Blog

Unconscious Bias in Medicine and How AI Can Help Recognize Social Determinants of Health

 Unconscious Bias in Medicine and How AI  Can Help Recognize Social Determinants of Health
Our care can be biased, and those biases can be subconsciously learned and baked into the way that we deliver care. For instance, a traditional radiology pain standard used to interpret pain severity (and which is also a factor in determining arthroplasty eligibility) was derived from markers in a white patient population, and those markers appear to be blind to other non-diagnostic markers which reflect greater knee pain in women, in Black patients, in patients with lower income and educational status.
green apostrophes Deepscribe

Artificial Intelligence has the potential to offer us tools which can help combat the effects of social determinants of health, particularly those based on systemic racism. For clinicians this can take form in checking our often unconscious biases when it comes to providing care.

Researchers just published a study in Nature Medicine describing a machine learning algorithm that they developed to help interpret pain based on radiology imagery.  As healthcare professionals know, two patients presenting with very similar symptoms can have very different pain severity even though they may not have different diagnoses.  This algorithm was using markers in radiology imagery for knee osteoarthritis not used to diagnose to help interpret pain severity, and the algorithm’s interpretations were much closer to patients’ expressed pain levels than standard measures historically used to interpret pain based on radiology imagery.

Importantly, this algorithm was interpreting higher pain levels for historically marginalized populations -- populations that were experiencing more pain, but whose more severe pain was not being reflected in standard radiology measurements to interpret pain.  We know that underserved populations disproportionately report experiencing higher levels of pain--which might be attributed to added social determinants of health and systemic barriers compounding with diagnoses.  The study looked at gender, race, income, education attainment and the algorithm’s interpretation of these patients’ pain based on imagery reflected this disproportionate pain variation.  

The authors celebrate this algorithm as a tool for healthcare providers to better treat their patients. Under this revised pain severity measure, the percent of Black patients who might be eligible for arthroplasty, a treatment which is known to reduce pain, doubled (from 11% to 22%).  This exciting innovation points to the potential for Artificial Intelligence to support equitable medical care in a way that is conscious of race and other social determinants of health.

Our care can be biased, and those biases can be subconsciously learned and baked into the way that we deliver care.  For instance, a traditional radiology pain standard used to interpret pain severity (and which is also a factor in determining arthroplasty eligibility) was derived from markers in a white patient population, and those markers appear to be blind to other non-diagnostic markers which reflect greater knee pain in women, in Black patients, in patients with lower income and educational status.

We know that recognizing our biases can help us to deliver better care to our patients, and real time data can help us to check ourselves. Real-time data analysis using AI has the potential to give us metrics to compare our care based on race, etc.  There are concerning studies which demonstrate that healthcare professionals are more likely to prescribe stronger painkillers for white patients than they are for other patients.  Perhaps real time analytics derived from Artificial Intelligence can flag if or when that is happening so that we can have checks against ourselves.  So while we should continue to tailor our treatment for each individual, analytics derived from Artificial Intelligence can offer checks for our care.

These real time analysis of our patient populations can help us to identify trends in our patient populations--providing us with data to identify and proactively combat other ways that social determinants are affecting our patients.

A med student’s blog this summer asked the medical profession to:

“Change the equation. Let’s start to view race as a social construct, not a biological one. Let’s develop plans to address the social determinants of health and prioritize equal access to health care for everyone. Let’s raise awareness about systemic racism in health care and confront racist policies. Let’s educate medical students about racial biases and the impact they can have on patient care.”

We agree, and see hope in tools like Artificial Intelligence to help us address the impact of racial bias on medical care and social determinants of health.

Transition Gray Line
Free telemedicine platform

Workflow designed
by doctors, for doctors.

See it in action
Resources

Learn more from others

Watch videoLearn MoreLearn More
Watch videoLearn MoreLearn More
Watch videoLearn MoreLearn More
Line transition
Blog

Unconscious Bias in Medicine and How AI Can Help Recognize Social Determinants of Health

Input your email below to see DeepScribe work during a patient encounter.

Thanks! You unlocked the video!
Watch Video Now
Oops! Something went wrong while submitting the form.
Have a question?

Get in Touch

Bring the joy of care back to your practice today.

Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.
Line transition
Resources

Learn more from others

Watch videoLearn MoreLearn More
Watch videoLearn MoreLearn More
Watch videoLearn MoreLearn More
Line transition