This study systematically reviewed studies to determine the frequency and nature of medication administration errors in Latin American hospitals. We systematically searched the medical literature of seven electronic databases to identify studies on medication administration errors in Latin American hospitals using the direct observation method. Studies published in English, Spanish, or Portuguese between 1946 and March 2021 were included. A total of 10 studies conducted at 22 hospitals were included in the review. Nursing professionals were the most frequently observed during medication administration and were observers in four of the ten included studies. Total number of error opportunities was used as a parameter to calculate error rates. The administration error rate had a median of 32% (interquartile range 16%-35.8%) with high variability in the described frequencies (9%-64%). Excluding time errors, the median error rate was 9.7% (interquartile range 7.4%-29.5%). Four different definitions of medication errors were used in these studies. The most frequently observed errors were time, dose, and omission. Only four studies described the therapeutic classes or groups involved in the errors, with systemic anti-infectives being the most reported. None of the studies assessed the severity or outcome of the errors. The assessment of the overall risk bias revealed that one study had low risk, three had moderate risk, and three had high risk. In the assessment of the exploratory, observational, and before-after studies, two were classified as having fair quality and one as having poor quality. The administration error rate in Latin America was high, even when time errors were excluded. The variation observed in the frequencies can be explained by the different contexts in which the study was conducted. Future research using direct observation techniques is necessary to more accurately estimate the nature and severity of medication administration errors.