Importance: The association of surgeons' and hospitals' operative volumes with postoperative patient outcomes has been studied for decades and holds important policy implications; however, in many volume-outcome analyses, this association is described without the envisioning of a clear intervention, which often introduces unintentional bias. Acting on such results may lead to unintended consequences from policy interventions or patient recommendations.
Objective: To specify how (hypothetical) target trials would be designed to estimate the association between postoperative mortality of patients undergoing operations and a range of surgeon and hospital volume conditions and then to emulate these trials by using observational data.
Design, setting, and participants: This observational data analysis emulated 4 hypothetical target trials of increasing complexity, ranging from a poorly defined trial that would randomly assign participants only to surgeon volume to one that would randomly assign participants to surgeon volume, hospital volume, and specific surgeon and hospital. This population-based cohort study included 9136 Medicare beneficiaries with a first diagnosis of pancreatic malignant neoplasm who did not require neoadjuvant therapy and underwent pancreatectomy between January 1, 2012, and September 30, 2016. Data analysis was performed between September 1, 2019, and October 8, 2021.
Exposures: Number of pancreatectomies performed by surgeon and hospital during the prior year.
Main outcomes and measures: Ninety-day mortality.
Results: The analyses included 9136 Medicare beneficiaries treated by 1358 surgeons at 697 hospitals; median age was 73.3 years (IQR, 69.1-78.1 years), and 4642 were men (51%). When trials with poorly defined interventions on surgeon volume were emulated, the estimated 90-day mortality was 7.9% (95% CI, 6.4%-9.4%) for lower-volume surgeons and 5.2% (95% CI, 2.7%-10.9%) for higher-volume surgeons. When trials with better-defined interventions were emulated, the difference was reduced: 7.8% (95% CI, 6.3%-9.3%) for lower-volume surgeons and 7.2% (95% CI, 6.0%-8.7%) for higher-volume surgeons.
Conclusions and relevance: In this cohort study that emulated 4 different target trials with data from Medicare beneficiaries undergoing pancreatectomy, mortality differences across surgical volume levels were attenuated when the interventions were well defined. The application of the hypothetical target trial framework to this specific volume-outcomes scenario revealed the complexities of this research question and the unintentional biases introduced in prior studies, which emulated poorly defined trials whose results are therefore difficult to interpret. The target trial framework may be of value to outcomes researchers asking questions that correspond to well-defined interventions for the real world.