Improving the representativeness of the tribal behavioral risk factor surveillance system through data integration.

BACKGROUND: Previous literature showed significant health disparities between Native American population and other populations such as Non-Hispanic White. Most existing studies for Native American Health were based on non-probability samples which suffer with selection bias. In this paper, we are the first to evaluate the effectiveness of data integration methods, including calibration and sequential mass imputation, to improve the representativeness of the Tribal Behavioral Risk Factor Surveillance System (TBRFSS) in terms of reducing the biases of the raw estimates. METHODS: We evaluated the benefits of our proposed data integration methods, including calibration and sequential mass imputation, by using the 2019 TBRFSS and the 2018 and 2019 Behavioral Risk Factor Surveillance System (BRFSS). We combined the data from the 2018 and 2019 BRFSS by composite weighting. Demographic variables and general health variables were used as predictors for data integration. The following health-related variables were used for evaluation in terms of biases: Smoking status, Arthritis status, Cardiovascular Disease status, Chronic Obstructive Pulmonary Disease status, Asthma status, Cancer status, Stroke status, Diabetes status, and Health Coverage status. RESULTS: For most health-related variables, data integration methods showed smaller biases compared with unadjusted TBRFSS estimates. After calibration, the demographic and general health variables benchmarked with those for the BRFSS. CONCLUSION: Data integration procedures, including calibration and sequential mass imputation methods, hold promise for improving the representativeness of the TBRFSS.

Effect Modification of the Association Between Race and Stage at Colorectal Cancer Diagnosis by Socioeconomic Status.

OBJECTIVES: To compare risks of distant-stage colorectal cancer (CRC) diagnosis between whites and American Indian/Alaska Natives (AI/ANs) and to explore effect modification by area-based socioeconomic status (SES). DESIGN: Retrospective cohort study using data from the Oklahoma Central Cancer Registry. SETTING: Oklahoma. PARTICIPANTS: White and AI/AN cases of CRC diagnosed in Oklahoma between 2001 and 2008 (N = 8 438). A subanalysis was performed on the cohort of those aged 50 years and older (N = 7 728). MAIN OUTCOME MEASURE: Risk of distant-stage CRC diagnosis stratified by SES score. RESULTS: Race and SES were independently associated with distant-stage diagnosis. In SES-stratified analyses, AI/ANs in the 2 lowest SES groups experienced increased risks in the overall cohort and among those aged 50 years and older. In multivariable models, risks remained significant among those aged 50 years and older in the lowest SES groups (Adjusted risk ratio SES score of 2: 1.31, 95% confidence interval: 1.06-1.63 and adjusted risk ratio SES score of 1: 1.21, 95% confidence interval: 1.01-1.44). CONCLUSION: Socioeconomic status is an effect modifier in the association between race/ethnicity and stage at CRC diagnosis. Disparities in stage at CRC diagnosis exist between AI/ANs and whites with lower estimated SES. Efforts are needed to increase CRC screening among lower SES AI/ANs.

Prevalence and Mortality of Melanoma in Oklahoma Among Racial Groups, 2000-2008.

INTRODUCTION: This study assessed the period prevalence (2000-2008) and mortality rates of melanoma, in Oklahoma, among different racial/ethnic strata. METHODS: We analyzed incident cases of melanoma from 2000-2008 from the Oklahoma Central Cancer Registry and determined disease duration using Kaplan-Meier survival analysis to calculate period prevalence of melanoma in Oklahoma. Using a series of Chi-Square tests, we compared period prevalence and mortality rates among the racial groups and compared mortality between Oklahoma and the US. RESULTS: White non-Hispanics in Oklahoma have the highest period prevalence (p<0.0001) among the racial strata. American Indian or Alaska Native (AI/AN) individuals have the second highest period prevalence in Oklahoma (p<0.0001). Furthermore, white non-Hispanics (p<0.0001) and AI/AN individuals (p=0.0003) in Oklahoma had higher mortality rates compared to the US. CONCLUSIONS: There are disparities in the prevalence and mortality of melanoma among the AI/AN population in Oklahoma, and prevention and education programs should focus on this population.

Cancer incidence and staging among American Indians in Oklahoma.

BACKGROUND: This study describes overall and site specific cancer incidence among AI/ANs compared to whites in Oklahoma and differences in cancer staging. METHODS: Age-adjusted incidence rates obtained from the Oklahoma Central Cancer Registry are presented for all cancer sites combined and for the most common cancer sites among AI/ANs with comparisons to whites. Percentages of late stage cancers for breast, colorectal, and melanoma cancers are also presented. RESULTS: AI/ANs had a significantly higher overall cancer incidence rate compared to whites (629.8/100,000 vs. 503.3/100,000), with a rate ratio of 1.25 (95% CI: 1.22, 1.28). There was a significant disparity in the percentage of late stage melanoma cancers between 2005 and 2009, with 14.0% late stage melanoma for whites and 20.0% for AI/ANs (p-value:0.03). CONCLUSIONS: Overall, there were cancer disparities between AI/ANs and whites in Oklahoma. Incidence rates were higher among AI/ANs for all cancers and many site specific cancers.

Combining Probability and Nonprobability Samples by Using Multivariate Mass Imputation Approaches with Application to Biomedical Research.

Nonprobability samples have been used frequently in practice including public health study, economics, education, and political polls. Naïve estimates based on nonprobability samples without any further adjustments may suffer from serious selection bias. Mass imputation has been shown to be effective in practice to improve the representativeness of nonprobability samples. It builds an imputation model based on nonprobability samples and generates imputed values for all units in the probability samples. In this paper, we compare two mass imputation approaches including latent joint multivariate normal model mass imputation (e.g., Generalized Efficient Regression-Based Imputation with Latent Processes (GERBIL)) and fully conditional specification (FCS) procedures for integrating multiple outcome variables simultaneously. The Monte Carlo simulation study shows the benefits of GERBIL and FCS with predictive mean matching in terms of balancing the Monte Carlo bias and variance. We further evaluate our proposed method by combining the information from Tribal Behavioral Risk Factor Surveillance System and Behavioral Risk Factor Surveillance System data files.