Thread by @iwashyna

@MattWhite_95 @WesElyMD @DrDaleNeedham @SapnaKmd @tmprowell @HUMANIZALAUCI @A_MacLullich @Krockdoc @geri_doc @drdangayach @NidaQadirMD This is another one of those things harder to discuss over twitter than in-person

@MattWhite_95 @WesElyMD @DrDaleNeedham @SapnaKmd @tmprowell @HUMANIZALAUCI @A_MacLullich @Krockdoc @geri_doc @drdangayach @NidaQadirMD When we ask "is race associated with mortality?", we want to be clear what we are asking. Dr @RheaBoydMD has written about this better than I will, so I just refer you to her article

healthaffairs.org/do/10.1377/hbl…

@MattWhite_95 @WesElyMD @DrDaleNeedham @SapnaKmd @tmprowell @HUMANIZALAUCI @A_MacLullich @Krockdoc @geri_doc @drdangayach @NidaQadirMD @RheaBoydMD Bottomline: we rarely think the melanin content of the skin is directly leading to mortality. most of the time we really mean "is likely exposure to racism leading to mortality?"

@MattWhite_95 @WesElyMD @DrDaleNeedham @SapnaKmd @tmprowell @HUMANIZALAUCI @A_MacLullich @Krockdoc @geri_doc @drdangayach @NidaQadirMD @RheaBoydMD so one would want, in this case, to be clear about how it could act, and decide if you want to "adjust that away"

@MattWhite_95 @WesElyMD @DrDaleNeedham @SapnaKmd @tmprowell @HUMANIZALAUCI @A_MacLullich @Krockdoc @geri_doc @drdangayach @NidaQadirMD @RheaBoydMD Every model included a "hospital random effect", which means any effects of racism that are mediated by differences in which hospital people go to -- because neighborhood segregation, for example -- are "adjusted out"

@MattWhite_95 @WesElyMD @DrDaleNeedham @SapnaKmd @tmprowell @HUMANIZALAUCI @A_MacLullich @Krockdoc @geri_doc @drdangayach @NidaQadirMD @RheaBoydMD to the extent that there are differences in average quality of care between hospitals for all their patients -- if Black patients disportionately go to less resourced hospitals -- than that is deemed "not an effect of race" and removed from their models

@MattWhite_95 @WesElyMD @DrDaleNeedham @SapnaKmd @tmprowell @HUMANIZALAUCI @A_MacLullich @Krockdoc @geri_doc @drdangayach @NidaQadirMD @RheaBoydMD They seem to always adjust for how ill people are "on arrival" (after triage in some sense), so differences in how effective hospitals are in getting people in and welcoming people -- also "not an effect of race"

@MattWhite_95 @WesElyMD @DrDaleNeedham @SapnaKmd @tmprowell @HUMANIZALAUCI @A_MacLullich @Krockdoc @geri_doc @drdangayach @NidaQadirMD @RheaBoydMD The differences in the neighborhoods from which people originated -- and residential segregation in the US definitely means there will be differences that are associated with outcomes -- also adjusted away

@MattWhite_95 @WesElyMD @DrDaleNeedham @SapnaKmd @tmprowell @HUMANIZALAUCI @A_MacLullich @Krockdoc @geri_doc @drdangayach @NidaQadirMD @RheaBoydMD Differences in how sick patients were because of racially differential previous access to care -- adjusted away

@MattWhite_95 @WesElyMD @DrDaleNeedham @SapnaKmd @tmprowell @HUMANIZALAUCI @A_MacLullich @Krockdoc @geri_doc @drdangayach @NidaQadirMD @RheaBoydMD and I can't quite tell, but I think they ran the analyses adjusted for or stratified by ICU use and mechanical ventilation use -- which means any differences in care in the hospital that lead to differential decompensation, are also adjusted away