Using Structural Equation Modelling in Routine Clinical Data on Diabetes and Depression: Observational Cohort Study

Area of Science:

• Health Services Research
• Biostatistics
• Clinical Informatics

Background:

• Routine clinical data is increasingly available for health research.
• Structural Equation Modelling (SEM) can create research-friendly constructs from complex clinical variables.
• SEM offers a potential analytic method for large-scale routine clinical data.

Purpose of the Study:

• To apply SEM to routine clinical data in East London to model established associations in type 2 diabetes.
• To investigate the relationship between depression, diabetic control, and healthcare utilization.
• To test the utility of SEM in routine clinical data by modeling known clinical associations.

Main Methods:

• SEM was employed on a large dataset of routine clinical data from East London.
• The analysis investigated associations between depression, diabetic control, diabetes care, mental health treatment, and Accident & Emergency (A&E) use.
• Latent variables and their associations were defined based on existing clinical knowledge.

Main Results:

• Depression was associated with worse diabetic control and increased A&E use.
• Worse diabetic control was unexpectedly linked to lower A&E use.
• Mental health treatment did not significantly impact diabetic control, but diabetes care receipt was linked to better control and other positive outcomes.

Conclusions:

• SEM partially supported the modeling of established clinical associations in type 2 diabetes patients using routine data.
• The study highlighted data quality issues that may have affected SEM's utility.
• Further data improvement could enhance the application of SEM in routine clinical datasets.