Examining differential item functioning in self-reported health survey data: via multilevel modeling

Area of Science:

• Health Services Research
• Biostatistics
• Psychometrics

Background:

• Measurement equivalence is critical for health survey data, but traditional methods struggle with hierarchical structures.
• Differential item functioning (DIF) analysis is essential for assessing measurement equivalence in health surveys.
• Hierarchical data structures in health surveys necessitate advanced analytical approaches.

Purpose of the Study:

• To evaluate the benefits of multilevel modeling for DIF analysis in health surveys.
• To compare multilevel DIF analysis with traditional single-level methods.
• To demonstrate the application of multilevel models for binary and polytomous health survey data.

Main Methods:

• Applied multilevel binary logistic regression for binary response data.
• Applied multilevel multinomial logistic regression for polytomous response data.
• Compared multilevel models against their single-level counterparts using health survey data.

Main Results:

• Multilevel models demonstrated a better fit compared to single-level models.
• Multilevel models explained significantly more variance in the data.
• The study confirmed the superiority of multilevel approaches for DIF analysis in hierarchical health surveys.

Conclusions:

• Multilevel modeling is a more effective approach for DIF analysis in hierarchical health survey data.
• Healthcare researchers and practitioners can benefit from understanding and applying multilevel modeling for improved measurement equivalence.
• This study advocates for the adoption of multilevel models to enhance the reliability and validity of health survey research.