Malcolm Barrett

Would likely need to expand our DAGs to include these

This is an interesting problem, and I bumped into [this paper](https://arxiv.org/abs/2111.04513) on [this twitter thread](https://twitter.com/alexpghayes/status/1505944287221063687)

Immortal time bias

1

* in the target trials section * in data prep section * in sensitivity analysis

The recent update to gtsummary includes easy-to-add robust SEs, which may be a nice addition to what we show for bootstraps: http://www.danieldsjoberg.com/gtsummary/reference/custom_tidiers.html

The answers here are pretty good https://stats.stackexchange.com/questions/481110/propensity-score-matching-what-is-the-problem