Statistical methods for overcoming residual confounding with administrative health data

I have extensive administrative health data (general patient demographics + medical information, place of residence, many study years, few outcome measures) that have limited variables for effectively controlling for confounding. Thus, usual methods like multiple regression are not sufficient for getting unbiased results from analyses. However, I have read/hear about more advanced statistical techniques that allow better controlling of confounding (e.g. using a conditional regional or temporal variance of an X on Y).

Could you please provide names of statistical techniques that I should google for?

My own research lead to these techniques:

  • Use of causal analysis (DAGs) (I know it only works out when my DAG is correct).
  • Front-door criterion
  • Instrumental variable

Not really about Stan, but a relevant presentation The rise of R in public health research institutes - RBelgium open content

1 Like