OBJECTIVE: To determine which model best simulates the actual IO procedure in children. METHODS: Forty emergency and critical care physicians with significant IO experience (6 or more IO procedures) in real children were recruited at 4 academic centers. Study subjects were provided with a kit containing 15 gauge IO needles and 5 IO models; a plastic IO doll leg (PL), a turkey femur/thigh (TT), a turkey tibia/drumstick (TD), a chicken femur/thigh (CT), and a pork rib (PR). Study participants scored the similarity of the bone model to that of placing an IO in a child (0 = no experience, 1 = perfect simulation, 2 = excellent, but not perfect, 3 = good, 4 = moderate, 5 = poor) and the hardness of the bone (H = too hard, J = just right, S = too soft) for each age group (preterm, newborn, 1-4 months, 5-12 months, 12-36 months, 3-6 years, 6-12 years, and >12 years old) of whom they had previous experience. Mean scores were calculated and compared to determine which model provided the best simulation. RESULTS: After excluding zero scores, overall mean scores for the CT, TD, TT, PL, and PR models were 3.2, 3.3, 3.3, 3.3, and 4.4 (P < 0.001), respectively. The pork rib had the worst score in all age groups, while the other 4 models had roughly similar scores. All models had substantial percentages of the models classified as too hard or too soft, except for the pork rib, which was predominantly too hard. CONCLUSIONS: For IO research and teaching purposes, bone models should be age appropriate. This study suggests that there is great variability in preference with the chicken, turkey, and plastic models.