Our objective was to evaluate how prevalence estimates for health conditions are affected by the number of cows sampled and the selection method used. On 10 California dairy farms, we assessed all cows in the high-producing pen (DIM < 100; range 81–241 cows) using measures from the Welfare Quality Protocol for Dairy Cattle. Cows were evaluated for body condition, hygiene, skin alterations (hairless patches, lesions, or swelling), discharge (ocular, nasal, vulvar), diarrhea, and impaired respiration while restrained in headlocks. Lameness was scored upon release from the feed bunk. Prevalence for each condition was calculated as a percentage of cows in the pen. The most common conditions were dirty hindquarters (33.5 ± 10.7%, mean ± SD) and lesions or swelling on the knee (34.4 ± 17.0%) and hock (26.4 ± 16.7%). Diarrhea (8.0 ± 5.8%), lameness (mild: 7.3 ± 4.7%, severe: 2.2 ± 2.2%), and neck (5.8 ± 12.6%), flank (4.5 ± 5.0%), or hindquarter alterations (5.5 ± 3.9%) were less common. Very fat cows, vulvar discharge, and impaired respiration were excluded from further analysis (prevalence ≤ 1%). To evaluate how many cows are needed to accurately estimate prevalence, 7 subsets of data were created by selecting every 10th, 5th, 4th, 3rd, 2nd, 2 of 3, or 3 of 4 cows using their position at the bunk. In addition, 7 matching proportions of the pen were randomly computer-selected (14 subsets total). Estimates were compared with true values using regression analysis and were considered accurate if they met 3 criteria: R² > 0.9 and the slope and intercept did not significantly differ (P > 0.05) from 1 and 0, respectively. All estimates met the slope and intercept criteria, whereas R² increased when more cows were sampled. Regardless of how many cows were used, both selection methods were accurate for ocular discharge (22.2 ± 27.4%), neck alterations (5.8 ± 12.6%), and knee hairless patches (14.1 ± 17.4%). For the other conditions, all estimates based on feed bunk position were accurate when ≥ 2/3 of the pen was used, and all but severe lameness were accurately estimated when sampling randomly. Selecting cows using feed bunk position did not differ systematically from computer-selecting random ear tag numbers, suggesting the former may be an appropriate method for welfare assessments.