The aim of this review was to determine the effectiveness of patient based feedback, brief training and their combination on the interpersonal skills of physicians. Only a small number of trials were identified and thus any conclusions about the effectiveness of these interventions is preliminary.
The effectiveness of patient based feedback
The search identified only two feedback studies, which have been identified previously in a review of instruments and feedback methods for the assessment of physicians using patient surveys . It is unclear from this review whether patient based feedback is an effective quality improvement tool for changing physicians' interpersonal care behaviour. The study involving trainees showed a significant positive effect for patient feedback on patient satisfaction scores , whereas the study involving experienced physicians showed no effect [23, 24]. A study of feedback excluded from the review (conducted in a hospital setting) also reported a significant positive effect on the interpersonal skills of trainee internal medicine residents . Clinically experienced physicians may have more enduring interpersonal care behaviours that have developed over many years of practice, whereas trainees may be more easily able to adapt their behaviours in line with feedback.
An alternative explanation for the difference in results may relate to the intensity of the feedback, with the study reporting positive effects applying the intervention at five time points (3 months apart) over a two year period . In contrast the study reporting no effect, only gave patient feedback at one time point (3–6 months after the start of the study) within the 15 month study period [23, 24].
The use of patient feedback assumes that patients can judge the quality of interpersonal care and that the current assessment technology is capable of capturing patient views. Although doctors and patients have been shown to disagree about what constitutes technical quality of care [36–38], it could be argued that no one is better placed than patients to judge interpersonal performance. There is evidence that patients are able to detect improvements in the quality of the physician-patient interaction .
A recent systematic review examined instruments designed to evaluate patients' experiences with individual practicing physicians and whether they are able to provide performance feedback at the individual level . Although many had some evidence of validity, it was generally limited, and it was not clear how well they correlated with other measures of doctor performance. One particular problem with using patient assessment instruments is the so called 'ceiling effect' due to the majority of patients express high levels of satisfaction with care i.e. there is little variation in responses [40–42]. The failure of these instruments to capture negative feedback is another issue that may reduce their effectiveness.
Studies of feedback are unique in that the intervention is very similar to the outcome assessment (i.e. both use patient assessments, although only in the former is the data fed back). If the mere act of measurement (without feedback) were sufficient to change behaviour, then these studies may underestimate the effect of the intervention.
Finally, the authors of the review discussed above  found that the aim of feedback was often vague, the exact procedures to be used lacked specificity, and there was a lack of detail about the mechanism by which feedback was expected to lead to improvement, beyond an implicit suggestion of the impact of normative comparisons. The format of feedback may also be important. There was limited detail about the exact form of feedback given in the two studies, although one fed back data on individual questions and nine dimensions of care, with individual data for the GP and reference figures for all GPs . Studies suggest that the style and content of feedback is important , and there may be potential in testing different methods of presenting the data and the use of qualitative information from patients to complement quantitative data. Further work on the 'active ingredients' of feedback is clearly required.
The effectiveness of brief training
Brief training has previously been found to be effective in changing physician behaviour in general [16, 44] and reviews focussing specifically on training for interpersonal skills, have also suggested that training can be effective . For example, a Cochrane review of training to improve patient-centredness reported positive effects on a number of outcomes . The difference between the results of the Cochrane review and the current study may reflect differences in outcome measures. The Cochrane review included multiple measures, including process measures of physician behaviour and health outcomes. When restricted to the seven trials using a patient based assessment of interpersonal care skills (the inclusion criteria for the present review) only two of the seven studies in the Cochrane review showed a positive significant effect, a result not substantively different from the results reported here [35, 46].
The only positive study was the oldest of all the training studies. This may be due to lower baseline levels of physician interpersonal skills. The medical training undertaken by the participating physicians (whose average age was 41.7 years when the study was undertaken and published in 1987) may have placed less of an emphasis on teaching interpersonal skills as trainees and practitioners were assumed to acquire interpersonal skills incidentally, simply via interacting with patients . Physicians in more recent studies would have undertaken more formal instruction and assessment.
Although the review was restricted to primary care physicians, findings from the wider literature on communication skills for health professionals may be informative in developing more effective interventions. Reviews suggest that effective interventions require combinations of didactic components with practice rehearsal and feedback from peers . Interventions may also need to focus on attitudes that may clash with the interpersonal skills being taught . Another key issue is the length of training. The study used a maximum of one week training as an inclusion criterion, based on discussions with GP colleagues as to what was likely to be feasible in relation to practising GPs. The limited effects of training may simply reflect the limited duration of the interventions, and reflect the paradox that in primary care, effective training may be unfeasible, whereas feasible training may be ineffective .
Limitations of the study
We offer several cautions about the interpretation of these results, over and above caveats concerning the number of identified studies. Firstly, as in all such reviews there is the potential for publication bias. Such bias can lead to an overestimation of an intervention's effect on the outcomes i.e. a false positive [51, 52]. Secondly, due to time constraints, we were unable to contact authors for additional information therefore we included only published data. A second consequence of time constraints excluded searching via other means e.g. hand-searching of journals and conference proceedings etc. Thirdly, if studies showing an intervention to be effective are more likely to be published in English, then any summary of only the English language reports retrieved through a database search may result in an overestimate of effectiveness due to a language bias [53, 54].
Both of the included studies indicating positive effects did not adjust for clustering. There is a risk of inflating statistical significance when analysing patient level data without adjusting for clustering .
The study included trials where the outcome measure was a patient assessment. This criteria was used because interventions that change in patient assessments are likely to be of greater interest to policy makers. However, it should be noted that it may be more appropriate to use a range of assessment technologies (such as process measures of behaviour in the consultation) as well as patient outcomes .
The current review was restricted to primary care physicians as they currently provide the majority of care in this setting . Future reviews into these interventions should take into account the potential shift towards increased nurse-led delivery of primary care .
Implications for research
Although the trials identified in the review were of reasonable quality, their limited number means that confident conclusions about the efficacy of these interventions must await the publication of new studies. A more substantial evidence base is also required to explore the various factors that may affect the efficacy of patient based feedback. Such factors may include the frequency, content and style of feedback and training, and physician and patient characteristics.
The theoretical basis of feedback and training interventions was sometimes unclear. More explicit statements of theory underlying interventions and qualitative research conducted as part of the trials may provide insights into why these interventions succeed or fail.
Thirdly, the effectiveness of patient based feedback in combination with other interventions should be investigated (e.g. the combination of patient based feedback with brief training, or with financial incentives). Financial incentives are known to be effective (external) motivators . This type of arrangement is already utilised in the United States. General practice in the United Kingdom has become accustomed to conducting patient surveys on an annual basis for financial incentives, but the current incentive structure pays physicians primarily on the basis of conducting the survey rather than making changes.
Finally, the cost effectiveness of these interventions need to be assessed. The National Health Service in the United Kingdom has already made a significant financial investment in the process of patient assessments in primary care, and it is critical that this investment can be proven to be a good use of resources compared to other competing priorities.