Using nationally representative survey data for external adjustment of unmeasured confounders: An example using the...

Area of Science:

• Epidemiology
• Health Services Research
• Biostatistics

Background:

• Healthcare databases often have incomplete data on confounders like smoking and obesity.
• Accurate risk estimation in observational studies requires addressing unmeasured confounding factors.
• Population-based surveys offer valuable supplementary data for improving health research validity.

Purpose of the Study:

• To evaluate the utility of national survey data, such as NHANES, for external adjustment of imperfectly measured confounders in US healthcare databases.
• To assess the impact of incorporating external data on the reliability of risk estimates from observational health studies.

Main Methods:

• Utilized Medicaid Analytic eXtract (MAX) data to estimate the relative risk (RR) of prenatal serotonin-norepinephrine reuptake inhibitors (SNRIs) exposure and cardiac defects.
• Employed National Health and Nutrition Examination Survey (NHANES) data for prevalence of smoking and obesity among women with depression.
• Performed sensitivity analyses to correct the RR using literature-based estimates of confounder-outcome associations.

Main Results:

• The unadjusted RR for prenatal SNRI exposure and cardiac defects in MAX was 1.51; adjusted RR was 1.20.
• NHANES data indicated higher prevalence of smoking (60.2%) and obesity (59.2%) in SNRI users with depression compared to non-users.
• External adjustment for smoking and obesity could potentially reduce the RR to approximately 1.10, assuming independence of effects.

Conclusions:

• National surveys like NHANES are accessible resources for obtaining information on potential confounders.
• This approach can enhance the validity of relative risk estimates in observational studies lacking crucial risk factor data.
• External adjustment using population survey data is a viable strategy to mitigate bias from unmeasured confounding in health research.