We are all exposed to multiple risk factors in variable quantities throughout our lives and, currently, these are virtually impossible to measure.
Moreover, for all the cohorts except African American males, age has multiple interactions with other risk variables.
The possibility of including multiple risk factors in an instrumental-variable analysis is discussed in many econometric textbooks (8), and applied instrumental-variable analyses involving multiple risk factors have been performed (9, 10), but we are unaware of any application of the approach in genetic epidemiology.
We used longitudinal Poisson regression, within a generalized estimating equations (GEE) framework, to assess the longitudinal associations between selected individual/social distal and ultimate variables and the multiple risk factor score.
A multivariable analysis was conducted with in-hospital mortality as the outcome variable and with multiple risk factors to include atorvastatin and simvastatin use.
Cox regression analysis with forward stepwise variable selection was employed to build a model with multiple risk factors.
Risk variables that had significant association were included as independent variables in multiple linear regression analysis.
Logistic regression analysis examined the association between EI and risk variables, adjusted for by baseline characteristics (gestational age < 28 weeks, multiple birth and male gender).
The final model on multiple risks had the following variables: poor knowledge on night sweat, chest pain, having no employment, residing > 5 km from a nearby health facility with DOTS services and a belief that TB is always associated with HIV infection.
Statistical tests to select discriminate variables also present the problem of controlling multiple risks for which many solutions have been developed in biostatistics [ 40].
