of the course; the follow-up assessment 12 months later, halfway through the two-year PCPT course. At baseline, half of the GPs from each of the four PCPT courses had a consultation with a trained actor who role-played a patient with advanced stage lung cancer. The other 50% had a consultation with an actor playing the role of a patient with advanced colon cancer. At the follow-up assessment, the simulated patient to whom the GPs were assigned was reversed from the baseline assessment. The setting in which the consultation took place was standardized to avoid any environmental variability.

The participating GPs were aware of their group allocation, but the actors involved in role-playing a palliative care patient and those who rated the videotaped GP-simulated patient encounters were not.

Coding was carried out directly from videotape by four trained raters using The Observer® software (http://​www.​noldus.​com/​). Average coding time was three to four times the duration of the consultation. Throughout the coding period, a random sample of 11.5% of the tapes was rated by all coders to assess interrater reliability. Interrater reliability averaged for the ACA issues 0.85 (range 0.68-0.99) and for the percentages of utterances with a mean occurrence greater than 2% 0.71 (range 0.56-0.89), respectively [15, 16]. These reliability estimates are comparable to those achieved in other studies [18‐21].

We assessed the comparability of the GPs in the intervention and the control condition with regard to socio-demographic and professional characteristics using the Chi-square statistic for categorical variables and the Mann–Whitney test for continuous variables. Variables on which the two groups were not comparable at baseline were entered as covariates in subsequent multivariable analyses.

We summed the number of 13 current and anticipated issues that were discussed by the GP during the simulated consultation. Consequently, the scale ranged from 0 to 13. For each issue we calculated the percentage of consultations in which that issue was discussed. For the outcome quality of communicative behaviour we calculated mean numbers and percentages of the several sub-scores. Verbal dominance was calculated by dividing the sum of all GP utterances by the sum of all patient utterances.

We used linear mixed models and accompanying effect sizes to evaluate between-group differences over time for interval level outcome variables (e.g. mean numbers and percentages). For dichotomous outcome variables (e.g. whether a given issue was discussed) we used the logistic regression method of generalized estimating equations (GEE) to account for dependence of data due to repeated measures, yielding odds ratios. In all analyses we used the GP’s sex, years of experience as GP, urban versus rural or semi-rural practice location, the actor, and duration of the consultation longer than 15 minutes as covariates. In order to adjust for multiple testing, the level of significance was set at 0.01. All data were entered and analysed in SPSS 20.0 (SPSS, Inc., Chicago, IL).

All 126 GPs eligible for this study agreed to participate. Sixty-two were assigned to the intervention and 64 to the control group (see Figure 1). GPs in the intervention group were less likely to practise in an urban location and had a few more years of experience than those in the control group. No further significant between-group differences were observed (Table 3).

We found no statistically significant differences over time between the intervention and control group in the mean total number of ACA issues, the mean number of current issues or the mean number of anticipated issues discussed (Table 4). In the total study sample, GPs raised on average eight of the 13 ACA issues during the consultation with the simulated palliative care patient (4.5 current and 3.5 anticipated issues).

A significant between-group difference over time was found only in the percentage of consultations in which the anticipated issue ‘end-of-life decisions’ was discussed: the percentage of consultations in which this issue was discussed decreased from 50% at baseline to 27% at follow-up in the intervention group, while an increase from 31% to 41% was seen in the control group (Table 4).

The four ACA issues physical complaints, psychosocial complaints, discussing treatment and care options and offering follow-up appointments were addressed in 90-100% of the consultations in both groups at baseline and follow-up measurements. Spiritual/existential issues and ‘unfinished business’ were infrequently addressed by the GPs.

No statistically significant between-group differences over time were observed in any of the outcomes related to availability, with the exception of the task-focused utterance ‘check’ (Table 5).

Verbal dominance showed no significant between-group difference over time (P=0.6), with or without inclusion of the rather frequently scored back channels (=utterances indicating attentive listening, such as ‘mmm-huh’). In both groups the verbal dominance was about 1 and decreased slightly from baseline to follow-up (i.e. GPs became slightly less dominant in terms of proportion of given utterances).

To our knowledge, this is the first study on effectiveness of a communication training programme specifically targeted at GP-patient communication in palliative care [12]. Our intervention largely meets the recommendations for communication skills training in oncology as formulated at a recent consensus meeting by Stiefel et al. [23]. Both educational approach and content of the intervention are evidence-based [14]. The outcomes of our trial were based on behavioural observations of simulated GP-patient consultations assessed by a validated quantitative instrument (RIAS).

As we had to assign participating GPs to either the intervention or the control condition without randomization, we carefully compared both groups and included significant between-group differences on background characteristics as covariates in the subsequent analyses. The GPs were not blind to their training condition. As a trial with videotaped consultations of GPs with real palliative care patients was not deemed to be feasible, we used trained actors to simulate patients with advanced stage cancer. Our study was based on the four levels of competence according to the pyramid model of Miller; 1. knows (knowledge), i.e. recall of basic facts, principles, and theories; 2. knows how (applied knowledge), i.e. ability to solve problems, make decisions, and describe procedures; 3. shows how (performance), i.e. demonstration of skills in a controlled setting; and 4. does (action), i.e. behaviour in real practice [24]. We focused our effectiveness evaluation on the third level. Moreover, we measured one 15-minute consultation, while in daily practice, Dutch GPs visit their palliative care patients frequently at home and thus discussion of the 13 issues will be spread over several visits.

We found no effectiveness studies that specifically address GP-patient communication in palliative care [12]. Two systematic reviews on effectiveness of communication training programmes for health professionals in cancer care reported positive effects (e.g. more open questions, expressions of empathy) from such training programmes [8, 9]. These health professionals (not GPs) had probably received less extensive training in doctor-patient communication as part of their educational curriculum, and therefore the baseline level of their communication skills might have allowed more room for improvement compared with the GPs in our trial. Furthermore, these studies focused primarily on ‘breaking bad news’ and ‘dealing with patients’ feelings’ surrounding diagnosis, prognosis, and treatment options, while the ACA programme is targeted at issues in palliative care and anticipating the patient’s end-of-life concerns. In previous studies the primary outcomes were typically basic communication skills such as the availability aspects of the ACA checklist, while our primary outcome included the number of current and anticipated issues discussed by GPs. In their monograph on patient-centred communication in cancer care, Epstein and Street emphasize communication skills (i.e., how to provide information) more than specific issues to be addressed [25]. In their systematic review, Parker et al. discuss in detail the specific content as well as the style of end-of-life communication; the content areas they cover are similar to those of the ACA checklist [26]. However, the ACA checklist lays more emphasis on the patient’s personal wishes, unfinished business and bringing life to a close.