With use of a user-centered approach, the intervention development followed a series of iterative steps to ensure that all content was understandable and evidence-based. Guided by the C-SHIP framework and health communication best practices, the text messaging intervention was developed through a systematic process that included needs assessment interviews, followed by an iterative cycling through text message drafting, revision through health literacy evaluation, review by target community interviews via cognitive response interviews, scientific advisory panel evaluation and feedback, and message revision, followed in turn by usability testing of the entire program.

The order of the development processes was established from our research team’s multimedia intervention development experience [47,48] and the relevant literature [43,49] to ensure the integration of formative research, pretesting content, pilot testing,and the involvement of stakeholders. The following sections detail each phase of the developmental process. The final intervention product (Txt2commit) is also described. This study was approved by the Institutional Review Board at Fox Chase Cancer Center.

Constructs from the C-SHIP model were used to guide the intervention development strategy and content. The C-SHIP model offers a unifying theoretical framework for assessing and addressing the psychosocial relapse factors typically experienced by postpartum women [50-52]. The model is built on cumulative findings and theorizing from diverse relevant sub-areas of cognitive-behavioral science and evidence-based psychosocial interventions, and integrates key cognitive and affective processes. Guided by this model, we identified 5 areas of psychosocial relapse risk factors among low-income, minority women: (1) low knowledge and perceived risk for relapse, (2) inadequate decisional balance (low pros and high cons of cessation), (3) high affect (distress), (4) negative beliefs (low self-efficacy and fatalistic beliefs), and (5) lack of self-regulatory, practical, and social support strategies. Over the past decade, a sizable literature has accumulated to support the role of these factors in undermining smoking behavior, acting as barriers of self-initiated cessation and enrollment in smoking programs, and decreasing quit rates and maintenance of smoking abstinence [18,53-56]

Participants (N=30) were recruited via flyers and staff referrals at 3 Women, Infants, and Children (WIC) clinics located in the inner city Philadelphia region. Interested women were invited to participate if they: (1) were 18 years of age or older, (2) had quit smoking for at least 1 pregnancy, and (3) had given birth to at least 1 child in the past 3 years. Interviews lasted approximately 20 minutes and included specific questions addressing the experiences, motivators, and techniques that had helped women maintain smoking abstinence after pregnancy, as well as personal, social, and environmental factors that triggered a postpartum relapse. Their mobile phone and text message use and preferences were also obtained. Upon completion of the interview, participants received a $20 gift card as compensation for their time and input.

Following an initial review of the interview transcripts, the research team developed a coding scheme based on the C-SHIP model. Responses categorized under C-SHIP were coded into 1 of the 5 psychosocial relapse risk factors. Subcategories in each domain were established as relevant themes emerged from the data. The intention behind coding responses into the C-SHIP processes was to systematically assess the major psychosocial relapse risk factors and identify the unique pattern of challenges faced by low-income, inner city women during the postpartum period.This information was used to guide the development of the Txt2commit intervention text message content.

As reported above, text message content development was grounded in the C-SHIP theoretical model and guided by a formative research approach through a needs assessment interview phase with the target population. In addition, we conducted a comprehensive literature review, as well as an evidence-based guideline search for the potential adoption of existing materials to adapt to the target population. Content component constructs included factors such as messages for increasing awareness of the risks associated with smoking on infant’s health (knowledge), techniques for controlling exposure to stressful cues that trigger the urge to lapse/relapse (affective distress), reinforcing beliefs about the woman’s ability to sustain smoking absence (self-efficacy), considering the consequences of continued smoking (decisional balance), and identifying new behaviors to be substituted for smoking-related activities (self-regulatory skills). To respond effectively to situations in which the participant required an instant message to deal with craving or lapse situations, text messages were developed to offer strategies and emotional support to cope with craving/lapse situations and strategies for how to get back on track.

To ensure that the messages were designed for a low health-literate audience, 2 health literacy experts systematically evaluated all drafted messages using software and their health literacy expertise. The program Health Literacy Advisor (Health Literacy Innovations, LLC) was used to scan and highlight the text for complex terms, complex health terms, polysyllabic words (ie, words with more than 3 syllables), and long sentences (ie, sentences with 12 or more words).The findings from the health literacy analysis were then evaluated by the 2 health literacy experts and the research team to further revise the text messages to improve readability.

Following the revisions for health literacy, the messages were reviewed and vetted by participants (N=30) recruited at the 3 WIC offices with the same eligibility criteria identical to those of the needs assessment phase.Each participant reviewed 21 messages and was asked to think aloud about the messages, paraphrase the content, and respond to other inquiries and probes using the cognitive response technique [57]. Specifically, participants were asked if messages were relevant to them, understandable, and helpful in a stressful situation, as well as whether and how they would change the wording of the messages. Additionally, they rated each message from 1 (very poor) to 5 (very good) based on how helpful the message would be if they were inclined to smoke. Upon completion of the interview, participants received a $20 gift card as compensation for their time and input. The interview data and field notes were reviewed by the research team and messages were subsequently revised according to the participants’ ratings and comments through a consensus process. Low-rated messages were flagged for possible removal.

A multidisciplinary scientific advisory panel was convened to provide further insight into the cultural sensitivity, appropriateness, and message appeal of the intervention for the target population. The panel (N=7) was composed of smoking cessation professionals, with expertise in psychosocial behavior, health communication, WIC management, and community perspectives. Panel members were provided with a list of messages (with information on revisions made in the health literacy evaluation process and suggestions made by participants), and the C-SHIP coding guide for each message. Members of the advisory panel evaluated the appropriateness of the content of each of the messages and provided their recommended revisions to the wording and readability of each message from their individual perspective and expertise. The research team then reviewed the advisors’ comments and made necessary modifications to produce the final program.

Ten participants, with the same eligibility criteria identical to those of the needs assessment phase, were recruited to pretest the program for usability testing following the scientific advisory review phase. The program was delivered over 1 week. Participants were provided with phones equipped with text messaging service and were instructed as follows: complete (open and confirm receipt of) 7 system-initiated messages (1 per day), and initiate 3 crave (described below) and 3 lapse message requests (described below) during the 1-week testing phase. Participants were debriefed individually by research staff at the conclusion of the usability testing and received a $20 gift card as compensation for their time and input. They were asked to provide the following overall feedback: system-initiated text message understandability, cultural appropriateness of text message wording, ease of use of crave and lapse functions, and problems encountered in use of the program. Each participant was also asked to provide additional comments concerning the intrusiveness, timing, and general burden imposed by using the program. Feedback was compiled and evaluated by the research team to determine whether adjustments in the messages, system, or study procedures warranted correction or adjustment, with subsequent implementation of necessary corrections and adjustments.

The findings of the needs assessment interviews are reported elsewhere; participant characteristic information is presented in Table 1. In summary, participants (N=30) expressed the following reasons for why they resumed or refrained from smoking following childbirth: (1) motherhood demands (26/30, 87%), (2) partner and family relationships (22/30, 73%), and (3) the presence of other smokers in the environment (15/30, 50%). Participants reported 4 main strategies that helped them cope successfullywith postpartum cravings and relapses, including being informed of smoking risks (26/30, 86%), maintaining goal-oriented thoughts (18/30, 60%), thinking about the pros of quitting (26/30, 87%), and receiving positive social support from families and friends (27/30, 90%).

With regard to phone habits, most participants (21/30, 70%) had a home phone, but even more were mobile phone users (25/30, 83%). In addition, 93% (28/30) of participants reported that they used their mobile phones for texting, and 83% (25/30) reported that they texted on a daily basis. However, none of the participants in this population owned a smartphone. Therefore, our intervention only used text-messaging delivery, which is compatible with all models of mobile phones. Furthermore, when inquiring about their attitudes toward text message styles for a smoking cessation/relapse program, a more formal tone of content without the use of common text messaging abbreviations was preferred because it was thought to increase the creditability of the program. Information gathered from the needs assessment interviews was synthesized by the research team to guide the drafting of the text message content.

A pool of 204 messages was initially developed based on the findings of the needs assessment and a review of the relevant literature.The emerging themes from the needs assessment content analysis were also integrated into the message drafting. The C-SHIP model provided a theoretical guide for message creation because this model indicates specific areas that are especially relevant for health interventions. Smoking facts from several evidence-based resources, including the National Cancer Institute’s fact sheet concerning the harms of smoking, were identified and extracted to be included within the messages where appropriate.

The text messaging intervention developed included 3 main components: low-frequency messages to be sent during the third trimester (1 per week for engagement before the start of the main intervention), high-frequency messages to be sent after participants had given birth (1 message per day for 1 month), and messages available upon participant request when experiencing a craving for a cigarette or a lapse into having smoked one. The messages were systematically and comprehensively developed to address each component, so that some messages were specific to the prenatal or postpartum period and some to the situation of craving or lapsing. The drafted messages were coded across the 5 C-SHIP constructs by the research team.

The initial assessment of the 204 text messages identified a number of issues, including the use of polysyllabic words, complex health terms, complex nonmedical terms, and long sentences. After the 2 health literacy experts reviewed and modified each message in collaboration with the research team, a second assessment using Health Literacy Advisor revealed a significant reduction in the following health literacy problems: a 72% reduction in polysyllabic words, 79% reduction in complex health terms, 91% reduction in complex nonmedical terms, and 47% reduction in long sentences. Most of the complex terms remaining in the text messages were words that pertained to the message content and the study, such as tobacco, smoking, and cancer. During the message review interview phase described below, these words were specifically tested and found to be readily recognized by study participants, indicating that they were not difficult to read or understand.

Messages were reviewed by 30 women at 3 different WIC clinic sites using the cognitive response interview technique. Table 1 displays the sample characteristics. Participants provided specific comments regarding improvements that could be made to the messages, as well as their general reaction to each message. This feedback was used to guide message revisions. For all but 1message, at least 67% (20/30) of participants found the messages to be understandable. Messages were generally viewed as helpful, obtaining a mean rating of 4.2 on a 5-point scale, and 50% of the messages were highly rated (>4.5). However, 25 messages received lower ratings, as summarized in Figure 1.These low-rated messages were subsequently either revised (n=8) or eliminated (n=17). We also received universal agreement from the participants that 2-3 messages per day would be appropriate during the postpartum period.

All messages were evaluated by the advisory panel in relation to the relevance of the messages to the population, as well as their coverage of factors involved in the C-SHIP framework. A final review of the advisory panel input by the research team resulted in slight modifications to 50% of the messages and 21 additional messages being eliminated.Table 2 provides examples of the message revision process and Table 3 provides samples of final text messages corresponding to the 5 C-SHIP constructs. For most messages, the basic content of the original message was retained, but revisions improved the readability and specificity of the message.

Ten participants were recruited from the same 3 WIC clinic sites as in the interview phases to participate in a 1-week usability testing session. Each participant was provided with a mobile phone and instructions for mobile phone and message use. Each participant received 7 system-initiated messages over a 1-week period and was instructed to initiate 3 crave and 3 lapse message requests. Among the 10 enrolled participants, 8 successfully completed the testing and responded to the evaluation interview. All 8 participants identified the messages as helpful and reported being satisfied with the messages, as well as with the crave and lapse functions. They all stated that they would use the program and recommend it to others. They also suggested changes, including making the system-initiated messages more frequent (3 per day) and adding voice and video messages to the program. Based on participants’ feedback, the research team finalized the “Txt2commit” program, which provided: (1) 1 text message per week during the prenatal period (27 weeks gestation and on) to enhance and sustain program engagement, (2) 3 text messages daily during the first postpartum month, and (3) crave and lapse functions so that participants could request additional messages when feeling the need. The final program was tested in a larger feasibility pilot study, results of which will be reported elsewhere for ease of communication. Figure 1 provides an illustration of the order and findings throughout different phases of the development, including the changes of the numbers of messages over time.