(ACS-NSQIP). Other large databases include the National

Cancer Database, the National Trauma Database, and administrative datasets, such as the national Medicare dataset or the Nationwide Inpatient Sample (NIS). Advanced statistical techniques, such as hierarchical modeling, are becoming increasingly important in this work. Hierarchical modeling can be used to minimize problems with small sample size (discussed above) by using empirical Bayes techniques to “adjust for reliability.” Finally, risk-adjustment techniques are used to ensure that hospital outcome comparisons account for differences in patient severity of illness.

Implementation Science

There is often a fine line between research and operational improvement initiatives when it comes to improving the quality of the care that we provide to our patients. In fact, for those of us involved in this type of research, the line is often blurred. The term “quality improvement” usually applies to the operational aspect of improving care, while “implementation science” is a discipline that is devoted to studying the optimal approach to organizational change and performance improvement. A landmark study by Pronovost published in the New England Journal of Medicine in 2006 highlights these issues.9 Pronovost designed and implemented a checklist for placement of central lines across the state of Michigan. The authors were able to show a decrease in the mean rate of catheter-related infections from 7.7 to 2.3 per 1,000 catheter-days and in fact dropping the median from 2.7 to 0 infections per 1,000 catheter-days within 3 months. The results were sustained over the 18-month study period. Controversy erupted around this work when it was discovered that the IRB at Johns Hopkins did not consider this work to be human subjects research, a decision with which the Office for Human Research Protection (OHRP) did not agree. This highlights the importance of being clear about the goals of your work. According to OHRP, if the goal is to improve local performance, then it is not research; however, if you intend to study the impact of your intervention and disseminate your

104 C.C. Greenberg and J.B. Dimick

results so that care can be improved more broadly, it is considered human subjects research. Regardless of these difficulties, the success of this study has been followed by several other high-profile checklist-based interventions including the Surgical Safety Checklist for intraoperative safety and the SURgical PAtient Safety System (SURPASS) checklist for the entire perioperative period.10, 11 Both interventions led to a documented decrease in mortality and were published in the New England Journal of Medicine in the last 2 years.

Patient Safety

The National Patient Safety Foundation defines patient safety as the avoidance, prevention, and amelioration of adverse outcomes or injuries stemming from the processes of health care. Patient safety research is therefore the academic discipline dedicated to the study of these unintended negative consequences that health care, whether that is an individual intervention or the design of the health care system, can have for patients. Surgery is ripe for patient safety interventions as numerous population-based studies have repeatedly shown that surgical adverse events account for approximately half of the injuries patients encounter while in the hospital and that most of these injuries originate in the operating room. Several studies by Drs. Atul Gawande and Selwyn Rogers published in Surgery utilized a variety of techniques including analysis of closed malpractice claims and focused interviews with surgeons to identify the most common contributing factors to errors leading to injury in surgical patients.12, 13 Perhaps the most pivotal patient safety work in surgery originated in the UK with the work of Carthey and de Leval, a human factors expert and a surgeon, who teamed up to investigate how major and minor errors in the operating room can have significant consequences for survival following aortic switch operations.14

Both implementation science and patient safety require a more granular approach to research than any of the other intellectual disciplines described in this chapter. The