A Tweet is worth a thousand words: using Twitter for epidemiologic research, a joint webinar from two research groups
February 25, 2020 – 12pm EDT
Part 1: Twitter-derived measures of sentiment towards minorities and birth outcomes.
Quynh Nguyen, PhD, MSPH is an Assistant Professor of Epidemiology and Biostatistics at the University of Maryland School of Public Health. Twitter handle: @quynhcnguyen
Interpersonal and structural racial bias are leading explanations for the continuing racial disparities in birth outcomes but research to confirm the role of racism has been hampered by challenges in both measuring racial bias and evaluating its impact. We use Twitter data to characterize area-level racial hostility and examine the associations with birth weight and preterm birth. In this webinar, we cover Twitter data collection and processing, sentiment analysis, and use of machine learning to classify tweets for racist content. Use of nontraditional data sources like Twitter has the potential to lead to greater tracking of area-level racial bias and to provide essential information needed to develop interventions to reduce the impact of racial bias on health.
Part 2: Discussions of Miscarriage and Preterm Birth on Twitter.
Nina Cesare is a Postdoctoral Associate at Boston University School of Public Health. Twitter handle: @nlcesare
Studies suggest that there is a trend towards expressing disenfranchised grief on social media. However, no large studies have investigated trends and discussions around miscarriages and preterm births on Twitter. Our presentation will review findings from a study analyzing disclosure of miscarriage and preterm birth on Twitter. First, we will show that there are multiple conversation topics related to miscarriages and preterm births. Second, we demonstrate that specific events usually drive surges in discussions. Lastly, in addition to grief, we illustrate that women who have experienced a miscarriage may use social media to share feelings towards insensitive comments by clinicians, friends and family; healthcare costs; legislatures affecting women’s health etc. Our findings are intended to inform both researchers utilizing digital data for healthcare experience research, as well as clinicians seeking to guide conversations about miscarriage and preterm birth and improve patient care.
December 11, 2019 – 12:00-1:00pm EDT
Dr. Jeanette A Stingone
Incorporating machine learning approaches into perinatal and pediatric epidemiology: opportunities and challenges
The use of machine learning, broadly defined as analytic techniques that fit models algorithmically by adapting to patterns in data, is growing in use across many areas within public health and epidemiology. This talk will provide attendees with broad exposure to the elements of machine learning and its practical applications within perinatal and pediatric epidemiology. The talk will include discussion of technical aspects of machine learning, as well as important ethical and scientific considerations of using data-driven methods for epidemiologic research. A number of examples from the scientific literature will be presented and a general listing of resources for additional information and training will be provided.
Jeanette A Stingone PhD MPH
Assistant Professor, Department of Epidemiology
Mailman School of Public Health, Columbia University
Dr. Jeanette Stingone is an environmental epidemiologist with a focus on perinatal and pediatric health. She received her BS in Biomedical Engineering from Boston University, an MPH from the Mount Sinai School of Medicine and a PhD in Epidemiology from the University of North Carolina, Chapel Hill. Now an Assistant Professor in the Department of Epidemiology at Columbia University’s Mailman School of Public Health, she conducts research that couples data science techniques with epidemiologic methods to address research questions in children’s environmental health. Supported by an NIEHS-funded career development award, her current research seeks to uncover the combinations of air pollutants associated with adverse child health outcomes within high dimensional public health data. Read more
It is recommended to read the overview by Bi et al and then skimming the others, as Dr. Stingone will refer to these in the talk when providing examples.
1. Overview of ML approaches: Bi Q, Goodman KE, Kaminsky J, Lessler J. What is machine learning? A Primer for the epidemiologist. AJE 2019; https://doi.org/10.1093/1je/kwz189 [epub ahead of print]
2. Examples of implementation:
a. Pan I, Nolan LB, Brown RR, Khan R et al Machine learning for social services: a study of prenatal case management in Illinois. AJPH 2017; 107:938-944.
b. Chiavegatto Filho ADP, Dos Santos HG, do Nasciemento CF, Massa K, Kawachi I Overachieving municipalities in public health: a machine learning approach. Epidemiology 2018; 29:836-840.
c. Das LT, Abramson EL, Stone AE, Kondrich JE, Kern LM, Grinspan ZM. Predicting frequent emergency department visits among children with asthma using HER data. Pediatr Pulmonol 2017; 52:880-890.