UC San Diego

Women and minority genders have been excluded from biomedical research due to the presumption that males are less variable, easier to study, and results from male subjects will generalize effectively to everyone. However, many conditions are known to present differently in females or only in people with uteruses, e.g. pregnancy, endometriosis, certain cancers, etc. Presumptions about female variance should therefore be examined and formally described to support health equity. Traditional low-interval measurements of dynamic human biometrics, such as those taken at health care appointments, do not easily support exploration of variance across timescales. This low time resolution and reliance on subjective feedback instead of direct physiological measurement could be augmented by remote monitoring. Wearable devices have been shown to enable detection and prediction of acute illness, often faster than subjective symptom reporting. They have been employed for years to monitor and/or manage chronic conditions, such as continuous glucose monitors for individuals with diabetes. Commercial wearables have become ubiquitous in our society with latest reports expecting more than 500 million devices to be sold in 2024 alone. Despite the resultant abundance of data, only recently have efforts emerged translating wearables data into actionable health insights. Here I explore longitudinal physiological data from commercial wearable devices to understand variability between sexes in two related physiological measurements: temperature and activity. I also present the first published dataset of physiological measurements that characterized measurements across all of pregnancy revealing clear trajectories of pregnancy from menstruation to conception through postpartum recovery. In our assessment of individuals in whom pregnancy did not progress past the first trimester, we found associated deviations, corroborating that continuous monitoring adds new information that could support decision-making in the early stages of pregnancy. Lastly, I present high-performing explainable models that detect onset of pregnancy, demonstrating feasibility for developing related services for real-world use. Through this research, I have demonstrated that low-cost, high-resolution monitoring in real-world settings supports development of women’s health algorithms, as well as inclusion and equity in research.