“Practice makes perfect” – an old adage in support of the current emphasis on how volumes appear to impact healthcare outcomes. Variations in healthcare volume were first noted by Wennberg and Gittlesohn in Science in 1973 and are often considered as part of the Dartmouth Atlas projects that examined variation in healthcare volumes in many different settings. This variation has been shown to be associated with practices at local institutions and with healthcare costs. More interestingly, with respect to patient outcomes, volumes appear to matter. For example, a study of interventional cardiologists divided them into 3 volume-specific groups and found that those with the higher volumes had greater clinical success of impacting stenotic lesions and lower rates of needing to go to coronary artery bypass graft. A follow-up study of cardiologists demonstrated that clinicians with the highest volumes had the lowest mortality rates. Despite these and other findings, however, volume may not always be a predictor of surgical outcomes. In another Dartmouth Atlas study, they found no impact of an individual surgeon’s surgical volume on mortality rates in the setting of carotid endarterectomy.

In addition to the individual surgeons’ volumes, it appears that total hospital volume also appears to impact outcomes. In 1 of the largest studies to examine this question, the operative mortality rates in 14 different surgical procedures were examined among hospitals of differing volumes. It was demonstrated consistently that those hospitals with higher volumes had lower operative mortality rates. Certainly, hospital volumes have been demonstrated to be associated with a variety of obstetric and neonatal outcomes. In a large study of neonatal intensive care unit (NICU) volume, even among level III NICUs, the volume mattered. Among those NICUs with >100 very low birthweight (VLBW) neonates per year, the mortality rate was 10% lower than those units with 50–100 VLBW neonates and 20% lower than those units with <25 VLBW neonates per year. When it comes to obstetric care, there clearly are different rates of cesarean deliveries by hospitals that range from 15-50% in the United States alone. Labor and delivery units that perform lower volumes of deliveries have been shown to have higher rates of complications, in particular, perinatal asphyxia.

Research that examines the effect of individual obstetric provider volume is a bit more challenging to conduct because of the attribution issue in obstetrics. However, it does appear that, particularly related to obstetric procedures, volume matters. In amniocentesis, both success and complication rates improved with increased experience, and providers who perform <50 amniocenteses per year have been found to have greater rates of maternal cell contamination. These findings support the California requirement for all providers who perform genetic amniocentesis to meet specific volume thresholds to be part of the state’s prenatal diagnosis program. In another interesting study of obstetric volume, it was shown that obstetricians with lower annual patient volume had higher cesarean delivery rates. Thus, it does appear that there are volume effects in obstetrics.

Given this background, Drukker et al conducted a study to examine the impact of cesarean volume on obstetric outcomes in a large teaching hospital in Israel, which is published in this month’s American Journal of Obstetrics and Gynecology . The authors initially dichotomized surgical volume by surgeon by above and below an annual median volume that was 48 cesarean deliveries per year. They then further stratified the cesarean delivery volume into quartiles. They examined composite maternal outcome, which incorporated a host of outcomes that included urinary and gastrointestinal tract injuries, hemoglobin drop >3 gm/dL, blood transfusion, relaparotomy, puerperal febrile morbidity, prolonged maternal hospitalization, or readmission to hospital. Additionally, they examined the procedural length of time for the cesarean deliveries.

In their initial dichotomous analysis, surgeons with high volume (above the median) had a lower rate of the overall composite outcome (15.8% vs 18.9%; P < .001). Interestingly, the rate of urinary or gastrointestinal injuries was lower in women specifically whose cesarean delivery was performed by high-volume surgeons (0.2% vs 0.6%; P < .001). These results held when an array of potential confounders that included indication for cesarean delivery and other patient characteristics were controlled for with multivariable analysis.

Next, they examined quartiles of volumes and found that, in the top quartile, the rate of the primary composite outcome was 15.6%, that the middle quartiles were 17.4% and 17.9%, and that the bottom quartile was 21.8% ( P = .004). Similarly, when they controlled for potential confounders, these differences between the top and bottom quartiles persisted.

The authors point out several limitations. In a retrospective study such as this, it is important to control for potential confounding, which they did. However, one cannot control for potential confounders that were not measured. One potentially important confounder is patient body mass index (BMI). Certainly, increased BMI is associated with higher rates of surgical complications, and BMI varies among patients. It may be that the scheduled cesarean deliveries, which were higher among the high-volume clinicians, were in patients that also had lower BMIs.

One important consideration in the current study is that it is not a randomized trial. The strength of randomized trials is that they reduce the probability that there are baseline differences in the study population. In this case, we are interested in both the patient population and the surgeons themselves. Certainly, it was problematic that there were more scheduled cesarean deliveries in the high-volume group. Such patients often will have lower rates of complications compared with women who undergo cesarean deliveries during labor. However, the authors were able to control for these factors to some degree and still demonstrated some differences. However, the difference between high-volume and low-volume surgeons may be even more important. It may be that a high-volume surgeon is simply someone who is more interested in doing surgery and, in turn, is potentially a better surgeon at baseline than those surgeons with less interest or skill who have lower volumes. Thus, this may not be due to the volumes but may be due to the surgeons’ innate skills and preferences. If that is the case, then it still would be a good idea to have high-volume surgeon involved in cases; however, it does not support the idea that surgeons should be held to a particular volume in practice. A study to examine that question would need to randomize the physicians to being high-volume vs lower-volume surgeons.

Considering the results from this study and the others cited earlier, it is clear that procedural volume impacts outcomes, generally with higher volume providers and hospitals having lower rates of procedure-related complications. Given these findings, should there be strict volume requirements to maintain privileges, to maintain reimbursement from payers, or perhaps to maintain board certification? The first is already happening, although the volumes set to maintain privileges are often far below the volumes being described earlier with regards to impact on outcomes. The biggest challenge is that, although the studies mentioned earlier are intriguing, there is not an adequate research base in any of these specific procedures to define what an adequate threshold is. In the study of amniocenteses, although the researchers used a cut-off of 50 procedures per year, they did not examine different thresholds. Similarly, in the current study of cesarean deliveries, although they did examine some different volume quartiles, they saw improvements even in the very top quartile over the 2 middle quartiles, which suggests that there are volume benefits in even the very top group as described. Before specific thresholds should be required, much more rigorous research must be accomplished.

Additionally, if such volume thresholds were applied, what would be the impact on access to care? In many clinical settings, the volume to maintain a critical threshold would simply not be available. In cesarean delivery, although it is a more common procedure, we certainly would not want an incentive to do more cesarean deliveries to maintain credentialing or reimbursement. Thus, instead of identifying a threshold that minimizes the number of complications, the goal should be to determine at what volume is an acceptable rate of complications generally achieved. In this way, access to providers for such procedures will be maintained.

Thus, I do not believe that strict volume requirements should exist at the current time. However, we should all be aware that maintaining higher volumes improves outcomes both in health systems and at the level of the individual clinician. The role of simulation training in achieving similar effects should also receive much greater attention and research. Perhaps simulation can supplement the available procedural experience in lower volume environments and allow the providers who practice in these settings to achieve the outcomes that are seen in higher volume organizations. In the end, we have a duty to our patients to improve outcomes while maintaining access to care. Managing the tradeoffs between these 2, while considering the science of clinical volume, will be the challenge for the next several decades, particularly for procedural-based healthcare.