The Promise and Pitfalls of Big Data Studies in Perioperative Medicine

“Big Data” Studies and Perioperative Clinical Equipoise

Although the quantity and quality of randomized controlled trials (RCTs) in perioperative medicine continues to rise, many fundamental clinical decisions continue to lack a robust evidence base. For example, inhaled volatile versus total intravenous (IV) general anesthesia remains a controversial clinical decision without large-scale RCTs. In the surgical realm, the choice between laparoscopic versus robot-assisted minimally invasive surgery continues to challenge consensus. In addition, the limited number of high-quality, reproducible clinical trials in perioperative medicine often lack the generalizability to inform the care of most patients or focus on the average treatment effects, which may hide risks to individual patients.

In the perioperative setting, big data studies can offer three forms of value: (1) describing clinical care variation; (2) forming hypothesis-generating inferences to inform the design of clinical trials; and (3) providing incremental evidence when prospective RCTs are infeasible or impractical. Simply describing national, regional, or clinician variation in care can provide vital knowledge and be the first step to reliable evidence. A given clinician may not realize that their own practice varies from local, regional, or national standards of care; however, clinician-specific feedback combined with peer-reviewed studies of variation in care can inform and improve clinician adherence to accepted standards of care. Additionally, as clinical care evolves over time, clinician feedback and high-quality studies with recent and longitudinal data provide insight into dynamic rather than static care patterns as new evidence is applied. Next, observational research using large national databases can identify possible areas for future prospective trials. The field of “failure to rescue” research was inspired by rigorous observational analyses of administrative data and surgical registries. Since these seminal hypothesis-generating papers identifying the concept of failure to rescue, many interventions have been proposed and are undergoing rigorous prospective testing. Finally, some processes of care are difficult or prohibitively costly to evaluate using prospective clinical trials. For example, the relationship between overlapping surgery and postoperative adverse events remains unclear; there are ethical and practical challenges in randomizing patients to receive care from a surgeon involved in two overlapping procedures. As a result, research using data from millions of patients has provided some evidence to guide the vigorously debated topic.