Impact of intercensal population projections and error of closure on breast cancer surveillance: examples from 10 California counties

Abstract

IntroductionIn 2001, data from the California Cancer Registry suggested that breast cancer incidence rates among non-Hispanic white (nHW) women in Marin County, California, had increased almost 60% between 1991 and 1999. This analysis examines the extent to which these and other breast cancer incidence trends could have been impacted by bias in intercensal population projections.MethodWe obtained population projections for the year 2000 projected from the 1990 census from the California Department of Finance (DOF) and population counts from the 2000 US Census for nHW women living in 10 California counties and quantified age-specific differences in counts. We also computed age-adjusted incidence rates of invasive breast cancer in order to examine and quantify the impact of differences between the population data sources.ResultsDifferences between year 2000 DOF projections and year 2000 census counts varied by county and age and ranged from underestimates of 60% to overestimates of 64%. For Marin County, the DOF underestimated the number of nHW women aged 45 to 64 years by 32% compared to the 2000 US census. This difference produced a significant 22% discrepancy between breast cancer incidence rates calculated using the two population data sources. In Los Angeles and Santa Clara counties, DOF-based incidence rates were significantly lower than rates based on census data. Rates did not differ significantly by population data source in the remaining seven counties examined.ConclusionAlthough year 2000 population estimates from the DOF did not differ markedly from census counts at the state or county levels, greater discrepancies were observed for race-stratified, age-specific groups within counties. Because breast cancer incidence rates must be calculated with age-specific data, differences between population data sources at the age-race level may lead to mis-estimation of breast cancer incidence rates in county populations affected by these differences, as was observed in Marin County. Although intercensal rates based on population projections are important for timely breast cancer surveillance, these rates are prone to bias due to the error of closure between population projections and decennial census population counts. Intercensal rates should be interpreted with this potential bias in mind.