We searched MEDLINE (Ovid Technologies, 1966 to June 2004; English language) for terms describing physician experience (keywords: physician age, clinician age, physician experience, clinician experience), physician demographic characteristics (keywords: physician characteristics, clinician characteristics), practice variation (subject heading: physician's practice patterns), and performance in various domains (subject headings: clinical competence, health knowledge, attitudes and practice, outcomes assessment[health care]; keywords: knowledge, guideline adherence, appropriateness, outcomes). We retrieved potentially relevant articles and reviewed their reference lists to identify studies that our search strategy may have missed (Figure 1). We also searched our personal archives to identify additional studies. We included studies if they 1) were original reports providing empirical results; 2) measured knowledge, guideline adherence, mortality, or some other quality-of-care process or outcome; and 3) included years since graduation from medical school, years since certification, or physician age as a potential explanatory variable. We excluded studies if they described practice variation that is not known to affect quality of care (for example, assessed test-ordering behavior in clinical situations where optimal practice is unknown) or evaluated the performance of fewer than 20 physicians. For studies that examined several different end points, we included only those outcomes that are linked to knowledge or quality of care.
We used a standardized data extraction form to obtain data on study design and relevant results. We categorized studies into 4 groups on the basis of whether they evaluated knowledge (for example, knowledge of indications for blood transfusion), adherence to standards of care for diagnosis, screening, or prevention (for example, adherence to preventive care guidelines), adherence to standards of care for therapy (for example, appropriate prescribing), or health outcomes (for example, mortality). We classified the results of each study into 6 groups on the basis of the nature of the association between length of time in practice or age and performance: consistently negative, partially negative, no effect, mixed effect, partially positive, and consistently positive. “Consistently negative” studies were those for which all reported outcomes demonstrated a statistically significant decrease in performance with increasing years in practice or age. “Partially negative” studies showed decreasing performance with increasing experience for some outcomes and no association for others. We used similar definitions for “consistently positive” and “partially positive” studies. “Concave” studies found performance to initially improve with years in practice or age, then peak, and subsequently decrease.