OBJECTIVES: Cluster analysis (CA) is a widely used statistical technique that helps reveal classifications of entities with similar characteristics in large data sets. However, little is known about whether it can be applied to healthcare claims data with highly skewed cost information. This study applied different clustering methods to changes in all-cause cost data from a group of patients with end stage renal disease (ESRD) who initiated hemodialysis (HD). METHODS: A retrospective, cross-sectional, observational study was conducted using the MarketScan Commercial Claims database. Patients aged ≥18 years with ≥2 ESRD diagnoses who initiated HD between 2008 and 2010 were included. The K-means CA method and hierarchical CA with various linkage methods were applied to all-cause costs within baseline (12-month pre-HD) and follow-up periods (12-month post-HD) to identify clusters. Demographic, clinical, and cost information were extracted from both periods, and then examined by cluster. RESULTS:   A total of 18,380 patients were identified. Meaningful all-cause cost clusters were generated using K-means and hierarchical CA with either flexible beta or Ward's methods. Based on cluster sample sizes and change of cost patterns, the K-means CA method and 4 clusters were selected:  those with average costs in both periods (n=16,624); high costs followed by very high costs (n=113); high and increasing costs (n=1,554); or very high costs reduced to high cost (n=89). Relatively stable costs after starting HD were associated with more stable scores on comorbidity index scores from the pre- and post-HD periods, while increasing costs were associated with more sharply increasing comorbidity scores. CONCLUSIONS: The K-means CA method appeared optimal in healthcare claims data with highly skewed cost information when taking into account both change of cost patterns and sample size in smallest cluster.