OBJECTIVES: We successfully implemented a multi-factorial intervention targeting general practitioners (GPs) in the Parma Local Health Authority (LHA), Italy, to decrease inappropriate prescribing in the elderly. It remains unclear as to whether the improvement was long-lasting.  We sought to determine whether reduction in exposure to potentially inappropriate medications (PIM) in the older population (≥65 years) continued after discontinuation of intervention, and, if so, what factors contributed to effectiveness.  METHODS: Data on all outpatient pharmacy claims  for  the pre-intervention (2005 Q1-2007 Q3), intervention (2007 Q4-2009 Q4), and post-intervention (2010 Q1-Q4) were retrieved. We assessed changes in quarterly incidence rates of PIM exposure for the intervention and post-intervention in both all PIM users and newly exposed users. Generalized estimating equations were used to model the odds of PIM exposure. RESULTS: A total of 299 GPs (98.7%) in the Parma LHA serving 111,282 older patients were included in this study.  PIM exposure incidence rates declined for all users from 7.1% prior to the intervention to 4.9% at the end of the intervention and continued to decline to 4.3% at the end of the post-intervention. Adjusted models attributed the intervention with an immediate reduction (p<0.001) and a steady decline in the odds of PIM exposure (p<0.001). No measured patient and GP characteristics modified the effect of the intervention.  Results among newly exposed users were similar to results for all users. CONCLUSIONS: Inappropriate prescribing in the elderly contributes to negative clinical and economic outcomes.  Specific interventions have been utilized to decrease incidence of inappropriate prescribing, but there is little evidence on the long-term sustainability of such project effects.   The significant reductions in PIM exposure after our intervention appear to have continued after its discontinuation, supporting the idea that the intervention concretely impacted GPs’ prescribing behavior. Patient and GP characteristics did not contribute to the effect of the intervention.