Research publications related to the novel coronavirus disease COVID-19 are rapidly growing in number. However, current online literature hubs are limited in examining the complexity of COVID-19 research topics. Hence, our methodology aims to implement a comprehensive Latent Dirichlet Allocation (LDA) topic model using natural language processing (NLP) techniques, provide visualizations for temporal trends, and apply our model to improve existing online literature hubs. Using the search term “COVID”, research abstracts were extracted from PubMed®. An LDA topic model was trained on 81% of abstracts. Weekly temporal trends in topics were visualized as a heatmap. Then, our methodology was applied to abstracts from LitCovid, a literature hub for COVID-19 research from the National Center for Biotechnology Information. The topic model was used to subdivide LitCovid’s eight categories into the corresponding LDA topics. Our results for temporal evolution demonstrate interesting trends in COVID-19 research publications, for example, the prominence of “Mental Health” and “Socioeconomic Impact” increased, “Genome Sequence” decreased, and “Epidemiology” remained relatively constant. By applying our methodology to LitCovid, we improved the breadth and depth of research topics by subdividing their pre-existing extensive categories. We identified inadequate representation of the topic “Airborne Transmission Protection” in COVID-19 research publications, demonstrating that research on masks and PPE is skewed towards clinical applications with a lack of population-based epidemiological research. Overall, our generalizable model responds to the need for automated organization and analysis of themes in COVID-19 research publications.