Predicting Incongruence between Self-reported and Documented Colorectal Cancer Screening in a Sample of African American Medicare Recipients

Abstract

Assessments of colorectal cancer (CRC) screening rates typically rely on self-reported screening data, which are often incongruent with medical records. We used multilevel models to examine health-related, socio-demographic and psychological predictors of incongruent self-reports for CRC screening among Medicare-insured African Americans (N = 3,740). Results indicated that living alone decreased, and income increased, the odds of congruently self-reporting endoscopic CRC screening. Being male and having greater number of comorbidities decreased, and having less than a high school education increased, the odds of congruently self-reported fecal occult blood tests. Living alone, age and income had the most robust effects across classifications into one of four mutually exclusive categories defined by screening status (screened/unscreened) and congruence of self-reports. The results underscore the clinical importance of gathering socio-demographic data via patient interviews, and the relevance of these data for judging the veracity of self-reported CRC screenings behaviors.

Keywords:

• African Americans
• colorectal cancer screening
• self-report incongruence
• socio-demographics

ACKNOWLEDGMENTS

We are grateful to Drs. Louis Penner and Thomas Templin for their assistance with previous versions of this manuscript.

FOOTNOTE

Funding

This study was supported in part by grants from the National Institutes of Health, 5P30 AG015281, the Michigan Center for Urban African American Aging Research, and CMS Cancer prevention and treatment demonstration for ethnic and racial minorities (Award 1 AO CMS 3000068) and by NCI grant (U54CA153606, T. Albrecht, PI).

Notes

* Data for 17.5% of self-reported FOBT and 13.1% of self-reported SIGCOL were missing. Using medical record and available self-report data, we were able to classify CRC screening self-report congruence for 4645 respondents for CRC endoscopies, and 4409 respondents for FOBT before listwise deletion for missing covariates. Because we were hesitant to use our substantial theoretical covariates to both impute missing outcome data and test hypotheses, we did not impute missing outcome data. While missingness was negligible for most covariates (0.5% for self-reported health status, 0.4% for gender, 2.9% for number of comorbidities, 4.6% for family history CRC, 1.3% for living alone, 0.5% or education, 0.4% for partner status, 0.4% for age and 0.5% for income), 20.1% of respondents either refused to indicate their income or indicated that they did not know what it was (ie, the data were not missing; rather, they were consciously withheld). Given that this did not constitute random missing data, we did not employ imputation techniques for this covariate.