Information from this qualitative phase will lay the basis for developing general content modules that are consistent with the ASK model of health literacy [26]. This process will draw specific content from established domains of chronic disease management by teaching health literacy skills and health condition conceptual knowledge.

Interventions will be developed that are consistent with recommended strategies for persons with limited literacy skills [64,65]. All materials will use health care-related content as a vehicle (eg, prose about diseases or a table about desirable blood sugars). Intervention elements will focus on the three basic literacy types, as defined by the Educational Testing Service: prose, document, and quantitative [66]. Consistent with research on the effectiveness of direct training in reading comprehension for adult learners [67], training in prose comprehension will focus on strategies to develop functional skills such as identifying key facts, inductive reasoning from multiple facts, and self-monitoring of comprehension. Training in document literacy will use tables, charts, and health-related forms (eg, insurance, consent) and emphasize strategies for identifying key information and strategies for using the information to answer relevant questions (eg, “Which insurance program is better for me?”). Training for quantitative skills will integrate basic instruction on quantitative concepts and procedures for understanding risk, probability, and cost [68,69].

Information from the qualitative phase will allow us to develop general content modules that will focus on the key problems in chronic disease management listed above. Key strategies will be to teach health literacy skills and improve health condition-related knowledge. Problems, needed and useful skills, and disease-related conceptual knowledge revealed in qualitative work will then be used by investigators to create a content outline for each module. Content-relevant skills will be integrated with condition-related knowledge in each module (eg, by teaching reading comprehension skills using materials that are directly related to health). We will use a similar strategy in teaching listening comprehension by focusing on encounters with healthcare providers. This might involve teaching strategies for preparing a list of questions before an appointment, taking notes during the encounter, and being appropriately assertive in interactions.

Cognitive load theory [70,71] dictates that for optimal learning, instructional material must be presented in segments of content that do not exceed the learner’s ability to take in and retain each element. Consideration of this issue may be especially important in persons with limited health literacy [72]. Consistent with this issue, material will be presented in small segments via text and audio, with graphic elements reinforcing themes but not distracting from participants’ learning [70]. Interactions with the computer will only require that participants tap on the screen to press images of buttons, thereby keeping required computer skills to a minimum.

In a series of team meetings, content created by individual team members will be reviewed and alternate versions will be created at the three reading levels planned for each intervention group (3^rd, 6^th, and 8^th grade levels) based on the Flesch-Kincaid formula for English [73] and its Fernandez-Huerta modification for Spanish [74]. No clear criterion for the difficulty of health-related materials is available for use in tailoring, so we will use the criterion of grade equivalent levels. In this context, however, it will measure the difficulty of health-related materials and thus materials will be tailored to participants’ levels of health literacy.

Intervention content will be developed in either English or Spanish, with versions translated by team members fluent in the target language, with an emphasis on appropriate cultural adaptation rather than literal translation [75]. This strategy was previously effective in creating a health literacy measure in both English and Spanish [25], and helped to ensure that both language versions were intelligible to individuals from diverse linguistic and cultural backgrounds. Given the importance of regional usages of Spanish (eg, the same word may mean different things in Puerto Rico or Mexico), the Spanish version will be developed with an emphasis on standard Spanish to avoid comprehension problems. Intervention content will be assessed during usability testing, in addition to a focus on interface design and ease of use.

After creation, acceptability and usability testing of the new materials will be completed using a procedure that we successfully used in other projects [76,77]. Individuals will be observed interacting with the computer app and will be asked to think out loud [78] so that their cognitive processes can be tracked, and areas of difficulty can be identified [76]. After five participants have completed this process, the intervention will be revised and tested again with a new group of five participants [77]. This process will be repeated until no further difficulties are identified, potentially eliciting responses from 15 to 20 participants [79,80].

Implementation will allow participants to monitor their own progress and provide centralized data collection. The intervention will be created in standard off-the-shelf software, Captivate (Adobe Corporation; San Jose, California), which is widely used in developing educational media. This software has been used in previous projects due to its ease of use, integration of multimedia elements, and simple deployment to most learning management systems. The current version (Captivate 9) creates apps in HTML5/CSS/JavaScript formats that can be readily deployed on Windows, iOS, and Android devices, and on the Internet. Output from this program can be readily packaged as downloadable apps for the operating systems of most mobile devices, including iOS (iPhone, iPad), Android, and Windows Phone.

Delivery of FLIGHT/VIDAS on tablet computers will require modifications to the size of graphics, layout of page elements, and changes in font sizes. The file size for graphic elements will be reduced to improve download times over the Internet. Studies continue to show a dramatic increase in Internet use among older persons [81] and multiple studies show that elders evaluate the usability of tablet computers positively [82-84]. We will ask all participants to complete FLIGHT/VIDAS and the health literacy intervention on tablet computers, but it is possible that some of the oldest participants may not be able to do so. All participants in our previous study were able to complete assessments on touch screen computers (94 of whom were 70 years of age or older), so participants who have difficulty (ie, express frustration or state that they are not able to accurately record responses) with tablet computers will use the larger-format touch screen.

We will recruit Spanish- and English-speaking persons aged 40 years and older whose health literacy is at the 8^th grade level or lower. Literacy levels will be assessed via validated scores on the Rapid Estimate of Adult Literacy in Medicine (REALM) [85], or the Short Assessment of Health Literacy in Spanish Speaking Adults (SAHLSA) [86]. This level was chosen to include persons in the lower end of the basic proficiency level and below, as defined by the US National Assessment of Adult Literacy [1,87-89]. The REALM and SAHLSA were selected based on their demonstrated relation to health status in other studies and data from our previous study that enabled us to use them as screening tools. The intervention will thus target those most likely to profit from instruction at a moderate difficulty level approximately two grade levels lower than this (6^th grade), or those who may need low-literacy level instruction (3^rd grade) based on their instructional reading level being approximately two grade equivalents below the tested level. English-speaking participants will include approximately equal numbers of African-Americans and non-Hispanic whites, with recruitment stratified by age decades (to enhance existing FLIGHT/VIDAS norms) from age 40, with no upper limit. Considering the importance of broadening the geographic basis of our normative population and ensuring the usefulness of the health literacy intervention for diverse groups, in the proposed study we will expand data collection to include Grady Memorial Hospital in cooperation with our collaborators at the Morehouse School of Medicine and Emory University (all in Atlanta, Georgia). We will continue to recruit African-American, white non-Hispanic and Hispanic participants at our Nova Southeastern University site, as we did in our earlier project. During this project, we plan to expand our recruiting efforts in South Florida to include larger numbers of Spanish speakers with low educational backgrounds who are originally from Mexico. We will randomize 432 participants, as needed (216 at each site; see power analyses for sample size justification, below). Numbers of participants in Figure 1 reflect planned recruitment based on power analysis and potential attrition.

Participants will be eligible for the study if they are 40 years of age or older, are treated for at least one of the conditions in the Functional Comorbidity Index [90] (cardiovascular disease, arthritis, cancer, lung disease, osteoporosis, depression, and others) with at least one medication, have health literacy below the 8^th grade level as indicated by the REALM (English) or SAHLSA (Spanish), and are willing to participate for the length of the study. Potential participants will be excluded if they are not able to provide informed consent, have a cognitive or psychiatric disorder that impairs their ability to safely participate in the study, or are not able to remain involved over the full length of the study.

A screening procedure for the new project was developed by assessing the ability of the REALM [85] or SAHLSA [91] to predict if participants had an 8^th grade or lower level of reading on the Woodcock-Johnson or Muñoz reading tests. Using the R statistical package pROC [92], a score of 62 on the REALM predicted reading skills less than an 8^th grade level on the Woodcock-Johnson (area under the curve [AUC]=0.79, 95% CI 0.72-0.87; sensitivity=0.80, specificity=0.67) with similar results for the SAHLSA. Earlier analyses showed that the FLIGHT/VIDAS General Health Literacy (HL) scale will also be effective for use in detecting participants with lower levels of reading skills (all P- values <.001). Analyses did not differ for language groups (AUC for Spanish sample=0.86, P<.001).

Recruitment at the Atlanta sites will take place in the Primary Care clinics of Grady Health Services and Emory Healthcare. Grady Health Services is the largest public health provider in Georgia, and fifth largest in the nation, with more than 130,000 outpatient visits per year. Emory Healthcare has satellite clinics throughout metro-Atlanta and serves a broad array of patient demographics and backgrounds. In current studies, it is possible to access the electronic medical record to prescreen study participants. We will continue this practice after obtaining Institutional Review Board (IRB) permission. The study recruiter will identify potential participants meeting inclusion criteria; he or she will then meet with clinic staff and providers, supply them with study information (flyers, brochures), and ask the clinic staff/provider to give the study information to these potential participants. Passive recruitment through study flyers and information about the study on clinic phone-wait systems will also be used. All studies utilizing these procedures are currently on target with recruitment goals.

In addition to a pool of subjects who participated in prior studies and have given permission to contact them in future studies, Spanish-speaking Hispanics will be recruited from Nova Southeastern University internal, family, and geriatric medicine clinics. Special efforts will be made to recruit a diverse group of Hispanics representative of South Florida and the rest of country’s Hispanic population, per US census data, particularly with respect to country of reference and educational attainment. In our previous study, we recruited more than 320 Spanish-speaking Hispanics over 3 years of active recruitment [25-28].

Participants in the experimental groups will use modules that are tailored on three key dimensions: (1) preferred language (English or Spanish), (2) level of health literacy (3^rd or 6^th grade), and (3) racial or ethnic relevance (using verbal and graphic elements that will enhance the relevance of the material to African-Americans, Hispanics, or non-Hispanic whites). In the experimental groups, the intervention will add a fourth dimension of tailoring by providing individual feedback on understanding, via ongoing assessment of learning and review of material that is not yet mastered. The intervention content will be closely linked to the ASK model of health literacy [26] by targeting skills and knowledge related to CDSM, as informed by results of the qualitative study in Phase 1.

Participants in the control group will view informational screens and answer questions similar to the experimental group, but materials will be presented at the standard and widely recommended 8^th grade reading level (the approximate average reading level of adults in the United States [93]) to simulate the literacy demands that participants typically encounter when using health care information. Materials will not be tailored for race or ethnicity, except that Spanish speakers will complete the intervention in their preferred language.

A typical intervention session will begin with the participant being greeted by study staff and directed to a touchscreen computer station. The study assistant will ensure that the patient has begun the correct study module (eg, self-management of sleep problems). The participant will view screens that present basic information on types of sleep disorders and their causes. After presentation of a discrete amount of information, a review screen is presented to consolidate learning (eg, “Common causes of insomnia are…”). The next several screens will present questions to check participants’ learning; if questions are answered incorrectly, the program loops to a review of the topic with a second presentation and check question. Specific sections of the intervention will focus on improving information search skills via computer-coached materials that facilitate comprehension of written materials on insomnia (eg, “What is the major point of this paragraph?”) and Internet information search skills (ie, how to find more information on the Internet). In addition, coaching on interacting with health care providers (eg, writing down questions before visits, or taking a friend with you who can help you remember what the doctor says) will be woven throughout the modules.

After confirmation of eligibility, participants will be randomized to treatment condition according to a computer-generated schedule using the randomization procedures available in Research Electronic Data Capture (REDCap) electronic data capture tools hosted at Nova Southeastern University [94]. The principal investigator at each site (RLO, DWV) will be responsible for each treatment assignment.

Primary outcome measures will be health literacy (FLIGHT/VIDAS), self-reported general health (Medical Outcomes Study 36-item Short Form; SF-36 [95]), health self-efficacy (Chronic Disease Self-Efficacy Scale [96]), patient activation (Patient Activation Measure [97]), and treatment adherence. Additional data will provide the ability to consider race, ethnicity, age, and other variables as potential confounders. These data will include demographic variables, mood [98], stress [99], patient-provider relationships [100], treatment engagement, general cognitive abilities, academic skills, health conditions, current symptoms, health care attitudes, self-efficacy [101], and health care utilization.

Severity and number of health conditions will be assessed with the Functional Comorbidity Index [90], a previously validated measure linked to physical status, and an index based on the Midlife in the United States study [102] protocol, as used in our earlier study [27], to allow comparisons with data from that study. Assessments will include information on the specific symptoms of key problems in chronic disease management, and will allow a pre/postevaluation of the effects of the intervention on each. Simple measures of physical status (walking speed [103], body mass index [104], and waist-hip ratio [105]) will also be obtained to assess the effects of health literacy, cognition, and comorbidity on these variables, as they have all been related to risks for mortality [6,106,107].

Most self-report instruments will be administered via Automated Computer-Assisted Self-Interview (ACASI) software with assessment materials presented on touchscreen computers. This strategy enhances data collection by allowing participants anonymity in their responses and reduces issues with hand scoring of measures and manual data entry. Both sites will use the ACASI software Questionnaire Development System (QDS; Nova Research, Bethesda, Maryland). Individually administered measures, including assessments of cognitive, academic, and health literacy status, will be hand scored and verified. QDS output files will be transferred between sites using encrypted file transfer technology also available in REDCap.

The intervention will be standardized by its computer-based presentation across sites, but assessments and interactions with site workers must also be standardized to ensure fidelity across sites. Procedures will be standardized via visit flow sheets accompanied by training and monitoring visits by the investigators at each site. To ensure secure retention delivery of data across sites, all data will be entered into the REDCap software twice [94]. Each entry will be compared using utilities in REDCap and discrepancies will be resolved via inspection of original paper-based data collection instruments.

All assessment procedures planned in this study are either standard psychological or educational procedures (or are similar to standards), and all interventions are similar to typical educational interventions, so we believe that all interventions are associated with minimal risk to participants. We do not anticipate the need for a formal data safety monitoring board for this study, but we will follow a data safety monitoring plan. The plan will require ongoing monitoring of participants’ reactions to assessment and intervention activities with formal logs recording any events suggestive of participant distress, frustration, or boredom. All study personnel will be instructed on the importance of recording any relevant events in study logs, the contents of which will be reported on a monthly basis to the principal investigator for minor incidents, and immediately for any serious event (eg, substantial indication of participant distress such as crying, expression of desire to discontinue participation due to emotional upset, or request for emotional assistance due to of feelings or thoughts triggered by assessments). All events will be tracked and reviewed via regular conference calls with site principal investigators for determination of what changes in study procedures should be made to ensure the safety of participants.

Given the length of the study and frequency of study visits, we will be able to maintain close contact with all participants in an ongoing fashion. We will be able to develop and maintain a relationship with each participant that facilitates communication about their reactions to the study. Study personnel will be trained to be alert for signs of emotional upset or frustration in participants as a reaction to study procedures, and will be proactive in approaching participants about their emotional states. Participants will be offered supportive counseling if indicated, and referral for other services if needed.

Finally, we will routinely assess (both formally and informally) participants' reactions to the study using interviews and an ACASI questionnaire that will allow them to privately express any concerns or problems related to the study. We will actively monitor participants' experience in the study, both informally in exit interviews and formally via their responses to ACASI questionnaires, to determine if the study is excessively stressful or upsetting.

Informed consent is a process that will begin with IRB review of all recruitment materials. All study personnel will complete mandatory training in the protection of human subjects in research, including the courses of study contained in the Collaborative Institutional Training Initiative program [108], as required by the IRBs of Nova Southeastern University and Emory University, and only trained personnel will be involved in the process. The informed consent process will continue as study personnel explain study procedures during the education process preceding the request for written informed consent.

Participants will be protected from the risk of inadvertent disclosure of their health status or other health-related information by identifying all study-related data collection instruments by the participant's study number. Participants' identities will only be linked to data via a list that includes participant names and contact information and their participant identification number. All study databases will be kept on password-protected computers or secure servers (REDCap). Paper copies of all materials will be kept in locked filing cabinets in the investigators’ offices.

Initial review will evaluate data for completeness and valid characteristics through inspection of descriptive statistics. Although every effort will be made to minimize loss to follow-up, we recognize that some attrition is likely to occur. We will evaluate patterns of missingness in the data and follow widely-recommended procedures for dealing with this issue [109]. If our evaluation shows that missing data can be considered as missing at random (MAR) then our use of full-information maximum likelihood routines in MPlus statistical software (Muthén & Muthén; Los Angeles, California) will result in unbiased estimates of parameters [110]. If data are not MAR, we will include pattern of missingness in random effects pattern mixture analyses [111]. Group equivalence on demographic characteristics (except those defining group membership) and baseline measures not used in assignment will be assessed, and if differences are found these variables will be used as covariates in analyses. Other analyses will evaluate the distributional properties of dependent variables to assess the suitability of planned statistical models.

Hypotheses will be tested via three aims, detailed below.

Power analyses were evaluated using procedures available in the software program PASS 11 (NCSS, LLC; Kaysville, Utah). The effect sizes used were drawn from studies of reading interventions for low- and high-ability readers, and on the effects of our computer-delivered intervention on participant knowledge. Results show that a total sample size of 30 per group (high vs low vs control health literacy; African-American vs Hispanic vs non-Hispanic white), after possible attrition in any group of up to 38% (yielding at least 30 persons per group), will provide powers greater than 0.90 to test main effects and 0.80 or greater to test the interaction of time by group or by treatment condition. Power is adequate for all two-way interactions; the power to detect all three-way interactions would require a very large sample size that would be excessively costly. We will evaluate higher-level interactions even when not significant to assess the possibility of their presence.