Background: Medication errors have potential to cause harm and death; especially children who are three times more vulnerable than adults. Risk of medication errors is higher in out- patient settings due to a stressful work environment with less familiarity of individual patients. This problem in sub-Saharan Africa is however largely undetermined. A Voice Recognition System that converts verbal messages into text and stores it in a database in a retrievable format could impact on reduction of medication errors. Objectives: The primary objective was to compare medication prescription and dispensing errors in written prescriptions with those from a Voice Recognition System. Secondary objectives were to determine the types and frequency of medication errors, determinants of medication errors and acceptability of routine use of a Voice Recognition System to make medication prescriptions. Study design: A before -after Intervention study to determine the impact of introduction of a Voice Recognition System on the occurrence of medication errors. Methods: Prescriptions issued from the Paediatric Accident and Emergency Department at Aga Khan University Hospital Nairobi over a six month period were randomly selected and analyzed for errors. Patient‟s bio-data, diagnosis, prescriber‟s specialization and time of prescription were retrieved from outpatient medical records and documented in a standard study tool. A Voice Recognition System was installed and doctors and pharmacists consenting to use Voice Recognition were trained to enhance proficiency in its use. During consultations, doctors enrolled patients who provided written informed consent to have their prescriptions made using Voice Recognition. Prescription and dispensing records were analysed to determine the occurrence of medication errors. Questionnaires were issued to pharmacists and doctors to rate the use of Voice Recognition in the medication process. Results: During the VRS phase the proportion of female patients reviewed were 56.9% compared to 40% in the pre VRS phase. (OR= 0.5 (95% CI 0.37-0.69), P<0.001). The top five conditions diagnosed at the pediatric A&E department were upper respiratory tract infections, urinary tract infections, tonsillitis, pharyngitis and gastroenteritis. Incidence was similar in both pre VRS and VRS phases. (51.5% and 58.3% OR=0.74 (95% CI 0.53-1.01), P=0.063.) Overall, there was a 19.5% reduction in prescription errors from 86.1% in the pre Voice Recognition phase to 69.3% in the Voice Recognition phase (P<0.001). Among prescription errors analysed, there was a 31.9% reduction in omitted drug route (P <0.001) and a 64.8 % reduction in incorrect drug dose (P<0.001). Analysis of dispensing errors revealed the greatest