This section discusses the lessons we learned from the search for ongoing multi-level obesity prevention interventions for the SPOTLIGHT project [15].
The number of interventions found was limited. There are several explanations for this. First of all, there simply are relatively few multi-level obesity prevention interventions. In the systematic review of multi-level interventions published in peer-reviewed journals [18], only seven were identified in Europe. Many of the interventions addressing obesity that were initially identified during the current data collection process either used a single component approach (e.g. solely adjusting the environment or only targeting individual level determinants) instead of a multi-level approach; or they focussed on children and adolescents, or targeted people already suffering from obesity (i.e. applying treatment instead of prevention). An additional reason for the low number of interventions found is that there were few ongoing interventions. An problem with interventions that are no longer active is that the contact persons cannot always be reached. Several interventions stopped just before or during the data collection. In addition, few interventions were long-term; they are often implemented for a fixed time period of a few years. This was also found in the systematic review of multi-level interventions [18], which found low levels of sustainability and dissemination of interventions.
An alternative explanation for the low number of interventions is that our search methods were inadequate, and we failed to find them. We encountered various problems during the data collection that might result in missing out on interventions. However, given the variety of data collection methods used, and the efforts made to identify interventions, we believe the examples reported are representative of the interventions being undertaken.
An uneven distribution between countries regarding the number of interventions was observed. This might be attributed to a bias introduced by the researcher and methods, but we think it is probable that there are actual differences between countries, for various reasons. Fewer obesity prevention interventions were identified in the Eastern (i.e. Bulgaria, Czech Republic, Hungary, Poland, Romania, Slovakia, Slovenia) and Southern (i.e. Cyprus, Greece, Italy, Portugal, Spain) European countries. This is in line with the systematic review of published interventions within the SPOTLIGHT project: the seven multi-level interventions identified in Europe were all from Northern and Western European countries (Norway, Belgium, the Netherlands, and the UK) [18]. Also a previous study on ongoing European community interventions against childhood obesity showed fewer interventions in Southern and Eastern Europe, with the exception of Spain [22]. Such regional differences could be attributed to health promotion itself receiving less priority in some countries, or being the victim of economic cut-backs: although cuts in budgets for public health and health promotion have been noticed throughout Europe [29], they may be more severe in some countries, such as Greece [30].
In line with previous reporting [18], the majority of the interventions focused on a combination of both physical activity and nutrition. EBRBs are known to often co-occur (so-called clustering), with inactive individuals often also consuming unhealthy diets [31]. By addressing physical activity and nutrition simultaneously, the synergy between clustered behaviours can be utilised, which is hypothesised to maximise intervention results [32]. In line with the SPOTLIGHT systematic review [18], the large majority of the interventions further targeted multiple types of environmental determinants of EBRB in addition to individual determinants. The socio-cultural environment was addressed as part of almost all of our 78 interventions, which is also in line with the systematic review [18].
An important feature of the data collection was the diversity of ways in which the data was collected. Each country required a different approach. For instance, for Luxembourg, a relatively small country, the researchers went to the Ministry of Health for a face-to-face interview. For larger countries, communication mostly went via email or telephone. There were also differences in the data collection between interventions within each country, depending on the contact information available for each intervention, and the preferences of the informants and respondents.
Several of the data collection approaches were successful for the identification of relevant interventions and information regarding the interventions: Using a stepwise approach, keeping response burden low, and using tailored and personal approaches. The stepwise approach ensured feasibility and made the data collection manageable, not only for the researchers, but also for the respondents. By asking them information in small ‘portions’, they were less likely to drop out during the data collection process. In line with this, the amount of work that had to be put in by respondents was minimalized, using short communications and utilising any information that had been collected previously. Previous research among health care professionals (physicians) has also shown the importance of a low research burden, shifting as much of the work as possible away from the participants to the research staff [33]. The phased data collection procedure in the current study is similar to the procedure used in a previous study that made an inventory of community-based initiatives to prevent childhood obesity [22]. In that study, the researchers also started by asking key informants and searching databases for interventions and contact persons, after which in a second step they collected detailed information from the contact persons. Another strategy that worked well was to use personal and tailored approaches when contacting respondents. Personal contact is one of the known success factors for recruiting health care professionals for research studies [33]. Tailored communications are more likely to be read and to be perceived as relevant [34]. A disadvantage of the personal approach using the researchers’ own professional networks was that the results might have been influenced by the researchers: a larger network might result in more contacts than a smaller network, and the researchers’ network is likely to be centred in the field of research and geographical region in which the researcher is working. However, the researchers also called upon respondents to spread the calls for interventions and key informants to their own networks. Such peer-to-peer recruitment has previously been indicated as a successful strategy for recruiting health care professionals for research [35].
There were also several strategies that were less successful. In the first phase of the data collection the researchers used mass emailing to approach possible key informants. Several problems were encountered with this approach, the first being who should be contacted. Key informants and contact persons frequently changed during the data collection, and for some interventions or organisations contact persons or contact information for these persons could not be found. The second problem with mass emailing was the very low response. The same problem arose when contacting networks of professionals, as described in the methods section. To illustrate, at some point a mass emailing was sent out to a total of 313 possible contact persons. A follow-up message was also sent. Out of those 313, only 11 people responded (3.5%).
Approaching people through existing networks in the current study often provided a response rate that was even lower. A major challenge is therefore to build motivation and commitment from professionals in the public health field to contribute to the development of a better evidence base to evaluate ongoing practice. Leaders in public health should take responsibility to establish such commitment.
Another approach which met difficulty was searching the internet. Search terms requiring the term ‘intervention’ and ‘obesity’ could easily miss important interventions, as they might be called programs, projects, approaches, initiatives or any other alternative for ‘intervention’, and might use terms such as lifestyle, health, nutrition, (physical) activity and well-being in their title, rather than obesity. A second problem with searching the internet was that the information found online was mostly outdated, as websites are often not regularly updated or reviewed by the search engines.
A further problem with using the internet to find interventions was that many interventions do not have their own website, or are not referred to on other websites. This is especially the case for the more local, small-scale interventions. Similar problems were encountered when using reports, databases and existing overviews of interventions. Such reports are quickly outdated, and did not result in the identification of many ongoing interventions. Moreover, these overviews often only list interventions that are already published or in some way connected to research, again excluding the small-scale interventions. However, although the internet and existing databases and reports were not successful approaches for finding interventions, they were often helpful for finding detailed information regarding some of the interventions that were already identified.
There were also some language issues. Most communication took place in English, with the exception of some communication in Dutch and Norwegian/Swedish (i.e. the mother tongue of the researchers). There was no budget to hire translators. The extent to which this resulted in problems differed between countries. Although English is by far the most commonly used language in Europe [36], it severely impeded the communication with the contact persons and key informants in some countries. For instance in France and the French speaking part of Belgium, communication in English was often problematic, with a low response rate, or responses being sent in French. Research has shown large differences in the English skills in the different European countries, ranging from about a quarter of citizens being able to have a conversation in English in Hungary, the Czech Republic, Slovakia, Spain and Portugal, up to almost nine out of ten in the Netherlands, Malta, Sweden and Denmark [36].
Gathering information on the RE-AIM characteristics also posed particular problems. The interventions identified in the current study were often not coupled to scientific research, and the people working with the interventions often did not have an academic background. We do not want to want to imply that interventions should always originate from an academic background. Instead we would recommend having academics support community led intervention development. A good example of this is ‘Woerden Actief’ (Woerden Active), an intervention which was initiated by a group of residents of Woerden (a municipality in the Netherlands) as a civic initiative. Later on, researchers got involved through the Municipal Health Service and other professional organisations. To date the intervention is still ongoing and successful [19].
Parallel to the involvement of researchers in community led initiatives, we also need to look at the other side of the spectrum. More focus needs to be on the implementation and dissemination of researcher led interventions. Lately, attention to these issues has been increasing, for instance by requiring researchers to outline their plans for dissemination and continuation when first applying for funding. Especially the involvement of linkage systems with representatives from the community throughout the intervention development and implementation is recommended (e.g. [37]).
Information was difficult to extract, with respondents saying they had insufficient time to respond to the questionnaires and emails or that they were not motivated to respond as they were not interested in publicity for their intervention, or that they did not perceive the relevance of participating in the study. These reasons for non-response were most often mentioned when little information from other sources could be pre-filled into the questionnaire and more information was thus requested from the respondent, increasing the respondent burden. Other respondents indicated that they did not feel competent to provide the needed information, or implied that they feared criticism or shame for not being able to provide certain information. Some of them further felt the questionnaires were like an inspection, in which the respondents were being graded or examined. These feelings of respondents should be taken into account in future studies to optimise response rates. The aims and use of the reported data should be made clear to the respondents. Another problem was that certain information regarding interventions was confidential, because of the involvement of commercial organisations that owned copyright of the intervention process or materials. In other cases, the information was confidential because the research studies coupled to the interventions had not been published yet. An embargo on the data until a certain date could perhaps help in such cases.
There are some other data collection methods which could have been used and which we would like to recommend for future projects, but which were not considered feasible within the current project. Local researchers could have been employed for the data collection in each specific country. This would resolve the cultural and language barriers that we faced in the current study. In addition, such local researchers could use their own national networks for the data collection, as the researchers did in the current study for the Scandinavian countries (for the Norwegian members of the research team), the Netherlands and Belgium (for the Dutch team members) and the UK (for the English team members). We therefore recommend funding for local researchers or translators to be taken into account in multi-country studies, to ensure a better response rate.
As a result of these factors, the completeness of the intervention descriptions was low. Information was especially lacking for smaller and local interventions. In particular, information regarding the RE-AIM characteristics of interventions was often lacking, which has also been noted for interventions with a scientific background [18]. More information on the external validity and sustainability of interventions is needed in order to take informed decisions about intervention development and implementation.