Abstract:Comparative effectiveness research frequently employs the instrumental variable design since randomized trials can be infeasible for many reasons. In this study, we investigate and compare treatments for emergency cholecystitis -- inflammation of the gallbladder. A standard treatment for cholecystitis is surgical removal of the gallbladder, while alternative non-surgical treatments include managed care and pharmaceutical options. As randomized trials are judged to violate the principle of equipoise, we consider an instrument for operative care: the surgeon's tendency to operate. Standard instrumental variable estimation methods, however, often rely on parametric models that are prone to bias from model misspecification. We outline instrumental variable estimation methods based on the doubly robust machine learning framework. These methods enable us to employ various machine learning techniques for nuisance parameter estimation and deliver consistent estimates and fast rates of convergence for valid inference. We use these methods to estimate the primary target causal estimand in an IV design. Additionally, we expand these methods to develop estimators for heterogeneous causal effects, profiling principal strata, and a sensitivity analyses for a key instrumental variable assumption. We conduct a simulation study to demonstrate scenarios where more flexible estimation methods outperform standard methods. Our findings indicate that operative care is generally more effective for cholecystitis patients, although the benefits of surgery can be less pronounced for key patient subgroups.

What problem does this paper attempt to address?

The main problem this paper attempts to address is the evaluation of the effectiveness of surgical treatment for acute cholecystitis. Specifically, due to the ethical difficulties of conducting randomized controlled trials, researchers have had to rely on observational studies to gather evidence. However, observational studies are prone to indication bias, where the choice of treatment for patients may depend on prognostic factors that also affect patient outcomes. To overcome this challenge, the study employs an instrumental variable (IV) design and proposes an estimation method based on a doubly robust machine learning framework to more accurately assess the effect of surgical treatment. ### Main Issues: 1. **Evaluating the effectiveness of surgical treatment for acute cholecystitis**: The study aims to assess the effect of surgical treatment for acute cholecystitis compared to non-surgical treatment. 2. **Overcoming confounding bias in observational studies**: By using an instrumental variable design, specifically using the surgeon's tendency to operate (TTO) as an instrumental variable, to reduce the impact of confounding bias. 3. **Improving the flexibility and accuracy of estimates**: Proposing an estimation method based on a doubly robust machine learning framework to more flexibly and accurately estimate causal effects, including the Local Average Treatment Effect (LATE) and Conditional Local Average Treatment Effect (CLATE). ### Background and Motivation: - **Treatment options for acute cholecystitis**: The standard treatment for acute cholecystitis is surgical removal of the gallbladder (cholecystectomy), but there are also non-surgical treatment options such as observation, minimally invasive surgery, medication, and supportive care. - **Limitations of randomized controlled trials**: Due to ethical reasons, randomized controlled trials cannot be conducted in some cases, necessitating reliance on observational studies. - **Problem of confounding bias**: In observational studies, the choice of treatment for patients may be influenced by prognostic factors, leading to confounding bias. - **Design of instrumental variables**: Instrumental variable design can be used to address unobserved confounding factors, thereby more accurately estimating causal effects. ### Methodological Innovations: - **Doubly robust machine learning framework**: This method combines the flexibility of machine learning with the robustness of statistical theory, providing consistent and fast-converging estimates in the presence of unobserved confounding factors. - **Estimation of Conditional Local Average Treatment Effect**: In addition to estimating the Local Average Treatment Effect, new methods have been developed to estimate the Conditional Local Average Treatment Effect for different patient subgroups to explore the heterogeneity of treatment effects. ### Conclusions: - **Effectiveness of surgical treatment**: The study shows that surgical treatment is generally more effective for patients with acute cholecystitis, but the benefits of surgery may be less pronounced in certain key patient subgroups. - **Advantages of the method**: The proposed doubly robust machine learning method performs well in simulation studies, providing more accurate estimates of causal effects, especially in the presence of complex confounding factors. In summary, this paper improves the accuracy of evaluating the effectiveness of surgical treatment for acute cholecystitis by introducing new estimation methods, providing strong support for clinical decision-making.

Doubly robust machine learning for an instrumental variable study of surgical care for cholecystitis

Biased Encouragements and Heterogeneous Effects in an Instrumental Variable Study of Emergency General Surgical Outcomes

Flexibly Estimating and Interpreting Heterogeneous Treatment Effects of Laparoscopic Surgery for Cholecystitis Patients

Comparison of Laparoscopic Cholecystectomy Combined with Intraoperative Endoscopic Sphincterotomy and Laparoscopic Exploration of the Common Bile Duct for Cholecystocholedocholithiasis

Doubly Robust Nonparametric Instrumental Variable Estimators for Survival Outcomes

Instrumental variable analysis in the presence of unmeasured confounding

Doubly robust estimation of optimal treatment regimes for survival data using an instrumental variable

Estimating optimal treatment regimes in survival contexts using an instrumental variable

Selective Machine Learning of the Average Treatment Effect with an Invalid Instrumental Variable

An Approach for Combining Clinical Judgment with Machine Learning to Inform Medical Decision Making: Analysis of Nonemergency Surgery Strategies for Acute Appendicitis in Patients with Multiple Long-Term Conditions

Bridging preference-based instrumental variable studies and cluster-randomized encouragement experiments: Study design, noncompliance, and average cluster effect ratio

Instrumental variable analysis with categorical treatment

Triangulating Instrumental Variable, confounder adjustment and Difference-in-Difference methods for comparative effectiveness research in observational data

A semiparametric instrumental variable approach to optimal treatment regimes under endogeneity

Robust causal inference with continuous instruments using the local instrumental variable curve

Nonparametric Bayesian Instrumental Variable Analysis: Evaluating Heterogeneous Effects of Coronary Arterial Access Site Strategies

Machine learning methods applied to audit of surgical outcomes after treatment for cancer of the head and neck

Bayesian Feature Selection for Multi-valued Treatment Comparisons: An Electronic Health Records Study of Vasopressor Effectiveness

Mostly Harmless Machine Learning: Learning Optimal Instruments in Linear IV Models

Causal Proportional Hazards Estimation with a Binary Instrumental Variable

Comparing Long-term Mortality After Carotid Endarterectomy vs Carotid Stenting Using a Novel Instrumental Variable Method for Risk Adjustment in Observational Time-to-Event Data