Participants were enrolled in the Cancer Care Outcomes Research and Surveillance Consortium (CanCORS), a prospective, population-based, multisite observational study of participants with lung and colon cancers that has been previously described.8 The population was diverse with respect to race, socioeconomic status, and geography. The purpose of the study was to assess the impact of system, care provider, and patient factors on cancer outcomes. Participants were at least 21 years of age at the time of CRC diagnosis and were enrolled within 3 months of diagnosis during 2003 to 2006. The study collected participant surveys, surrogate surveys for participants who were deceased or too ill to participate, and medical records data.
Abstractors at each site collected information on tumor characteristics and cancer treatments received. Participant and, when necessary, surrogate surveys were completed using computer-assisted telephone interviews. Surveys included items about demographic and socioeconomic factors (age, insurance coverage, income), communication with health care professionals, and beliefs and preferences regarding cancer treatments. The surveys have been previously described.9
CanCORS included patients with colorectal and lung cancers who were enrolled by 7 groups of investigators.8 North Carolina recruited 990 patients with CRC, and it was the only site to administer health literacy assessments.
The study population was a random sample of 347 participants from the North Carolina–based CanCORS study. The sample was stratified by self-reported years of education with oversampling of lower strata to achieve similar-sized strata of adequate vs inadequate or marginal health literacy, ultimately with the goal of enhancing power for planned analyses. We used the following sampling fractions for the respective ranges of years of education: 0 to 8 years (100%), 9 to 11 years (100%), 12 years (70%), and more than 12 years (40%).
The sample included all cancer stages. The Institutional Review Board at the University of North Carolina at Chapel Hill approved the protocol. All participants provided informed consent.
Measure of Health Literacy
Functional health literacy was assessed using the Short Test of Functional Health Literacy in Adults (S-TOFHLA).10,11 A trained interviewer visited participant homes to administer the assessment in person. The assessment tested reading comprehension using 36 questions in response to 2 prose passages.
Across the entire sample, 4 different interviewers were used, but 1 interviewer alone administered the assessment for any given participant. All interviewers were trained in how to administer the S-TOFHLA. The interviewer read a scripted introduction and instructions to the participant, and then remained silent while the participant completed the questionnaire. Participants were given up to 7 minutes to complete the questionnaire, but they were not told beforehand that the assessment would be timed.
Scores were categorized as inadequate (0–16 correct), marginal (17–22 correct), or adequate (23–36 correct). For analysis, we combined marginal and inadequate scores into 1 category.
As part of the surveys, participants in CanCORS answered questions about whether they received adjuvant chemotherapy, how the chemotherapy decision was made, and their beliefs and preferences regarding chemotherapy. Chemotherapy is generally recommended for patients with colon cancer diagnosed with stage 3 or 4 disease.12 We examined responses about chemotherapy decisions among participants with CRC in whom health literacy was assessed and who were diagnosed with either stage 3 or 4 disease. Survey responses of “Do not know,” “Declined to answer,” “Not applicable,” or were missing were considered non-informative. Non-informative responses were excluded when conducting Fisher exact tests to compare survey responses by level of health literacy.
CanCORS tumors were staged according to the TNM classification system. We considered stage 1/2 to be early-stage disease and stage 3/4 to be late-stage disease.
Participants were followed for survival after baseline data collection. Vital status for all participants was verified using the Social Security Death Index on May 4, 2010, providing at least 42 months of follow-up observation time for each person. We defined participant survival as dichotomous vital status (alive or dead) at last observation.
Among participants in the literacy sample with any stage of cancer (N = 347), we calculated overall and health literacy–stratified descriptive statistics for demographic and socioeconomic characteristics. Chi-square tests of association were conducted to examine differences in participant characteristics by level of health literacy. We performed logistic regression analyses to estimate associations between health literacy and (1) whether participants received chemotherapy (for stage 3/4 disease), (2) cancer stage at diagnosis (for all participants), and (3) all-cause mortality at last observation after baseline (for all participants). Across all stages, we estimated the marginal effect of health literacy on survival as well as its conditional effect on demographic and socioeconomic covariates.
Among 130 participants with stage 3/4 disease in the health literacy sample, we calculated overall and health literacy–stratified descriptive statistics for responses to survey questions about their beliefs and preferences regarding chemotherapy, communication with health care professionals about chemotherapy, and their role in making the decision whether to receive chemotherapy. We used Fisher exact tests to evaluate whether survey responses differed by participant health literacy level.