**4. Discussion**

With the chosen iterative participatory mixed methods approach, we identified social media as the best access path to the target population and developed an m-health campaign to increase knowledge about the consequences of SHS in children and about simple measures to reduce exposure. The adequacy of the campaign for the target group was confirmed in a quantitative online survey. Translating the material, the campaign might also be used in other settings and locations.

The main challenge of the project was the recruitment of participants from the target population. This is in agreemen<sup>t</sup> with almost all studies targeting sensitive topics like SHS exposure in children among disadvantaged families [11,35]. Due to these difficulties in recruitment, with respect to socio-economic status, the relevant criterion for inclusion in the study was the job status of the parent that took part in the study. Thus, we cannot rule out that their partners might have had a better employment status. However, it is very likely that this applied to no more than a very small proportion of the participants as in Germany spouses largely have the same socio-economic position. In order to improve participation, we worked with community partners and recruited/interviewed participants using staff whose native languages were Turkish, Spanish or Russian. Nevertheless, the target population felt uneasy to discuss this sensitive topic with members of a university hospital. The reason for this was partly that most of the invited members of the target group were aware that SHS exposure has negative health effects. They were thus concerned that researchers would stigmatise them even more. This was confirmed in the quantitative part of the study in which most participants were aware of the harms of SHS. However, as in other studies [10,15], knowledge was incomplete.

One limitation of the quantitative part of the study was the lack of a list of the target population of which a random sample could have been chosen. In addition, based on our own experience and the experience of other researchers aiming at asking disadvantaged families about a sensitive topic, we only expected a response of 10–20% [35,36]. For recruitment, we followed recommendations to contact the target population in their native languages by interculturally trained staff and to distribute the study information through multiple channels (counselling centres, clubs, doctor's practices, Internet) [36]. Furthermore, we offered financial compensation for participation. We assume that by using these measures we were able to recruit a diverse, although more motivated than average, sample. At the same time, the validity of our conclusions was ensured by our participatory mixed methods approach [37]. The consistency of the results of the qualitative and quantitative phases supports this assumption.

The subjects' concern to be confronted with their own smoking, or the smoking of their partner, as well as the possible effects on their children's health, was also evident in the quantitative evaluation of the campaign messages; the first impression of the animated illustrations was rather negative for many of the participants. On the other hand, the participants indicated that they would take a closer look at the illustrations in social media—but would rather not like or share them with others because of the unpleasant content. This converges with the results of the focus groups regarding possible obstacles to dissemination via social media. Hamill and colleagues reported similar experiences regarding the spread of an anti-smoking campaign via Facebook [38]. In their project, they followed the general recommendations for 'anti-smoking campaigns' using deterrent photos of the consequences of smoking. By this, they reached only 10% of the invited users. In accordance with our respondents, their participants stated that the photos were offensive and could provoke others if they were shared [38]. In our focus group discussions, participants suggested that the campaign should be spread by trustworthy persons, such as doctors, rather than sharing them themselves. This, in turn, coincides with the results of an anti-smoking campaign in Egypt, where the campaign was posted and advertised on Facebook and other media [38]. Such social media campaigns are relatively inexpensive and can be specifically tailored to the target audience [38,39].

In summary, our project developed an m-health campaign in close collaboration with the target group. The campaign was finally assessed positively in an independent evaluation, also carried out by members of the target group. The quality of the images and key messages was found to be satisfactory, the effectiveness and credibility rated as high, and it was confirmed that it was good to address this issue in social media. However, the topic has caused some degree of consternation among the participating smoking parents. Hence, it cannot be assumed that the pictures will be shared in social media actively by smoking parents or their partners. As suggested by the participants, dissemination should therefore be done by doctors, scientists or authorities. This could be achieved through the dissemination by paediatricians via social media, as well as through paid advertisements within social media [38,39]. These access routes need to be assessed in a follow-up study. In addition, it would be important to study whether the campaign has a lasting effect on the behaviour of smoking parents; i.e., the extent to which the proposed simple measures to reduce children's exposure to passive smoking are actually implemented. Further light on the efficacy of such campaigns may be shed by the results of a similar project in two other major German cities [40]. A recent representative population-based survey showed considerable support for tobacco control measures in Germany independent of socio-economic status (although, not surprisingly, different among smoker and non-smokers), however, indicated that such campaigns may fall on fertile ground [7].
