Exploring the Social Drivers of Health During a Pandemic: Leveraging Knowledge Graphs and Population Trends in COVID-19

Abstract

Social determinants of health (SDoH) are the factors which lie outside of the traditional health system, such as employment or access to nutritious foods, that influence health outcomes. Some efforts have focused on identifying vulnerable populations during the COVID-19 pandemic, however, both the short- and long-term social impacts of the pandemic on individuals and populations are not well understood. This paper presents a pipeline to discover health outcomes and related social factors based on trending SDoH at population-level using Google Trends. A knowledge graph was built from a corpus of research literature (PubMed) and the social determinants that trended high at the start of the pandemic were examined. This paper reports on related social and health concepts which may be impacted by the COVID-19 outbreak and may be important to monitor as the pandemic evolves. The proposed pipeline should have wider applicability in surfacing related social or clinical characteristics of interest, outbreak surveillance, or to mine relations between social and health concepts that can, in turn, help inform and support citizen-centred services.

Social Determinant Trends of COVID-19: An Analysis Using Knowledge Graphs from Published Evidence and Online Trends

“…The approach consists of two steps, described in detail below: (1) a knowledge graph of relations between concepts is mined from PubMed using NLP techniques [5] and (2) population trends are mapped to the knowledge graph in order to study online search trends and how they might relate to the insights found from published evidence.…”

“…MetaMap [8] was used to tokenize and identify UMLS concepts in the sentences of the abstracts. Sentences containing pairs of UMLS concepts were retrieved and the relationship between concepts was assessed using the same fine-tuned BERT transformer model as in [5]. UMLS 4 is a very extensive and diverse ontology, so we restricted the concepts to SDoH and health concepts of the semantic types most relevant to our use case: Disease or Syndrome, Individual Behavior, Mental or Behavioral Dysfunction, Findings.…”

“…UMLS 4 is a very extensive and diverse ontology, so we restricted the concepts to SDoH and health concepts of the semantic types most relevant to our use case: Disease or Syndrome, Individual Behavior, Mental or Behavioral Dysfunction, Findings. The classes used in the annotations and as output of the BERT model were the binary restriction of the 5-class system described in [5], and simply put, they model if the concepts in the sentence are stated to be in a correlation relation or not.…”

“…Indeed, internet sources may be employed to inform public health and policy and can be valuable to monitor and forecast outbreaks and epidemics [4]. Our previous work [5] focused on monitoring concepts of interest in Google Trends and exploring, in PubMed, their connection to other concepts in the health and social domains. A knowledge graph was built by drawing an edge between two concepts found in correlation using a state-of-the-art neural NLP model [5] and other approaches are possible [6,7].…”

“…Our previous work [5] focused on monitoring concepts of interest in Google Trends and exploring, in PubMed, their connection to other concepts in the health and social domains. A knowledge graph was built by drawing an edge between two concepts found in correlation using a state-of-the-art neural NLP model [5] and other approaches are possible [6,7]. This paper reports the subsequent analysis of a later span of Google Trends data and gives a first glimpse of the predictive ability of such a graph.…”