Medical Technology: Intersectional Approaches

The Challenge

Medical technology—in this case study, especially devices, diagnostics, products, and services—harms twice as many females and other patients who do not fit a typical male profile, such as gender-diverse individuals, as males. These technologies can embed ethnic biases as well. Pulse oximeters, for example, fail to correctly analyze oxygen levels in people with darker skin far more often than in people with lighter skin.

Method: Intersectional Approaches

Medical technology can be biased in multiple and compounding ways. Technologists should consider relevant intersectional analyses, for example, when developing and calibrating medical devices.

Gendered Innovation 1: Pulse Oximeters Adjusted for Skin Tone

Pulse oximeters may not work well for users with darker skin, and these patients may not get the supplemental oxygen they need. Pulse oximeters can be developed to take tissue color (e.g., skin tone or melanin content) into account.

Gendered Innovation 2:Total Hip transplants and Bioprinted Kidneys

Hip transplants have very high success rates; but for those that fail, they fail more often for females. Sex differences will also be important for organ bioprinting: For female patients, technologists must decide which X-chromosome activated cells to use. For better outcomes, sex can be considered when identifying the problem, designing research, and collecting and analyzing data.

Gendered Innovation 3: Smaller Artificial Hearts

Only 12% of the 1,596 artificial heart implants worldwide since 1982 have been implanted in women and only 5% in children—because the heart is typically too big for these populations. Recently a smaller heart has been developed for smaller patients.

Gendered Innovation 4: Correcting Dataset Bias in X-Ray Imaging

Researchers have found that diagnostic algorithms are less accurate for females because females are often underrepresented in datasets. Recent studies show that sex-balanced datasets achieve the best results for females and males.

Gendered Innovation 5: FemTech

“FemTech” (female technology) refers to software, diagnostics, products, and services that use technology to provide better treatments for females and greater gender equity in the healthcare system. FemTech is currently expanding to include non-binary and trans people not born biologically female.

The Challenge

Medical technology—in this case study, especially devices, diagnostics, products, and services—harms twice as many females and other patients who do not fit a typical male profile, such as gender-diverse individuals, as males. A 2019 study showed that females made up 67% of 340,000 patients injured by medical devices. Because technology is developed and tested more often in males, females are more likely to suffer complications from medical technology than males (Duffy et al., 2020). These technologies can embed ethnic biases as well. Pulse oximeters, for example, fail more often for darker skin tones. Intersectional approaches are needed to promote fairness in medical devices.

Portions of this case study were first published in Zou, J., & Schiebinger, L. (2021). Ensuring that biomedical AI benefits diverse populations. EBioMedicine, 67, 1-6.

Gendered Innovation 1: Pulse Oximeters Adjusted for Skin Tone

COVID-19 has brought new scrutiny to a common medical device: the pulse oximeter, first invented in Japan in 1972 (Severinghaus & Honda, 1987). Pulse oximeters—able to measure oxygen levels without drawing arterial blood—were among the first defenders against the impacts of the COVID-19 pandemic.

Pulse oximeters, however, may not work well for patients with darker skin. (Keller & Harrison-Smith, 2022). Pulse oximeters measure oxygen saturation in the blood by shining infrared and red light through the finger. The problem is that both the deoxyhemoglobin in the blood and the melanin in the skin absorb light, meaning that pulse oximeters may overestimate oxygen saturation in patients with darker skin (Feiner et al., 2007). The results is that these patients may not get the supplemental oxygen needed to avoid damage to vital organs (Moran-Thomas, 2020).

A recent study compared oxygen saturation measures taken with pulse oximeters with those taken from arterial blood gas. Analysis of over 47,000 paired readings found that oximeters misread blood gases 12% of the time in Black patients compared to 4% of the time in white patients (Sjoding et al., 2020). A further study confirmed that pulse oximetry inaccuracies were associated with significantly delayed or unrecognized eligibility for COVID-19 therapies among Black and Hispanic patients (Fawzy et al., 2022). Similar problems affect consumer wearables that use infrared, red, or green light signaling (Colvonen et al., 2020).

Medical researchers have known since 1989 that both deoxyhemoglobin and melanin in skin are primary light absorbers (Ries et al., 1989). An early oximeter patent to adjust for skin tone was filed in 1999 (Chin, 1999). Further patents were filed in 2019 and 2020 (Barker et al., 2019; Bechtel et al., 2020), both of which take tissue color (e.g., skin tone or melanin content) into account. Device developers, however, have been slow to take action (Colvonen et al., 2020; Moran-Thomas, 2020).

The pulse oximeter case exemplifies a broader phenomenon whereby data from individuals with darker skin are often underrepresented in evaluations of medical devices, findings, and algorithms. A recent survey of 36 papers describing cutaneous manifestation associated with COVID-19 showed that none of those papers included photos of darker skins (Lester et al., 2020).

Method: Intersectional Analysis

Pulse oximeter outcomes are worse for patients with darker skin. But this data has not been disaggregated by sex (Zou & Schiebinger, 2021). How does sex influence results? Does sex intersect with race, meaning that results are significantly worse for darker-skinned women or gender-diverse individuals? (Interestingly, to find articles about sex in pulse oximetry, one often needed to search “gender”; medical researchers tend to use these distinct terms interchangeably and often incorrectly [Madsen et al., 2017].) Findings are highly inconclusive. Feiner et al. suggested in 2007 that women and others with smaller fingers exhibit greater variability in oximeter performance, especially at low arterial saturation levels. One study that took an intersectional approach found that peripheral oxygen saturation values less than 97% were 6 times more frequent in light-skinned males than in dark-skinned females (Witting & Scharf, 2008). A very small 2020 study pointed out, however, that low hemoglobin levels were prevalent in females; thus it was not possible to statistically separate the contributions of sex and low hemoglobin to oximeter readings (Choi et al., 2020). The 2020 study of over 47,000 patients found that oximeters misread hypoxemia more often in females than in males—but the difference was slight compared to differences related to skin tone. Nonetheless, this means that accuracy for females with darker skin would be the most compromised (Sjoding, 2020).

Intersectionality is a technique that needs to be incorporated into analyses to overcome health inequalities. Device makers should consider relevant intersectional analyses when calibrating medical devices.

While the impact of sex on pulse oximetry remains unclear, that of gender may be significant: a randomized, blind study found that nail polish (worn more often by people who identify as women than as men) interfered with oximetry, with black, blue, and green lowering the accuracy of reading more than purple or red (Coté et al., 1988). The authors recommend medical professionals remove patient nail polish before employing an oximeter. Oximeter readings are further influenced by anemia, jaundice, poor circulation, cold fingers, current tobacco use, age body shape, and device placement (Brytanova et al., 2022).

More research is needed to understand intersecting human characteristics of sex, gender, race/ethnicity, socioeconomic status, age, etc. to enhance health outcomes across the whole of society.

View General Method

 

Gendered Innovation 2: Total Hip Transplants and Bioprinted Kidneys

Hip transplants have very high success rates, but for those that fail, they fail more often for females or gender-diverse individuals who do not fit a typical male profile. After considering patient-, surgery-, surgeon-, volume-, and implant-specific risk factors, females had a 29% higher risk of implant failure than males after a total hip replacement (Inacio et al., 2013). One reason is that females tend to have stronge immunological reactions to metal-containing devices. These reactions impact 49% of females, compared to 38% of males (Caicedo et al., 2017). Differences in bone anatomy and the prevalence of osteoporosis can also affect outcomes (Hartman et al., 2009). Females also have a higher risk for surgical revision, adverse local tissue reaction, dislocation, and aseptic loosening (U.S. FDA, 2019).

It is important to understand sex differences. Within the next decade, physicians may bioprint a kidney for organ transplant using a frame seeded with a patient’s own cells. If that patient is female, however, the surgeon must decide which X-chromosome activated cells to use. And, because of the complex X-inactivation in females, the 3-D bioprinted graph will likely never be an exact copy of the original kidney, which, of course, is the goal (van Daal et al., 2020).

Gendered Innovation 3. Smaller Artificial Hearts

Sex-based malfunctions of cardiovascular devices, such as ventricular assist devices, implantable cardioverter-type defibrillators, and drug-eluting coronary stents, are well known (U.S. FDA, 2019). Here we discuss artificial heart implants.

Only 12% of the 1,596 artificial heart implants worldwide since 1982 have been implanted in women and only 5% in children—because the heart is too big for these populations. The original, male heart was 70cc. Recently, SynCardia Systems, headquartered in Tucson, AZ, has developed a 50cc heart. This 29% reduction in size allows smaller patients—numerous females, smaller males, and children typically ten years of age and up—to receive the heart. Trials to date show a 50% success rate (Wells et al., 2017).

Gendered Innovation 4. Correcting Dataset Bias in X-Ray Imaging

Increasingly, medical devices are used to collect large datasets that are then processed using artificial intelligence systems. This can introduce human-generated bias. Algorithms can learn, for example, diagnostic patterns from x-ray imaging datasets. Female patients have been underrepresented in these datasets: only 40% of images in common chest x-ray datasets are from female patients. This imbalance results in diagnostic algorithms that are less accurate for females than for males. Interestingly, researchers have found that a gender-balanced dataset achieved the best results for females and males (Kadambi, 2021; Larrazabal et al., 2020). The strategies machine learning researchers are using to avoid and correct bias may become important for medical researchers as diagnostic algorithms become more common in medicine.

Gendered Innovation 5. FemTech

Women file only 13 percent of patents in the U.S. Rem Koning at Harvard Business School has estimated that if all biomedical patents filed between 1976 and 2010 had been produced equally by women and men, there would be some 6500 more female-focused biomedical inventions (Koning et al., 2021).

We are beginning to see these benefits in the FemTech revolution. “FemTech” (female technology) refers to “software, diagnostics, products, and services that use technology to focus on females’ health” (Wiederhold, 2021). The term was coined in 2016 by Ida Tin, the Danish-born founder of Clue, a period and ovulation tracking app established in Germany in 2013 (FemTech Live, 2021). And FemTech is taking off: Estimates for the industry’s current market size range from $500 million to $1 billion (McKinsey & Company, 2022).

The goal of much of FemTech is better treatments for females and greater gender equity in the healthcare system. Take Evvy, a U.S. startup that provides at-home vaginal microbiome tests. People with vaginas may suffer from infections and discomfort that are mis- or undiagnosed. Evvy helps by allowing users to balance their vaginal microbiome to avoid infections, thus putting the power to improve their health in their own hands.

In Japan, Ring Echo is a new technology developed by Lily MedTech to replace the mammogram, a test that many people find extremely painful. The Ring Echo system is designed to be highly accurate and painless. An examinee lies facedown on a bed-type device and puts their breast into a hole filled with water. The ring-type ultrasonic transducer then moves up and down, taking a three-dimensional scan of the breast to locate any potential cancers.

Some criticize FemTech for excluding non-binary and trans people not born biologically female. Queerly Health is one startup exclusively dedicated to the lesbian, gay, bisexual, transgender, queer or questioning, intersex, asexual, two-spirited, or others. It allows participants in the U.S. to book LGBTQ+ friendly health and wellness practitioners digitally. Another startup, Plume, provides gender-affirming hormone replacement therapies and medical consultations tailored to the trans community (Agarwal, 2021). In Berlin, Germany, Clue, the original FemTech company, defines its work as focused on “women and people with cycles”; it has also developed a focus on LGBTQ+ health (Lovett, 2020).

Next Steps: Funding and Regulatory Agencies

Funding and regulatory agencies have an important role to play to ensure that medical technology works for everyone. The U.S. National Institute of Health has taken a leading role by requiring that all public-funded research consider sex as a biological variable since 2016 (Arnegard et al., 2020). The Canadian Institute of Health Research (2010), the European Commission (2014), and the German Research Foundation (2020) have all implemented guidelines for sex, gender, and diversity analysis in proposals where relevant, in efforts to enhance equity in research outcomes. Implementation of such policies, however, depends on training researchers, evaluators, and staff in inclusive, intersectional research methodologies as well as regular reviews of the efficacy of these policies (Hunt & Schiebinger, 2022).

Regulatory agencies hold the keys to medical technology safety. In 2014, the U.S. FDA issued a guidance outlining FDA’s expectations regarding sex-specific patient enrollment, data analysis, and reporting of study information for medical device applications (U.S. FDA, 2019). These, however, are recommendations and not mandated requirements. In addition, in 2016 the Center for Devices and Radiological Health (CDRH) established the Health of Women Program to eliminate potential sex and gender biases in the design of medical devices and clinical trial. The CDRH vision is that with an integrated approach, innovators, researchers, and FDA staff have the tools necessary to consider how sex and gender are factored into research designs, device development, and analyses of studies in humans, vertebrate animals, tissue culture and primary cell lines, where relevant.

In 2022, the CDRH Health of Women Program released a revised strategic plan to achieve these goals, outlining aims to expand communication strategies, increase educational training for data collection and analytical practices, and collate a portfolio of teams building women-specific medical devices, among other goals (U.S. FDA, 2022).

Not all health and medical research, however, is funded by public funding agencies or regulated by the U.S. FDA or similar groups. Only 6% of trials registered on ClinialTrials.gov were funded by the NIH in 2014; non-federally funded studies were significantly less likely to include sex analysis (Duffy et al., 2020). In cases where regulatory agencies such as the U.S. FDA, International Medical Device Regulators Forum, or European Medicines Agency have no jurisdiction, Duffy et al. (2020) recommend that Institutional Review Boards (IRB) provide oversight to ensure that sex, gender, race/ethnicity analyses are appropriately integrated into research.