A theoretical sample,23 i.e. criteria was defined to ensure that contexts and profiles that could provide information which is different and relevant to the study's objectives are included, was selected through a two-stage process. In the first stage, the contexts healthcare networks were selected to represent the diversity of management models in Catalonia: Baix Empordà, the city of Girona and the Ciutat Vella district of Barcelona (Table 1). A single entity manages both PC and SC in Baix Empordà and in Girona. In Ciutat Vella, two public entities manage PC and a different public entity manages SC.

In the second stage, in each network PC and SC doctors who perform clinical activities and with a minimum labour linkage to the organization of a year and a half were selected. Maximum variation with regard to age and sex was sought, and for SC doctors also to speciality. The sample was selected in a sequential way, so profiles that emerged as relevant in initial interviews were also included in the study. No contacted doctor declined or showed reluctance to participate in the study. The final sample size was between 15 and 18 doctors per network (Table 2).

Individual, semi-structured interviews were carried out using a topic guide adapted from previous studies.17,24 The topic guide addressed doctors’ opinions on clinical coordination between PC and SC and factors influencing it through open questions. Interviews were conducted by the first author, an anthropologist/pharmacist with a good knowledge of qualitative methods, the research topic and the context, who worked in close collaboration with the second and last authors. The interviews lasted between 45 and 80minutes and were recorded and transcribed.

Doctors were contacted by telephone or email either by a contact in the health organization or by the researcher. Interviews were performed in the healthcare facilities. Data collection stopped when saturation was reached in each study network. Fieldwork took place between July and September 2012 (Baix Empordà) and between December 2013 and May 2014 (Ciutat Vella and Girona).

A thematic content analysis25 was conducted with the support of the Atlas-ti software. Data were segmented by healthcare network and level of care. The process of category generation was mainly inductive, i.e. it was oriented towards the identification of emergent patterns in the data.23 Themes were identified, coded, re-coded and classified, identifying common patterns by looking at regularities, convergences and divergences in data, through a process of constant comparisons, going back and forth between data. Data quality was ensured through triangulation, by comparing different networks, groups of informants, and sources (literature).

The first author was responsible for the analysis, and worked in close collaboration with the second (a health economist) and last authors (public health doctor). Differences were discussed and resolved by going back to the data. The rest of the authors contributed to the interpretation of data.

This study was approved by the ethics committee of Parc de Salut Mar (2010/412/1). Written consent was obtained from every participant prior to interview. Confidentiality and anonymity were guaranteed.

In all three healthcare networks, doctors considered that care provided to patients was generally coordinated across that levels, and described it in terms of the availability and uptake of clinical information generated at the other care level through the shared EMR, the resolution of doubts regarding diagnosis and treatment through direct communication, and rapid access to SC following a referral in urgent cases. However, certain limits to clinical coordination emerged, which varied depending on the network and care level.

According to the informants, the availability and uptake of clinical data generated at the other level on the EMR meant that tests and treatments are not duplicated or contraindicated (Table 3 a). This information also facilitated an adequate follow up of the patient when they pass from one level to another, for example following a referral. However, doctors pointed out that some doctors from the other care level failed to transfer all the information needed in patient referral (reason for referral) and reply letters to PC (diagnoses or recommendations for follow-up) (Table 3 b).

Doctors considered that direct communication, through virtual consultations via EMR or email, telephone, and joint clinical case conferences encouraged greater consistency in medical instructions, prevented unnecessary patient referrals and helped to speed up the diagnostic process and treatment (Table 3 c). Likewise, difficulties in communicating with certain SC doctors, translated into delays in diagnosis and treatment (Table 3 d). Disagreements regarding prescriptions also emerged, which were described as frequent changes in prescriptions from the other care level, creating confusion in patients and conflict between doctors of the two levels (Table 3 e).

The rapidity of access to SC when the reason for referral was urgent, was considered to contributes to a timely diagnosis and treatment (Table 3 f). However, PC doctors and some SC doctors considered that the waiting times for non-urgent SC and, in Girona, for hospital tests, were long, which caused delays in diagnosis and treatment and sometimes forced the patients to seek help for their problems in inappropriate places, such as hospital emergency services or PC (Table 3 g-h). In this sense, in Girona and Ciutat Vella, PC doctors considered that the patient did not always receive care in the most appropriate place due to being prematurely discharged from hospital or emergency services (Table 3 i) and, in Girona, due to the rejection of certain referrals, a factor which they also associated with delays in diagnosis and treatment (Table 3 j). However, SC doctors considered referral rejections a consequence of unnecessary referrals (Table 3 k).

In the informants’ discourse, various types of factors related to the organization and professionals emerged that influenced clinical coordination across care levels. The majority of these factors emerged in all networks, although with some differences in organizational factors. Differences were also observed depending on the care level, mainly in factors related to professionals.

The existence of coordination mechanisms between care levels emerged as one of the main facilitators of clinical coordination between levels. They identified mechanisms that facilitate access to the information generated at the other level (shared EMR) or enable problem-solving communication and agreement on clinical approaches (clinical case conferences between PC and SC doctors [Girona], virtual consultations via EMR [Ciutat Vella and Baix Empordà] or email, and telephone [Table 4 a-b]). Joint clinical case conferences also contribute to improve mutual knowledge, both in terms of the skills and roles of the other level, fostering a more positive attitude towards collaboration. In addition, they identified the rapid diagnosis pathway for suspected cancer to guarantee rapid access to urgent SC (Table 4 c).

Lastly, the lack of shared clinical criteria for prescription and incentives only for PC doctors emerged as a barrier, as it makes it difficult to reach an agreement on a treatment plan for the patient (Table 4 d).

Physical proximity to each other was highlighted by doctors from both care levels as a factor that facilitates clinical coordination. Various organizational elements, which differed according to the network, ensured proximity: the co-location of SC doctors in PC centres in Ciutat Vella, working in a small organization in Baix Empordà and in centres that are close to each other in Baix Empordà and Girona. Physical proximity increases contact and mutual knowledge, fostering a more favourable attitude towards collaboration. It also facilitates informal communication and a greater use of coordination mechanisms (Table 4 e).

Insufficient time for coordination due to work overload was mentioned as the main barrier to clinical coordination by doctors from all networks. Doctors highlighted having insufficient time to use coordination mechanisms to communicate and to participate in joint meetings (Table 4 f). Furthermore, PC doctors pointed out that inappropriate referrals were a result of the insufficient consultation time per patient.

Changes in the organizational model in the context of the economic crisis emerged from the discourse of PC doctors and, with less intensity, in that of the SC doctors, in Ciutat Vella and Girona. They attributed these changes to cuts in the healthcare budget, which translated into a cut in resources (reducing the number of hospital beds). They considered that coordination mechanisms which present an alternative to conventional referrals, such as virtual consultations, were implemented with the purpose to reducing the number of patients who access SC. In addition, PC doctors perceived that hospital stays and emergency admissions have been curtailed (early discharges) and the volume and complexity of patients treated in PC had increased. Moreover, sometimes the patient was referred back to PC without the problem having been resolved (Table 4 g).

Insufficient training and clinical skills of PC doctors was related to clinically inappropriate referrals by some SC doctors, especially those who are not in regular contact with PC doctors (Table 4 h). However, according to PC doctors and other SC doctors, this perception is built on SC doctors prejudices and their scant knowledge of the function and the resources available at PC.

Attitudes of doctors towards coordination emerged as a factor which influences the use of coordination mechanisms: specifically, telephone and email to communicate, the EMR to share medical data and their participation in joint meetings (Table 4 i). It also influences their willingness to perform tasks aimed at making the other care level more effective, such as requesting tests so that the SC doctor can see the results on the patient's first visit (Table 4 j). The informants considered that these values are conditioned by mutual knowledge and contact between professionals, which foster a more positive attitude towards coordination.

In general, doctors considered that care provided to patients is coordinated across PC and SC in all three networks, with some limitations and their perception of clinical coordination was similar across networks, although some differences were identified between networks and professionals. In two networks, Ciutat Vella and Girona, problems in the adequacy of patient care transitions in terms of early discharges and the rejection of referrals were described, mainly by PC doctors. However, these problems were attributed to changes introduced in the context of the economic crisis, thus, this difference could be due to the fact that the field work was carried out at different times: at the start of the economic crisis in Baix Empordà and over a year later in the other two networks, when the main reductions in public health spending budget took place.

Doctors of all three networks also identify similar organizational factors; most importantly, physical proximity between doctors and the implementation of similar coordination mechanisms.

Among organizational factors highlighted by informants were the implementation of coordination mechanisms to foster information transfer and problem-solving communication. In addition, it became evident that physical proximity was important, since it promotes mutual knowledge and trust among doctors that encourage direct communication and foster a better attitude towards coordination. A positive attitude towards coordination emerges in this and other studies12,26 as one of the most relevant factors influencing collaboration among professionals, as it contributes to the better use of coordination mechanisms and makes them more willing to perform tasks destined to make care more effective at the other level.

On the other hand, the results reveal various factors, which hinder coordination across care levels. Most importantly, lack of time for coordination activities emerges as the main organizational obstacle to clinical coordination, also pointed out in the scientific literature.10,27,28. This problem seems to have become more severe as a result of the change in the organizational model, aimed at reducing the activity of SC, and change that has gone hand in hand with a higher level of referral rejections and early discharges.

The discourse highlights incentives in drug prescription, which are heavily weighted towards the PC. These incentives, along with a lack of shared prescription guidelines or a lack of awareness, lead to frequent changes in treatments. This problem points to the need to implement a more global focus involving both care levels by making doctors of both levels equally responsible for prescribing drugs to the population they serve.29,30

Finally, the three networks form part of a national health system, and thus share many of the contextual factors that can influence coordination, such as the same healthcare model; almost universal coverage and the same type of financing and incentive systems. This might be explaining the coincidence in discourses.

Despite the heterogeneity of the population interviewed in terms of certain characteristics, such as medical speciality and type of centre, we cannot rule out the possibility that information saturation was not fully reached, although the main arguments are represented in the results. Furthermore, a contact in the organizations participated in the selection process, so they could have introduced a bias towards a more positive discourse. However, the research team also participated in the selection process, discussing the profiles of the informants, and some informants were identified in a sequential way as part of the sampling strategy. Finally, the fact that the field work in one network was performed at a different time to the other two may be the reason for the differences observed between study areas, especially those related to changes in the socioeconomic context seen in recent years.