Semiparametric estimation of structural nested mean models with irregularly spaced longitudinal observations

Area of Science:

• Biostatistics
• Epidemiology
• Causal Inference

Background:

• Longitudinal observational studies often involve irregularly spaced data, posing challenges for traditional causal inference methods.
• Existing structural nested mean models (SNMMs) typically assume fixed time points, limiting their applicability.

Purpose of the Study:

• To develop a framework for causal inference using continuous-time SNMMs to handle irregularly spaced observations.
• To identify causal parameters under a no unmeasured confounding assumption in continuous-time settings.

Main Methods:

• Developed semiparametric efficiency theory for continuous-time SNMMs.
• Proposed locally efficient estimators robust to model misspecification (double robustness).
• Investigated estimators under ignorable censoring, identifying the complete-case estimator as optimal among weighting methods.

Main Results:

• The proposed continuous-time SNMM framework enables causal analysis that respects the continuous nature of data.
• The complete-case estimator demonstrated double robustness and optimality within a class of weighting estimators.
• Simulation studies indicated superior performance of the new estimator compared to existing methods.

Conclusions:

• Continuous-time SNMMs provide a flexible and robust approach for causal inference in longitudinal studies with irregular data.
• The developed methods are applicable to real-world scenarios, such as estimating treatment effects in HIV research.