Attributable risk estimation for adjusted disability multistate models: application to nosocomial infections

Area of Science:

• Clinical Epidemiology
• Biostatistics
• Health Services Research

Background:

• Attributable risk is crucial in clinical epidemiology for understanding disease burden.
• Existing multistate models for nosocomial infections often overlook population heterogeneity.
• Competing risks like death and discharge are present in healthcare settings.

Purpose of the Study:

• To extend existing multistate models by incorporating covariates to estimate attributable risk.
• To define and estimate overall and profiled attributable risk using an adjusted model.
• To assess the impact of covariates on attributable risk estimation for nosocomial infections.

Main Methods:

• Developed an adjusted disability multistate model including covariates in each transition.
• Employed a semiparametric approach for model and attributable risk estimation.
• Conducted a simulation study to evaluate model performance and bias.

Main Results:

• The adjusted multistate model accounts for population heterogeneity and time-dependent risk factors.
• Neglecting covariates in the model leads to significant bias in attributable risk estimation.
• The methodology was applied to ventilator-associated pneumonia data from French intensive care units.

Conclusions:

• The adjusted disability multistate model provides a more accurate estimation of attributable risk for nosocomial infections.
• Incorporating covariates is essential to avoid bias in epidemiological studies.
• This approach enhances the understanding of risk factors in intensive care settings.