Is your pulse oximeter biased? Explaining racial bias and its impact in medical technology | News explained

“Equity in Medical Devices: An Independent Review,” published March 11, examined three types of medical devices. These include: 1) Optical medical devices including pulse oximeters. 2) Artificial intelligence (AI) supported medical devices. 3) Polygenic risk score (PRS), which measures genetic risk for susceptibility to disease.

Notably, the study found that “racial bias is highly likely to exist and could pose harm to patients if not identified and addressed.”

What is behind this prejudice?

The main reason for bias, according to the study, is the modeling, development and testing of the instrument on a 'typical patient', where patients are 'typically white, male, relatively affluent and British. Born.”

Regarding optical medical devices such as oximeters, which measure light through a patient's skin, the report says, “This technology can be used to detect a number of factors, including skin pigmentation, nail polish, movement, reduced peripheral blood flow, false tanning, and henna. “It may be influenced by these factors.” , tattoos, and sickle cell disease. ”

Additionally, the study points to multiple instances where biased data from testing devices led to incorrect or delayed diagnoses for women, racial minorities, and people in poor communities. For example, oximeters were only tested on white skin and their readings were considered the “standard,” resulting in inaccurate oxygen readings for Black people and other minorities.

The problem is similar for AI-assisted devices, with the inability to represent diverse datasets when designing measurement scales. Furthermore, ignoring the regional, sociocultural, and economic factors behind health data further reduces the likelihood of accurate interpretation.

“Data is often indirect measurements that reflect patient interactions with health care systems and health status. Therefore, analyzing medical records requires being aware of the context in which they were generated. “Without context, health record data are inappropriate for many research questions,” the study states.

Commenting on polygenic risk scores, review co-author Enitan Carroll, professor of pediatric infectious diseases at the University of Liverpool, said: “The main genetic datasets used by polygenic risk scores are overwhelmingly about people of European descent. This also applies to people with other ancestry.”

Is this bias in technology new?

No, discrimination in technological devices and programs is not new. The fact that optical medical devices perform worse in people with darker skin has been known since 1992.

However, the true impact of this difference on health and healthcare has only recently become clear. This study found evidence of harm caused by poor performance of medical technology in the U.S. health care system. “There is a strong association between racial bias in pulse oximeter performance and delayed recognition of illness, refusal or delay in treatment, and worse organ function,” and black patients compared to white patients. death. ”

Evidence that racism pervades technology even in non-medical cases, from facial recognition software that incorrectly identifies people to no-touch faucets and soap dispensers that fail to recognize non-white hands and palms. There are many.

What are the stakeholders doing about it?

The UK Government has accepted all 18 recommendations, 51 sub-recommendations and three calls to action contained in the report. The recommendations primarily call for prioritizing more diverse data in terms of skin tone, genetics, lifestyle, and other cultural factors.

We also ask developers to be transparent about device restrictions. Regulators will continue to monitor the deployment of such tools. The NHS and other healthcare organizations also regularly seek feedback from the public, experts and healthcare providers to help comply with the above regulations and close any gaps.