We conducted an unmatched case-control study of 5,992 infant mortality cases and 60,000 randomly selected controls from a North Carolina Birth Cohort (2003-2015). PM2.5 during critical exposure periods (trimesters, pregnancy, first month alive) were estimated using residential address and a national spatiotemporal model at census block centroid. Here we describe data sources for outcome (i.e., infant mortality) and exposure (i.e., PM2.5) data. This dataset is not publicly accessible because: EPA cannot release personally identifiable information regarding living individuals, according to the Privacy Act and the Freedom of Information Act (FOIA). This dataset contains information about human research subjects. Because there is potential to identify individual participants and disclose personal information, either alone or in combination with other datasets, individual level data are not appropriate to post for public access. Restricted access may be granted to authorized persons by contacting the party listed. It can be accessed through the following means: The North Carolina Birth Cohort data are not publicly available as it contains personal identifiable information. Data may be requested through the NCDHHS, Division of Public Health with proper approvals. Air pollutant concentrations for PM2.5 from the national spatiotemporal model are available upon request and may require a processing fee. Air monitoring data questions can be directed to Ms. Amanda Gassett at the University of Washington. Format: Birth certificate data from the State Center for Health Statistics of the NC Department of Health and Human Services linked with data from the Birth Defects Monitoring Program (NC BDMP) to create a birth cohort of all infants born in NC between 2003-2015. The NC BDMP is an active surveillance system that follows NC births to obtain birth defect diagnoses up to 1 year after the date of birth as well as identify infant deaths during the first year of life and include relevant information from the death certificate. A national spatiotemporal model provided data on predicted PM2.5 concentrations over critical prenatal and postnatal time periods. The prediction model used data from research and regulatory monitors as well as a large (>200) array of geographic covariates to create fine scale spatial and temporal predictions. The model has a cross-validated R2 of 0.89 for PM2.5. Concentrations were predicted for every 2 weeks in the study period at the centroid of each 2010 census block in NC. This dataset is associated with the following publication: Jampel, S., J. Kaufman, D. Enquobahrie, A. Wilkie, A. Gassett, and T. Luben. Association between fine particulate matter (PM2.5) and infant mortality in a North Carolina Birth Cohort (2003-2015). Environmental Epidemiology. Wolters Kluwer, Alphen aan den Rijn, NETHERLANDS, 8(6): e350, (2024).