Adding a second surprise question triggers general practitioners to increase the thoroughness of palliative care planning: results of a pilot RCT with case vignettes

Between April and July 2016, we invited a random sample of 140 GPs in the south-eastern part of the Netherlands to participate in our pilot vignette study. Because the second SQ had never been studied before, we performed a pilot RCT. We estimated that about 30% of invited GPs would be willing to participate. Thus, by inviting 140 GPs to participate, we would have about 10 participating GPs per group. This number was expected to be sufficient to answer the research questions, keeping in mind the explorative character of this innovate study. In order to prevent socially desirable answering, we told participating GPs that the purpose of the study was to explore the influence of the type of patient illness on GP’s care planning, without mentioning that we were also interested in the influence of the SQs. For the same reason, we did not mention that they would be randomized. Forty-three GPs provided written informed consent, and were randomized to one of the four arms, after stratification for being a “specialized GP” (a GP who has completed an extensive additional course in Palliative care or in Elderly care). In each group, the GPs received a questionnaire with four vignettes [Additional file 1]. Group 1 was used as a control group. Group 2 received only SQ1 after each vignette and in group 3 both SQs (SQ1 and SQ2) were asked after each vignette.

Group 4 received an aid for anticipatory, multidimensional care planning after each vignette in addition to both SQs. However, this concerned another research question; for that reason, in this paper we will only report the results of groups 1–3.

All participating GPs received the randomization-assigned questionnaire online through Castor EDC (a valid research database software program), or by mail, according to their preference.

Each questionnaire contained four vignettes representing four different patient cases. Vignette A portrays a patient at an advanced stage of a chronic condition (chronic heart failure and chronic obstructive pulmonary disease), vignette B a patient with terminal metastatic pancreatic cancer, vignette C a frail lady with dementia and vignette D a relatively young patient with a recently diagnosed metastatic colon carcinoma. These vignettes were inspired by real anonymized cases from one of the authors (CV). [Additional file 1].

With each vignette, the respondent was asked if he would plan any kind of care for this particular patient. If so, they were asked to describe what instructions they would give to their young and inexperienced GP-trainee (who therefore needs explicit instructions) and about what kind of care needs to be planned for this patient in the upcoming period. At the end of the questionnaire, GPs in the corresponding groups were asked their opinions about the applicability and usefulness of, respectively, SQ1 and SQ2 [Table 1].

All invited GPs were asked to provide socio-demographics and palliative care experience-related data [Table 2]. All information provided in the questionnaire was collected anonymously in Castor EDC and blindly evaluated. As a scoring instrument did not exist to assess whether all requirements of the WHO palliative care definition are fulfilled, we created a scoring instrument in order to analyze the open text fields and quantify them, dubbed the RADboud Indicators of ANTicipatory care (RADIANT) score [Table 3]. Therefore, we searched for guidelines from Pallialine (a Dutch website with an evidence and consensus based collection of palliative care guidelines, powered by the Netherlands Comprehensive Cancer Organization; IKNL), requirements from the WHO concerning Palliative Care and information from thuisarts.nl (a Dutch website for patients, powered by the Dutch College of General Practitioners, with clear information about various subjects including treatment limitations and end-of-life preferences) and a list of discussion points about patients’ end of life, powered by the Royal Dutch Medical Association [3, 24‐26].

The primary outcome was the correlation between answer combinations for SQ1 and SQ2 (“Yes/No I would / would not be surprised if this patient were to die in the next 12 months” and “Yes/No I would / would not be surprised if this patient is still alive after 12 months”) and RADIANT scores. If an answer combination of SQ1: “No” and SQ2: “Yes” correlates with higher RADIANT scores than answer combination of SQ1: “No” and SQ2: “No”, it suggests that this answer combination prompts GPs to plan anticipatory care more thoroughly.

Continuous, normally distributed data were expressed as means and percentages. Because of the small numbers, no effect sizes were calculated.

The study was approved by the research ethics committee of Radboud university medical center in accordance with the Medical Research Involving Human Subjects Acts (WMO), case number 2016–2716. We obtained written consent from all participants to report individual anonymized data.

Of the 140 GPs approached, 57 did not want to participate and 40 never answered our invitation; thus in total 97 GPs did not provide informed consent [Fig. 1]. Fourty-three GPs agreed to participate and were randomized into groups 1–4 and sent the corresponding questionnaire. Of those 33 were in groups 1–3. Twenty-eight GPs completed the questionnaire and 5 did not, mostly due to high workload or personal circumstances [Fig. 1]. Of these 28 GPs, nine were in group 1, ten in group 2, nine in group 3. Their characteristics are included in Table 2.

Characteristics of participants in the different groups were mostly similar. However, the percentage of males in group 2 was lower and the percentage of GP-trainers in group 1 was lower. The mean age was 49 (SD 9.2) and they had an average GP work experience of 18 (SD 9.4) years. Four GPs worked in an individual practice (14%), 11 GPs (39%) in a two-man practice and 13 (46%) in a group practice. Only one of them (4%) completed an additional course in palliative care or in elderly care and 14 (50%) were GP-trainers. Participants treated a mean of six palliative patients a year, and GPs rated their interest in palliative care as an 8.0 (scale 0–10; 0: not at all-10: extremely interested) and their self-assessed expertise in palliative care as a 7.2 (scale 0–10; 0: no expertise-10: extremely high expertise).

In group 3, GPs who answered that they would not be surprised if this patient were to die in 12 months, and would be surprised if this patient is still alive after 12 months, had higher RADIANT scores than GPs in the same group who answered that they would not be surprised if this patient were to die in 12 months but they would also not be surprised if this patient is still alive after 12 months [Table 4].

We did not find differences between GPs groups 1–3 in Radiant scores. Regarding vignettes A, B and D, group 3 has lower scores than group 1 and 2 [Table 5]. In all groups, vignette B (end-stage metastatic pancreatic cancer) had the highest mean score, followed by vignettes A (advanced organ failure) and C (frailty), and the lowest on vignette D (recently diagnosed metastatic colon carcinoma). In group 3, those vignettes of whom the GPs would be surprised if they were still alive after one year had mostly higher Radiant scores than the others. [Table 5].

GPs mentioned specific treatment limitations least often in their care plans and in only three care plans (3%) GPs mentioned that they would discuss how to achieve patients’ personal goals. There were no striking differences between groups 1–3, except for involving other disciplines (50% of the case vignettes in group 1, compared to 20% for groups 2 and 3) and involving family and loved ones in planning care (36% of the case vignettes in group 1, compared to 22% and 25% in groups 2–3 respectively).

Many GPs mentioned that, in general, the diagnosis of a disease without curative options prompts them to start a palliative care approach. Their intuition, or information from a medical specialist, that death is approaching are considered important triggers. They also mentioned that they are prompted by frailty, complex situations, or when a patient suddenly or gradually has increased care needs. These latter situations give a GP the feeling of lagging behind events and of no longer being able to act proactively: “the feeling of being overtaken by events.” Also, the absence of a family caregiver or lack of a social network prompts them to initiate palliative care. Some GPs especially want to adhere to the patient’s wishes and start palliative care when the patient himself asks for it or gives other signals of being open to a conversation about palliative care: “the tangibility of the finiteness of life and the patient’s capacity to feel this too.” One GP in group 1 described SQ1 as an important trigger, another GP in group 1 mentioned SQ2, although they were not offered these questions and SQ2 has not been described before.

This is the first study in which the SQ is used in combination with case vignettes to see whether it prompts GPs to thoroughly plan proactive palliative care, instead of studying its prognostic value, which has been questioned before [31]. Adding a more specific trigger, SQ2, is also innovative.

However, the study also has several limitations. In this pilot study, the number of included GPs was limited and not sufficient to discover statistical significance. It was meant as a first exploration and therefore all results should be interpreted with caution. Although we aimed to use the results for a full-scale study, the relatively small percentage of GPs that wanted to participate in this vignette study led us to abandon this plan. Instead, we have started a prospective study in which GPs screened their patient list with both SQs. One year later, a retrospective medical record review will take place to determine sensitivity and specificity of both SQs and whether the different answers to these SQs are related to anticipatory care planning and care needs.

The response rate of 22% is about the same as in Schnakenberg et al.’s study (27.5%). In that study, GPs in Germany and Sweden were sent surveys about palliative care, which took no more than ten minutes to fill out [32]. Although we consider 22% fairly low, given the time required to complete the questionnaire (30–40 min) and Dutch GPs’ high workload, it is satisfactory.

Vignette 4, concerning the patient with recently diagnosed metastatic colon carcinoma, was not completed by two GPs and also had the lowest RADIANT score. We think that mainly the case itself will have influenced the score, but maybe the order in which the vignettes were presented might have influenced this outcome as well.

Although the RADIANT score is not a validated tool, it enabled us to quantify the open text fields and thus to compare groups not only qualitatively but also quantitatively. We chose anticipatory care to have a high share of the total RADIANT score, because we believe a medical professional can only provide good palliative care when, in addition to addressing actual problems, he also anticipates possible future problems. Besides, a medical professional can only be proactive when he is made aware of his patient’s condition, which is what we aimed to achieve with the SQs.

Whether the RADIANT score is related to the quality of a palliative care plan (the obtained scores are a fraction of the total possible score per vignette of 21 points) still needs to be studied in real patient care, as well as the validity and specificity of the tool.

Finally, we do not know whether all GPs in group 3 have understood the subtle but essential difference between SQ1 and SQ2. After all, we did not provide any explanation about the intention or use of this new SQ2.

This simple, additional SQ2 was considered helpful by GPs and showed promising pilot results. In four prospective studies in, respectively, general practice, a medical oncology outpatient clinic, the Intensive Care Unit and nursing homes, we will further explore its usefulness and validity.