Design

We examined intervention-context interactions in a cross-sectional observational study (2012–2014). Included were 30 municipalities or alliances of municipalities participating in a ministerial programme on intersectoral health policymaking. Four other projects in this programme were not eligible: one prematurely ended its participation in the programme, one did not implement interventions in the years concerned, and two refused permission to approach the partners responsible for the implementation of the interventions.

Study setting

The ministerial programme (2009–2015) was initiated by the Dutch Ministry of Health, Welfare and Sport . The programme gave municipalities the opportunity to experiment with intersectoral health policymaking over a period of 24–48 months. Municipalities or alliances thereof could apply for participation in the programme. One requirement was the appointment of a project leader who had to adopt a coordinating role in establishing local partnerships and implementing health promotion interventions. The employment of the project leader was covered by the financial support provided by the ministerial programme. This financial support also partly covered the implementation of the health promotion interventions. The ministerial programme additionally provided professional support addressing, for instance, the selection and implementation of evidence-based health promotion interventions.

As previously reported [25], the local partnerships in the projects encompassed an average of seven different sectors (e.g., public health, education and transportation). The health promotion interventions applied a variety of behaviour change methods (e.g., education, facilitation and regulation), to address overweight, alcohol use (sometimes in combination with drugs and smoking) or other health themes, in a range of local settings (e.g., school settings and outdoor public sites). The intervention-context combinations that most often were being implemented in the projects are characterized in Supplementary file 1.

The implementation of interventions was mostly carried out by one of the partners in the project (i.e. the implementer) and supported by one or more other partner organizations (i.e. co-implementers working at co-implementing organizations). Most of the implementers worked for a municipal government organization, and almost half of them for a health organization. On average, the implementers had 10 years of relevant work experience.

Data collection

Details about the data collection have been reported elsewhere [25]. In brief, the data was collected from 2012 to 2014 (inclusive). Two questionnaires were used: one for project leaders (n = 30) and one for implementers of the interventions (n = 181). For the present study, both the project leaders and the implementers were asked to complete questions regarding the characteristics of the intervention systems (n = 424). The implementers had to complete additional questions about the conditions acting as bottlenecks for implementation.

Intervention system Questionnaire items

The project leaders were asked to report the names of the health promotion interventions being implemented in their project. The implementers were asked, for each intervention they were responsible for, to concisely describe its aim, topic, content/components and implementation setting.

Data processing

We operationalized the intervention system using three proxy measures for its function: the core behaviour change method employed, the main health theme addressed, and the primary setting of implementation [10, 26]. The core method of behaviour change was retrieved from the aim and content of the health promotion intervention, and categorized into [6]: education (e.g., school learning module), regulation (e.g., legislation regarding the sale of alcohol products in sports ground cafeterias), facilitation (e.g., environmental changes, such as new play gardens), citizen participation (e.g., citizens organizing a walking event), and case finding (e.g., spotting drunk youngsters in nightlife). The main health theme was inferred from the topic, aim and content of the intervention, and categorized into overweight (e.g., nutrition and physical activity), alcohol (sometimes in combination with drugs and smoking) and other health themes (e.g., fall prevention or self-defence). The primary implementation setting was derived from the description by the prime implementer, and categorized into [4]: schools or preschools, outdoor public sites (e.g., playgrounds, nature areas), sports facilities, homes (including websites to be consulted at home), commercial buildings (e.g., supermarkets, bars, restaurants), health or welfare buildings (e.g., hospitals, welfare organizations, addiction centres), and public buildings (e.g., libraries, community centres).

Bottlenecks for implementation Selecting conditions

An extensive review of the literature resulted in a list of 125 conditions necessary for the implementation of health promotion interventions in local settings [8, 27,28,29,30,31,32]. To select the conditions relevant to our study, we held 17 semi-structured telephone interviews: five with Dutch implementation experts and twelve with Dutch health promotion professionals responsible for local implementation. None of the interviewees was participating in the ministerial programme. Guided by an implementation framework [27], but without being provided with the prepared list, they were asked to name those conditions that were most important in the context of intersectoral policymaking. The 47 conditions that were mentioned most were included in the questionnaire for the prime implementers.

Questionnaire

The relevant conditions were organized into seven categories (i-vii) [27], that we adapted to the context of intersectoral policymaking, e.g., by referring to an integrated approach, and by making separate categories of conditions for the co-implementer(s) and co-implementing organization(s). Conditions were framed as statements: (i) five related to the prime implementer (e.g., ‘I have sufficient skills to implement the intervention’); (ii) five to the co-implementer(s) (e.g., ‘Other professionals are capable enough to implement the intervention’); (iii) ten to the intervention (e.g., ‘The intervention is easy to carry out’); (iv) ten to the prime implementer’s organization (e.g., ‘The intervention fits my organization’s policy’); (v) eleven to co-implementer’s organization(s) (e.g., ‘Other organizations sufficiently support the intervention’s health theme’); (vi) four to the broader context (e.g., ‘There is enough administrative and political support for the intervention’); and (vii) two to the implementation strategy employed (e.g. ‘Good materials required for implementation are available’). For the complete questionnaire, see Supplementary file 2.

To assess the extent to which the conditions for the implementation of the intervention under consideration were regarded as being present, the prime implementers had to score each statement on a five-point scale (from strongly disagree to strongly agree). To assess the perceived importance of the conditions, the prime implementers were asked to select the five conditions they regarded as most important for the successful implementation of the intervention. We opted for this top-5 of importance as to discriminate the expected limited number of crucial conditions [16, 17] from the myriad of potential conditions for implementation [8, 20]. For their top-5, the implementers could refer to the 47 conditions in the list or add a condition not included in the list. Of the added conditions, half could be recoded as a prelisted condition. The other half, making up 11% of all answers, were not specific enough to be categorized (e.g., a lack of time, insufficient skills or short of manpower in general), and were not further taken into account.

Data processing

For each individual intervention, the perceived presence of each of the conditions for implementation was dichotomized into being ‘optimal’ if a prime implementer indicated strong agreement with the corresponding statement, and being ‘sub-optimal’ for all alternative answers. This cut-off point was chosen because of the skewed distribution of perceived presence: any other division would have minimized the percentage of interventions for which a condition was marked as ‘sub-optimal’, leaving many bottlenecks undetected. Next, conditions were marked as ‘important’ if assigned to the top 5, irrespective of their position therein. Finally, conditions were marked as a bottleneck if they were perceived as being both ‘important’ and ‘sub-optimal’.

Data analysis

Descriptive statistics were used to characterize the included intervention systems, and to calculate the percentage of systems in which a condition for implementation was marked as sub-optimal, important and a bottleneck.

We tested our study hypotheses using stratified analyses (see below). To warrant the availability of sufficient observations for hypotheses testing, we selected the intervention systems that were most frequently present in our sample (n > 10; see Supplementary file 1). After stratification, a condition was regarded a bottleneck for implementation if it was marked as such in more than 10% of the intervention systems that made up a certain stratum, i.e. a certain subset of systems.

To test our first hypothesis (H-1), we stratified the percentage of bottlenecks by frequent intervention system. To assess whether each distinct intervention system came across a unique set of bottlenecks for implementation, we compared the number and the nature of the conditions that emerged as bottlenecks in the different strata, i.e. in the different subsets of intervention systems. All comparisons were observational.

To test our second hypothesis (H-2), we additionally stratified the percentage of bottlenecks by intervention system characteristics: the behaviour change method, health theme and implementation setting. We then compared the conditions that were regarded a bottleneck after each of the stratification procedures. To indicate that a bottleneck was associated with the characteristics of a particular intervention system (H-2a), we labelled it ‘expectedly present’ in that system if the condition involved also acted as a bottleneck in all systems having a characteristic in common. To indicate that a bottleneck emerged independent of the characteristics of a particular intervention system (H-2b), we labelled it ‘unexpectedly present’ in that system if the condition involved did not act as a bottleneck in all systems having a characteristic in common. In addition, a bottleneck was labelled ‘unexpectedly absent’ if the invers incongruence was true, i.e. if a condition did not act as bottleneck in a particular system, while it did so in all intervention systems having a characteristic in common.