Feasibility of multi-sector policy measures that create activity-friendly environments for children: results of a Delphi study

The study was conducted in four medium-sized Dutch municipalities that were participating in a large-scale research project described in more detail elsewhere [22]. To guarantee complete anonymity of the respondents in the study, city names are blinded throughout the text. Table 1 summarizes the main characteristics of the cities that were enrolled. Despite the fact that municipality D was somewhat smaller compared to the other municipalities, the municipalities were similar regarding the composition of their population.

The Delphi method is a well-founded method for reaching consensus among stakeholders in complex (policy) problems [23] and has been widely used in the field of health policies related to obesity [24‐27]. Within the Delphi method, respondents are provided with the opportunity to adjust their opinion based on group's mean or median scores in two or more consecutive Delphi rounds and the procedure stops when consensus is reached or response rates decrease [28]. In this study, four separate Delphi studies were conducted (one in each municipality), to provide the municipalities with locally relevant results, which increases the applicability of the research in the municipal policy development process.

In the Netherlands, three levels of government exist: national, regional, and local/municipal. The municipal government consists of a bureaucratic system staffed by policy officers, a political level of alderman and mayor, and a municipal council. The policy officers support the aldermen and mayor in administrating the municipality. The members of the municipal council supervise the aldermen and mayor and hold power of decision. Whereas the aldermen and municipal council members are re-elected every four years, the pool of policy officers remains more stable over time. Respondents in this study were chosen from the policy officers because they are best informed about the content of the policies within their sector. Respondents were selected by means of 'snowball sampling', starting with existing contacts with policy officers in the public health domain, who referred to their colleagues from other policy sectors. In each municipality, six policy sectors (public health, sports, youth and education, spatial planning/public space, traffic and transportation, and safety) were invited for participation because of the potential influence on the environmental determinants of children's physical activity [18]. On the respondents' initiative, an additional policy sector was invited in municipality B (environmental affairs), municipality C (economic affairs), and municipality D (play facilities). In addition to policy officers from the municipal organization, the regular policy advisors from the Regional Public Health Services for each municipality were invited to participate in the Delphi study of their particular municipality. In order to prevent overrepresentation of particular policy sectors, a maximum of two respondents per policy sector within each municipality was set. In total, 36 respondents were invited for participation.

The first Delphi round took place at the venue of the city hall and took approximately 90 minutes. The main results of our survey on environmental correlates of physical activity among children (conducted in the participating municipalities) were presented by the principal researcher of the project [5] and discussed in relation to the state of the art knowledge from scientific reviews [6, 29]. During this discussion, it emerged that physical as well as social neighborhood characteristics are related to different components of children's physical activity behavior (e.g., outdoor play, active commuting to school, sports participation). Based on these insights, social cohesion, accessibility of facilities, and traffic safety were considered important environmental correlates of different aspects of children's physical activity [5, 6, 29] that are affected by policy measures of different policy sectors (e.g., youth and education, spatial planning, traffic, and transportation). Participants were asked to identify possible municipal policy measures that address these three determinants in a plenary brainstorm chaired by a professional discussion leader (20 minutes per determinant, or one hour for the total brainstorm). Respondents were explicitly asked not to consider the feasibility of policy measures during this first round. At the end of the first round, respondents summarized the identified policy measures from this round by compiling a list of at least four policy measures per determinant, and these were further explored during the second Delphi round. The discussion leader made certain that there was a common understanding of the exact content of the policy measures among the participants of the Delphi meetings (for a detailed description of each policy measure see Additional file 1).

The second Delphi round (which took approximately two hours) followed immediately after the first Delphi round and took place in the venue of the city hall as well. Five feasibility aspects derived from the literature (political feasibility, cultural/community acceptability, technical feasibility, cost feasibility, and legal feasibility [19, 20]) were briefly introduced. Thereupon, each respondent individually weighted the general importance of each of these different aspects of feasibility in the policy development process by dividing 100 points over the five feasibility aspects. Subsequently, each respondent was provided with a printed questionnaire and scored the policy measures derived from the first Delphi round on the five aspects of feasibility (seven-point Likert-type scale, higher scores indicated higher feasibility). These questionnaires were completed and handed in during the meeting at the city hall.

The aim of the third Delphi round was to develop group consensus, and this round consisted of a printed questionnaire sent to the respondents by post approximately one week after the first and second Delhi round. Respondents were provided with their own scores, as well as the median group scores from the second Delphi round and were asked to re-evaluate their individual feasibility scores. Respondents unable to attend the first and second Delphi round were invited to evaluate the feasibility of the policy measures during the third Delphi round as well. These respondents were asked to first weigh the five aspects of feasibility (similar to the other respondents) and were provided with the median group scores from the second Delphi round as well. Detailed descriptions of each policy measure (including practical examples mentioned during the discussion meetings) were provided for each respondent (see also Additional file 1). Respondents could return the completed questionnaire of the third Delphi round by an enclosed prepaid envelope.

By multiplying the individual weighting scores by the feasibility scores on each feasibility aspect and summing the five feasibility scores for each policy measure, one overall (weighted) feasibility score per policy measure was calculated for each respondent for the second and third Delphi round separately. The median weighted overall feasibility score was then computed per policy measure for each municipality for the second and third Delphi round separately. In addition, the standard deviation (SD) was calculated as an indicator for consensus within each municipality for the second and third Delphi round separately (higher SD scores indicate less consensus). Two respondents had missing values on their weighting scores, and therefore these were imputed by the average weighting scores of the other respondents within the same municipality. Three respondents that participated in the first and second Delphi round did not return or had missing values on the questionnaire during the third Delphi round. In those cases, the missing scores on the third Delphi round were replaced by the respondent's scores from the second round.

Policy measures were considered 'consistently feasible' if they met the following conditions in the third Delphi round: weighted median overall feasibility score ≥ 5.00; standard deviation ≤ 1.00; and the minimum overall feasibility score given by any individual respondent within that municipality ≥ 3.50. Policy measures were considered 'consistently less feasible' if they met the following conditions: weighted median overall feasibility score ≤ 4.00; and standard deviation ≤ 1.00 in the third Delphi round. These criteria were chosen as natural cut off points based on the scales used and data obtained. To further compare results across municipalities, each policy measure was classified into one or more of the following categories: communicative policy measures such as health education and advertisements; juridical policy measures such as laws and prohibitions; economic policy measures such as subsidies, grants, charges, and taxes; and environmental policy measures such as changes in facilities, infrastructure, or neighborhood design. All policy measures were classified by two authors independently, and in case of inconsistencies, consensus on classification was reached by discussion.

In line with the findings of this study, the importance of economic and political factors was also mentioned by municipal employees in previous research [18, 30]. Although respondents initially were less familiar with the possibilities to improve social cohesion within their municipality, during the Delphi process they became aware of several feasible policy measures that address this determinant of children's physical activity. Policy measures aimed at improving traffic safety were also perceived as feasible. However, policy measures aimed at improving accessibility of facilities were considered less feasible, probably because this requires drastic modifications in the built environment. Policy measures that have a more authoritative character (e.g., obliging parents to choose a primary school within their own neighborhood) were also rated less feasible. These findings are in line with the cross-national results from the European PorGrow project, gathered among a broad spectrum of governmental and non-governmental stakeholders. The PorGrow results indicate that although the need for an integrated approach aimed at environmental changes is recognized, less drastic policy options aimed at education and information for parents and children are generally ranked highest [31]. The PorGrow results further show that economic policy measures such as subsidies and taxes are given low appraisal scores [31].

In their Delphi study among Dutch experts on opportunities for monetary incentives to stimulate healthy eating, Waterlander et al. have shown that experts tend to rate policy options outside their own area of responsibility more positively [26]. Although respondents did emphasize the own responsibility of parents for their child's activity behavior, there were no indications that respondents from any policy sector tended to pass the responsibility to other policy sectors in the present study. This was probably due to the fact that the respondents were mostly colleagues within the same organization and already discussed the policy options together during the first Delphi round.

Although the aim of this study was to cover roughly the same policy sectors within each municipality, the exact job description of the respondents did vary between municipalities, and diversity of respondents between municipalities could not be completely eliminated. The fact that this study was aimed at municipality-specific recommendations somewhat limits the possibilities to generalize the results. However, some of the findings (such as the fact that cultural/community acceptability, political feasibility, and cost feasibility were consistently considered of greatest importance) can be generalized to other settings as well. The use of a multi-criteria mapping technique--in which respondents have the opportunity to bring up policy options themselves and to weigh the different evaluation criteria--is a useful strategy to provide respondents with adequate freedom of expression, but nevertheless retain the possibility to compare results across municipalities or countries [21, 32]. In addition to assessing the feasibility of different policy measures by assigning scores on Likert-type scales, ranking policy measures could further stimulate participants to single out the different policy alternatives [24, 26, 28]. Although this study provided the respondents with information on which environmental characteristics could possibly affect children's physical activity behavior during the first Delphi round, no detailed information was available on the (theoretical) effectiveness of the proposed policy measures. Calculating the potential health gains of different policy measures could be of great value because this could help to persuade policy makers to seriously consider the implementation of less feasible, but possibly more effective policy alternatives as well. Future research should evaluate if local policy makers themselves see the Delphi technique as a valuable tool in the development of multi-sector policy measures aimed at health promotion at the local level and if it facilitates their actual adoption and implementation. It would also be interesting to study whether elected officials have the same view on feasibility of the identified policy options identified by policy officers in this study.