We examined the association of pulmonary nodule characteristics with adherence to follow-up low-dose computed tomography (LDCT) after the initial screening in lung cancer screening. Using 2014-2021 electronic health record data from a large integrated health system, we analyzed adherence to Lung Imaging Reporting and Data System (Lung-RADS) follow-up recommendations, considering socio-demographic, clinical factors, and natural language processing-extracted nodule characteristics. Multivariable logistic regression models assessed the impact of these factors on adherence to follow-up LDCT. Among 2,673 individuals (mean age = 66.8 ± 5.9 years), overall adherence was 27.6%, with rates of 24.2%, 27.5%, 26.7%, and 64.0% for Lung-RADS categories 1-4 A. A race-ethnicity disparity in adherence was observed among category 1, with non-Hispanic blacks less likely to adhere than non-Hispanic whites (OR[95% CI] = 0.59[0.41-0.85]). Among patients in categories 2 to 4 A, category 4 A was significantly more likely to adhere (OR[95% CI] = 3.18[1.86-5.40]) and having more nodules increased adherence (OR[95% CI] = 1.12[1.09-1.14]). Adherence to follow-up LDCT is suboptimal, driven by patient and nodule characteristics, and influenced by how physicians communicated initial CT results. These findings underscore the need for structured screening programs and consistent follow-up protocols to improve adherence and ensure effective lung cancer screening.