Our study has several limitations. First, we relied on self-report of the receipt and timing of screening. However, although the desire to report behaviors in a favorable light and telescoping (recalling events as occurring more recently than they actually did) probably result in overestimates of screening mammography and Pap smears (13,18), several studies have shown that self-report data validly measure screening use (42 - 44). Second, we lacked data on specific comorbid conditions other than diabetes, hypertension, and heart disease. However, the PCS-12 includes self-reported health and function, which predict mortality regardless of specific comorbid conditions (45). Third, our most significant limitation was the 63.7% response rate, and we lacked data to directly compare respondents with nonrespondents. However, comparisons of CHIS with 2000 U.S. Census data confirm that ethnic and income profiles of study participants closely mirror those of the California population (46). This suggests that despite nonresponse bias, our sample is generalizable to women in California. Nonetheless, it is possible that nonresponse bias resulted in an overestimate of the overall rate of screening. It is very unlikely, however, that nonresponse bias would change our robust central finding that age has a much greater effect on receipt of cancer screening than health status. For nonresponse bias to affect this central finding, it would have to be acting in dramatically opposite directions for older people compared with people in poor health, which seems implausible.