Systematic review of the literature on assessment, feedback and physicians' clinical performance: BEME Guide No. 7

There is a basis for the assumption that feedback can be used to enhance physicians' performance. Nevertheless, the findings of empirical studies of the impact of feedback on clinical performance have been equivocal.

To summarize evidence related to the impact of assessment and feedback on physicians' clinical performance.

The authors searched the literature from 1966 to 2003 using MEDLINE, HealthSTAR, the Science Citation Index and eight other electronic databases. A total of 3702 citations were identified.

Empirical studies were selected involving the baseline measurement of physicians' performance and follow-up measurement after they received summaries of their performance.

Data were extracted on research design, sample, dependent and independent variables using a written protocol.

A group of 220 studies involving primary data collection was identified. However, only 41 met all selection criteria and evaluated the independent effect of feedback on physician performance. Of these, 32 (74%) demonstrated a positive impact. Feedback was more likely to be effective when provided by an authoritative source over an extended period of time. Another subset of 132 studies examined the effect of feedback combined with other interventions such as educational programmes, practice guidelines and reminders. Of these, 106 studies (77%) demonstrated a positive impact. Two additional subsets of 29 feedback studies involving resident physicians in training and 18 studies examining proxy measures of physician performance across clinical sites or groups of patients were reviewed. The majority of these two subsets also reported that feedback had positive effects on performance.

Feedback can change physicians' clinical performance when provided systematically over multiple years by an authoritative, credible source.

The effects of formal assessment and feedback on physician performance are influenced by the source and duration of feedback. Other factors, such as physicians' active involvement in the process, the amount of information reported, the timing and amount of feedback, and other concurrent interventions, such as education, guidelines, reminder systems and incentives, also appear to be important. However, the independent contributions of these interventions have not been well documented in controlled studies. It is recommended that the designers of future theoretical as well as practical studies of feedback separate the effects of feedback from other concurrent interventions.