Addressing missing data in real-world administrative health datasets

Area of Science:

• Health Services Research
• Biostatistics
• Data Science

Background:

• Administrative health data are crucial for healthcare insights but often suffer from missing data, compromising research validity.
• Missing data is a significant obstacle in ensuring the accuracy and reliability of findings derived from administrative health datasets.

Purpose of the Study:

• To present a structured methodology for addressing missing data in administrative health datasets.
• To enhance the reliability and policy relevance of healthcare research utilizing administrative data.

Main Methods:

• Utilizing causal diagrams to understand data missingness.
• Applying statistical methods, including multiple imputation, within the Treatment And Reporting of Missing data in Observational Studies (TARMOS) framework.
• Demonstrating the approach with a real-world example in large-scale health research.

Main Results:

• A structured approach to data assessment and statistical methods for handling missingness was developed.
• Multiple imputation techniques were effectively demonstrated on a large-scale administrative health dataset.
• The proposed methods enhance the transparency and rigor of analyzing administrative health data.

Conclusions:

• The structured approach improves the veracity of findings from administrative health data.
• Transparent and rigorous methods are essential for reliable and policy-relevant healthcare research using administrative data.
• This methodology addresses a key barrier in administrative health data analysis, promoting more trustworthy research outcomes.