Inpatient costs are often assigned using reimbursement rates corresponding to diagnosis related groups (DRGs) or similar coding system. Other approaches to cost assignment may provide more valid estimates. OBJECTIVE: To evaluate different inpatient cost estimation methods utilizing data from the 2001-2005 National Inpatient Survey (NIS) applied to event-level data from a randomized trial of patients with cystic fibrosis and to compare these approaches within the NIS sample. METHODS: Hospitalizations in a Phase 3 clinical trial were matched (1:many) to NIS discharges representing cystic fibrosis patients based on ICD-9 diagnostic and surgical procedure codes. Costs for hospitalizations in the trial were estimated using these NIS discharges applying six different methods; mean cost, median cost, mean daily cost, median daily cost, DRG-based costs, and regression analysis. Cost estimation within the NIS sample was evaluated for each method by comparing root mean squared error (RMSE) and mean absolute percent error (MAPE) between predicted and actual discharge costs. RESULTS: All but two of the 98 hospital admissions from the trial could be matched with 10 or more NIS discharges. Mean estimates of inpatient costs in the trial ranged from $5,368 (SD=2,071) with the standard DRG method to $16,635 (SD=69,822) with the regression method. Of the six methods, median daily cost resulted in the smallest MAPE (51%±11% SE) with a RMSE of 12,597 within the NIS matched sample (N=8,485), followed by the mean daily cost approach MAPE (60%±13% SE) with a RMSE of 12,248 and regression analysis (69%±12% SE). CONCLUSION: Different methods are available for estimating inpatient expenditures which may provide advantages over existing, more generalized cost estimation procedures such as DRGs. Further evaluation of such methods is warranted to improve the validity of costs assigned to hospitalizations in studies of patients with serious underlying conditions.