OBJECTIVES/GOALS: A growing number of older adults in the United States have multiple social determinants of health (SDoH) that are barriers to effective medical care. We used generalizable machine learning methods to identify and visualize subtypes based on participant-reported SDoH profiles, and their association with delayed medical care (self-reported yes/no). METHODS/STUDY POPULATION: Data. All participants aged >=65 in All of Us with complete data on 18 SDoH self-reported variables, selected through consensus by 2 experienced health services researchers, and guided by Andersen’s behavioral model. Covariates included demographics, and the outcome was delayed medical care . Cases (n=4090) consisted of participants with at least one of the 18 SDoH variables, and controls (n=7414) consisted of participants with none of them. Method. (1) Used bipartite network analysis and modularity maximization to identify participant-SDoH biclusters, and visualize them through ExplodeLayout. (2) Used multivariable logistic regression (adjusted for demographics and corrected through Bonferroni) to measure the odds ratio (OR) of each participant bicluster to the outcome, compared with the controls. RESULTS/ANTICIPATED RESULTS: The analysis identified 7 SDoH subtypes (https://postimg.cc/Vd7Pg4xZ) with statistically significant modularity compared with 100 random permutations of the data (All of Us=.51, Random Mean=.38, z=20, P DISCUSSION/SIGNIFICANCE: The results identified 7 distinct subtypes based on SDoH profiles and their risk for delayed medical care, highlighting the importance of addressing specific combinations of barriers, with affordability having the highest risk. Furthermore, the analytical methods used are generalizable and have been made publicly available on CRAN and All of Us.