Avaliação de intervenções educativas na prescrição de medicamentos potencialmente perigosos, em três hospitais de Belo Horizonte

Abstract

Prescribing errors are a cause of high morbidity and mortality worldwide, thereby representing a serious public health problem. OBJECTIVES: To evaluate the impact on the incidence of errors in prescribing of high-alert medication (HAM) in three hospitals in Belo Horizonte, Brazil. METHODOLOGY: Based on the incidence of prescribing errors and the possibility of harm to the patient, data were collected in the first stage (2007), from the prescriptions of unfractionated heparins (UFH) and potassium chloride for intravenous use (KCL) which having been selected for the educational interventions. The first study was experimental or an intervention-type community trial, applying the same printed educational methods (PEM): leaflets, charts and banners in three hospitals (2009). The second phase was experimental or a randomized study trial carried out in 2010, using multifaceted measures (MM): opinion local leaders, outreach visits, tailored educational materials and use of computer alerts in one hospital (case) and no interventions in two hospitals (controls). The incidence of prescribing errors related with UFH and KCL was evaluated before and after educational interventions. Univariate analysis assessed the differences between hospitals and phases, using the chi-square and Fisher-Freeman-Halton tests. The accuracy of the evaluation of legibility made by judges was tested by Kappa of Fleiss. The influence of each independent variable (hospital, ward, age and sex of the patient, type of intervention, study stage, prescribing errors with UFH/KCL, and adherence to the suggested prescription standards) was evaluated by multivariate logistic regression. RESULTS: A total of 2,667 prescriptions from UFH (75,4%) and KCL(25,5%) have been examined, with 2,160 errors having occurred with an incidence of 80.9%, or 809 errors per 1,000 prescriptions of UFH/KCL, and this incidence was considered high when compared with other previous studies. The incidence of prescribing related with KCL was 97,3% and 75,3% for UFH The Fleiss Kappa index showed strong agreement among the judges (>0.75). The univariate analysis demonstrated an association of reduced errors with all independent variables. Multivariate logistic regression showed statistical significance for all independent variables except for age and sex. The standard prescription model proposed for UFH was accepted in 24.8% of prescriptions with a significant reduction in writing errors after implementation. The standard prescription model proposed for the KCL and handwritten prescriptions (UFH and KCL) were not accepted by prescribers. CONCLUSIONS: The PEM and MM had a significant impact in reducing prescribing errors related with UFH, and no effects were observed in the KCL´s prescriptions.