We had to rely on the readily-geocoded diagnosis address which doesn’t guarantee a same level of geocoding accuracy as that of birth addresses geocoded using our automated approach. However, the use of 100m as cutoff as well as alternative 200m should take account into the possible positional errors during geocoding. Besides, there might be intended or unintended misreporting in birth residence and diagnosis residence among individuals seeking health care services. Address information from LexisNexis has several intrinsic issues including multiple unique addresses for the same time period, inaccurate or missing first seen and last seen dates associated with certain addresses , disagreement between LexisNexis addresses and registry-obtained addresses at the time of self-reporting, inconsistent quality in different sub-groups of general population, and time-varying sources of addresses over years, therefore limiting the power to rely solely on it for large-scale records-based epidemiological studies. Future research attempting to reconstruct residential histories for study subjects using LexisNexis should also be aware of the potential selection bias, introduced by varying availability of public records by sub-groups of populations. Despite that, LexisNexis’ rich data could augment existing address information and assist in reconstructing residential histories. Future approaches with the advancing technology in the big data era may help to obtain a more accurate residential history for populations with or without diseases and evaluate the impact on estimating associations between environmental exposures and childhood cancers.
Though such data sources would be ideal for studies in environmental epidemiology,grow bag gardening confidentiality is a concern. In conclusion, residential mobility among childhood cancer cases diagnosed in California and their matched controls was associated with a number of child and maternal factors. Unlike adult cancers or other childhood outcomes such as asthma, which have a range of known demographic risk factors, the etiologies of childhood cancers remain largely unknown and therefore they do not have as many established risk factors, so this may be less of an issue. The overall agreement between exposures to agricultural pesticides in early life of children assessed using a 2-km buffer of residences at birth and alternative addresses was moderate to good. These findings suggest that birth residence should be used with caution when estimating environmental exposures in early childhood, especially after the first year of life. LexisNexis data, or other similar methods for reconstructing residential histories, may be useful for augmenting existing address information and constructing residential histories in estimating environmental exposures for large records-based epidemiological studies. Future research should consider factors that might help correct for the misclassification introduced by residential mobility. In large-scale records-based epidemiologic studies of environmental exposures in early childhood, relying solely on maternal address at birth and birth or diagnosis address may be insufficient to produce accurate exposure measures in studies of childhood cancers. Recently large data linkage studies of childhood cancer have been conducted in which exposure measures were derived for residential addresses at birth or diagnosis; and alternative becoming available now is a public records database, LexisNexis, that provides all known addresses for a set of individuals upon request. It might be a useful source of augmenting address histories when estimating environmental exposures for such studies.
Residential mobility in early childhood that is associated with maternal or child demographics may cause exposure misclassification for agricultural pesticides applied in proximity to the residences . For the estimation of pesticide exposures from agricultural applicationnear homes, we previously showed that the overall agreement between exposures assessed within a 2-km buffer of residences in early life of children based on birth, diagnosis, or LexisNexis addresses was moderate to good, the agreement decreased with in increased distance between the two residences. Degree of misclassification also depends on the length of stay in each residence, as well as the accuracy of birth and diagnosis addresses.Numerous studies that have examined the potential impact of household use of pesticides in early childhood on brain tumors suggest overall positive associations . There are relatively fewer studies of the childhood brain tumor risk from residential proximity to agricultural pesticide applications, but they have suggested mostly null associations with low precision except one ecologic study showed increased risk. These earlier inconsistent findings might partly be explained by the lack of spatiotemporally accurate exposure assessment, which has been largely improved by GIS-based estimation systems. Here we explore the impact of exposure misclassification due to residential mobility between birth and diagnosis. The primary objective of the present study was to estimate and compare the effect estimates for agricultural pesticide exposures during children’s early life based on birth residence only and birth or diagnosis address with those based on reconstructed address histories using LexisNexis, as the ‘alloyed gold standard’. Earlier validation studies have shown that addresses acquired from LexisNexis are useful for reconstructing residential histories with an overall match rate of ~70-85% with detailed address histories obtained from personal interviews .
However, not all residents’ records can be matched to records in the LexisNexis public records. Thus a secondary objective of our study is to examine whether selection bias is likely to occur when we restrict our study population to individuals with available LexisNexis addresses only. The Air Pollution and Childhood Cancers study population and the three sources of addresses have previously been described in section 3.3. Cases with missing diagnosis address were excluded. In brief, for each individual requested, LexisNexis provided all known addresses and the first and last dates associated with them. Following previously developed methods , we removed all P.O. Box addresses, identified and removed duplicate addresses compiled from multiple sources, and created a residential history timeline from birth to diagnosis for each case and to the reference date for each control. We define the time from birth to diagnosis as the children’s lifetime, and consider those whose California LexisNexis addresses covered more than 80% of their lifetime as having LexisNexis addresses available, and the rest as unavailable. A 80% cut-off was chosen to account for the uncertainties and random errors of first and last dates reported for addresses in LexisNexis. Cases diagnosed with brain tumors were defined as International Classification of Childhood Cancer, Third edition ) code 031-036. Included in the primary analysis were cases of childhood brain tumors diagnosed ≤ 5 years of age and controls born in 2001-2008 with California addresses available from birth certificates, at cancer diagnosis, and from LexisNexis records. Our cases were limited to children age 5 andyounger who may inherently be more susceptible to prenatal and early life exposures than children or adolescents diagnosed with cancer at an older age. Because our pesticide estimates were only available through December, 2012, children born 2009 onward only have pesticide exposure estimates from their birth until a maximum of 4 years of age. Among children born in 2001-2008, 397 of 634 cases of brain tumors and 6,614 of 12,502 controls had at least one California address available on LexisNexis records which covered more than 80% of their lifetime. In the secondary analysis, we compared the former subset with the rest of the cases and controls with only birth and diagnosis addresses but without available LexisNexis records. We reviewed previous published articles which examined associations between specific chemical agents and childhood brain tumors,plastic grow bag and selected those carcinogens that have been reported to be positively associated with these outcomes of interest. Following two previously California based studies that examined the associations between childhood cancers and residential pesticide exposure, we selected the following potentially high-risk pesticides: propargite, methyl bromide, metam-sodium, trifluralin, simazine, dicofol, and chlorothalonil. These agents also have wide usage in California, high genotoxicity, carcinogenic potency based on U.S. Environmental Protection Agency , field volatilization flux, and persistence . Pesticide estimation was described in detail in in section 3.3. For each of the pesticides examined in this study, we summed the annual pounds applied per acre to obtain exposure values for eachcalendar year using the 2km buffer surrounding each address. These annual exposure estimates were then summed across the multi-year range for cases and controls. From the date of birth to the date of diagnosis for cases and reference date for matched controls, we used weighted averages with weights representing the proportions of the relevant exposure period falling into each calendar year. Three estimates were calculated accordingly, using 1) birth address for both cases and controls, 2) diagnosis address for cases and birth residence for controls , and 3) LexisNexis addresses for both cases and controls address throughout the entire period.
After obtaining the pesticide exposure estimates using annual PUR records, we further categorized children’s exposure status into lifetime “ever exposed” versus “never exposed”. We conducted univariate and multivariate conditional logistic regression analyses adjusting for year of birth and estimated odds ratios and 95% confidence intervals . Based on the literature as well as our own explorations of associations in our data , we considered the following potential confounders in the adjusted models: maternal and paternal age at birth , maternal race/ethnicity , maternal education , child’s sex, parity , payment source for prenatal care as a proxy for family income , maternal place of birth , and rural/urban status of residence at birth according to the Rural-Urban Commuting Area Codes , for which we dichotomized 1-3 as urban and 4-10 as rural, in addition to the matching factor year of birth, and left out the covariates that introduced < 10% change in estimate. Final models were adjusted for year of birth , child’s sex, maternal age, maternal race/ethnicity, and maternal education. Table 4.2 shows the child and maternal characteristics by availability of maternal LexisNexis records in their early life from birth through diagnosis for cases and reference for controls. The availability of LexisNexis addresses showed a clear temporal trend, with substantially less missing records in more recent years, indicating an improvement in data collection over time. Cases diagnosed at an older age compared with those diagnosed within their first year of life were more likely to have LexisNexis addresses available through their lifetime; similar pattern was observed for controls when age was calculated based on the reference date. Among both cases and controls, mothers of older age, with higher education, a non-Hispanic background, a higher parity, those who used private insurance, born in California, or resided in metropolitan areas or neighborhoods with higher SES at delivery were more likely to have available LexisNexis records . In adjusted models, there were modestly elevated ORs for childhood brain tumors associated with several pesticides including methyl bromide, simazine and propagate when assigning exposures to birth address for both cases and controls, and chlorothalonil, trifluralin, simazine, and propargite propagate when assigning exposures to birth address for controls and to diagnosis address for cases . However, when assigning exposures according to LexisNexis addresses for both, most associations were attenuated and the point estimates were close to the null . Restricting to children for whom only birth and diagnosis addresses but no LexisNexis data were available, we observed null associations for childhood brain tumors and all select pesticides using birth addresses but positive associations for trifluralin, metam-sodium, and propargite using diagnosis for cases and birth address for controls . Adjusted ORs were generally attenuated when additionally adjusting for child’s sex, maternal age, maternal education, and maternal race/ethnicity, except for propargite which had a slightly higher adjusted than crude OR. Among all five types of estimates, LexisNexis-based exposures tend to produce the most conservative effect estimates for most pesticides . Comparing the sub-population with available LexisNexis addresses with those without available LexisNexis addresses, ORs based on birth or diagnosis address were generally higher in those who did not have LexisNexis information available. This study aimed to assess the extent to which using a single source of address from birth certificates or cancer diagnosis records, or a reconstructed address history using public records, to estimate children’s early life exposures to agricultural pesticides might impact effect estimates for childhood brain tumors. We relied upon an innovative method to reconstruct residential histories as a “alloyed gold standard” accounting for mobility in children’s early life using a public database , and compared the effect estimates based on exposures assigned to birth and diagnosis addresses with those based on LexisNexis addresses.