This study aimed to gain insight into the needs and mechanism to improve the collaboration in OAT. Therefore, this study identified the bottlenecks in collaboration between health care professionals in OAT, the proposed solutions to overcome these bottlenecks, and the perceived barriers to implement these successfully. Our study revealed that professionals experienced several problems in the collaboration regarding OAT that need to be improved. The most prevalent experienced bottlenecks were lack of knowledge, lack of consensus about OAT among health professionals, and inadequate information exchange between health professionals. The proposed solutions to overcome these bottlenecks were related primarily to the improvement of professionals' decision support, mainly by education. Finally, several attitudinal and behavioural barriers to improve collaboration for OAT were identified, most frequently mentioned were lack of non-AC professionals motivation and lack of time.
As far as we are aware, our qualitative study is the first that explicitly identified the experienced bottlenecks in collaboration in daily care practice for OAT. In line with previous studies on chronic care, our results reveal that collaboration needs to be improved to manage the OAT successfully [15
]. This study showed that the three identified bottlenecks in collaboration, i.e. lack of knowledge, lack of consensus, and lack of information exchange were experienced to affect the quality of care negatively. The quality of care was seriously affected since these bottlenecks resulted more than once in the use of a suboptimal dosage of oral anticoagulants, inefficient time-consuming gathering of patient information, and patients being confused due to conflicting dosage advices from different professionals.
The proposed solutions given by the interviewees are in line with CCM's suggestions for improvement, such as multidisciplinary meetings and quality management [18
]. Especially decision support elements (i.e. education, meetings, and agreements and protocols), were proposed to improve the collaboration. Remarkably, a discrepancy between experiences and expectations revealed regarding the proposed solutions. Formal approaches like protocols and regular meetings were frequently proposed as solution by our interviewees as well as in the CCM literature, while in ACs where the formal approaches were used, these formal approaches were scarcely experienced to improve the collaboration. In addition, solutions regarding the health care organization were frequently proposed, but only if the professional experienced that their organization influenced the collaboration positively. Hence, the potential of the health care organization to improve collaboration seemed to be underestimated by professionals. Professionals are not inclined to reform their own health care organization to improve the collaboration. These discrepancies show the importance to exchange experiences between professionals prior to implementing proposed solutions.
Moreover, AC professionals considered education of other professionals a prerequisite for a successful implementation of other proposed solutions. This is in line with the implementation sequence of other complex interventions [14
]. As a consequence, although CCM components are supposed to be more effective if they are applied as comprehensive interventions [11
], we argue that these interventions can be even more effective if the sequence of implementation of these components is taken into account. This is in line with other models of behavioural change that identified knowledge as a prerequisite for acceptance and attitude change [20
]. Taking the sequence of implementation into account will result in a reduction of the number of both priorities and changes at the same time which has also been shown to improve the implementation and thereby the effectiveness of the interventions [19
Education for other professionals as well as the implementation of other proposed solutions were also hindered by attitudinal and environmental constraints as is in line with previous studies [14
]. Our results revealed that the attitudinal constraints could mainly be related to a lack of motivation in non-AC professionals. Their lack of motivation to improve the collaboration was mentioned as a cause for partial or slow implementation of all proposed solutions. Although the lack of motivation could be due to factors unknown by AC professionals, it is related to a lack of alignment regarding the collaboration between AC professionals and non-AC professionals. Especially alignment regarding chronic care management is needed for successful collaboration [10
]. Alignment of all involved professionals should also encounter the lack of status which can be interpreted as lack of trust. Trust is needed as it is identified as one of the fundaments to collaborate [9
However, reaching alignment is probably also hampered by the lack of status and knowledge. Based on our results it seems likely that easily approachable informal contacts could improve ACs' status and professionals' knowledge as was also suggested by others [9
]. Furthermore, our results showed that a lack of time and money hampered the improvement of collaboration, since the professionals are focused on specific tasks (e.g. measuring INR, giving dosage advice) instead of the entire chronic care process. This is in line with the financial constraints that were experienced by professionals in improving the collaboration for other chronic diseases [19
Besides general conclusions regarding the needs and barriers to improve the collaboration in OAT in the Netherlands, this study elicited regional differences that should accounted for. First, regions can inspire each other by their differences. For instance, a few ACs were hampered to improve the collaboration because general practitioners were not united. One of the ACs did overcome this barrier by involving the out-of-hours primary care centers in OAT in which all general practitioners of the region are unified. Second, solutions should be adapted to a regional level since complex interventions work best if they are tailored to local contexts rather than completely standardized [26
]. Third, our results showed that more discussion and knowledge is needed regarding the variation in treatment protocols between regions, disciplines, and hospitals to improve the quality of care.
The strength of this study is that we systematically identified professionals' experiences with improving collaboration in OAT, since previous research emphasized that the implementation process should be understood to develop an effective and sustainable intervention [4
]. However, several limitations of this study should be noticed. First, we only included the perspective of AC professionals because they are the main care providers playing a pivotal role in OAT in the Netherlands. Since regional differences are prevalent we decided to interview AC professionals from different regions instead of interviewing all actors. This could well be the next focus of research as it is of interest to identify the factors that can explain non-AC professionals' lack of motivation that was experienced by AC professionals.
Second, we used Cabana's framework to identify barriers to behavioural change by health care professionals. However, other models could also be utilized and it is up for discussion what model would be best [22
]. Since our data were first inductively analyzed and only subsequently categorized using Cabana's framework, the framework did not restrict our analysis compared to other models. Based on our inductive analysis, we added a barrier, lack of status, to Cabana's framework. Third, we used the CCM as theoretical framework since CCM components enable to structure the proposed solutions at practice level, while taking the multidimensional character of collaboration into account. Other collaboration models such as the model of D'Amour could be of additional value in further research when the extent of collaboration needs to be measured and further detailed analysis is preferred. Finally, we only interviewed professionals of 30 percent of the Dutch ACs. Nevertheless, data saturation seemed to be achieved since we did not achieve new insights in the lasts interviews.