Sean Kim

MS4, Weill Cornell Medicine

[Sean Kim @SeanHunKim]

@MClerkship I like the step by step approach the infographic shows. It's easy to feel like the interpreter service is an added burden to the workflow but it's so important to keep in mind that it is a vital tool that enables patient communication to begin with! @languagelinePR

[Anjana Rajan @AnjanaRajanMD]

Limited English Proficiency (LEP) patients receive worse palliative care compared to English-speaking counterparts - they are more likely to die in the hospital (instead of at home) and experience more pain and anxiety at the end of life.

[Sean Kim @SeanHunKim]

Algorithm trained on a homogenous data set will further reinforce the existing disparities in the community and hospital that the training data came from. It would be dangerous to assume that such tech is generalizable.
@MClerkship https://t.co/v454UiqE8F

[Sean Kim @SeanHunKim]

Just as many RCTs have historically failed to incorporate heterogenous samples into their studies, AI in healthcare is prone to the same intrinsic bias that further reinforces disparities. #healthequity, #socialdeterminantsofhealth, #SDOH, #socialjustice, #healthdisparities

[Sean Kim @SeanHunKim]

@Anjana_Rajan1 @MClerkship Would be interesting to study the reasons for such difference. Will medicaid rates and charity care implicitly or even explicitly affect hospital admins and downstream physicians' decision making in acute settings?

[Sean Kim @SeanHunKim]

@AliceZhaoS Would be interesting to see how much the price cap translated into higher premiums for the enrollees of the plans. Insurers can also easily shift the cost by charging more for other services. Another reason we need a systemic change.

[Anjana Rajan @AnjanaRajanMD]

1/ According to @JAMAInternalMed, homeless adults admitted for cardiovascular disease are more likely to die in the hospital than nonhomeless adults. #socialdeterminantsofhealth #healthequity #socialjustice #healthdisparities

https://t.co/Rx3OFRLZyY

[Sean Kim @SeanHunKim]

This presents a new emerging structural bias in healthcare. This time, the bias is not only implicit but also automated. Conscious effort to diversify training data sets and carefully narrow applications of AI to specialized use cases will lessen the bias.