Background: FDA black box warning, Risk Evaluation and Mitigation Strategies (REMS), and CMS national coverage determination (NCD) aim to reduce inappropriate use of erythropoiesis-stimulating agents (ESAs) that are widely used in anemic patients. Previous studies have not linked specific safety interventions to changes in ESA utilization patterns in the inpatient settings nor assessed such interventions on off-label use of the drugs. Ineffectiveness of the intervention and lag time between such interventions and the observed change in clinical practice could lead to serious clinical outcomes. In addition, such interventions may unintentionally reduce on-label and some off-label use of ESAs considered “appropriate” in patients who could otherwise benefit. Objectives: The primary aim of the study is to quantify the impacts of the (1) addition of black box warning, (2) implementation of NCD, and (3) institution of REMS on ESA on-label and off-label utilization patterns of adult inpatients. Demographic, clinical condition, physician, and hospital characteristics of ESAs users by their use category are also described in detail. Methods: Electronic health records in Cerner Database from January 1, 2005 to June 30, 2011 were used. The use of the two erythropoietic drugs: epoetin alfa and darbepoetin alfa were categorized into three groups using ICD-9-CM diagnoses and procedures codes and patients’ medication information. The three categories were (1) on-label use (approved by the FDA); (2) off-label use supported (use for the indications not approved by the FDA, but there is strong clinical evidence to support its use); and (3) off-label use unsupported (use for the indications not approved by the FDA and lacking clinical evidence). The immediate and trend impacts of the interventions on the proportion of ESAs prescribed for each usage category between 2005 and 2011 were assessed using an interrupted time series technique. The likelihood of receiving ESAs among patients with on-label, off-label supported, off-label unsupported indications was assessed using a generalized estimating equation approach with binary logistic regression technique, clustering for hospitals and controlling for potential confounders such as patient characteristics, patient clinical conditions, physician specialty, and hospital characteristics. Results: During the study period, there were 111,363 encounters of ESA use. These encounters represented 86,763 patients admitted to Cerner health system between January 1, 2005 and June 30, 2011. Of these patients, 66,121 were prescribed epoetin alfa only (76.2%); 20,088 darbepoetin alfa only (23.2%); and 554 were prescribed both epoetin alfa and darbepoetin alfa (0.6%). Forty-nine percent of the patients used ESAs for the on-label indications, 8.6% for off-label supported indications, and 42.7% for the off-label unsupported indications. The main uses of ESAs in our sample were for CKD (ONS, 41.1%) and chronic anemia (OFU, 31.8%). From 2005 to 2010, the proportion of visits with ESA ONS and OFS use decreased 53.2% and 81.9%, while ESA OFU increased 112.6%. Results from binary logistic regression using GEE model showed overall decreasing trends in ESA use for the on-label and off-label supported indications, but not off-label unsupported indications. REMS had no impact on the odds of receiving ESAs among patients with on-label and off-label conditions. Black box warning reduced the odds of being prescribed with epoetin alfa in patients with off-label unsupported conditions by 40%. It was also associated with 4% and 15% per month reduction in the odds of using darbepoetin alfa in patients with off-label supported and unsupported conditions. Lastly, there was a significant decline in all categories of ESA use the month after Medicare national coverage determination was implemented. The impact of NCD ranged from a 20% reduction in the odds of off-label supported use to a 37% reduction in on-label use. Age, gender, race, source of payment, admission type, clinical complexity, discharge disposition, and hospital size were significant associated with ESA use on-label and off-label. Conclusion: This study was the first to determine the impact of safety interventions on ESA on-label and off-label utilization patterns in the inpatient settings using the Cerner database. We demonstrated lag between the interventions and observed change in clinical practice, and the relative impacts of three types of safety interventions on on-label and off-label ESA use in the hospital settings. The indirect impact of the reimbursement change was the potential unintended consequence of reducing the likelihood of receiving ESAs for a patient with indicated conditions who could have otherwise benefited from the drugs.