Stigmatizing language, inclusive solutions and telehealth big topics in health IT

Ensuring diversity, equity and inclusion has been a major theme at medical conferences this year, and the American Medical Informatics Association’s recent Virtual Clinical Informatics Conference was no exception. Hospitals and health systems shared some of their projects to promote diversity and equity across the field of health IT.

Here are three examples:

Stigmatizing language in electronic health records

Hospitalist and informaticist Subha Airan-Javia of the University of Pennsylvania was caring for a patient in a long-term acute care unit last summer when she read in the chart that the patient had refused his morning medications. When she talked to the patient, he said he didn’t refuse but declined to take his medications because he didn’t understand why some of them were ordered. He wanted to talk to the doctor first.

Airan-Javia began thinking about how language in documentation matters. So she and her colleagues set out to understand the stigmatizing language used in clinical documentation. They reviewed information from the CDC, American Psychological Association, and Twitter chats, compiling a list of 122 words and phrases that should be avoided, such as “elderly,” “fragile,” “crazy,” “poor,” and “refused.”

Studying 136,410 provider notes in her health system’s electronic health record (EHR), as well as 214,787 free text handoff notes from her institution and 407,096 free text handoff notes from another institution from January 2004 through November 2020, the team found 48 of the 122 stigmatizing terms used at least once, most in the patient-blaming category. This included terms such as “noncompliant,” “abuser,” “failed” a therapy, etc. Overall, about 19% of notes had at least one stigmatizing term.

The work, ongoing and not yet published, indicates that more research is needed in this area, she said. They plan to refine the list and continue the analysis with the ultimate goal of designing interventions to reduce the use of this language and incorporate lessons into medical education.

Disclosure: Airan-Javia is the founder and CEO of TrekIT Health Inc. doing business as CareAlign.

Inclusive gynecologic health solutions for diverse users  

Many people who seek gynecologic health support may face discrimination and dismissal from clinicians, or barriers to care because of their reproductive needs, said Uba Bačkonja, a nurse and informaticist with the University of Washington. Health IT can provide a way to help bridge the gap, she said.

Looking at the gynecologic ecosystem in health technologies, however, she noted a focus on mobile apps that track menstruation, with a focus on fertility. “That’s a very small, narrow way of looking at gynecological health,” Bačkonja said. Many apps she found assumed users were heterosexual, were not evidence-based and did not support gynecological health literacy.

Her team has held user-centered design sessions with practitioners including doctors, nurses, massage therapists, and acupuncturists, asking them to create technology to support the transition to menopause. The team also has queried diverse patients, finding a range of ideas that could be addressed through existing and emerging technologies like artificial intelligence. The work is ongoing but you can read about her review of menstrual apps here and her work evaluating potential technologies to support menopause here.

Disparities in telehealth use during COVID-19  

A few presentations looked at the use of telehealth during the COVID-19 pandemic. Ellerie Weber, an assistant professor of health economics and health services research at the Icahn School of Medicine at Mount Sinai, in New York, and colleagues sought to study the use of telehealth versus in-person clinician visits among patients seeking care for COVID-19 symptoms at her health system during the peak pandemic period of March 20-May 18, 2020.

Of 39,229 patient encounters studied, about 15,000 were in the ED or via telehealth, and about 9,000 were in an office. Patients who identified as Black or Hispanic/Latinx were more likely to use the ED for their first encounter versus telehealth or an office, whereas patients who identified as white or Asian were more likely to use telehealth for their first encounter versus the ED or an office. Patients over 65 also were more likely to use the ED for their first encounter versus telehealth or an office, and patients with preferred languages other than English were more likely to use in-person visits versus telehealth compared to English speakers. These results were published in the Journal of the American Medical Informatics Association (JAMIA).

The findings likely reflect a combination of disparities in both telehealth use and prevalence/severity of COVID-19 symptoms, Weber said. However, they were consistent with other early papers indicating telehealth disparities. The outstanding questions are why, and what the barriers are like awareness, cost/coverage or distrust of digital appointments. Good data is needed about patient characteristics and their use/experience with telehealth, she said.

Among other efforts, her team is partnering with a community-based organization to evaluate state residents’ telehealth needs and consider interventions to increase uptake and using EHR data to explore additional aspects of telehealth disparities, such as the impact on health outcomes.

Tips for journalists

Always question the data, health IT experts say.

Ask about sources of data behind EHR and informatics projects, Weber said. Why is it often sparse, or of poor quality? Data related to COVID-19 has varied greatly based on whether and how states collected information about race and gender in test results and vaccinations. “It’s timely to ask why our systems are not collecting the needed data,” she said.

Another area often overlooked in health IT stories is the reimbursement/cost component, she said: “This touches both on telehealth specifically (unequal reimbursement systems contribute to telehealth disparities) but also to the data issue that health systems may not have the incentive to collect good quality data because it doesn’t affect their reimbursement.”

When writing about any of the “sexy, hot topics in IT” – artificial intelligence, machine learning, decision support systems or algorithm development – ask about data used for developing and training these technologies, and whether bias in all forms was evaluated, Bačkonja advised. “There’s an assumption that if you have large volumes of data, that the large volume will wash out any bias in the data,” she said. “That assumption is false.”

For example, if researchers are developing an algorithm to use for a cardiac health predictor to be integrated into the EHR, using data from a large organization that only inputs gender as male or female potentially could lead to a wrong and potentially dangerous algorithm for people who are transgender or taking hormone therapies, she said.

“There is currently a reckoning happening within clinical medicine with biased algorithms, especially corrections for race,” she added, referring to a recent article from the New England Journal of Medicine. “There needs to be more questions asked about what data are being used to develop and train algorithms, what assumptions are ingrained in the data and algorithm, as well as what sociopolitical and cultural factors need to be considered when evaluating data and algorithms, to ensure that the data are truly measuring the phenomena of interest.”

For access to recordings or slide decks from these and other meeting sessions, contact Lisa Gibson at AMIA at For general resources on covering health IT, check out AHCJ’s health IT section.

Leave a Reply