AI can do what?

In this week’s blog, I’d like to share a personal story prior to moving into the content that will be covered here shortly. In the Summer of 2018, my brother came back from Medical School, and we decided to play pick-up basketball at our gym. Long story short, I ended up tearing my ACL and meniscus going up for a game winning dunk and didn’t get surgery for an entire year. Let’s walk through some of the reasons why I didn’t receive the care I needed by comparing my experience with Black patients and how AI is helping to reduce racial disparities in healthcare.  Shall we begin?

KNEES HURT MEME 1 | Peter Hurts His Knee | Know Your Meme

We’re well aware of the racial inequalities that exist in our healthcare system, and the pandemic this past year has given healthcare professionals and patients like us more reason to question why some communities suffer disproportionately from diseases such as Covid-19. History tells that access to healthcare and insurance, as well as the quality of care one may receive is unequal and the process of reaching a solution seems arbitrary. Some of the critical factors used to determine access to medical care are biological differences, socioeconomical differences, and geographical differences amongst a long list of many more.

Those that remember the Affordable Care Act, know that the passage resulted in narrowing the gaps in coverage between Black, Hispanic and white patients, but minorities are still less likely to be insured. As a result, minority groups are more likely to avoid seeking medical help because they have a history of medical mistreatment, or have poorer patient outcomes, and even concerns about cost. Mental health, which has been brought to the front of attention in recent years, is another form of treatment people of color have a harder time accessing. According to the National Institute of Mental Health, “while 50% of white people suffering from mental illness receive treatment, roughly a third of Black or Hispanic people with mental illness receive treatment.” This is an extremely large number when applied to a profound diagnosable disease that we as a country are trying to minimize.

Many mental health conditions are often underdiagnosed. For example, “black mothers are more likely to suffer from postpartum mental illness and depression, but less likely to receive treatment than non-Black mothers.” Black patients are also more likely to be inaccurately diagnosed with mental health conditions. I think we see the point, a large disparity. So, what can AI to solve these issues?

There was a study on both Black and white patients who visited orthopedic doctors and complained about having arthritis related knee pain. Though Black patients reported more severe knee pain, they were less likely to get knee surgery than white patients. When AI was first introduced to this study, it replicated the disparities rather than combat them. The algorithms used to identify the disparities highlighted the cause of why Black patients weren’t being treated, which is why it resulted in initial replication studies. To notice a more accurate result diagnosis of disease patterns, the AI algorithms must be trained to focus on specific variables that affect the outcome. As an example, one AI was trained to focus on cost, but Black patients collected lower lifetimes healthcare costs because they had lower incomes and less robust insurance. By switching the variable from cost to comorbid conditions or patient reports of pain, the algorithm could accurately provide impartial solutions.

Disparities in Health and Health Care: 5 Key Questions and Answers | KFF

Although AI tools have the potential to reduce racial disparities in the healthcare space, we need humans to interact with the systems for more robust and accurate solutions. Only then can we continue to improve on the platforms we’ve already created. Teletherapy has helped increase access to all sorts of patients who struggled to have access before, so hopefully AI and other tools developed down the line can do just that, reduce racial disparities in healthcare.

To tie my store back into all of this, I then realized some of my white friends had torn their ACL, had an MRI, were examined by an orthopedic surgeon, and even had surgery all before I could even get an MRI. I’m obviously comparing apples to oranges here, but you can see my point. A torn ACL is a torn ACL, and the timeline to be checked and treated should be the same for all patients, no matter what color is represented by their skin. I’ve seen lots of change in recent years to address these issues, and with time, I certainly have the confidence that one day we’ll all be on one level.

8 comments

  1. This is a really really important blog, I appreciate you drawing attention to these big discrepancies exist in the way that different groups get access to treatment. It’s one of those things that I feel like often goes overlooked (unless it affects you or someone you know directly – like your ACL injury), and one guy who does a good job of bringing those types of issues back into the social consciousness for people is Jon Oliver. He did an episode on racial and gender bias in medicine a couple of years ago, and the whole episode is viewable for free here: https://www.youtube.com/watch?v=TATSAHJKRd8 . It’s worth a watch!

  2. This is a great post that highlights some critical issues in our society. I think AI can be a great tool to address these types of disparities, if used correctly. As you mentioned, some underlying facts are not obvious. Still, I hope that by understanding the kinds of biases we face every day, we will achieve a better and far equal society.

  3. Really interesting blog and ties in really well with Parker’s blog from last week where he talked about machine learning system’s ability to unlearn. There are obvious biases that the engineers have inadvertedly placed within these algorithms and we need more discussion and attention to be brought onto them, in order to fix the underlying issues with the systems. As I was reading this, I was reminded of a vox video I watched a few weeks ago that discussed what role race should play in the medical field. It seems like there are still misunderstandings of the biological differences (or lack there of) in the medical space; it seems obvious that we need to clear up those misconceptions with our healthcare providers before we begin relying on our AI systems in this space. https://www.youtube.com/watch?v=0cjSEWZ8LM8

  4. Much like @dropitlikeithox discusses above, I, too, wondered how AI can backfire this important initiative. If AI is learning from bad habits/data, it will perpetuate such. I imagine this is what you were talking about what you said that “we need humans to interact with the systems for more robust and accurate solutions.”

    Teletherapy’s ability to increase access is really important and exciting. Granted, sort of like the discussion we had in class regarding whether public transit can go full digital and thus exclude those without smart phones, I worry about the same thing here. Nonetheless, if it has “already helped increase access to all sorts of patients who struggled to have access before,” then it’s doing its job.

    Thanks for the interesting and forward-thinking topic. And for ‘walking’ us through the material, with a previously torn ACL!

  5. I think this is a great post and really highlights some of the major issues in our health services that people like to ignore or just are not aware. Utilizing AI to assist people into making smarter and better choices will definitely be a great way to improve the system for everyone that uses it an not just the ones that can afford it.

  6. Really great and thoughtful post. I think one of my coauthors mentioned last week when we interviewed the CIO of Anthem, they were using AI to identify the patients who were most likely to put off care as a result of COVID, and target them for interventions. I am hopeful AI could also help address some of the disparities you note here. Becuase, the bottom line is that sometimes in healthcare, the cheapest solution is to make sure people get good care early on, rather than waiting for problems to develop. A win-win. Thank you for sharing your personal story.

  7. Firstly, thanks for sharing this personal story and connecting it to a larger issue. Secondly, did you win the game? Lastly, I think AI can certainly help in the healthcare industry, especially with preventative care. I think you make an important point about telehealth being a critical factor for increasing access to healthcare. Being able to connect with a live healthcare professional can connect a patient with a doctor that they know and trust (or start the process of building trust). There is some balance of AI, telehealth, and personal connection that could help improve access to and quality of healthcare.

  8. Thank you for sharing – this is so important to discuss. I completely agree that AI has the potential to reduce racial disparities. At the same time it has the potential to exacerbate pre-existing disparities by using historical data. This is something that I have discussed in many of my other classes at BC which is so important. AI without human interaction and adjustments can lead to biases in hiring, healthcare, housing, and many other scenarios.

Leave a Reply

Please log in using one of these methods to post your comment:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: