All studies are subject to bias and error caused by their inherent design, or by unforeseen methodological limitations. The INTERMEDE study is no exception, and the potential bias and errors that may be important are discussed below.
Selection bias is caused by the existence of systematic differences between those who take part in a study and those who do not ( p.166). This means that selection bias can cause a misrepresentation of the population within the study sample, and therefore potentially render results and findings invalid, as they do not represent the results that would have been found in the wider population. In this study selection bias may have been introduced at two points; first, at initial sampling, when individuals were assessed against the eligibility criteria, and second, when patients declined to participate or dropped out of the study. In the first instance, the patients could have been wrongly deemed eligible, or ineligible, and in the second instance, patients with specific characteristics could be systematically refusing to participate or dropping out, or conversely, patients with specific characteristics could be systematically agreeing to participate. These factors could potentially lead to selection bias, whereby the sample of patients was not representative of the 'real' body of patients due to the selection procedure. However, data were collected on patients who refused to participate and analyses will be conducted to determine if selection bias is likely to be present, and allow for results to be adjusted accordingly.
Selection bias could also be present among the sample of GPs who volunteered to participate. Bias could be introduced via two mechanisms, firstly through the baseline catchment of GPs contacted via the GP networks (the Toulouse Department for general medicine and the French society for general medicine), and secondly through the voluntary nature of their participation in the study. The GP networks could influence the types of individuals contacted. It is possible that certain types of GPs are more likely to be members of such organisations, and in this instance, many of the GPs who volunteered were implicated in teaching or training junior colleagues and had a specific interest in improving the quality of their role as community or family doctors. This could render the GP more likely to accept participation in a study, and mean that the GPs involved are more attentive to the quality or advancement of their work. Furthermore, the participants who all volunteered may have had a specific interest in this type of multidisciplinary study, or a special affinity for the subject matter of patient-physician interaction. These aspects of samples based on volunteers may also influence the types of GPs participating. Both potential type of selection bias could mean that the sample of GPs participating in the study is not a true representation of the GP population in France, and notably that GPs particularly interested in improving and advancing the nature of their work are over-represented in this study. These factors need to be kept in mind when interpreting results from the qualitative or quantitative analyses.
Knowledge and awareness that the study was on the patient-physician interaction could influence patients' as well as GPs' behaviours in several ways. It could determine a patient's decision to participate in either the qualitative or quantitative phases of the study, as well as alter the relationship between the two actors during the consultation. However, in order to preserve the ethical integrity of the study informed consent had to be obtained, and therefore the minimum amount of information was made available to patients in the waiting room allowing them to decide whether or not to participate. GPs volunteered to take part having been informed as to the nature of the study.
The impact of observers on ethnographic data is a subject much debated and discussed in the social sciences. Nevertheless, it is an inevitable consequence of this type of research. Patients and physicians alike may alter their behaviour in the presence of a third party. In this study, several of the GPs participated regularly in training medical students, and therefore were accustomed to having another person present during consultations. Nevertheless, the aim here is to compare the consultation as experienced by the GP to the consultation as experienced by the patient. Even if their behaviours were altered by the presence of a third party, each actor participated in and experienced the same consultation. Furthermore, the goal of the study was to analyse the interaction between the two participants during the consultation. Thus, whatever the content of the consultation, it is important to analyse the concordance between each actor's understanding of what had happened and was said. Since it is unlikely that the presence of a silent observer should alter the mutual comprehension between two persons, the goal of the study should not be affected in depth by such a presence.
Data analysis scheme
There are three objectives to the qualitative phase: the first, to verify the feasibility of this type of ethnographic observational study in a general practice setting; the second, to generate new and original results; and the third, to use the qualitative data to develop hypotheses on the interaction between patients and physicians and how the latter may be linked to health inequalities. An inductive sociological approach will be taken using grounded theory analysis whereby the data itself guides the researcher to develop theories . No specific data analysis scheme was set a priori in the aim of allowing the qualitative researchers to develop theories and schemes of analysis as they worked their data. Due to the open and non-prescriptive nature of the data analysis, the researchers will work their material individually, as well as in group sessions, in order to limit the impact of subjectivity on interpreting the data.
Some of the areas that will be explored by the qualitative team will reflect themes present in the literature. The nature of the 'doctor' effect on patient-physician interactions, and therefore the impact of variability between different physicians' manners of working will be an important area to explore. Of particular interest is how GPs might alter their approach to adapt to each individual consultation, taking into account patient characteristics, prior relationships and the medical specificity of each case. With reference to the patient, it will be important to ascertain which factors affect how a patient relates to their GP during the consultation, and whether they feel they have control over their health in general. Differences in the way patients with similar medical conditions are diagnosed and treated will also be an area to explore, more specifically in the context of overweight/obesity being raised during the consultation. Factors that influence the relationship between GPs and patients, and the way in which the consultation may be negotiated between the two parties will be of interest. The notion of a shared identity between physicians and patients as being a key factor in positive health care exchanges, as evoked by Street et al (2008), will be a key theme to explore in both qualitative and quantitative phases of analyses .
The quantitative analyses will be developed in more detail based on the hypotheses generated through the qualitative phase when the above themes will be thoroughly explored and new areas of interest identified. The quantitative analyses will be used to compliment the qualitative work, but also to inform it further and suggest new pathways of analysis.
A preliminary quantitative analysis scheme will aim to explore: a) concordances and discordances between the patient and GP about what happened during the consultation, as well as how each party rates the patient's health and whether they were satisfied with the consultation; b) the social context of the consultation by comparing objective and subjective measures of social position as described by the GP versus the patient; c) the relationship between the different social position measures and consultation outcomes, such as treatment, advice and health promotion; d) how gender differences and similarities between GP and patient may affect their interaction, e) the variability in interactions during the consultation and whether this depends primarily on within- or between- GP differences; f) the treatment and follow-up care of obese versus non-obese patients, due to the strong links between obesity and social position.
The feasibility and positive feedback from patients and GPs alike have been encouraging outcomes thus far observed in the study process. Indeed, both qualitative and quantitative phases were readily accepted by both parties. Patients and GPs were prepared to have a third person present during the consultations, and GPs were willing to accept a qualitative protocol where patients have in-depth post-consultation interviews analysing the nature and type of relationship they have with their doctor. Equally, in the quantitative phase, both parties were willing to fill in a large number of questionnaires, patients being asked to fill-in three separate questionnaires, and GPs having to fill-in a questionnaire for every patient they saw. These aspects that could have posed constraints on the sampling were readily accepted by the participants, and the refusal rate was low.
The development of mirrored semi-structured interviews and mirrored questionnaires in the qualitative and quantitative phases respectively is a methodological success of this study so far. As research tools they will be used to extract valuable data allowing us to compare the patient and GP's perspectives on the consultation, their mutual relationship as well as the patient's social circumstances and their respective assessment of the patient's health status.
Thus far, the multidisciplinary nature of this study has been successful, with all disciplines working independently as well as together to develop the study protocol, collect data and secure funding. Nevertheless the real challenge will be in how well the different parties can collaborate during the analysis phase. Overall, the different disciplines working on the project will develop their respective schemes of analysis together, updating their results and analytical pathways in a circular manner each informing and complimenting the other. A series of workshops will be organised in order to set the work pace, to decide upon important definitions and procedures and to identify common themes. Though no easy feat, this manner of working will maximise the potential of this type of multidisciplinary study and optimise the value of mixed methods research.
To our knowledge this study is original in its design and will contribute to understanding what goes on during a consultation between patients and GPs, and identify what elements of the consultation could contribute to generating health inequalities. Through exploring the interaction between patients and GPs, recommendations can be made to improving primary and secondary health care in France, with the aim of reducing health inequalities and ameliorating the health care system.