Background: Most adult intensive care units (ICUs) worldwide adopt restrictive family visitation models (RFVMs). However, evidence, mostly from non-randomized studies, suggests that flexible adult ICU visiting hours are safe policies that can result in benefits such as prevention of delirium and increase in satisfaction with care. Accordingly, the ICU Visits Study was designed to compare the effectiveness and safety of a flexible family visitation model (FFVM) vs. an RFVM on delirium prevention among ICU patients, and also to analyze its potential effects on family members and ICU professionals. Methods/design: The ICU Visits Study is a cluster-randomized crossover trial which compares an FFVM (12 consecutive ICU visiting hours per day) with an RFVM (< 4.5 ICU visiting hours per day) in 40 Brazilian adult ICUs. Participant ICUs are randomly assigned to either an FFVM or RFVM in a 1:1 ratio. After enrollment and follow-up of 25 patients, each ICU is crossed over to the other visitation model, until 25 more patients per site are enrolled and followed. The primary outcome is the cumulative incidence of delirium measured by the Confusion Assessment Method for the ICU. Secondary and tertiary outcomes include relevant measures of effectiveness and safety of ICU visiting policies among patients, family members, and ICU professionals. Herein, we describe all primary statistical procedures that will be used to evaluate the results and perform exploratory and sensitivity analyses of this study. This pre-specified statistical analysis plan was written and submitted without knowledge of the study data. Discussion: This a priori statistical analysis plan aims to enhance the transparency of our study, facilitating unbiased analyses of ICU visit study data, and provide guidance for statistical analysis for groups conducting studies in the same field.