Smoking continues to be the leading preventable cause of death. Digital interventions for smoking cessation (DISCs) are health communication programs accessible via the internet and smartphones and allow for greater reach and effectiveness of tobacco cessation programs. DISCs, including ours, called Decide2Quit (D2Q; www.decide2quit.org), have led to high 6-month cessation rates and also reach vulnerable populations. Despite this, the impact of DISCs has been limited, and new ways to increase access and effectiveness are needed. A widely used and proven behavioral intervention tool is computer-tailored health communication (CTHC). We will test 2 different ways of implementing CTHC: standard CTHC, which selects quit email messages based on participants' readiness to quit smoking measured by a single self-report question; and the recommender CTHC, which uses a machine learning algorithm that incorporates user ratings to select quit email messages. We also used a peer-recruitment tool on the D2Q website. Our pilot data showed that the recommender CTHC can potentially be more engaging and effective than is standard CTHC. Peer recruitment engages users to recruit their friends and family to the intervention. Our pilot study showed that providing smokers with access to peer recruitment increased their motivation to quit smoking, quadrupled the sample size, and increased the proportion of African American smokers in the sample.
To test 3 hypotheses: : access to peer-recruitment tools embedded in the recommender CTHC increases the proportion of African American individuals who are smokers recruited to our study. : use of the D2Q website improves with exposure to (a) recommender CTHC, (b) peer-recruitment tools embedded in CTHC, or (c) both recommender CTHC and peer-recruitment tools. : receiving tailored email motivational messages from the recommender CTHC improves smoking cessation outcomes compared with receiving standard email messages from the CTHC.
The interventions were standard CTHC, the recommender CTHC, and a peer-recruitment tool on the D2Q website. Smokers were recruited online and randomized to (1) the recommender system CTHC or (2) a standard CTHC system, followed in each arm by partially random allocation to either; (3) access to a tool on the D2Q website that enabled them to peer recruit their friends and family group to the study and email reminders to recruit their friends and family members; or (4) no access to the peer-recruitment tool. All were followed for 6 months. Outcomes were (1) the proportion of recruited study participants who were African American; (2) whether a participant was recruited to the study via peer recruitment or other means, which was tracked via self-report at baseline data collection; (3) repeated use of the D2Q website; and (4) 6-month point prevalence of smoking and number of cigarettes smoked. D2Q use was tracked via online scripts, and 6-month smoking cessation was assessed via self-report.
Out of the 1487 smokers recruited to the study, 273 smokers (18%) were peer recruited and allocated to receive the peer-recruitment tools. Our loss to follow-up rate was 47%. Having access to the peer-recruitment tools did not increase the proportion of African American smokers: Access to peer recruitment: n = 96 (13%), no access n = 93 (13%); = .84, difference = −0.004; 95% CI, −0.04 to 0.03). Access to neither the peer-recruitment tools nor the recommender CTHC was independently associated with repeated use of the recruitment tool ( = .16 and .07, respectively). Using a worst-case-scenario analysis, where all participants missing data were considered to be smokers, we found no difference between the recommender and standard CTHC for the effectiveness outcomes, as follows: for 6-month 7-day point prevalence, recommender CTHC, 20%; standard CTHC, 21% ( = .56; difference = 0.012; 95% CI, −0.03 to 0.05); for reduction in number of cigarettes smoked, recommender CTHC mean (SD), −8.9 (11); standard CTHC mean (SD), −9.9 (11) ( = .20; difference = −1.07; 95% CI, −2.54 to 0.41). We found that those with access to peer-recruitment tools significantly improved their own smoking cessation outcomes, as follows: for 6-month 7-day point prevalence, access to peer recruitment, 45%; no access, 31% ( < .0001; difference = 0.14; 95% CI, 0.07-0.21); for reduction in number of cigarettes smoked per day, access to peer recruitment mean (SD), −10 (10); no access mean (SD), −8.6 (11) ( = .040; difference = −1.59; 95% CI, −3.06 to −0.12).
The message selection (recommender or standard CTHC) did not significantly improve D2Q engagement or smoking cessation, but having access to the peer-recruitment tools significantly improved smoking cessation. Future studies should explore the mechanisms by which access to the peer-recruitment tools improved outcomes.
Our smoking outcomes were assessed via self-report. We used a 2-stage allocation process that included a primary randomization (recommender or standard CTHC), followed by a partially random allocation to access peer-recruitment tools. We used this approach to preserve the primary randomization since we allocated all peer recruited smokers to the group having access to the peer-recruitment tools. We chose this approach to enhance blinding because these peer-recruited smokers may communicate with those who recruited them about D2Q. However, this resulted in the inclusion of nonrandomized users for the peer-recruitment comparison. This also raised concerns about unmeasured confounders that affected both allocation to the website tool and response to smoking-cessation messages.
