Implementing healthcare interventions in clinical practice is challenging (Greenhalgh, 2020). Most healthcare interventions are considered to be complex interventions as they usually comprise several interacting components or have a variety of outcomes. These components impact the length and complexity of the causal chain from intervention to outcome and the influence of features of the local context (Craig et al., 2013). Complex interventions also require new behaviours by those who are involved in delivering the intervention (Craig et al., 2013; Skivington, 2021). When implementing such complex interventions, it is important to assess and understand the context. Hereby, the intervention needs to be adapted to the local context and therefore implementation strategies have to be selected, tailored to and executed in the context to get the intervention into routine use of multiple involved professionals (Craig et al., 2018; Damschroder, 2020). This makes implementation complex, and even more complex when different contexts are involved (Damschroder, 2020).

Healthcare professionals have a responsibility for the delivery of interventions and their implementation. Implementing a complex healthcare intervention requires adaptations in the work processes of different healthcare professionals. They have to change their behaviour to enable and execute this new intervention from current to desired behaviour. Conditions for behavioural change are related to three areas: capability, opportunity and motivation (Michie et al., 2014; West et al., 2020). In coming to the desired behaviour, professionals must have sufficient knowledge and skills (i.e. capability). They must have the time, space, resources and a positive attitude of colleagues (i.e. physical and social opportunity). They must be motivated to make the new intervention part of their daily work and convinced of the added value (i.e. automatic or reflective motivation) (Atkins et al., 2017; Michie et al., 2014). Insight into the capability, opportunity and motivation of healthcare professionals provides input for the implementation strategy and thus contributes to the implementation process of complex interventions.

This study on exploring behaviour change of healthcare professionals was part of a larger study in which we implemented a complex nursing intervention in twelve of the sixteen Dutch cardiac surgery centres: the ‘PREvention Decline in Older Cardiac Surgery patients’ (PREDOCS) consultation (see Appendix A). As every participating hospital has specific contextual factors with different barriers and facilitators, we developed tailored implementation strategies for adapting PREDOCS to the local context of each centre (Ettema et al., 2014; Jordens et al., 2022). To support changing current behaviour to desired behaviour during implementation, in each context we adapted implementation strategies based on results of measurements. For this study, we hypothesised that the behaviour of all involved healthcare professionals would change over time due to the use of locally tailored implementation strategies. Most studies examine behaviour at a single point in time (Barker et al., 2016; Coupe et al., 2022; Murphey et al., 2023). However, we know little about how professionals’ behaviour changes over time when implementing complex interventions. Measuring two points in time provides insight into behavioural change over time. Therefore, this study explored the change in behaviour in terms of capability, opportunity and motivation of all healthcare professionals involved in the implementation of the complex nursing intervention PREDOCS in twelve cardiac surgery centres in the Netherlands.

Methods

Study Design

In this multicentre survey, we identified capabilities, opportunities and motivations related to the current behaviour of professionals involved in enabling and providing the PREDOCS consultation, as well as determinants which may influence the actual use of PREDOCS in clinical practice (Fleuren et al., 2014). We measured the behaviour of healthcare professionals at two points in time by using a questionnaire with statements regarding capability, opportunity and motivation on which respondents could score from 1 to 7 for the extent to which it applied to them, leading to quantitative results. Respondents also had the opportunity to post comments in response to the statements in an open field, leading to qualitative results. This study was embedded within the implementation study of the PREDOCS consultation, which started in January 2016 in twelve cardiac surgery centres in the Netherlands and ended in December 2019 (Jordens et al., 2022).

The PREDOCS Consultation

The complex nursing intervention we implemented was ‘the PREDOCS consultation’. The PREDOCS consultation is an effective evidence-based multi-component nursing intervention. It aims to improve older patients’ physical and psychosocial conditions to reduce their risk of delirium, depression, pressure ulcers, and infection during hospital admission after cardiac surgery (Ettema et al., 2014). The consultation is based on three guiding principles: 1) using waiting time for hospital admittance as preparation time, 2) performing a risk assessment of potential increased risk of postoperative complications directly after the decision for surgery and 3) self-management support by providing specific preventive interventions for application in the preparation time (Appendix A).

Implementation Approach

The larger implementation study, in which we implemented the PREDOCS consultation, consisted of four general steps including (1) adapting evidence, (2) identifying barriers and facilitators, (3) tailoring implementation activities and (4) monitoring and evaluating. For monitoring, we measured three implementation outcomes: (a) barriers and facilitators of implementation, (b) behaviours of involved healthcare professionals, which were guided by the COM-B model, and (c) the fidelity of executing the intervention. Based on the analysis of these measurements, we provided customized feedback to local project teams to optimize locally tailored implementation strategies (Jordens et al., 2022).

In this study, we measured the behaviours of involved healthcare professionals twice through a self-assessment questionnaire. The analysis of the self-assessment questionnaire contributed to the optimization of local tailor-made implementation strategies.

Population and Setting

The source population of this study consisted of all healthcare professionals from the twelve cardiac surgery centres who fulfilled at least one role in organising, enabling and/or providing the PREDOCS consultation. Prior to this study, the national project leaders of the larger implementation study (YJ and RE) identified seven roles concerning implementing and executing the PREDOCS consultation. Professionals in these roles have to work together to enable the PREDOCS consult. These roles include project leaders, PREDOCS nurses, managers, general ward nurses, cardiothoracic surgeons, planners and data managers. The roles were in some cases represented by different professionals across the participating cardiac surgery centres. In each centre, professionals in these roles had to adapt to new behaviour to enable performing the intervention. An overview of the roles, their corresponding tasks for enabling the PREDOCS consultation and possibly representing professionals is presented in Appendix B.

The participating cardiac surgery centres started with the implementation of the PREDOCS consultation in 2016 and 2017. Five of them were university hospitals focusing mainly on special cases, the other seven were regional hospitals with a special national function on cardiac surgery focusing mainly on mainstream cardiac surgery procedures such as bypasses and valve surgery. The cardiac surgery centres are geographically spread across the Netherlands.

Sample

In this study, different sampling methods were used to reach a representative sample of every role. The same population was approached for both the measurement in 2018 and the measurement in 2019. For PREDOCS nurses and project leaders, total population sampling was used, as they were of high importance in the implementation process. During the implementation of the PREDOCS consultation, the roles of dedicated planner, cardiothoracic surgeon, manager, and data manager were mostly represented by one participant in every cardiac surgery centre. For these roles, convenience sampling was used and we approached one participant per role in every centre. We assumed that all ward nurses in the cardiothoracic ward potentially took care of PREDOCS patients in the postoperative period. To obtain a representative insight into the current behaviour of cardiothoracic ward nurses, we used convenience sampling for this role and determined to include five ward nurses per cardiac surgery centre.

Measurements

We measured the behaviours of involved healthcare professionals twice through a self-assessment questionnaire: COM-B-Questionnaire for PREDOCS (COM-B-QP), in which the analysis contributed to the optimization of local tailor-made implementation strategies. The questionnaire was based on the Self-Evaluation Questionnaire V1 which is based on the COM-B model (Michie et al., 2014). This questionnaire provides insight into psychological capabilities, social and physical opportunities and automatic and reflective motivations of involved professionals.

Capability

Capability refers to an individual’s psychological and physical ability to participate in an activity. Psychological capability is the capability that involves a person’s mental functioning (e.g. the professional must have the knowledge, skills, psychological strength and stamina to organise, provide and enable the PREDOCS consultation in a desired way). Physical capability refers to bodily functions, such as having the strength, stamina or dexterity needed to perform a behaviour. Since physical capability was considered to be a present condition, we did not measure this.

Opportunity

Opportunity refers to the external factors – physical and social – which make the execution of a particular behaviour possible. Physical opportunity is the opportunity that involves inanimate parts of the environmental system and time (e.g. what the environment allows or facilitates in terms of time, triggers, resources, locations, personnel and integration in the hospital information system). Social opportunity involves other people and organisations (e.g. culture and social norms; the feeling of a shared responsibility regarding one’s tasks in the PREDOCS consultation).

Motivation

Motivation refers to the internal processes which influence our decision making and behaviours. Its two components are automatic- and reflective motivation. Automatic motivation involves habitual, instinctive, drive-related, and affective processes (e.g. desires and habits; patient safety during admission is seen as important by the professional and the PREDOCS consult supports that). Reflective motivation involves conscious thought processes (e.g. plans and evaluations; preparing older patients for a hospital admission with an intervention such as PREDOCS is seen as logical for the involved professional).

We added eight items of the Measurement Instrument for Determinants of Innovation (MIDI) (Fleuren et al., 2014) to the questionnaire. MIDI items consist of determinants that can hinder or promote the introduction of an innovation. Based on these determinants, targeted implementation strategies can be designed. Respondents were also asked to post comments in open boxes, for more in-depth understanding and clarification about capability, opportunity and motivation.

Adaptation of the COM-B-Questionnaire for PREDOCS (COM-B-QP)

We followed six steps to adapt the original COM-B Self-Evaluation Questionnaire V1 (Michie et al., 2014) to the PREDOCS intervention. First, the original COM-B Self-Evaluation Questionnaire went through a process of forward and backward translation (Degroot et al., 1994). Two Dutch native translators performed the forward translation (one with a nursing background and YJ), compared both translations and reached a consensus in case of differences. A formal English translator back-translated the synthesis questionnaires into English. YJ and the translator discussed the differences. Members of the research team (YJ, RE and MS) finalized the questionnaire. Second, the existing COM-B statements were adapted to professionals (by YJ) and were adapted to the phase of the change process. COM-B items are aimed at the start of a change process. Since we did not measure at the beginning of the change process, we adjusted the questions to, for example, ‘I have enough…’ instead of ‘I need more…’. In addition, the start of implementation differed per hospital, which meant that the implementation phase differed per hospital at the time of data collection.

In the third step, the relevance of every capability, opportunity and motivation item was determined by two independent implementation specialists and YJ to ensure that the items of the COM-B-QP represent a comprehensive range of factors affecting professional behaviour involved in an implementation process. Based on both judgements, two researchers (YJ, RE) made a final decision on the items to be included. These items were related to capability, such as knowledge, skills, and mental strength/stamina; opportunity, such as availability of time, space and materials and shared responsibility; and motivation, such as beliefs, self-identity, intentions, plans, emotions, feelings, habits and drives. We eliminated six items which were not relevant to our population and setting. Four items focus on physical capability which was not applicable for professionals implementing the PREDOCS consultation. Two items are related to the availability of preconditions, other than time, space and material. In addition, we merged four items into two items, because these items both focused on mental strength and support from other professionals. So we combined ‘sufficient mental strength’ with ‘overcoming mental obstacles’ and ‘enough people around me who do it’ with ‘support from others’.

As the fourth step, to improve our understanding of the determinants that may affect implementation and subsequent better targeting the implementation strategy, we added eight MIDI items concerning: knowledge about the innovation, availability of sufficient personnel, and changes within the organization in addition to the implementation of the intervention (e.g. reorganization, staff turnover, other innovations), connection with current working methods and whether the effect of the intervention is visible in practice. An overview of all items is presented in Appendix C.

The fifth step was to determine the desired behaviour of all involved professionals. To address every participant’s specific context in implementing the PREDOCS consultation, project leaders were first asked to deliver information about the division of roles in their hospital and the corresponding tasks of professionals representing these roles. Subsequently, the questionnaires were adjusted and specified to every role and every hospital. Several participants fulfilled multiple roles. In the latter case, participants received a questionnaire with items specified and corresponding to their different roles and hospital/setting.

At last, two professionals: a project leader and a PREDOCS nurse, tested the final questionnaire for understandability. This led to minor improvements, mainly aimed at clarifying the examples. In conversation with these professionals, we found that they had the correct interpretation of the questions. See the questionnaire in Appendix C.

The final COM-B-QP was adjusted for each professional role and varied between fourteen and 26 items depending on the professional’s role. The questionnaires were also adapted to the context of each centre by citing examples, e.g., the method of planning the PREDOCS consultation and the person who plans it, the way of embedding the PREDOCS consultation in the hospital process, discussing the outcomes of the PREDOCS consultation in the medical team which decides for surgery. With each item, the extent to which the respondent agreed was measured on a 7-point Likert scale (1 = ’strongly disagree’; 7 = ’strongly agree’). A score of 7 (‘strongly agree’) was presented as the desired score for every item. We collected demographic data such as age, gender, role, level of education and number of years of work experience.

Data Collection and Procedures

Data collection of the COM-B-QP took place at two points in time through this self-administered questionnaire. The first measurement was conducted between March 2018 and May 2018. The second measurement was between July 2019 and August 2019. Project leaders were informed about this study by email and were instructed to send the questionnaires to the right participants in their hospital, to collect the completed ones and send them back to the researchers. Instructions took place by email and by telephone. Two reminders were sent via email to non-responders two weeks after the initial questionnaire had been sent. Informed consent was obtained from the overarching PREDOCS implementation study. The introduction of every questionnaire stated that participants gave informed consent by filling in and returning the questionnaire.

Data Analysis

Demographics of the healthcare professionals were described for both groups involved in the two measurements. The median and interquartile range and the subsequent difference/delta in medians in two time points were calculated to describe the scores of the measured variables. Responses to the items of the COM-B-QP were used to construct composites of psychological capability, physical opportunity, social opportunity, automatic motivation, and reflective motivation scales. The number of the requested items differed per role, which is why the mean scores of sub-components of the composites were included in the analysis. Subsequently, medians and interquartile ranges were calculated by role for each of the five COM-B items: psychological capability, physical and social opportunity and reflective and automatic motivation. Medians and interquartile ranges were chosen because groups were mostly small (n < 10) and/or not normally distributed. To compare the overall median scores for capabilities, opportunities and motivations of all involved roles, boxplots were used. If the IQR concerning each item from 2018 showed an overlap with the IQR of 2019, we concluded that there was no significant difference. For MIDI items with a 7-point Likert scale, we used overall median scores and presented this in a boxplot. Missing values were sparse and did not show any perceptible patterns. No individual item was missed more than four times and only two respondents had missing information for more than two items. Median scores were based on the number of non-missing values for each item.

Data entry and descriptive analysis took place using IBM SPSS Statistics (IBM, Chicago, USA). For data visualization, we used Microsoft Excel (Microsoft, Washington, USA).

The qualitative analysis using inductive thematic analysis, took place in two phases. First, comments in open text fields were coded by YJ and an independent nursing scientist, after which a consensus was reached. Second, overarching themes were extracted from the codes by YJ and RE.

Ethical Considerations.

The Medical Research Ethical Committee of the Utrecht University Medical Centre confirmed in January 2018 that this study did not fall under the scope of the Medical Research Involving Human Subjects Acts. The study is subject to the European General Data Protection Regulation. Study participants were informed that they gave informed consent for their anonymized data to be used for research through completing the questionnaire.

Results

Demographic Characteristics and Response Rate

In 2018, a total of 92 healthcare professionals completed the first questionnaire, representing eleven out of twelve (91.7%) cardiac surgery centres. Seventeen (18.3%) healthcare professionals were responsible for multiple roles in enabling and providing the PREDOCS consultation in their hospital: thirteen professionals fulfilled two roles and four professionals fulfilled three roles. This resulted in a final sample of 113 roles, yielding a response rate of 67.7%. The type and number of items in the questionnaire varied by role. These seventeen professionals received a questionnaire that integrated multiple roles and as such counted on each role they responded to. E.g. if one professional fulfilled three roles, this profession counted for three. Six centres worked with two project leaders and six centres with one, giving eighteen project leaders. To reach a reliable and complete insight into their current behaviour, in this first measurement, we chose to include all eighteen project leaders and all 39 PREDOCS nurses, with a range of 1 to 7 PREDOCS nurses per hospital (Table 1).

Table 1 Demographic characteristics and response
Full size table

In 2019, a total of 73 healthcare professionals completed the second questionnaire, representing 10 out of twelve (83.3%) cardiac surgery centres. Ten professionals fulfilled multiple roles: nine professionals fulfilled two roles and one professional fulfilled thee roles. This resulted in a final sample of 84 roles, yielding a response rate of 63.2% in which twelve project leaders and 33 PREDOCS nurses were involved. The role of the project leader was often fulfilled by several people at the start of the implementation. Over time, the role was performed by one person, which partly explains the fewer respondents in 2019 (Table 1).

In 2018, the response rate for PREDOCS nurses was 82.1% (n = 32/39) and for project leaders 77.8% (n = 14/18). Cardiothoracic surgeons responded the least to the questionnaire with a response rate of 33.3% (n = 4/12).

In 2019, the response rate for PREDOCS nurses was 75.8% (25/33) and for project leaders 91.7% (11/12). Data managers responded the least with a response rate of 13.3% (2/15). Table 1 presents demographic data of healthcare professionals and response rate per role.

COM-B

The capability, opportunity and motivation of all involved healthcare professionals in eleven cardiac surgery centres are presented in Fig. 1. Except for physical opportunity, the median score of all components was five or higher indicating that professionals agreed with the statements. In addition, all medians were equal or increased in 2019 compared to 2018. Table 2 presents participant responses to the questionnaire statements on item level. The response on all items was high. In eleven of the sixteen items, no difference was observed between the scores at both measurement moments. On five items, the median increased with 1 or 2 points including ‘triggers/reminders to perform PREDOCS well’; ‘integration into existing process’; ‘shared responsibility of colleagues’; ‘part of the work process’; and ‘tasks have become habit’.

Fig. 1
figure 1

Boxplot of Capability, Opportunity and Motivation Note. Capability, opportunity and motivation in 2018 and 2019 of all involved healthcare professionals. Four scores are low extreme values and there are 16 low potential outliers. Median scores of social opportunity and reflective motivation coincide with Q3. Scores were from 1 to 7, meaning 1 = ‘strongly disagree’; 7 = ‘strongly agree’

Full size image
Table 2 Questionnaire statement responses of COM Items
Full size table

MIDI

Determinants of innovation were identified using eight MIDI items (Appendix D). Appendix E presents participants’ responses to three questionnaire statements: a) To what extent are you aware of the contents of PREDOCS?, b) Are there formal management agreements in your organization about the use of PREDOCS? and c) Besides the introduction of PREDOCS, are there any other changes that you are currently experiencing or expect to experience in the foreseeable future?

Figure 2 presents the remaining five MIDI items. Median scores were calculated on the following questions: 1) Measures have been taken in my organization so that employees who use PREDOCS and who leave the organization are replaced in a timely manner by new employees who were sufficiently trained in the PREDOCS program. 2) There is currently a policy of having sufficient staff in our organization to use the PREDOCS program as intended. 3) PREDOCS is in line with how I am used to working. 4) I consider my tasks within PREDOCS to be part of my professional position. 5) I find the effects of PREDOCS clearly visible.

Fig. 2
figure 2

Boxplot of MIDI Items Note. Median scores on MIDI items in 2018 and 2019 of all involved healthcare professionals. Three scores are low extreme values, there are seven low potential outliers and one high potential outlier. Scores were from 1 to 7, meaning 1 = ‘strongly disagree’; 7 = ‘strongly agree’

Full size image

Themes Derived from the open Fields

In addition to COM-B items and MIDI items, themes derived from the responses in the open fields are listed in Table 3. In 2018, the majority of healthcare professionals reported insufficient time and space to carry out a PREDOCS consultation. In addition, the degree of being informed among nurses in the nursing department differed, with the majority indicating that nurses are insufficiently informed about the PREDOCS consultation. In contrast, the nurses who provide PREDOCS are motivated and have a positive experience with the consultation:

In my opinion, the effects of PREDOCS within our institution are not sufficiently visible. The measurements we now take on days 1 and 3 contain too much missing data to provide a good insight into the effects of PREDOCS. I have not heard from nurses and team leaders that they now experience fewer complications. The past year has been a turbulent year due to increasing waiting lists, a lot of changes in assistants within the cardiothoracic surgery and high variability in patient planning, which means that I feel we have not been able to implement PREDOCS optimally. I consider its importance to be very important, but the time to actually improve this is often too little. (PREDOCS Nurse 1)

Table 3 Themes derived from the open fields
Full size table

Furthermore, nurses indicated in 2018 that there is a desire to optimize conversation techniques. The degree of being informed among professionals varies, as does a clear division of tasks. Healthcare professionals indicate that they receive sufficient support in the implementation of PREDOCS. “All PREDOCS nurses are very motivated. There is regular evaluation of the progress of the project. Patients find it very pleasant. This year, the PREDOCS nurses are following a conversation technique course” (PREDOCS nurse 2). “I have been involved in the introduction of PREDOCS from the start. I am well informed and enjoy doing it” (PREDOCS nurse 3).

Staff shortages were reported as an obstacle to implementing PREDOCS in both 2018 and 2019. “Due to the frequent changes of colleagues in the team, daily care is almost impossible due to a lack of time, which means that I am sometimes unable to carry out my tasks regarding PREDOCS” (PREDOCS nurse 2). The effect of PREDOCS was not sufficiently visible and there were many organizational changes, such as a new hospital information system and reorganization. In 2019, healthcare professionals report sufficient knowledge and perseverance for implementation, they report a shared responsibility among colleagues, and positive reactions from patients, but also a lot of work and a lot of overlap with other information systems. In addition, they indicate that tasks are not always performed by the right people, causing permanent integration of PREDOCS to fail:

For some of the data to be entered, I am completely dependent on the nurses’ paper registration. Due to a lack of time, data was not entered for all patients. I understand this completely. I cannot retrieve the missing data via the EPD. I regret that I cannot provide all the data. (Data manager)

Discussion

We conducted a multicentre survey study in which we explored the change in behaviour in terms of capability, opportunity and motivation of all healthcare professionals in their various roles, involved in the implementation of the complex nursing intervention PREDOCS in twelve cardiac surgery centres in the Netherlands. Although we hypothesized that the behaviour of the professionals changes over time due to the use of the locally tailored implementation strategies, we observed high scores at both moments in time with no significant difference in five COM-B items: psychological capability, physical and social opportunity and reflective and automatic motivation. We observed that professionals were generally well informed about the PREDOCS consultation and that it fits within their current work and profession. Staff shortages, staff turnover and insufficient insight into effect were reported as reasons that hinder further implementation.

We expected to observe significant differences in the quantitative data between the two measurement moments because locally tailored implementation strategies were developed and implemented based on the analysis of the first measurement to facilitate the implementation of the PREDOCS consultation. However, we observed high scores in both measurements in which further improvement was not probable. Despite the non-significant changes over time in the quantitative data, the qualitative data reported in open fields showed differences in the themes (see Table 3). For example, at the start of the implementation, mainly practical barriers and attitudes of stakeholders concerning the new intervention were reported. However, during the follow-up in 2019, themes related to embedding the intervention in daily practice were reported.

Some of the reported themes were consistent with other literature. Kormelinck and colleagues (2020) report the easiness of applying the intervention in practice, sufficient resources, educated staff and culture. Parker and colleagues (2020) report the need for ongoing education and support and the need to consider how the workload associated with the implementation is managed and reduced as the practice change is sustained and embedded as normal practice. Such inconsistencies can be explained by the fact that themes and issues change in the various phases of implementation (Couturier et al., 2018). A further explanation for the differences in the themes could be that our population was a dynamic cohort. Within the context of a high nursing staff turnover in cardiac surgery departments in that period, new professionals may have been questioned in the 2019 measurement. This could also be a cause for the fact that we see no difference in the quantitative data between the two measurement moments.

Strengths and Limitations

To appreciate our results, some aspects need to be discussed. First, we found conflicting results between the quantitative and the qualitative data on issues of ‘staff deployment’ and ‘visibility of effect’. In open fields, staff turnover and staff shortages were mentioned several times as barriers to the implementation of PREDOCS. In contrast, we found a two-point increase in the median on MIDI items ‘measures on replacement of new employees and training’ and ‘policy on sufficient staff’ (Fig. 2). It seems that at the organizational level, there is sufficient policy on the replacement and training of personnel, but professionals nevertheless report a shortage of personnel. We also observe this contrast in ‘visibility of effect’. Involved professionals scored on the one hand on the MIDI item ‘effect clearly visible’ a median score of 4 in 2018 and 5 in 2019 (Fig. 2). On the other hand, in the open text fields, several professionals indicated that they had insufficient insight into effect, in both 2018 and 2019.

These contrasting findings may be explained by how data was collected and analysed. Professionals received a personalized questionnaire, adapted to their activities and role or roles in the implementation. We analysed professional groups without distinguishing between centers. Therefore, we cannot attribute the contrasting findings to differences among centers.

Second, we found contradictory results between the items ‘visibility of effect of the intervention’ and ‘reflective motivation’. Providing feedback on the outcome of behaviour and thus providing insight into the effect of the intervention is related to reflective motivation (Michie et al., 2014). On the one hand, professionals in the open field indicated that they had insufficient insight into the effect of PREDOCS. Nevertheless, they reported a high score on “reflective motivation” (Table 2). They could see for themselves the benefits for patients of the PREDOCS consultation. This may be related to the fact that in our larger implementation study staff were informed about the improvement in patient outcomes during the learning community meetings. Despite insufficient insight into the direct effect of the intervention on patient outcomes, reflective motivation scored high in both 2018 and 2019.

Third, in our study, we used self-reported questionnaires and we observed high quantitative scores on several items. The likelihood of socially desirable responses cannot be ruled out (Paulhus, 2017). Possibly there is some overestimation of one’s capability and motivation. The aspect of ‘opportunity’ for performing the task necessary for executing the PREDOCS consults lies less in one’s sphere of influence, which could lead to problems that are then reported more often, e.g. availability of sufficient time and space and integration in the hospital information system.

Fourth, in this study, we analysed the levels of capability, opportunity and motivation of involved healthcare professionals and determinants of innovation at the individual level. We limited our data collection to the COM-B-QP questionnaire, which includes the COM-B items, MIDI items and the open fields. We did not explicitly include measurement of the context of each participating hospital, even though insight into context is essential in gaining insight into implementation success (Damschroder et al., 2019). In our larger implementation study, we continuously adapted implementation strategies to the context of each participating hospital. We may not have been able to measure and explain some notable contextual results.

Fifth, several professionals fulfilled multiple roles, for example, project leader, PREDOCS nurse and ward nurse at the same time. Professionals who were passionate about the project took often multiple roles. Therefore, we used tailor-made questionnaires, so professionals fulfilling multiple roles completed one questionnaire in which multiple roles were included. In our analysis, however, we included all roles separately, even if those multiple roles were fulfilled by only one person. The consequence is that the results must be seen at the level of the roles each professional fulfilled and not at the level of each professional. This should be taken into account when interpreting our results.

A strength of this study is that, in the context of the larger implementation study, we built several measurements and feedback moments to enhance the implementation strategy in each local context. We used the results of the COM-B-QP, which included COM-B and MIDI items, to tailor the implementation strategy to each local context. In addition, we distributed the questionnaires at two points in time. Literature shows that the COM-B questionnaire is often deployed at one moment. Mainly to gain insight into the current behaviour of, for example, professionals or patients, on which interventions can be used to bring about behavioural change (Ellis et al., 2019). We found no other studies concerning measuring behaviour at different points in time. Further research is needed to explore how COM-B can be used over time and to what extent this contributes to the successful implementation of complex interventions.

Conclusion

In this study, we found no significant change in behaviour in terms of capability, opportunity and motivation of healthcare professionals involved in the implementation of the complex nursing intervention PREDOCS. Involved professionals were generally well-informed about the complex intervention and it did fit within their current work and profession. It is unclear to what extent staff shortages, staff turnover and insufficient insight into effect influenced behaviour. Given the complexity of implementation, repeated monitoring of behaviour seems useful, so that continuous adjustment of the implementation strategy can take place.