Addressing Selection and Confounding Biases in Dental Claims Data: A Causal Inference Framework for...

Area of Science:

• Causal inference methodology
• Health services research
• Oral and systemic disease research

Background:

• Administrative health care data offer insights into oral and systemic disease links.
• Methodological challenges, particularly selection bias, can compromise causal inference in these datasets.
• Selection bias, arising from restricted insurance eligibility, can distort treatment effect estimates.

Purpose of the Study:

• To demonstrate how selection and confounding biases in administrative health care claims data affect causal inference in periodontal-systemic disease research.
• To introduce methodological approaches for addressing these biases.
• To highlight the need for improved methodology in dental research.

Main Methods:

• Utilized causal inference theory and directed acyclic graphs (DAGs) to distinguish and identify confounding and selection bias.
• Reviewed 7 studies on periodontal-systemic disease associations in claims data.
• Illustrated selection bias effects with a numerical example.
• Introduced causal inference strategies like G-methods and inverse probability of selection weighting.

Main Results:

• A review of existing studies revealed methodological gaps in addressing selection bias.
• Selection bias can distort or even reverse observed associations between periodontal treatment and systemic disease outcomes.
• Established causal inference strategies can correct for confounding and selection bias.

Conclusions:

• Strengthening causal inference methodology is crucial for robust oral-systemic health evidence.
• Routine use of DAGs, bias-correction techniques, and transparent reporting are advocated.
• Improved methodology supports clinical decision-making and oral health integration into overall healthcare.