person typing.

February Research Roundup

New study uses machine learning to analyze racial/ethnic bias, birth outcomes 

Assistant Professor Veronica Barcelona, PhD, has received a grant from the Columbia University Data Science Institute (CUDSI) to use machine learning to investigate why Black and brown birthing people in the United States face worse pregnancy outcomes than their white peers. 

Along with co-investigators Professor Maxim Topaz, PhD, and Assistant Professor Kenrick Cato, PhD '14, Barcelona is developing natural language processing (NLP) tools to identify patterns of bias in clinician notes and determine how these patterns relate to birth outcomes. The two-year, $150,000 CUDSI grant will support the analysis of 2017-2019 records for all patients admitted to the labor and birth units of two New York City hospitals, which together report about 7,000 births annually. 

Stigmatizing language in clinician notes can include statements questioning the patient’s veracity, as well as wording that expresses negative judgement about the patient, such as racial/ethnic stereotypes, Barcelona explains.  

“There is a growing body of qualitative literature describing obstetric racism,” she adds. “Black birthing people have reported experiences of lack of autonomy, poor communication, and feeling unheard. Unfortunately, generating solutions to improve this problem has proven difficult. Racism and discrimination are embedded in our society at large, and solutions must address racism in institutions and structures to improve the health of our most vulnerable.” 

Most of the research on racism and poor pregnancy outcomes has focused on patient-level and behavioral factors, resulting in blaming the patient, Barcelona adds. However, the harmful effects of systemic discrimination on health are increasingly being recognized, and the new study will be evaluating system-level factors.  

“The project addresses institutional bias in an actionable way to improve perinatal health,” Barcelona says. “By examining the association between linguistic bias and pregnancy-related morbidity, we can inform tailored interventions to improve care and patient outcomes, while addressing institutional racism.” 

Infographics improve outcomes for people living with HIV 

Using tailored infographics to assist in clinician-patient communication is associated with improved health outcomes for people living with HIV, Professor Rebecca Schnall, PhD '09, Professor Suzanne Bakken, PhD, Professor Maureen George, PhD, Assistant Professor Jianfang Liu, PhD, and their colleagues report in AIDS and Behavior. 

Samantha Stonbraker, PhD, a former pre-doctoral and post-doctoral trainee at Columbia Nursing and now an assistant professor at the University of Colorado, developed “Info Viz for Health,” which includes 15 culturally relevant illustrations, for use at their study site, a non-profit health center in the Dominican Republic specializing in treating sexually transmitted infections. In the new study, clinicians used the infographics during a patient’s first three clinic visits, selecting one to five based on the patient’s needs. 

At nine months’ follow-up, average viral load, self-efficacy, self-reported health, and HIV-related knowledge had significantly improved compared to baseline. “The improvements observed support the assertion that well-designed infographics may lead to better clinical communication and health outcomes,” the authors conclude. 

Job loss, depressive symptoms tied to risky drinking early in pandemic 

Heavy drinkers were much more likely to report increased alcohol consumption at the beginning of the COVID-19 pandemic, according to a new study in Preventive Medicine led by Cindy B. Veldhuis, PhD, an associate research scientist and research psychologist at Columbia Nursing. 

Veldhuis and her team analyzed responses to the baseline wave (April 5-May 5, 2020) of a longitudinal online survey they conducted (from April 2020-September 2021) on the mental health impacts of the pandemic.  Among 2,175 participants, 10% reported high-risk drinking, based on the World Health Organization definition of at least three drinks a day on average for women and four daily drinks for men. Thirty-six percent of participants said their alcohol intake increased at the beginning of the pandemic. 

Risky drinking was independently associated with household job loss and depressive symptoms and women were more than twice as likely to report high-risk drinking compared to men. Even after the authors took mental health and economic stressors into account, high-risk drinkers were still six times more likely to report their alcohol consumption had increased. 

“These findings suggest a need for targeted interventions to address the complex mental health and economic stressors that were associated with increased alcohol consumption and high-risk drinking during the COVID-19 pandemic and the ramifications of increased drinking in the post-pandemic era,” the authors conclude.