Horses kept in stables may exhibit undesirable behaviours and stress during police service. The objective of this study was to evaluate the influence of confinement, i.e., time in stalls, on horse behaviour and welfare employed by mounted patrols in urban areas. Eight adult horses were subjected to one of two types of confinement during 16 days, namely: total confinement (TC – 24 h in a stall) and partial confinement (PC – 12 h in a stall and 12 h free in a paddock); then the treatments were reversed. Mounted patrols were conducted in pairs, with one animal from each treatment group, for six hours every other day in a scheme of five 40-min rides interspersed with four 20-min rest periods. Animal behaviour was evaluated by applying direct coded focal observations (SA – standing alert, SC – standing calm, M – movement, EF – eating forage, DW – drinking water, V – vocalizing, SI – positive or negative social interactions, S – stereotypies, and O – others) for three time periods on non-work days, and during rest periods of the last patrol day for each experimental stage. Behaviours were assessed through a questionnaire (A – alert, C – curious, D – distracted, Dr – drowsy, SI – socially interactive, F – fearful, and N – nervous), by the military personnel during patrols. Heart rate (HR), HR variability (SDNN, VLF, LF, HF, and LF/HF), blood count, and cortisol levels were determined on the first and last non-work day and on the first, third, and seventh day of patrolling. The experiment was performed under a randomized block design, and the data were compared by Tukey’s test using SAS statistical software (P < 0.05). The type of confinement influenced some behaviours observed in horses on non-work days and rest periods of last patrol day (SA, SC, M, and SI), as well as on those reported by the military personnel during patrols (C, D, F, and N). Animal HR values in the first (P = 0.0478) and second (P = 0.0001) rest period, and the LF (P = 0.0238) and HF (P = 0.0162) values at the fourth rest period during the patrols were higher in animals under TC. Average cortisol values were higher for animals under TC (at rest: 8.18 ng/mL; after patrolling: 14.57 ng/mL) than in those under PC (at rest: 6.31 ng/mL; after patrolling: 6.46 ng/mL). Therefore, turnout is a simple and effective practice that promotes the welfare of horses and positively influences their behaviour during mounted patrols.