A systematic review and meta-regression of single group, pre-post studies evaluating food safety education and training interventions for food handlers

Background: Food handlers working in retail and food service establishments are a frequent source of foodborne disease outbreaks. Numerous studies have investigated different education and training approaches to improve their safe food handling knowledge, attitudes, and practices. Frequently, these studies use a single group, prepost design, measuring changes before and after an intervention without a separate control group. Objectives: We conducted a systematic review of these studies to identify their key characteristics and to evaluate possible predictors of between study heterogeneity in their estimates of intervention effect. Methods: the review steps included a comprehensive search; relevance screening; article characterization; risk-of-bias assessment; data extraction; and meta-regression on five outcome categories: attitudes and beliefs, knowledge, behaviours, food premise inspection scores, and aerobic plate counts. Results: Among 85 relevant studies identified, the most commonly investigated intervention type was in-person, group-based training courses (81%). Interventions primarily targeted food handlers in educational institutions and restaurants (32% and 31%, respectively). The most frequently measured outcome was food handler knowledge (66%). Studies mostly (89%) covered multiple food safety content areas, primarily targeting personal hygiene (88%) and avoiding cross-contamination (87%). All studies were rated as 'critical' risk of bias given the lack of an independent control group. Significant intervention effects were found for all outcome categories, but substantial heterogeneity was also identified. Studies that informed their intervention from formative research reported larger effect sizes for attitude and belief outcomes, those that based their intervention on a theory of behaviour change reported larger effect sizes for behaviour outcomes, and those published in grey literature sources reported larger effect sizes for behaviour and knowledge outcomes.