When subjects reported multiple DNA-PK pathway sources of health insurance without indicating the primary source, we imposed the following hierarchy in decreasing order of priority: Medicare >Medicaid > private group insurance > private individual insurance > other > none. For example, we classified subjects reporting Medicare and Medicaid coverage (i.e., dual eligibles) as having Medicare insurance. The private health insurance group used in our analysis included subjects reporting private

insurance of any type. The interviews gathered information on individual characteristics (e.g., socio-demographic and economic traits, health status, caretaker responsibilities, and technology access). All individual covariates used reference specific Pew survey questions and their responses (details available upon request). We included age as a continuous variable. Specific survey questions distinguished Internet users from non-Internet users as well as cell phone users from non-cell phone users; these questions provided

a filter in the survey for subsequent questions asked of only Internet users, only cell phone users, or combination users. We classified any subject indicating prior use of the Internet within the Pew survey as an Internet user, which provides a conservative estimate of Internet accessibility and use. The survey asked questions on text messaging behavior only among respondents who had previously indicated that they were cell phone users that sent/received text messages. Interview questions, response categories, and response data are all available on the Pew Web site (Pew Research Center, 2012). In all models, we dichotomized educational attainment, categorizing subjects as having any college degree or no degree. We were interested in the role that clinical need due to poor health might have on outcomes, thus in the main analyses, we dichotomized the self-reported health status variable (originally on a 5-point Likert scale) into “Fair/Poor health vs.

Not being in Fair/Poor health.” For the subjects who reported “Don’t Know” or who refused to answer, we coded them as “Not being in Fair/Poor Health.” We used similar definitions to dichotomize variables representing respondents’ having a chronic disease or any AV-951 recent emergency health event.2 We defined informal caregivers as anyone who reported providing unpaid care to an adult or child. To determine the categories of Federal Poverty Level (FPL), we followed the Health and Human Services 2012 Poverty Guidelines, assigning income as the mid-point of the category. If a respondent indicated they had children, we assumed two children lived in the household. We limited the number of adults per household to six and determined household size from the sum of the children and adults in that home. Based on income and household size, we determined the percent of federal poverty and created categorical poverty level variables.