Intervention targeting and retention, engagement and abstinence outcomes among Latino and White users of Smokefree. gov text messaging programmes: a cross-sectional study

Background We examined retention, engagement and abstinence among Latino users of SmokefreeTXT en Español (SFTXTesp), a Latino-targeted text messaging smoking cessation intervention, and Latino and White users of SmokefreeTXT (SFTXT), a non-targeted intervention. Methods Data came from 12 281 users (1562 Latino SFTXTesp users and 2301 Latino and 8418 White SFTXT users). We conducted time- to- drop-out analyses by race/ethnicity. Using logistic regression, we examined associations between intervention targeting, race/ ethnicity and responses to smoking status prompts, an engagement metric, and self-reported abstinence on quit day, intervention end and 1-month follow-up. Age, gender, census region, smoking frequency, cigarettes smoked per day, prequit time and number of quit attempts were covariates. Results Latinos in SFTXTesp (adjusted OR, aOR 0.63) and SFTXT (0.88) were less likely to drop out of the intervention than Whites. SFTXTesp Latino users had higher response rates to smoking prompts than SFTXT Latinos users (aORs 1.35, quit day; 1.84, intervention end; 1.82, 1-month follow-up). However, SFTXTesp and SFTXT Latino users had lower response rates than Whites (aORs 0.68, 0.45, quit day; 0.60, 0.30, intervention end; 0.64, 0.33, 1-month follow-up). Abstinence was equivalent among Latinos in SFTXTesp and SFTXT interventions, but Latinos using SFTXTesp and SFTXT were less likely to be abstinent than Whites (aORs 0.42, 0.41, quit day; 0.45, 0.37, intervention end and 0.53, 0.35, 1-month follow-up). Conclusion Linguistic intervention targeting improved retention and engagement among Latinos, but not abstinence. Latinos had higher retention but lower engagement and abstinence rates than Whites. Cultural targeting may engage Latinos in smoking cessation interventions and improve abstinence.


INTRODUCTION
Latino cigarette smokers are a unique population for tobacco treatment efforts.Constituting 8.8% of US adult smokers, 1 they are a heterogeneous group that shows considerable within-group variations in smoking prevalence, frequency, intensity and quit rates attributable to several factors (eg, nationality). 2Tobacco-related illnesses are leading causes of death among Latinos. 3Latinos have lower prevalence of being asked about tobacco use, receiving quit advice from healthcare providers and using smoking cessation pharmaceutical aids or counselling resources. 4 5AT IS ALREADY KNOWN

BMJ Public Health
][8][9][10][11][12] Text messaging is a viable intervention platform for Latinos among whom cellphone ownership is 100% (85% for smartphones, 14% for cellphones only). 13These interventions offer several benefits, including reduced cost, time and stigma associated with their use. 9Few text messaging smoking cessation interventions target Latinos. 14argeted interventions appeal to predefined subgroups with shared characteristics, usually static or slow-changing characteristics, such as race/ethnicity. 15Targeting is achieved through various strategies-from linguistic to cultural adaptations. 16Targeted interventions are more efficacious in reducing smoking than non-targeted interventions. 16Gaps remain in our understanding of the efficacy and effectiveness of Latino-targeting smoking cessation text messaging interventions.For example, Quitxt SMS, 17 an intervention targeting Latinos aged 18-29 in Southern Texas, Latino Kick Buts, 18 an adaptation of Txt2Stop intervention, 19 and Vive sin Tabaco… ¡Decídete!, 20 an intervention targeting Mexican smokers, have not been evaluated on a wide scale.Furthermore, evaluations of targeted interventions have been limited to examining whether targeting improves outcomes within their target population. 16Although this approach can show superiority of targeted (vs non-targeted) interventions among the target population, it does not show whether targeting reduces existent cessation disparities between minority and majority groups.
SmokefreeTXT en Español (SFTXTesp) is a free and publicly available smoking cessation intervention for Latinos through the Smokefree.gov Initiative of the National Cancer Institute (NCI).SFTXTesp is a linguistic adaptation of NCI's general text messaging smoking cessation intervention, SmokefreeTXT (SFTXT), which is available in English for US smokers. 21Latinos can enrol in SFTXTesp or SFTXT, whereas other races/ethnicities can enrol in SFTXT.Both interventions are nearly identical in intervention duration, message content and structure, which offers a unique opportunity to examine the effects of intervention targeting on intervention outcomes among Latinos and Whites.
In this observational study, we examine associations between (a) targeting and intervention outcomes among Latinos in SFTXTesp and SFTXT and (b) race/ ethnicity and intervention outcomes among Latinos in either SFTXTesp or SFTXT and Whites in SFTXT.The primary outcomes were retention by intervention end, engagement and point prevalence abstinence on quit day (ie, day 0), intervention end (ie, day 42), and 1-month postintervention end (ie, day 72).Retention is when a participant completes an intervention, 22 whereas engagement is an investment of physical, affective and cognitive energy in a stimulus/task. 23Retention and engagement are important determinants for intervention efficacy and achieving abstinence. 22 24-27

Interventions description
SFTXTesp and SFTXT, available only to smokers in the USA, are 6-8 weeks in duration and follow an identical structure.At signup, users select a quit day, which is recommended to be within 2 weeks of signup.During this optional preparation phase, users receive messages to prepare them to quit smoking. 28Starting on quit day, users receive messages that include cessation tips and motivational content over 42 days. 28Users receive 3-5 daily messages on average, with more messages delivered in the treatment phase (ie, weeks 1-2) than in the maintenance phase (ie, weeks 3-6).Users are unable to alter the number of messages they receive.At any time after signup, users can reset their quit day, opt-out of the intervention and text keywords to receive additional on-demand content.Users receive smoking status prompts, weekly throughout the intervention starting on quit day (ie, day 0) until intervention end (ie, day 42), and at 1, 3 and 6-month postintervention completion (ie, days 72, 132, 222).

Sample
Data came from eligible records of Latino SFTXTesp users and of Latino and White web enrollees in SFTXT who had complete race/ethnicity information, spanning from 1 January 2016 to 31 December 2018.Records represented real-world SFTXTesp and SFTXT users who self-selected to enrol in either intervention.We deleted duplicate, inconsistent and incomplete records (online supplemental table 1).Specifically, we used the most recent quit attempt for users who reset their quit days.We excluded users whose quit day preceded their signup date, who opted out of the intervention before the prequit period, and who set a quit day less than 42 days prior to end-of-study date (ie, 31 December 2018 when data were pulled).We applied these exclusion criteria to each dataset, resulting in an analytical sample of 12 281 unique records.

Measures
Users reported their age, gender, smoking frequency (daily vs non-daily), cigarettes smoked per day and zip code.SFTXT users had a 'non-binary' or 'prefer not to say' gender option plus male and female.Daily smokers reported smoking 'every day', whereas non-daily smokers reported smoking 'most days', 'some days', or 'less than that'.Cigarettes smoked per day were classified into light (≤10 cigarettes), moderate (11-20 cigarettes) and heavy (>20 cigarettes).Residential zip codes were converted into four US census regions plus Puerto Rico and the Virgin Islands.SFTXT users who enrolled online reported their race/ethnicity.SFTXT users who self-identified as Latinos were considered as such regardless of race.All SFTXTesp users were considered Latinos.
Users' signup date, quit day and, if applicable, opt-out date and quit day reset(s) were automatically captured.We used these dates to calculate prequit time and time-to-drop-out.For users whose signup to quit day was ≤14 days, prequit time equaled their time from signup to quit day.For users whose quit day was >14 days, we reset their prequit time to 14 days because users receive messages 14 days prior to quit day.Prequit time served as a covariate to adjust for potential effects of the preparation phase and intervention dose. 9 29Time-to-drop-out ranged from 0 to 42 and was used in the survival analysis among intervention initiators (ie, users who made it to quit day).
We created three dichotomous variables: intervention completion status, smoking status and response status.For completion status, users who opted out by texting 'STOP' or for whom text messages were undeliverable between day −14 (ie, 14 days before quit day) and intervention end (ie, day 42) were considered 'noncompleters', whereas users who remained in the intervention until day 42, regardless of intervention engagement, were 'completers'.For smoking status, users were considered 'smokers' if they responded 'yes', and 'nonsmokers' if they responded 'no' to each smoking status question (ie, 'Day 1 is almost over.Have you smoked today?' on quit day, 'Hi there!Have you smoked in the past 7 days?' thereafter, and 'Hi, just checking in.Have you smoked in the last 30 days?' on 1-month follow-up).They were considered 'responders' if they answered the smoking status question with either a 'yes' or a 'no' and 'non-responders' if they did not respond.Response to smoking status questions captured user engagement and abstinence.
To examine retention, we conducted a Cox regression of time from quit day to drop-out.For engagement and abstinence outcomes, we ran logistic regression models among Latino users in SFTXTesp or SFTXT (SFTXT Latino users were reference group) and among Latinos and Whites in SFTXTesp or SFTXT (SFTXT White users were reference group) at all assessment time points (ie, days 0, 7, 14, 21, 28, 35, 42 and 72).In each engagement model, n was the number of users who remained in the intervention on the day smoking status was assessed.In the abstinence models, users who failed to respond to smoking status prompts and/or had dropped out were considered smokers following an intention-to-treat approach.All analyses were limited to intervention initiators and were adjusted for age, gender, census region, smoking frequency, cigarettes smoked per day, prequit time and number of quit attempts.Day 72 engagement and abstinence models were further limited to those who could have made it to day 72 by end-of-study date and, thus, could have received the smoking status prompt.As sensitivity analyses, we ran all analyses among users who completed the intervention (ie, intervention completers) and among users without missing data (ie, complete case) where intervention outcomes are stronger. 8tient and public involvement Participants were not involved in the design, conduct, reporting or dissemination of this study.

RESULTS
Sample characteristics for full dataset and complete case appear in online supplemental tables 2 and 3.

Retention
Drop-out before quit day ranged from 28.69% to 52.72% among SFTXTesp and SFTXT Latino users (online supplemental table 2).Intervention non-initiators (ie, those who dropped out before quit day) were more likely to be males and SFTXT Latino users, smoke heavily and have 2+ quit attempts (online supplemental table 4).Time from quit day to drop-out averaged 10.40, 6.64 and 8.73 days among SFTXTesp Latino users, SFTXT Latino users and SFTXT White users (online supplemental table 2).Latinos in both SFTXTesp (adjusted OR, aOR 0.63) and SFTXT (aOR 0.88) were less likely to drop out compared with Whites in SFTXT (table 1, figure 1A).At intervention end, 68.9% and 59.6% of Latino SFTXTesp and SFTXT users were retained compared with 55.8% SFTXT White users.Similar patterns emerged under complete-case analysis (table 1, figure 1B).

Engagement
Overall response rate was 21.94% on quit day, 10.94% on day 42 and 14.88% on day 72 (table 2).SFTXTesp Latino users were more likely to respond to smoking status prompts than SFTXT Latino users on quit day, day 42 and day 72 (aORs 1.35, 1.84, 1.82) (table 3).Compared with SFTXT White users, SFTXTesp and SFTXT Latino users were less likely to respond to smoking status prompts across all time points (eg, aORs 0.68, 0.45 on quit day; 0.60, 0.30 on day 42 and 0.64, 0.33 on day 72) (table 4).Intervention completer (online supplemental tables 5-7) and complete case (online supplemental tables 8-10) analyses largely mirrored these results.

Abstinence
Overall point prevalence abstinence was 14.46% on quit day, 5.09% on day 42 and 5.73% on day 72 (table 2).SFTXTesp and SFTXT Latino users were equally likely to be abstinent across all assessment time points (table 3).Compared with SFTXT White users, SFTXTesp and SFTXT Latino users were less likely to be abstinent across all assessment time points (eg, aORs 0.42, 0.41 on  4).Intervention completer (online supplemental tables 5 and 7) and complete case (online supplemental tables 8 and 10) analyses mirrored these patterns where SFTXTesp and SFTXT Latino users were less likely to be abstinent than SFTXT White users across all assessment time points.Among intervention completers, SFTXTesp Latino users were less likely to abstinent than SFTXT Latino users on quit day and days 7, 14, 28 and 35 (online supplemental table 6).Similar patterns emerged under complete-case analysis for quit day and days 14, 28 and 35 (online supplemental table 9).

DISCUSSION
This is the first study, to our knowledge, examining if intervention targeting improves retention, engagement and abstinence outcomes among its intended audience and versus a majority racial/ethnic group.In a real-world assessment of Smokefree.gov text messaging smoking cessation interventions, targeting improved retention and engagement, but not abstinence, among Latinos.Whites outperformed Latino users of both targeted and nontargeted interventions in engagement and abstinence rates, indicating that targeting did not reduce disparities between Latinos and Whites.Results suggest that rather than relying solely on linguistic targeting, Latinotargeting interventions should incorporate cultural adaptations of their content to improve intervention outcomes and reduce smoking cessation disparities. 30igh cellphone penetration among Latinos presents an opportunity to deliver text messaging smoking cessation interventions to offset their limited access to traditional smoking cessation aids and resources. 4 5Text messaging interventions are especially necessary for hard-to-reach populations that do not own smartphones (eg, foreignborn Latinos) 31 or have limited digital literacy skills.

BMJ Public Health
Importantly, text messaging interventions support behavioural change techniques (eg, action cues) and allow sufficient tailoring and interactivity to engage users. 33owever, text messaging remains an underused intervention platform despite its simplicity, widespread use, user familiarity, 98% open rate and low-cost scalability. 33 34argeting improved retention and engagement.At 69% and 60%, retention was higher among SFTXTesp and SFTXT Latino users than among SFTXT White users (56%).Results reflect similar patterns where vulnerable groups (eg, Blacks, low socioeconomic smokers) remain in Smokefree.gov text messaging interventions. 35 36ur retention rates were lower than those reported for general behavioural health (eg, 86% in a meta-analysis of 19 randomised trials) 10 and targeted smoking cessation (eg, 85% at 12-week follow-up in small-scale study) 37 interventions.For engagement, response rates were higher among SFTXTesp Latino users than SFTXT Latino users, but Latinos were less engaged than Whites throughout the intervention.Measurement heterogeneity impedes comparisons across studies. 24For example, previous studies used programme completion, 21 user-initiated text messages 20 26 and survey responses 26 as engagement metrics.Since intervention efficacy depends on retention and engagement, 22 24-27 research is needed to identify user characteristics and intervention features associated with improved retention and engagement 38 39 and to unpack the associations between retention, engagement and abstinence, especially given that retention and engagement did not translate to abstinence among Latinos (vs Whites) in our study.
Several results are worth highlighting regarding abstinence.First, by intervention end, overall abstinence among SFTXTesp and SFTXT intervention initiators was slightly lower than previously reported among SFTXT users (5.09% vs 7.2%). 21Among intervention completers, overall abstinence was 8.77% consistent with evidence of improved abstinence among SFTXT users who get the full treatment dose (12.9%). 21Consistent with literature, 21 abstinence was attenuated at 1-month postintervention completion.Second, at 6.19% and 11.21%, abstinence rates among SFTXT hite intervention initiators and completers tracked closely those reported in the literature.Abstinence among SFTXTesp and SFTXT Latino users were roughly halved to 3.03% and 2.12% among intervention initiators and to 4.41% and 3.47% among intervention completers.Abstinence rates among SFTXTesp Latino users were lower than previously reported in pilot testing of targeted text messaging interventions for Latinos such as Quitxt (25.2%), 17 Latino Kick Buts (30%) 37 and Vive sin Tabaco… ¡Decídete!(40%). 20Third, abstinence rates were equivalent among SFTXTesp and SFTXT Latino users.Among intervention completers and those without missing data, SFTXTesp users underperformed their SFTXT counterparts on abstinence at various assessment times.Despite evidence that targeting improves intervention outcomes, 10 SFTX-Tesp showed that relying on linguistic targeting only did not improve abstinence among its users compared with SFTXT White users.
Observed abstinence rates between SFTXTesp and SFTXT Latino users and between Latino and White users may suggest that linguistic targeting is insufficient to improve SFTXTesp efficacy.With cultural intervention adaptations being well received by intended populations, 18 SFTXTesp content could emphasise Latino values (eg, familismo, respeto) and address Latino-specific factors related to smoking and cessation (eg, acculturative stress). 40The heterogeneity of Latino smokers presents an additional explanation for the observed abstinence outcomes.SFTXTesp Latino users were more likely to be older, male, from Puerto Rico and the Virgin Islands, and were less likely to smoke heavily (online supplemental table 11).This is particularly important given differences in smoking patterns among Latinos by sex, country of origin and acculturation, 2 suggesting a need for tailoring to improve intervention efficacy above and beyond targeting. 10Lower abstinence rates among SFTXTesp (vs SFTXT) intervention completers and under a completecase analysis could be attributed to the high percentage of abstinent Latinos SFTXTesp users who were excluded from the intervention completer and the complete-case datasets.Of those excluded from the completer dataset, 19.01%SFTXTesp Latino users (n=116 abstinent/ 610 excluded), 7.11% SFTXT Latino users (n=94/1322) and 23.54% SFTXT white users (n=1081/4591) were abstinent.Of those excluded from the complete-case dataset, 40.06% SFTXTesp Latino users (n=242 abstinent/ 604 excluded), 14% SFTXT Latino users (n=7/50) and 27.27% SFTXT White users (n=3/11) were abstinent.
Smokefree.gov interventions should allow each user to further tailor the intervention based on a unique set of outcome or behaviour related factors, which could further improve efficacy of targeted interventions. 10 15 (CF 29 Elements for content tailoring could include smoking frequency, intensity and prior quit attempts where Smokefree.gov interventions would allow for tailored message frequency, timing and content based on smokers' input at baseline.Future research should explore whether a preparation phase is beneficial given that we found longer prequit times were associated with worse intervention outcomes, 9 and solicit qualitative feedback from SFTXTesp and SFTXT Latino users who (un)successfully used the intervention on how to improve retention, engagement and abstinence. 30Finally, future research should re-evaluate intervention outcomes for SFTXTesp and SFTXT since both interventions are updated periodically (eg, in 2019 after data were pulled for this study).
Strengths of the study include reporting on a realworld assessment of Latino-targeting SFTXTesp text messaging smoking cessation intervention.The unique parallels between SFTXTesp and SFTXT afforded a quasi-experimental design to examine whether targeting improves intervention outcomes among its intended Latino population and when compared with Whites.Although Latinos were not randomly assigned to either

BMJ Public Health
SFTXTesp or SFTXT, comparisons of targeted versus non-targeted interventions have been typically limited to the target population (eg, Latinos) and have been researcher-controlled studies with incentivised participation.Limitations include unmeasured factors (eg, nicotine dependence) that could have impacted intervention outcomes.Drop-outs were users who actively texted 'STOP' or for whom text messages were undeliverable but did not include passive drop-out where users disengage from the intervention, potentially inflating retention.Drop-outs were considered smokers in intention-totreat analysis.However, users could have dropped out because they had already quit smoking. 37Engagement and abstinence rates for each assessment time point were independent of prior or subsequent assessment points.Accordingly, we could not capture disengagement/ reengagement, continued abstinence or relapses.Using response to smoking status prompts as an engagement metric could underestimate engagement among users who were not abstinent. 26Other factors could have been used as engagement metrics (eg, user-initiated text).Abstinence was self-reported and short-term.Users with >1 quit attempt had longer prequit time that reflected time from signup to quit day (rather than quit day reset date).SFTXTesp had a high percentage of missingness.However, we imputed missing data and ran sensitivity analyses using intervention completer and complete-case datasets.

Conclusion
SFTXTesp, a Latino-targeting text messaging smoking cessation intervention, resulted in higher retention and engagement rates among its users than among Latino users of SFTXT, a non-targeted intervention.Linguistic targeting did not improve abstinence among Latinos or reduce disparities between Latinos and Whites in engagement and abstinence rates.Cultural adaptation and tailoring are necessary to maximise the efficacy of targeted smoking cessation interventions.

on
November 22, 2023 by guest.Protected by copyright.http://bmjpublichealth.bmj.com/bmjph: first published as 10.1136/bmjph-2023-000222 on 8 November 2023.Downloaded from ON THIS TOPIC WHAT THIS STUDY ADDS⇒ The study is a real-world assessment of Smokefree.govtext messaging smoking cessation interventions.SmokefreeTXT en Español, a Latino-targeted text messaging intervention, improved retention and engagement among Latino users compared with Latino users of SmokefreeTXT, a non-targeted intervention.Intervention targeting did not affect abstinence.Despite their lower retention rates, Whites had better engagement and abstinence outcomes than Latinos who used either the targeted or nontargeted intervention.HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY ⇒ Beyond linguistic targeting, cultural adaptations of text messaging smoking cessation interventions are needed to improve retention, engagement and abstinence outcomes among Latino smokers and reduce smoking and cessation disparities with Whites.on November 22, 2023 by guest.Protected by copyright.

Table 3
Correlates of engagement and abstinence outcomes among SmokefreeTXT en Español (SFTXTesp) and SmokefreeTXT (SFTXT) Latino intervention initiators (full dataset)Logistic regression modelled the probability of 1=responder in response status models and 1=nonsmoker in smoking status models.Bolded cells represent significant results.*aOR was undefined.†All users who selected 'non-binary' or 'prefer not to say' had less than 72 days between quit day and end of study date and were excluded from day 72 analyses.

Table 3
Continued on November 22, 2023 by guest.Protected by copyright.

Table 4
Correlates of engagement and abstinence outcomes among SmokefreeTXT en Español (SFTXTesp) and SmokefreeTXT (SFTXT) Latino and White intervention initiators (full dataset) El-Toukhy S, Kamke K. BMJ Public Health 2023;1:e000222.doi:10.1136/bmjph-2023-000222Logisticregressionmodelled the probability of 1=responder in response status models and 1=non-smoker in smoking status models.Bolded cells represent significant results.*aOR was undefined.†All users who selected 'non-binary' or 'prefer not to say' had less than 72 days between quit day and end of study date and were excluded from day 72 analyses.aOR, adjusted OR.

Table 4
Continued on November 22, 2023 by guest.Protected by copyright.

Table 2 :
Supplementary Table 1: Exclusion criteria for SmokefreeTXT en Español (SFTXTesp) and SmokefreeTXT (SFTXT) Latino and White webIneligible for inclusion in the study were 33,385 SFTXT records of users (a) who signed up for the intervention via text message and were not asked the race/ethnicity questions (n= 20257), (b) who had missing race and/or ethnicity data (n= 9256), and (c) who belonged to racial/ethnic groups other than Whites and Latinos (n= 3872).Sample characteristics by intervention targeting, race, and ethnicity (full dataset,N= 12,281)Ns represent 1/20 th of a person rounded to the nearest whole number.All summary statistics were pooled across 20 imputations.SS: summary statistic, M: mean, SD: standard deviation, IQR: interquartile range.All time variables (i.e., time from signup date to quit day, time from signup to dropout, time from quit day to dropout, and prequit time) are reported in days.

Table 3 .
Sample characteristics by intervention targeting, race, and ethnicity (complete case dataset, N= 11,616) All time variables (i.e., time from signup date to quit day, time from signup to dropout, time from quit day to dropout, and prequit time) are reported in days.BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s) SS: summary statistic, M: mean, SD: standard deviation, IQR: interquartile range.

Table 4 :
Correlates of dropping out before quit day (i.e., being an intervention non-initiator) Full dataset (N= 12,281) Complete case dataset(N= 11,616)BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s) aOR: adjusted odds ratio, LCL: lower confidence level, UCL: upper confidence level.Bolded cells represent significant results.

Table 5 :
Response rates and point-prevalence abstinence among SmokefreeTXT en Español (SFTXTesp) and SmokefreeTXT (SFTXT) intervention initiators who completed the intervention (completers dataset, N= 5,758) BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s) BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s) BMJPH doi: 10.1136/bmjph-2023-000222 :e000222.1 2023; BMJPH El-Toukhy S, Kamke K.