Cancer death rates for older Asian Americans: Classification by race versus ethnicity

B1

Objective:
 Although Asian Americans are the fastest growing population in US, data describing their health are limited. For most health statistics, Asian Americans are grouped into a single race category because of obstacles to determining ethnic-specific rates, even though there is general agreement that ethnic subgroups may be so dissimilar that grouping them together is not meaningful for surveillance or informative for etiologic research. Furthermore, cancer rate estimates for Asian Americans were questioned because of recognized inconsistencies between ethnic designation from different sources: cancer registries or death certificates for rate numerators, and the Census for denominators. In this project we avoid pitfalls in using different data sources for numerators and denominators and determine site-specific cancer death rates for persons aged 65+ in six Asian American ethnic subgroups (Chinese, Indian, Japanese, Korean, Filipino, Vietnamese). We determine the cancer sites for which an aggregate race category is a misleading summary of subgroup cancer risk.
 Methods:
 Asian Americans are identified through administrative files at the Social Security Administration (SSA) and ethnicity is inferred, based on our previously developed algorithm using sex, date of birth, race, country of birth, surname, father’s surname (for women), and given name. To determine cause of death, we link SSA death records to death certificates from two sources: the 1990-1999 Death Statistical Master file for California and the National Death Index. After accounting for sampling weight and conversion from ICD-9 to ICD-10, we determine cause-specific death rates for 21 cancer sites and test for heterogeneity among subgroups using Poisson regression.
 Results:
 There are a total of 4,459,884 person-years of observation, over one million each for Chinese, Japanese and Filipino, and about 300,000 each for Indian, Korean and Vietnamese. The mortality rates of all cancer sites combined are significantly lower for each Asian ethnic group relative to Whites. However, there is significant heterogeneity among the Asian ethnic groups. Indian men and women have the lowest rates, about half the White rate. Korean men and Chinese women have the highest Asian rates.
 For over half of cancer sites, every Asian group has a lower rate than Whites: larynx; lung; skin; prostate; female breast; corpus & uterus; ovary; kidney & renal pelvis; bladder; brain & CNS; Non-Hodgkins lymphoma; and leukemia. Among these cancers, there is highly significant Asian heterogeneity (p < 0.001) for three sites: lung, breast and prostate. For three sites most subgroups have higher rates than Whites (stomach, liver and cervix), there is also significant heterogeneity. For other cancer sites, there is an inconsistent pattern, with some subgroups having rates lower than Whites and some having rates similar to Whites.
 Conclusion:
 Cancer death rates vary significantly among Asian subpopulations for over half of cancer sites, including the most prevalent cancers (colon, lung, breast and prostate) and those recognized as high among Asian Americans (stomach, liver and cervix), so the aggregate Asian American rate masks heterogeneity for these sites. Asian Indians are most often the subgroup with a divergent rate, suggesting that disaggregating Asian Indians would improve the interpretability of aggregate Asian rates.