Risk adjustment for cesarean delivery rates: how many variables do we need? An observational study using...

Area of Science:

• Obstetrics and Gynecology
• Health Services Research
• Biostatistics

Background:

• Inter-hospital comparisons require case-mix adjustment and risk adjustment for cesarean delivery (CD) predictors.
• Various data sources have been utilized to gather information on CD predictors.
• Assessing the impact of different data sources on predictive model performance is crucial.

Purpose of the Study:

• To compare the discrimination and fit of cesarean delivery (CD) predictive models using different data sources.
• To evaluate if more complex models enhance inter-hospital comparisons.
• To determine the added value of birth certificate and socio-demographic data in CD prediction.

Main Methods:

• Four predictive models for cesarean delivery (CD) were developed.
• Models incorporated variables from Hospital Discharge Records, previous hospitalizations, birth certificates (BC), and socio-demographic data.
• Model performance was assessed using Receiver Operator Curves and Akaike/Bayesian Information Criteria.

Main Results:

• Incorporating birth certificate data significantly improved model discrimination and fit.
• Inclusion of socio-demographic variables or past comorbidities did not enhance model performance.
• Hospital-specific CD rates derived from the models demonstrated high correlation.

Conclusions:

• Record linkage, particularly with birth certificates, enhances predictive model performance for cesarean delivery (CD).
• Despite improved model accuracy, record linkage does not substantially alter inter-hospital comparisons.
• The findings suggest that while individual model precision can be improved, the relative ranking of hospitals remains consistent.