Purpose: Recent studies have attributed declining prostate cancer treatment costs to improvements in care delivery. These studies have not accounted for changes in the type of patients who are now diagnosed with prostate cancer due to widespread adoption of PSA screening. We propose a method to account for the impact of healthy screenee bias when estimating costs of cancer care.

Methods: We used SEER-Medicare claims data to estimate initial care costs among 31,643 prostate cancer patients who were treated with radiotherapy. We first employed standard methods for assessing cancer attributable costs by estimating total costs for newly diagnosed patients during a 6-month initial treatment period then subtracting non-cancer costs from age-matched controls without cancer. We then performed a second analysis where we used cases as their own controls. Here, non-cancer costs were estimated based on claims among cases prior to diagnosis.

Results: Between 1991 and 1999, total costs per patient among cases declined from $10,900 to $8,700 on average, while non-cancer costs for matched controls increased from $2,800 to $5,700. Thus, cancer-attributable costs for initial treatment declined from $8,100 to $3,000 (8% decrease per year). Among cases, non-cancer costs measured prior to diagnosis actually declined slightly over the period from $3,300 to $3,200 indicating that by the end of the period cases were substantially healthier before their diagnosis than population controls without cancer. Matching on age and comorbidity in the original analysis was not sufficient to control for this healthy screenee effect. Based on our second analysis, we estimated that cancer-attributable costs of initial care changed from $8,200 in 1991 to $4,800 in 1999 (5% decrease per year).

Conclusion: Average radiotherapy costs for men newly diagnosed with prostate cancer have improved during the PSA era, although observed declines are inflated by healthy screenee bias. One third of the observed decline in initial treatment costs is because men diagnosed today include a select group of healthy subjects as indicated by their substantially lower non-cancer costs compared with age-matched controls without prostate cancer. Evaluations of economic outcomes in screened populations should consider the impact of healthy screenee bias.