OBJECTIVES: Parameters of central tendency (mean) are often estimated in budget impact assessment (BIA) with less emphasis on parameters of spread (variance); critical to fixed funding (FF) policies (e.g. envelope systems )is accuracy of the estimate and maximal precision. The complexity and diversity of the hospital pathology mix and its treatment are the main theoretical hurdles encountered. In this analysis, the variance of the drug treatment reimbursement cost was investigated on the Main Diagnostic Code (MDC) and the All Patients Refined Diagnosis Related Groups (APRDRG) -level. The aim was to select low cost variance diagnoses which would enhance precision in FF. METHODS: Belgian hospitals register admission data in minimum basic data sets (MBDS): we extracted anonymous data from stays of 21 peripheral Flemish hospitals (during 2002). MBDS contains ICD-9-CM codes, performed procedures, stay parameters (e.g. risk of mortality), patient characteristics (age, gender) and drug utilization data with the national insurance cost. Data were analyzed in SPSSWIN® 12.0. RESULTS: The database contained 368,618 unique stays. On MDC-level, the fit between mean cost and variance was merely a non-linear relationship; the MDC's cardio-vascular (CV) disease, myeloproliferation & neoplasms, infectious disease and liver disease exhibited a linear relationship with r2 > 0.90; in these MDC's low mean cost pathologies also allow fixed funding. On APRDRG-level, increased cost variance probably reflects increased drug utilisation variability and/or heterogenicity in drug pricing. llustrative examples: the ratio of mean cost for cesarean (95€) to vaginal (27€) delivery is about 4 but the ratio of variances exceeds 10. In the CV area, the corresponding ratios for heart failure (269€) to angina (104€) are respectively 2.5 and 8. CONCLUSIONS: Reducing mean cost is important to BIA but to FF reducing cost variance is essential in order for the funding to be in line with the resources used.