Women’s Health Research Uses AI to Eliminate Biases Once and for All

Negligence in women’s health care is a tale as old as time and still today women are spending 25% more of their lives in debilitating health than men. Gender inequity is something painfully present across many industries, but health care shouldn’t still be one of them considering the strides that have been made in technology and research capabilities over the past few decades.

As a result, women aren’t trusting their primary care and specialist physicians with one in five having felt completely ignored or dismissed by one in the past. They’re also not testing for diseases that are known to afflict them, or receiving the proper treatment they need. In the worst case, they’re even dying from causes otherwise completely preventable, which we see in areas like maternal health – where roughly 84% of pregnancy-related deaths were determined to be preventable.

These reasons, most of which could be avoided through a stronger foundation of research, are sure to have motivated President Joe Biden’s recent executive order pushing for the advancement of women’s health research and innovation.

His request for $12 billion to power research efforts could revolutionize health care by better understanding how to care for women specific to their needs. While it’s up to Congress to make the next move, the executive order alone serves to deem women’s health research a priority and spurs dialogue and innovation among the relevant sectors – a major one being artificial intelligence (AI).

In addition to ordering his administration to report on progress they would make to erase gender gaps in research, according to a White House summary of the order, Biden is also looking to study how AI can be used to improve women’s health research. While the outlook of AI use within health care has been mixed depending on the application, for research and actionable insights it offers a powerful potential to finally level the playing field. Not using it to benefit women’s health research would fail to do the program justice.

Having under-researched this demographic (and many others) for centuries, a stronger approach now must be taken to expedite progress. Critically, AI can understand causal factors contributing to a women’s overall health status well before traditional research can uncover them. In this way, AI can identify risk factors and test them according to scientific standards quicker and often times more accurately than through traditional research processes. Taking these valuable insights and taking action on existing care gaps means women can achieve better health outcomes earlier.

But it’s important to recognize that speed isn’t the No. 1 priority here, it’s being as comprehensive as possible. Women’s health is most certainly not one-size-fits-all and AI can help understand discrepancies in access and care among all subpopulations. These tools are able to provide a deep situational awareness for often overlooked health and social related needs (HSRN) which include cultural context, education level and health literacy, living environment, transportation capabilities, employment status, and many other components that impact women’s health outcomes.

However, many limitations exist with traditional, predictive AI models. Traditional AI models require training on preexisting data sets which aren’t adequately available for this group and most definitely lack representation. By using them, we risk perpetuating health inequity by putting a disproportionately low focus on higher-risk populations who face overlooked barriers to care.

For example, when managed care organizations upload basic claims data into traditional AI predictive models, it’s common for the highest volume of claims to be for white people. A traditional model might then suggest white people need additional support or services. In reality, it isn’t taking into account how race serves as a barrier to care, which contributes to the lower number of claims in the data in the first place.

To circumvent these roadblocks, the administration would be well advised to explore and prioritize causal AI – systems that analyze cause-and-effect to go beyond simple correlations and instead uncover the “why” behind complex problems. Causal AI is far more extensive and when used in tandem with traditional AI systems, can ensure all bias is being eliminated before providing those actionable, predictive insights.

These advanced systems allow us to identify direct health outcomes and adherence likelihood with health-related social needs in mind. For instance, a causal AI system might reveal that people living farther than 30 miles from the nearest care facility have a higher chance of suffering from a preventable disease. It might also find that that low-income areas with lack of healthy, fresh, food options might have a higher obese population. These insights inform not only health risks, but also how our health care ecosystem must adapt to meet each patient where she is.

These are the types of insights we must uncover through the program. Not doing so will prevent us from providing the targeted solutions women need in today’s health care system, failing them once again.

See the original article here.