In developed countries, smoking during pregnancy is a leading preventable cause of adverse prenatal outcomes, including miscarriage [1], stillbirth [2,3], and prematurity [4]. It is also associated with a wide range of infant health problems [5]. In the United Kingdom, around 11% of women are estimated to smoke throughout pregnancy [6], but rates rise considerably with increasing social deprivation [6,7], standing at around 5 times higher in the most deprived women than in the least [7]. Children born to smokers are also more likely to become smokers themselves [8]. Thus, smoking in pregnancy not only puts great financial burden on health services but also perpetuates and exacerbates health inequalities. Reducing its prevalence is a public health priority [9].

Most pregnant smokers want to quit [10], and effective interventions exist to help them [11,12]. Specialist Stop Smoking Services in England offer free pregnancy cessation support with proven efficacy [12]; however, uptake is low [13], with convenience and concerns about being judged reported as barriers to access [14]. In addition to addressing these barriers, research efforts have also been focused on developing effective and cost-effective “distance” alternatives that will appeal to pregnant smokers and be used sufficiently to yield a public health benefit. Self-help cessation support appeals to pregnant smokers [15], and delivering self-help by mobile phone text messaging may be helpful for this group, given its low cost, convenience, anonymity, and wide reach potential, with mobile phone ownership high across the socioeconomic spectrum [16]. Systematic review evidence shows that self-help cessation interventions for pregnant smokers can be effective [17] and that mobile phone–based cessation interventions are effective for nonpregnant smokers [18].

We have developed a low-cost, tailored, text-messaging intervention specifically for pregnant smokers (“MiQuit”) [19]. MiQuit is feasible to deliver and highly acceptable to pregnant smokers [19], and a recent randomized controlled trial (RCT; n=407) found that offering MiQuit in addition to usual care shows promising efficacy and cost-effectiveness [20]. As MiQuit is fully automated and user-initiated, women can start using it without the need for any health professional involvement, thus minimizing potential implementation costs. However, little is currently known about the likely real-world uptake of MiQuit should this become routinely available to pregnant smokers, or what the best implementation strategies might be to maximize its reach and initiate users into support as cost-effectively as possible. The public health impact of an intervention depends crucially on its real-world uptake, as well as its efficacy, but evaluations of smoking cessation interventions have largely neglected to estimate this [21].

The internet has obvious potential as a tool for reaching pregnant smokers and enrolling them into cessation programs, and evidence suggests that a digital intervention may particularly appeal to smokers offered cessation support through digital media [22]. Interventions can be promoted to potential users on the internet through commercial search-engine and banner-based (pop-up) adverts, as well as noncommercial websites. A previous real-world study in an antenatal setting estimated that 3% to 4% of pregnant smokers initiated MiQuit after a brief promotional leaflet was placed into their maternity booking pack without any introduction or endorsement from a health professional [23]. If pregnant women will initiate MiQuit after reading a brief leaflet, it seems likely that they may do so after reading an online advert. Offering MiQuit to pregnant smokers through search engines, in particular, might reach them earlier in pregnancy than they would typically be targeted in antenatal settings, thus maximizing the benefits of quitting to the fetus. Search-based adverts could also present an opportunity to offer support to women when they are motivated to quit, with cohort evidence suggesting that repeated quit attempts may be made throughout pregnancy [24]. As a tool for recruiting smokers into research trials of digital interventions, studies of nonpregnant groups [25-30] suggest that commercial online advertising can achieve high participant yield rapidly [26,27] and can recruit traditionally hard-to-reach smoker populations at relatively low cost [29,30]. However, although studies typically show similar participant characteristics and retention rates for smokers recruited through online versus traditional means [26-28], with the notable exception of younger age in those recruited via social media [27], others have found lower quitting confidence and lower study completion rates among smokers recruited to trials via the internet [31].

We are aware of only 2 previous studies to explore real-world uptake of digital smoking cessation support, rather than recruitment rates to cessation trials, as a consequence of online advertising, both of which targeted nonpregnant smokers [32,33]. To our knowledge, no published studies have explored uptake of cessation support among pregnant smokers via the internet. In addition, we know little about the characteristics of pregnant smokers who can be encouraged to take up digital interventions over the internet. In this study, therefore, we investigate whether pregnant smokers will initiate the MiQuit intervention after seeing paid-for or free online advertising; in addition, we monitor the costs incurred and, using a previously obtained estimate for MiQuit efficacy [20], assess the extent to which commercial online advertising might be cost-effective. Finally, we document the extent to which users engage with the support program, including the discontinuation rate, and describe key characteristics, exploring differences between pregnant smokers initiating MiQuit via different online routes.

This was an evaluation of the uptake of a digital (text-messaging) intervention advertised on the internet. Uptake rates of MiQuit were monitored while it was advertised via 4 concurrent online settings: 2 commercial websites and 2 noncommercial webpage links.

MiQuit provides a 12-week program of automated, theory-guided, interactive support for quitting smoking in pregnancy, delivered by text message. Support is tailored to 12 baseline user characteristics plus name, gestation, and smoking status, the latter collected at 3 and 7 weeks by text message. Tailoring characteristics include nicotine dependence, partner’s smoking status, and confidence, motivation, and determination to quit [19,20,23]. Women initiate MiQuit by texting a short code number. They are then invited to complete 12 baseline tailoring questions, including the option of setting a quit date, either by text or by website. Those tailoring by website must answer all other 11 questions. Those tailoring by text are given the option of answering either 6 or 12 tailoring questions but can stop responding at any point. If no tailoring questions are answered using either route, then generic support is delivered. MiQuit delivers 0-2 scheduled daily texts (“push” support), including motivational messages; advice about quit attempt preparation, managing cravings, or trigger situations; and information about fetal development and how smoking affects it. Those setting a quit date receive extra support oriented around their nominated date. Users can access on-demand, “pull” support for combatting cravings (“HELP”), returning to abstinence after a lapse (“SLIP”), or for distraction (“QUIZ”). The support lasts for 12 weeks unless discontinued prematurely by the user sending a “STOP” message. MiQuit texts are free to receive. Sending the initiation text and any subsequent texts sent by the user are either free or cost the user’s standard text message rate, depending on their phone “bundle.”

The 4 adverts were run concurrently, beginning late May 2015, either until their budget ended (commercial adverts) or for 6 months (free links). Initiations were permitted up to 3 months after each advert ceased to be shown. The 2 commercial adverts were monitored closely throughout the campaigns. The performance metrics these supply (shown below) were compiled on a weekly basis. Numbers of MiQuit submissions, initiations, and discontinuations were compiled weekly for all 4 sources.

Advice was sought from the National Research Ethics Central Queries Service as to whether the study should be classed as research requiring ethical review, and they confirmed that no ethical review was required. The study was conducted in accordance with the ethical principles that have their origin in the Declaration of Helsinki, 1996; the Principles of Good Clinical Practice; and the Department of Health Research Governance Framework for Health and Social Care, 2005. Participants were able to withdraw from the MiQuit support at any time.

Most commercial advert clicks came from mobile devices rather than desktops (Google: 560/812, 69.0%; Facebook: 1883/1889, 99.68%). Of those who accessed the initiation short code, 94.1% (301/320) did so from a mobile device (Google: 110/121, 90.9%; Facebook: 184/187, 98.4%; NHS Choices: 1/4, 25%; NCT: 6/8, 75%).

Figure 1 shows the flow of the targeted populations, through each online route, into initiation of MiQuit support. The Google advert was shown 29,022 times in its 1-month duration. Given our impression share of 70% for mobile-based searches and 50% for desktop-based searches, we estimated that over 46,000 Google searches had been made for our keywords, in the United Kingdom, during this time. The mean position of the Google advert from the top of the screen was 1.0 for mobiles and 1.2 for desktops (highly visible). The Facebook advert was shown to 248,618 broadly targeted women during its 3-month duration (mean 2.4 times each) and also had high visibility on screen (Multimedia Appendix 1). In 6 months, approximately 40,000 unique visits were made to the NHS Choices and NCT smoking-in-pregnancy webpages containing the MiQuit link, but the proportions who scrolled down to where the links were placed on these is unknown.

In total, 2754 individuals clicked on 1 of the 4 adverts to the MiQuit sign-up website (Google n=812, Facebook n=1889, NHS Choices n=33, NCT n=20). For Google, assuming 1 advert impression per person, this amounted to 2.80% of those served the advert (812/29,022) and, for Facebook, 0.76% of those served the advert (1889/248,618). For the NHS Choices and NCT websites, the numbers of clicks on the MiQuit links amounted to 0.09% (33/38,352) and 1.43% (20/1402), respectively, of the numbers of unique visits to the webpages containing the link.

MiQuit was initiated by 93 individuals in total, with the 2 commercial campaigns each yielding 42 initiations and the free links 9 in total. For our uptake rate calculation, the percentage who subsequently initiated MiQuit after clicking on an advert to the MiQuit website was 3.38% (93/2754, 95% CI 2.76%-4.12%); Google 5.2% (42/812, 95% CI 3.9%-6.9%), Facebook 2.22% (42/1889, 95% CI 1.65%-2.99%); NHS Choices 9% (3/33; 95% CI 3%-24%), NCT 30% (6/20; 95% CI 15%-52%). One in 691 Google advert impressions resulted in a MiQuit initiation (1 in 5919 women targeted by Facebook). One in 12,784 visits to the NHS Choices webpage, and 1 in 234 visits to the NCT webpage, resulted in an initiation.

The Facebook campaign cost £1000, whereas the Google campaign cost £1077, including £75 credited free to our account. Table 1 shows a breakdown of these costs in terms of how far MiQuit was accessed or activated. The mean cost per advert click to the MiQuit website was £1.33 for the Google advert and £0.53 for the Facebook advert. Both campaigns yielded equal initiations for their budget. Cost per MiQuit initiation was £25.64 via Google, £23.81 via Facebook, and £24.73 across both campaigns. Table 1 also shows the estimated cost per additional quitter of initiating pregnant smokers into MiQuit via commercial online advertising. Using the cost per initiation of the commercial adverts (£24.73) plus the mean cost of sending the MiQuit texts to those initiating support here (£2.73) gave a point estimate of £793.64 per additional quitter to both initiate and deliver the support. The mean cost per additional quitter from the bootstrap was £735.86 (95% CI £227.66-£5223.93).

A total of 53 of the 93 initiators in this study (57%, 95% CI 47%-67%) set at least 1 quit date with MiQuit during the program, including at baseline. Moreover, 63 (68%) of the initiators chose to answer all 12 tailoring questions by website (with quit date noncompulsory). Of those who chose to answer by text message instead (n=27), 16 (59%) responded to at least the first 6 questions. Only 3 initiators (3%) answered no tailoring questions and received generic support. “Pull” support features (“HELP,” “SLIP,” and “QUIZ” requests) were used by 35 (38%) initiators.

A total of 34 of the 93 initiators in this study (37%, 95% CI 27%-47%) stopped the 86-day program prematurely (mean days into program 18.6, SD 21.3). Although not formally tested, discontinuation rates, tailoring question completion rates, and use of interactive program features did not appear to differ between those taking up MiQuit via the 4 different online routes.

Table 2 shows key characteristics of those who initiated MiQuit via online advertising (n=93) and statistically compares those who initiated MiQuit via Google versus Facebook; numbers were insufficient to compare data from the 2 free links (shown combined). Readiness to quit smoking appeared high among initiators in this study, with 70 out of 93 (78%) seriously planning to quit within the next 2 weeks. Gestation differed substantially between women who initiated support via Google versus Facebook; no other characteristic differed significantly between them. Those from Google were typically very early in pregnancy, with 49% reporting a baseline gestation of 4 or 5 weeks (median 5 weeks; interquartile range [IQR] 10); those from Facebook were distributed more widely across pregnancy (median 16 weeks, IQR 14, Mann-Whitney U test P<.001).

When a low-cost, text messaging, pregnancy smoking cessation support program (MiQuit) was advertised via the internet, with no other form of promotion or recommendation, an overall uptake (initiation) rate of 3.4% was seen among those who clicked on any of the 4 adverts to the MiQuit website. Commercial adverts, which yielded the vast majority of initiations in this study, cost, on average, £24.73 per initiation. Although the initiation rate was higher among those who reached the MiQuit website via free webpage links, the total number of initiations generated through these was much lower than the number of initiations generated by commercial advertising and occurred over a longer time frame. Behavioral characteristics appeared similar for those initiating MiQuit from different online sources. User engagement appeared high as over half of initiators set a quit date with the system, and approximately two-thirds continued with MiQuit until the end of the 12-week program.

To our knowledge, this is the first study to explore the feasibility of using free and paid-for online advertising to reach pregnant smokers and promote their uptake of cessation support. We have shown that a significant minority of pregnant smokers are willing to initiate an automated text messaging intervention when offered this via the internet; given the high reach of the internet, this could translate into substantial numbers of pregnant smokers supported to quit. Unlike many previous evaluations of online advertising [25], we were able to trace the source of all those who arrived at the MiQuit website and all who initiated support, allowing us to compare uptake rates for different online strategies. By quantifying each step of the uptake process, from viewing an advert through to initiating support, we have explored a spectrum of engagement [40] with the offer of support and identified steps where barriers to uptake might be removed. Importantly, we tracked individuals’ behavior beyond support initiation, investigating their engagement with the program (quit date setting, use of “pull” support messages, tailoring question completion) as well as discontinuations (sending a “STOP” message). We were also able to use previous RCT data, with assumptions, to estimate the likely cost-effectiveness of MiQuit if implemented via commercial online adverts.

There are several limitations to this study, including some general challenges for real-world uptake studies. As with many studies recruiting people to interventions without person involvement, we cannot be sure that those who took up support were the intended group (ie, pregnant smokers), although our adverts and website made it clear that MiQuit is for pregnant smokers and almost all provided their gestation. It is also possible that adverts were clicked on out of curiosity by those who were neither pregnant nor smoking, particularly among women targeted by banner adverts on Facebook, or that adverts were passed on to others (eg, friends) by those originally targeted. In addition, it proved difficult to estimate the numbers of pregnant smokers exposed to each advert as a denominator: the number of pregnant smokers shown our Facebook advert is unknown and it is probable that not all Google searches for our keywords were made by pregnant smokers. Nonetheless, the high number of keyword searches made within the United Kingdom in a month (over 46,000) suggests that there may be many opportunities for engaging with this population on the internet. Given the low visibility of the 2 free links, our use of the number of visits to the webpages on which these were placed as a denominator may have greatly underestimated the proportions initiating MiQuit out of those exposed. Difficulties in specifying accurate and comparable denominators have been noted previously in uptake studies [23,41]. However, to health providers, having accurate information on the numbers of MiQuit initiations and on the cost of providing this is of the greatest importance.

It is important to be aware of assumptions inherent to the cost-effectiveness estimates presented. Our cost-effectiveness analyses assumed that, when initiated via the internet, MiQuit would have a similar quit rate to that observed in women recruited to an RCT from antenatal clinics. However, we found differences in characteristics between women in this study and those recruited to the RCT (discussed below), and it is possible that the intervention could vary in effectiveness for different groups of women. Both Facebook and Google had relatively opaque criteria used to determine their advertising charges; there is no guarantee that similar charges would be made for identical adverts in the future. Charges depend on concurrent competition from other advertisers, so are unlikely to replicate exactly from one campaign to another even if all other parameters are held constant. It is probably best to view our cost-effectiveness estimates as indicative rather than definitive; however, optimization of this type of “programmatic” advertising would likely reduce costs significantly.

Compared with pregnant smokers who were recruited to a large RCT of MiQuit via antenatal clinics in a previous study [20], women who initiated MiQuit via online advertising in this study had higher readiness to quit, with 78% (vs 32% RCT) seriously planning to quit within the next 2 weeks and 57% (vs 19% RCT) sending a quit date to MiQuit during the program, including at baseline. This suggests that online initiators may be more likely to make a quit attempt [42,43]. Conversely, online initiators appeared to be more nicotine-dependent than RCT recruits, with 34% (vs 14% RCT) classed as “moderate” or “high” dependence, and this appears to be a key determinant of failure to quit during pregnancy [43-46]. It is, therefore, possible that quit rates might be lower among individuals who engage with MiQuit via the internet compared with those who did so after being recruited to a trial in an NHS setting, but this is currently speculative. Others have not found heavier smoking among online recruits [31].This is an important avenue for future evaluation: MiQuit could have different effects depending on how it is implemented and, therefore, who makes use of it. Previous research has found high readiness to quit in smokers recruited to RCTs by search-based [29], but not banner, online advertising methods [28]; however, in this study, readiness to quit was similar between those initiating MiQuit via Google and Facebook.

A total of 37% of initiators in this study stopped the 12-week program prematurely (sent a “STOP” message). This was notably higher than the discontinuation rate among trial participants, where 13% randomized to the MiQuit condition discontinued the support [20], though more similar to the discontinuation rate when MiQuit was offered in an NHS real-world context, by leaflet (46%) [23]. It is possible that those receiving MiQuit as part of an RCT felt obliged to continue with it because of being a trial participant and receiving human contact as part of their involvement. Previous MiQuit research highlights that discontinuations are made for a variety of reasons, most of which are not related to irritation or dissatisfaction [19], and a separate study indicates that discontinuation can be an indicator of increased engagement in smoking cessation behavior [47]. In other ways, online initiators appeared more engaged with MiQuit than RCT participants, with 3 times as many sending a quit date to the system and almost twice as many (38% vs 21%) sending a “pull” support message.

Although our Facebook advert generated activations throughout pregnancy, those who initiated MiQuit via Google were often early in gestation (around 50% within their first 5 weeks). Adverts attached to internet search engines may, therefore, be a useful way to reach women when they are first pregnant and looking for support, and could potentially maximize health benefits by encouraging abstinence for more of their pregnancy. Currently, the earliest used cessation interventions tend to target pregnant smokers at their antenatal booking appointment, at around 8-12 weeks’ gestation.

We have shown that uptake of MiQuit via online advertising is feasible; our previous real-world study showed uptake of MiQuit to be feasible when offered via leaflets in maternity booking packs without health professional promotion. This suggests that there are promising routes to initiating pregnant smokers into support systems such as MiQuit, without the need for health professional involvement, in both clinical and nonclinical settings. We are aware of no other studies that have investigated using the internet for offering real-world cessation support to pregnant smokers, although 2 previous studies have explored this among nonpregnant smokers in the United States [32,33]. Our overall uptake rate (3.4%) was lower than that found among all smokers offered a national Web-based cessation program via commercial search-based and banner online adverts (6.8%) [32]; however, it was similar to the uptake rate found when banner adverts were used to promote Web-based cessation support specifically to Latino smokers, another hard-to-reach group (2.8%) [33]. Our average cost per initiation (£24.73) compares favorably with costs reported to initiate general smokers into real-world digital cessation support commercially via the internet (mean $35) [32] and very favorably with costs reported to initiate a hard-to-reach group via online banner methods (mean $209.34) [33]. Facebook may be more cost-effective than other banner-based methods for targeting specific populations, given that adverts can be restricted to a particular demographic [48].

Using an efficacy estimate of MiQuit from a previous RCT, we estimated a cost per additional quitter of £735.86 to initiate pregnant smokers into MiQuit through paid adverts, including text message delivery costs but excluding development costs. We did not include a formal cost-utility analysis here, and the caveats discussed in our limitations must be noted, but this is encouraging compared with costs reported for other smoking cessation interventions that are effective and cost-effective in pregnancy. For example, financial incentives are highly cost-effective, with a cost per additional quitter of £1127 [49].

Of those who clicked to the MiQuit website, a much greater percentage (11.6%) obtained the short code to initiate MiQuit than subsequently texted it to do so (3.4%). Obtaining the short code required a number of extra steps after clicking on an advert and landing on the MiQuit website, suggesting that these women were serious about taking up MiQuit. There may thus be potential for increasing uptake substantially, at no extra advertising cost, by ameliorating the drop-off between clicking to accept the support and texting to initiate it. Having to text a short code may be a barrier to initiating support for a number of reasons, including lack of credit among pay-as-you-go phone users, suspicion of hidden charges, and needing to act outside of the website; enabling women to sign up anonymously without the need to text a short code might increase uptake. Website content, tone, and appearance are also potential targets; clearly labeling the website as an NHS service might also reduce women’s barriers to sending an initiation text.

A definitive evaluation is planned for MiQuit. If it is shown to be effective, as our earlier trial suggests is likely [20], then an assessment of its efficacy in an online setting may be warranted. In this study, free-of-charge webpage links yielded relatively few initiations but might have performed better if given greater visibility. High initiation rates were found for women clicking from these links, suggesting that they were well targeted despite having lower reach than the commercial adverts, and pregnant smokers may have been more likely to initiate MiQuit if reaching it from a recognized health source. There may therefore be scope for future work to promote MiQuit via such websites. Future work could aim to minimize search engine–based advertising costs by investigating which specific “smoking in pregnancy” keyword phrases are associated with support initiations; this is possible if support is initiated by a webpage click rather than an external action such as texting a short code. Finally, it is important to establish whether uptake of text-based cessation support among pregnant smokers affects their uptake of traditional cessation support or whether it attracts those who would otherwise try to quit alone, if at all.

Commercial online advertising appears to be a promising method for initiating pregnant smokers into text message–based cessation support. Free and commercial adverts prompting pregnant smokers to click to a sign-up website resulted in an initiation rate of 3.4%. Search-based commercial advertising was able to reach women earlier in pregnancy than interventions delivered in clinical settings seem able to achieve, and those who initiated support in this study had high readiness to quit. Commercial online advertising to pregnant smokers is likely to be cost-effective and can probably be made more so. Given that pregnant smokers’ uptake of traditional support is low, it is important to find successful strategies for offering them effective alternatives.