General practitioners’ views and preferences about quality improvement feedback in preventive care: a cross-sectional study in Switzerland and France

We drew a random sample of 1100 community-based GPs (700 GPs practicing in Switzerland (Geneva and Vaud) and 400 in France (Alsace and Pays de la Loire)). The GPs were invited to participate in the study by post. Reminder messages (maximum twice by GP) were sent to non-responders. GPs who practiced complementary and alternative medicine were not eligible for the study. No other exclusion criteria were applied. The recruitment process, details about GPs’ selection, and sample size justification are described elsewhere [25].

Each randomly selected GP was contacted by post by a research assistant located in Geneva (for Switzerland) and by the local professional associations Union Régionale des Professionnels de Santé Alsace and Pays de la Loire (for France). GPs were informed about the aim of our study and the practical procedures for completing the questionnaire. The postal letters included a stamped return envelope. GPs were asked to fill out a questionnaire that included questions regarding socio-demographic characteristics (age group (<35, 35–44, 45–54, 55–64, >64), gender, location of the practice, certification, number of working days per week, number of working years in the current practice), and adherence to 12 preventive practices (see below).

These preventive practices were assessed with a 5-point Likert scale ranging from “never performed” to “always performed.” They were selected by consensus within the research team following a review of the literature. Ten preventive practices were included in the study because they had also been selected by previous authors: blood pressure, weight and height measurements, as well as screening for at-risk drinking (and advice to decrease drinking), for smoking (and advice to stop smoking), and for colon cancer, and finally, annual influenza immunization for patients >65 years and at-risk patients <65 years) [26]. We added two preventive measures to this list: cholesterol measurement because it is highly recommended and refraining from systematic screening for prostate cancer because several medical agencies recently recommended against systematic screening [9, 27, 28]. By contrast, we did not include diabetes screening because it was considered as targeted screening (i.e., limited to populations with particular risks factors such as obesity) or screening for breast cancer because this screening is often provided by gynecologists or in the context of screening programs [27, 29].

GPs were also asked to estimate the usefulness of a feedback regarding their preventive care practice (indispensable, very useful, rather useful, not very useful, useless) and to explain for which reason(s) they would find feedback useful (multiple answers allowed: to know GPs’ overall performance, to compare themselves with colleagues, to modify or improve their practice, to use regular feedback interventions to follow-up what is done in their practice, other reason). Finally, they were asked which type of feedback they would like (multiple answers allowed: a brief report (i.e., a report providing general information about GPs’ overall performance without individual and detailed results), a brief report and individual results, detailed results regarding their practice compared with the study results, a report and specific information regarding prevention best practice, a contact with the study investigators to discuss the results, a local quality circle meeting to discuss the results, other type of feedback). The selection of the feedback interventions was based on a consensus within the study team following a review of the literature [10, 11, 19, 21].

We pretested the questionnaire with seven GPs working in a primary care clinic (Centre Médical des Trois Chêne, Geneva, Switzerland) to ensure that the questionnaire was understandable and easy to complete. All collected data remained confidential throughout the study. It was assumed that tacit consent was given when a responder completed the questionnaire. No data was collected for the GPs who declined to participate.

We computed the proportion of GPs delivering each of these preventive care measures, defined as the proportion of GPs scoring 4 or 5/5 (i.e., often or always performed) on the Likert scale, as well as the mean number of measures by GP. We also assessed, from the GPs’ perspective, the preferred feedback intervention(s) and the reason(s) why feedback interventions could be useful, according to GPs’ gender, age, country, and adherence to recommendations (<10 vs. ≥ 10 preventive care measures: 10 was the third quartile). Frequency tables and chi-square tests were used. Finally, a multivariate logistic regression was used to investigate which were the main factors (among the GPs’ characteristics and compliance with prevention measures) associated with the perception that feedback is indispensable or very useful. The final model was chosen using a stepwise procedure based on the Akaike information criterion. All analyses were performed with TIBCO Spotfire S+® 8.1 for Windows (TIBCO Software Corporation, Palo Alto, CA, USA) or R version 3.2.2 (R Foundation for Statistical Computing, Vienna, Austria).

We showed that approximatively half of GPs found that feedback would be very useful for evaluating their clinical practice for preventive care and that younger GPs and those being more adherent to guidelines were more likely to consider feedback useful. We also showed that the two main reasons for feedback, as stated by GPs, were simply to have a view of GPs’ overall performance but also to modify or improve their practice. The two preferred feedback interventions were a brief report and a report with specific information regarding prevention best practice. Finally, we showed that very few GPs would like to discuss the results in local quality circle meetings and even less face-to-face with the study investigators.

Many authors showed that feedback may be useful to improve health care, including prevention [20, 32‐35]. However, to our knowledge, few data existed on whether GPs were interested in receiving feedback and how they would like to receive it. Our results are consistent with those of a recent cross-sectional survey about Dutch GPs’ preferences for interventions to improve guideline adherence (n = 703 GPs) [21]. In this study, Lugtenberg et al. showed that GPs preferred interactive small group meetings (84%), but feedback (53%) was also rated positively. By contrast, financial incentives, the use of educational materials, or big group meetings were mentioned by less than one fourth of GPs.

Lugtenberg’s study was designed to identify GPs’ preferences regarding a wide range of interventions, whereas our study was restricted to feedback interventions only. This is probably the reason why interactive educational small group meetings seemed to be highly appreciated by Dutch GPs in Lugtenberg’s study, whereas they were not part of the preferred interventions selected in our study. In other words, these findings suggest that GPs may appreciate interactive meetings when they address and discuss their difficulties in applying guidelines in their daily practice but not when the objective of these meetings is to discuss their own performance.

The finding that very few GPs would like to discuss the results face-to-face with the investigators, or in local quality circle meetings, could be explained in the same way: fear or inconvenience of highlighting their (lack of) performance. We may also consider two alternative explanations: belief that the objective of the study investigators is to transfer knowledge to GPs in a unidirectional way in a top-down relationship, and time constraints [32]. GPs in our study preferred written formats of feedback (brief or detailed report sent to the participants) over verbal feedback (face-to-face contacts with the investigators or local quality circle meetings). Though there may be a perception that feedback delivered face-to-face is more efficient compared to written feedback, the evidence shows little or no difference between the two formats [20].

It is reassuring that the second main reason cited by GPs for receiving feedback was to improve clinical practice whereas the second preferred type of intervention was a report with specific information regarding best practice: this is precisely why feedback is recommended [20]. This finding, and the fact that half of GPs found feedback useful, tends to suggest that implementation of a practice-based quality improvement feedback is feasible and acceptable to a large number of GPs. This is particularly true if feedback is written and not transmitted orally. Indeed, our results suggest that limited acceptability may be expected for face-to-face discussion with clinicians.

That younger GPs and those being more adherent to guidelines were more likely to consider feedback useful suggests that they are probably more open to criticism and more receptive to feedback seen as a way of improving their practice. Younger GPs (maybe because they are in general less experienced clinicians and, as a result, have not yet established their style of care) seem to have higher adherence to guidelines [36‐38]. For example, in a study assessing GPs’ attitudes to guidelines for elective surgical referral in England, it was shown that the odds of using guidelines decreased with increasing age, a 10-year increase in age being associated with halving odds of use [38]. By extension, it is hypothesized that younger GPs (and those being more adherent to guidelines) are also more receptive to feedback and therefore more likely to consider these interventions useful for improving their skills.

Only GPs practicing in Western Switzerland and two regions in France were invited to take part in the study, and only 47% of those who were contacted agreed to participate. Those who replied were likely to feel more concerned by the study question, which may have inflated the number of GPs who found feedback indispensable or very useful. However, response rates for surveys conducted among GPs are generally lower than surveys among other medical doctors, [39] and responses rates similar to ours are frequent in primary care research. As the answers were self-reported, our findings could partially be explained by the fact that responders may have a natural tendency to over-report positive, socially desirable behaviors (social desirability bias) [40]. In addition, our findings were obtained in a research context and may have been different if acquired in the context of a preventive guideline implementation project. As we did not collect any data on non-responders, we could not assess potential differences between responders and non-responders. However, our sample appears to be relatively representative in terms of age and gender of all community-based GPs practicing in Switzerland and France (data from Pays de la Loire). We believe that the risk of bias due to missing data is probably very small because there were only few missing data in our study. We used a stepwise selection procedure rather than a hypothesis-driven selection for the multivariate analysis of usefulness perceptions, which could lead to a model giving an over-optimistic impression despite the additional fitting error this may have added. Though many authors have shown that feedback may be useful to improve health care, including in prevention, we did not know to the best of our knowledge if GPs found feedback useful. Therefore, it would have been difficult to make theoretical hypotheses about the most likely predictors of usefulness perceptions to include in the model. Finally, after reviewing the literature, we decided to keep only five response options for the question assessing the reason(s) why GPs would find feedback useful, respectively seven options for the question about the type(s) of feedback they would like. These options seemed to be the most interesting, important, or pertinent to study in the context of primary care. We decided to limit these response options because with questions offering many possible response options, we were afraid that certain questions would have been completed incorrectly, for example at random. Though this restricted selection could result in a certain degree of information bias, we do not believe that this was in fact the case. Indeed, though “other reason” and “other type of feedback” were response options proposed to responders, they were never selected by GPs in our sample. In addition, no GP (n = 7) suggested in the pretest phase that we add other response options to the list.

GPs are increasingly placed under pressure to improve the quality of care. While the aim of guidelines is precisely to achieve a high quality of care (by providing specific recommendations for daily practice), their implementation in practice is highly challenging.

Several authors studied barriers to guideline adherences, but little is known on how to overcome these barriers [17, 18, 21]. Several interventions have been suggested and among these, feedback is increasingly used in primary care [10, 11, 20, 21]. Unfortunately, GPs are not usually involved in the choice of the format of the intervention chosen by decision makers to improve their practice [21]. However, it is likely that GPs’ adherence to guidelines could be improved if their views were taken into account [10]. In this context, we believe that our findings could help decision makers when they develop strategies to implement guidelines in primary care.

Such interventions are probably more efficient if they are repeated periodically [20]. Quality control circles (QCC) using PDCA (plan-do-check-act) iterative methods could be used for quality improvement feedback in preventive care [41]. QCC are used in healthcare to increase the quality of medical practice by establishing objectives (plan), implementing a plan (do), studying the results and comparing against the expected results (check), and making adjustments if necessary (act) [41‐44]. One of the main advantages of this method, when it is adequately carried out, is continuous monitoring, favoring sustained improvement. PDCA cycles can also be conducted by GPs themselves (potentially supported by a feedback system, e.g. a quality dashboard) without investigator involvement.