An essential summary and review of Maia Hightower’s compelling and inspiring keynote address at the recent AIMed Surgery, ICU and Neurosciences virtual event – addressing healthcare inequities and advocating a more inclusive culture

 

“COVID-19 is a wake-up call, especially in addressing health disparity,” stated Dr. Maia Hightower, Chief Medical Information Officer, Senior Director of Health Equity, Diversity, and Inclusion at the University of Utah Health, during her inspiring keynote address at the recent AIMed Clinician Series. “It’s hard to protect all communities as biases are embedded within algorithms used to manage the health of our populations. These algorithms are not necessarily AI-driven but statistical models that our healthcare systems have been using for quite some time. Many of them had used race as a factor which led to unintended consequences.”

Dr. Hightower believes labeling errors added to the challenge by preventing certain ethnic minority groups from receiving extra care. “There have been callings to disregard race in clinical algorithms,” she continued. “But systemic bias in healthcare is a complex problem. We hope digital transformation and AI can help us lose those inequality gaps and tackle disparate parties in healthcare. However, unless we intentionally address the existing biases within our datasets and those that are rooted in our systems, we risk widening, rather than closing, the gaps.”

Like many healthcare systems, the University of Utah Health recognized that systemic racism is a public health crisis. According to Dr. Hightower, the University’s clinical decision support committee had developed a process to review all its race-based algorithms at least once a year and not make them visible to the clinicians, so they can continue to rely on their clinical judgment. To formally attend to diversity, inclusion, and cultivate an equity mindset in the culture, Dr. Hightower and her team created a Healthcare IT Equity Maturity (HITEM) Model.

 

In general, a maturity model is a measurement of the ability for continuous improvement in a particular discipline over time. HITEM is an outcomes-driven maturity model. “At the top of the model are organizational outcomes that we wish to accomplish and evidence-based capabilities that will drive these outcomes,” Dr. Hightower explained. “We recommend organizations begin with a vision statement to show they have reflected and are committed to making a change in equity, diversity, and inclusion. We also need leaders to ensure there are policies in place to execute these visions.”

The middle part of the model focused on processes that measure the workforce and corporate culture. “We need to ensure the organization has a portfolio that supports equity and diversity,” Dr. Hightower added. “There’s also a need for a diverse workforce and representations across the continuum of leadership as well as an inclusive culture. Diversity cannot be fully exercised if team members do not have a sense of well-being or belonging. Moreover, we need to assess how are we engaging with our community partners. We need to have culturally appropriate and relevant tools to serve a broad spectrum of people.”

Dr. Hightower noted the model is a sort of self-assessment and the University of Utah Health has courses catered for those on the lower end of the maturity model for continuous improvement. She urged organizations to prioritize and leverage the model on certain efforts for the first year before rolling it out fully to other areas. The model has facilitated the University of Utah Health’s attempts to be more diverse and inclusive. “We have since developed a key performance indicator framework for healthcare equity, Dr. Hightower said. “We make sure we have the data to measure inequity that drives our internal processes and external partnerships and address them accordingly.”

In addition, the University of Utah Health also kick-started an equity dashboard project. “Presently, only 25% of my staff are women,” Dr. Hightower added. “I have sent nine of them to a leadership development program. As far as career progression is concerned, women and underrepresented minorities are less likely to get promoted, so by referring some of them to potential programs, opportunities and representations are provided across the pipeline and a culture of inclusion can be established. We also have activities exploring implicit bias and diversity champions who are trained in conducting crucial conversations on the topics of power, privilege and micro-aggressions.”

Dr. Hightower believes the next step is to target algorithmic bias and fair AI. “It’s important to understand that these two are different,” she said. “Algorithmic bias is ensuring when one is preparing the model, there are steps taken to minimize biases that already exist in the data or other forms of labeling biases and ensuring you are using representative samples for validity. Fair AI encompasses a wider picture of having stakeholders, governance structure and conversations in place to discuss what’s important for an organization when it comes to different principles within an algorithmic fairness framework.”