Protocol for developing a dashboard for interactive cohort analysis of oral health-related data

**Background** Open data science tools and data analytics skills are increasingly important in the digital health era, yet many oral health researchers lack training in data visualization and programming. Dashboards—defined as visual displays of key information consolidated on a single screen—are widely used across domains but have not been widely deployed in oral health epidemiology. National cohort surveys routinely collect data from children at different developmental stages, and interactive dashboards offer a means to present and disseminate these findings in an accessible, dynamic format. This protocol paper aims to demonstrate the development of an analytical, interactive dashboard using oral health-related data from multiple Irish national child surveys, targeting researchers and policymakers who may not have extensive coding skills.

**Methods** The dashboard was built using the R statistical programming language within R Studio. The flexdashboard package created the structural elements, and interactivity was added with the Shiny package. Data sources included the Growing Up in Ireland (GUI) infant cohort (initial n=11,134; waves at ages 3, 5, and 9), the GUI child cohort (initial n=8,500; waves at ages 9, 13, and 17/18), and the National Children's Food Survey II (NCFSII; n=594; children aged 5–12 years in 2017). Variables were selected based on known associations with oral health, including tooth brushing frequency, dental attendance, history of fillings and extractions, dietary intake of foods and drinks high in free sugars, and BMI status. Data wrangling was performed using tidyverse packages (dplyr), and visualizations were generated with ggplot2 and kableExtra. Specific functions were created to produce bar-plots and frequency tables. Ontology mapping minimized differences between survey questions across waves. Only subjects who participated in all waves of the GUI cohort surveys were included. The dashboard was deployed via shinyapps.io at https://dduh.shinyapps.io/dduh/.

**Key Results** The dashboard layout is structured by YAML metadata in the R Markdown document and Flexdashboard syntax. Survey type, wave of survey, and variable selector were set as filter options in an input sidebar. Shiny's render functions automatically update output when inputs change. The dashboard generates bar-plots and frequency tables for selected variables, automatically stratified by gender and socioeconomic status. Examples provided include: for the GUI child cohort at wave 3 (17/18 years old), the response is the number of fillings; at wave 2 (13 years old), the response was binary (Yes/No) for whether the child had dental fillings. The dashboard can be displayed on multiple devices including tablets and mobile phones. Version control was managed via Git and GitHub, and the R Markdown files and packages are available for download at https://github.com/oral-health-nutrition/oral-health-nutrition.github.io.git.

**Clinical Implications** Interactive dashboards allow viewers to dynamically explore oral health-related variables from multiple national cohort surveys without requiring multiple static plots and extensive documentation. This approach facilitates easy communication of descriptive statistics, data-driven hypothesis generation, faster dissemination of results, and computational reproducibility. For policymakers, readily accessible measures of trends in oral health behaviors—such as dental attendance, tooth brushing frequency, or consumption of sugar-sweetened beverages—can provide support for public health action. The authors note that data literacy should be a core competence in dental curricula, and that while basic R knowledge is sufficient for dashboard development, the learning curve for the broader skill set is steep. Strategies to improve uptake and training for both researchers and end-users are essential to achieve beneficial outcomes from dashboard-based dissemination.