class: center, middle # Contextualizing Sensitivity Analysis in Observational Studies Lucy D'Agostino McGowan Robert A Greevy, Jr --- ## data <span class="fa-stack fa-lg"> <i class="fa fa-circle fa-stack-2x"></i> <i class="fa fa-heart fa-stack-1x fa-inverse"></i> </span> Right Heart Catheterization <span class="fa-stack fa-lg"> <i class="fa fa-circle fa-stack-2x"></i> <i class="fa fa-frown-o fa-stack-1x fa-inverse"></i> </span> 30-day Mortality <span class="fa-stack fa-lg"> <i class="fa fa-circle fa-stack-2x"></i> <i class="fa fa-table fa-stack-1x fa-inverse"></i> </span> 50 Covariates .footnote[Connors, A F, T Speroff, N V Dawson, and C Thomas. 1996. "The effectiveness of right heart catheterization in the initial care of critically III patients. Jama.] --- layout: true .footnote[Lucy D'Agostino McGowan \#JSM2017] --- ## recommended components <span class="fa-stack fa-lg"> <i class="fa fa-circle fa-stack-2x"></i> <i class="fa fa-user-md fa-stack-1x fa-inverse"></i> </span> Include content-matter expertise <span class="fa-stack fa-lg"> <i class="fa fa-circle fa-stack-2x"></i> <i class="fa fa-anchor fa-stack-1x fa-inverse"></i> </span> Anchor a sensitivity analysis with your data <span class="fa-stack fa-lg"> <i class="fa fa-circle fa-stack-2x"></i> <i class="fa fa-balance-scale fa-stack-1x fa-inverse"></i> </span> Calculate a hypothetical "**tipping point**" confounder --- ## recommended components <span class="fa-stack fa-lg"> <i class="fa fa-circle fa-stack-2x fa-inverse-2"></i> <i class="fa fa-user-md fa-stack-1x fa-inverse"></i> </span> Include content matter expertise <span class="fa-stack fa-lg"> <i class="fa fa-circle fa-stack-2x"></i> <i class="fa fa-anchor fa-stack-1x fa-inverse"></i> </span> Anchor a sensitivity analysis with your data <span class="fa-stack fa-lg"> <i class="fa fa-circle fa-stack-2x fa-inverse-2"></i> <i class="fa fa-balance-scale fa-stack-1x fa-inverse"></i> </span> Calculate a hypothetical "**tipping point**" confounder --- ## recommended components <span class="fa-stack fa-lg"> <i class="fa fa-circle fa-stack-2x fa-inverse-2"></i> <i class="fa fa-user-md fa-stack-1x fa-inverse"></i> </span> Include content matter expertise <span class="fa-stack fa-lg"> <i class="fa fa-circle fa-stack-2x fa-inverse-2"></i> <i class="fa fa-anchor fa-stack-1x fa-inverse"></i> </span> Anchor a sensitivity analysis with your data <span class="fa-stack fa-lg"> <i class="fa fa-circle fa-stack-2x"></i> <i class="fa fa-balance-scale fa-stack-1x fa-inverse"></i> </span> Calculate a hypothetical "**tipping point**" confounder --- ## recommended components <span class="fa-stack fa-lg"> <i class="fa fa-circle fa-stack-2x fa-inverse-2"></i> <i class="fa fa-user-md fa-stack-1x fa-inverse"></i> </span> Include content matter expertise <span class="fa-stack fa-lg"> <i class="fa fa-circle fa-stack-2x fa-inverse-2"></i> <i class="fa fa-anchor fa-stack-1x fa-inverse"></i> </span> Anchor a sensitivity analysis with your data <span class="fa-stack fa-lg"> <i class="fa fa-circle fa-stack-2x"></i> <i class="fa fa-balance-scale fa-stack-1x fa-inverse"></i> </span> Calculate a hypothetical ["**tipping point**"](http://rpubs.com/lucymcgowan/enar2017) confounder ```r devtools::install_github("LucyMcGowan/tipr") tip_with_binary(p1 = .5, p0 = 0, lb = 1.2, ub = 1.5) ``` --- ## anchor your sensitivity analysis <span class="fa-stack fa-lg"> <i class="fa fa-circle fa-stack-2x"></i> <i class="fa fa-balance-scale fa-stack-1x fa-inverse"></i> </span> Imbalance of **exposure** <span class="fa-stack fa-lg"> <i class="fa fa-circle fa-stack-2x"></i> <i class="fa fa-line-chart fa-stack-1x fa-inverse"></i> </span> Predictive of **outcome** <span class="fa-stack fa-lg"> <i class="fa fa-circle fa-stack-2x"></i> <i class="fa fa-group fa-stack-1x fa-inverse"></i> </span> Independent of **other covariates** --- ## anchor your sensitivity analysis <span class="fa-stack fa-lg"> <i class="fa fa-circle fa-stack-2x"></i> <i class="fa fa-balance-scale fa-stack-1x fa-inverse"></i> </span> Imbalance of **exposure** <span class="fa-stack fa-lg"> <i class="fa fa-circle fa-stack-2x fa-inverse-2"></i> <i class="fa fa-line-chart fa-stack-1x fa-inverse"></i> </span> Predictive of **outcome** <span class="fa-stack fa-lg"> <i class="fa fa-circle fa-stack-2x fa-inverse-2"></i> <i class="fa fa-group fa-stack-1x fa-inverse"></i> </span> Independent of **other covariates** ---  ---  ---  --- ## anchor your sensitivity analysis <span class="fa-stack fa-lg"> <i class="fa fa-circle fa-stack-2x fa-inverse-2"></i> <i class="fa fa-balance-scale fa-stack-1x fa-inverse"></i> </span> Imbalance of **exposure** <span class="fa-stack fa-lg"> <i class="fa fa-circle fa-stack-2x "></i> <i class="fa fa-line-chart fa-stack-1x fa-inverse"></i> </span> Predictive of **outcome** <span class="fa-stack fa-lg"> <i class="fa fa-circle fa-stack-2x"></i> <i class="fa fa-group fa-stack-1x fa-inverse"></i> </span> Independent of **other covariates** ---  ---  ---  ---  ---  ---  ---  ---  ---  --- layout: true --- ## Lucy D'Agostino McGowan <span class="fa-stack fa-lg"> <i class="fa fa-circle fa-stack-2x"></i> <i class="fa fa-question fa-stack-1x fa-inverse"></i> </span> <b>Poster #34 Baltimore Convention Center Halls A&B 10:30a</b> <i class="fa fa-cog fa-spin fa-2x fa-fw"></i> [http://rpubs.com/lucymcgowan/jsm2017_poster](http://rpubs.com/lucymcgowan/jsm2017_poster) <span class="fa-stack fa-lg"> <i class="fa fa-circle fa-stack-2x"></i> <i class="fa fa-toggle-right fa-stack-1x fa-inverse"></i> </span> [http://rpubs.com/lucymcgowan/jsm2017](http://rpubs.com/lucymcgowan/jsm2017) <span class="fa-stack fa-lg"> <i class="fa fa-circle fa-stack-2x"></i> <i class="fa fa-twitter fa-stack-1x fa-inverse"></i> </span> [@LucyStats](https://twitter.com/LucyStats) .footnote[Slides created via the R package [**xaringan**](https://github.com/yihui/xaringan).]