Profile of e-Patients: Analysis of Their Cancer Information-Seeking From a National Survey

Abstract

Researchers have yet to fully understand how competent e-patients are in selecting and using health information sources, or, more importantly, who e-patients are. This study attempted to uncover how cancer e-patients differ from other cancer information seekers in terms of their sociodemographic background, social networks, information competence, and selection of cancer information sources. We analyzed data from the National Cancer Institute's 2005 Health Information National Trends Survey, and a series of chi-square tests showed that factors that distinguished cancer e-patients from other cancer information seekers were age, gender, education, employment status, health insurance, and membership in online support groups. They were not different in the other factors measured by the survey. Our logistic regression analysis revealed that the e-patients were older and talked about their health issues with friends or family more frequently compared with online health information seekers without cancer. While preferring information from their doctors over the Internet, e-patients used the Internet as their primary source. In contrast to previous literature, we found little evidence that e-patients were savvy health information consumers who could make informed decisions on their own health. The findings of this study addressed a need for a better design and delivery of health information literacy programs for cancer e-patients.

Preliminary findings of this research were presented at the Critical Issues in eHealth Research Conference in September 2006, and they were presented at the American Society for Information Science and Technology Conference in October 2007. We thank Dr. Lila J. Finney Rutten at NCI and Dr. Hyunji Kwon at King's College of London for their advice in statistical analysis of this study.

Notes

¹The jackknife variance estimation technique is the standard operating procedure for variance estimation in the HINTS survey. The sampling weights for each respondent are computed to generate statistically sound nationally representative estimators from the collected data. This procedure is necessary to adjust the risk of committing Type I error due to underestimation of standard errors caused by the multistage sampling design of the HINTS study. Consequently, this procedure helps to produce statistically valid standard errors for sampling estimators. The jackknife variance estimation technique takes selected subsets of the data for each “replicate” and determines a sampling weight for each respondent in the replicate subset, treating the replicate subset as the sample. The resulting weights are called replicate weights (Davis & Moser, 2004).

p value corresponds to chi-square for all categorical variables.

*p < .05; **p < .01; ***p < .001.

^a Model 1: Cancer e-patients vs. online seekers w/o cancer (n = 1,241).

^b Model 2: Cancer e-patients vs. offline cancer patients (n = 234).

^c Model 3: Cancer e-patients vs. offline seekers w/o cancer (n = 420).

^d All odds ratios (ORs) and corresponding 95% confidence intervals (CIs) shown in bold text are statistically significant at p < .05.

^e Tests were not completed due to the insufficient number of observations in each cell.

^a The original HINT item for “Used” is “The most recent time you wanted information on cancer, where did you go first?”

^b The original HINT item for “Preferred” is “The next time you have a strong need to get information about cancer, where will you go first?”

^c Printed materials include books, brochures, magazines, and newspapers.

^d Personal sources include friends, family, and someone with cancer.

^e Cancer organizations include cancer organizations and 1–800 numbers.