2. Some (n = 24) used clinical data to adjust for differences among patients in severity and comorbidity, but only 4 reported statistically robust models;

3. Few (n = 4) adjusted for differences in specific processes of care; and

4. Few studies (n = 2) addressed appropriateness of patient selection.

Several conclusions emerged from this review. There can be little doubt that for a wide variety of surgical procedures and medical conditions higher volume (whether assessed by hospital or by physician) is associated with better health outcomes. Although a publication bias against negative findings appearing in print cannot be excluded, the uniformity with which the published research documents or confirms the existence of the association is compelling. Fully 77% of the studies we reviewed found statistically significant associations between higher volume and better outcomes. The remaining 23% did not find statistically significant relationships. No study demonstrated a statistically significant association in the opposite direction. Finally, all 16 of the studies with the highest quality scores found statistically significant associations.

Overall, however, the methodological rigor of the studies was modest, limiting generalizability and leaving many questions about the nature and causes of the association unanswered. Studies of the same procedure or condition typically employed widely varying definitions of high and low volume, precluding definitive conclusions about the nature of its relationship to outcome. We do not know, for example, whether a volume threshold exists—or more than one—above which outcomes are better but do not continue to improve with further volume increases. Is there a threshold effect for physician or hospital volume? Both? Neither? Very few studies addressed the question of whether the only relevant volume measure is that of the specific procedure or condition under study. Thus, we do not know the extent to which experience with procedures (or conditions) that are closely related affects outcome.

Available research does not shed much light on what specific factors explain outcome differences between high and low volume providers. No longitudinal studies (and very few of any design) address the important question of how much of the variability, especially among low-volume providers, is due to chance. Few investigations have assessed differences in specific clinical processes of care, especially those known to affect outcomes. One intriguing exception showed that about one third of the mortality difference between high- and low-volume hospitals for acute myocardial infarction could be attributed to more frequent use of proven-effective medications at high-volume hospitals. We believe that this direction is the most promising avenue for future research: a detailed understanding of those critical clinical and organizational processes that affect outcomes and that differentiate high- and low-volume hospitals and physicians. Such knowledge could provide a blueprint for improvement.

These research findings have several potential implications for public policy in health care. Public policy might be directed at informing consumers and referring physicians about the nature of the relationship between volume and outcome or at improving outcomes, or both. Data on volume of procedures by hospital are publicly accessible in many states through hospital discharge abstract databases. Many of these databases also permit identification of treating physicians. Making these data widely available is one potential way of informing consumers and clinicians. Some of the challenges such efforts would face are the difficulties of explaining what the data mean (and what they do not), the technical challenges of carefully identifying conditions and procedures for which research is conclusive that higher volume is associated with better outcomes (and at what threshold), and the need to update the data regularly.