Drawing from previous studies that have demonstrated that women without private insurance and women with Medicaid are less likely to receive cancer screening, the authors assessed whether this pattern has measurable consequences. If so, they hypothesized that uninsured patients and Medicaid patient have more advanced breast cancer when the disease is first diagnosed and that they would die sooner. Women in New Jersey aged 35-64 who were diagnoses with invasive carcinoma of the breast during 1985-1987 were identified for this study.

Privately insured women were older, more likely to be white, more likely to be married, more likely to have a high median household income, and less likely to have a coexisting diagnosis. Uninsured women and those covered by Medicaid had distant staged disease when they were initially diagnosed (12.3% of uninsured and 17.4% of Medicaid beneficiaries versus 7.3% of privately insured). Stratifying by age did not eliminate these differences, although they were only statistically significant in some of the age/insurance categories.

Survival between 54 and 89 months after diagnosis was worse for uninsured and Medicaid patients. This pattern was true for those with local and regional staged disease, but not for those with distant metastases at presentation. In multivariate analysis with adjusted for sociodemographics, SES, health status and stage, both uninsured and Medicaid patients had a higher risk of death than privately insured patients (RR=1.49 for uninsured versus privately insured; RR=1.40 for Medicaid versus privately insured). Again, stratifying by age did not eliminate this pattern, but differences reached statistical significance for only some age/insurance groups.

The authors note that one limitation of this study is the possibility of incomplete staging for uninsured and Medicaid patients. Incomplete staging might lead to inadequate therapy, and therefore might have influenced the results of this study. Additionally, treatment was not measured, although it is an obvious confounder in these analyses. Finally, lead and length time biases are potential problems that were not addressed.