Directed Acyclic Graph Assisted Method For Estimating Average Treatment Effect.

Area of Science:

• Causal inference
• Observational data analysis
• Health services research

Background:

• Observational data is crucial for estimating Average Treatment Effect (ATE).
• Inverse Probability of Treatment Weighting (IPTW) is effective for ATE estimation if assumptions are met.
• Directed Acyclic Graphs (DAGs) help assess the exchangeability assumption.

Purpose of the Study:

• To propose an enhanced IPTW method for consistent and efficient ATE estimation.
• To incorporate confounding variables and predictors into propensity score models.
• To evaluate the causal association between tracheostomy and in-hospital infant mortality.

Main Methods:

• Utilized propensity scores and Inverse Probability of Treatment Weighting (IPTW).
• Employed Directed Acyclic Graphs (DAGs) to identify confounding variables.
• Incorporated a minimally sufficient adjustment set of confounders and predictors into the propensity score model.
• Conducted extensive simulations to validate the method's performance.

Main Results:

• Simulation results confirmed that including confounders and predictors improves ATE estimator consistency and efficiency.
• The proposed method was applied to 2016 Healthcare Cost and Utilization Project Kids' Inpatient Database data.
• The estimated Average Treatment Effect (ATE) for tracheostomy and in-hospital infant mortality was 2.30%-2.46% (p > 0.05).

Conclusions:

• The enhanced IPTW method provides a consistent and efficient approach for ATE estimation from observational data.
• Tracheostomy was not found to be significantly causally associated with in-hospital infant mortality in the studied cohort.
• Accurate ATE estimation requires careful consideration of confounding and predictive variables in propensity score models.