Pharmacoeconomic studies using claims data are frequently employed to compare the healthcare costs associated with competing drugs. Different methodological approaches with varying limitations for evaluating claims data are reviewed within the context of psychosis. Intent-to-treat paradigms and mirror-image designs have limitations that can bias comparisons of health resource use and can be addressed through the use of drug treatment episodes. Health resource use is better measured in financial rather than volume amounts, to account for service intensity. However, financial measures reported in claims records need to be carefully chosen. Between-group comparisons are frequently affected by selection bias, and within-group comparisons are often limited by period bias. While selection bias can be addressed by controlling for patient factors such as disease severity, and period bias by controlling for trend, devising appropriate control measures from the limited information in claims data can be challenging. Healthcare data are often skewed, which affects statistical testing. While skewed data are commonly handled through log transformation, this method has serious limitations, potentially distorting pharmacoeconomic comparisons. Determining the most appropriate methods for claims-based data involves careful evaluation of the alternate approaches to best achieve the goals of a pharmacoeconomic investigation.