Examining the effect of the individual characteristics of implementers and the interaction of multiple relationships on the structure of psychosocial intervention teams

RESEARCH Open Access Examining the effect of the individual characteristics of implementers and the interaction of multiple relationships on the structure of psychosocial intervention teams Ignacio Ramos-Vidal 1,3* , Jorge Palacio 2 , Ilse Villamil 3 and Alicia Uribe 3 Abstract Background: Teams’structure may undergo modifications due to the individual attributes of actors and collectivelevel variables. This research aims to understand the effect of extensive experience working in the program and the simultaneous interaction among different relationships in the network structure of a team of implementers. The Psychosocial Care Program for Victims of Conflict is implemented by psychologists, social workers, and community advocates. Methods: A cross-sectional study was carried out. Multivariate analysis, quadratic assignment procedures, and graphic visualization are used to (a) determine how seniority affects the professionals’level of centrality in the program and (b) clarify how the interaction among professionals favors new relationships. Results: Longer-lasting professionals in the program report stronger network bonding, predisposition to work, and information exchange. The nonparametric permutation test indicates an intense association between the information requests submitted and received and between the predisposition to work network and the network of received information requests. The results are discussed to optimize the teams implementing the intervention programs. Conclusions: Network analysis is a powerfull tool to evaluate program implementation processes. Analyzing the interactions among multiples relationships that emerge between members of multidisciplinary teams allows knowing how certain relationships (e.g., information exchange) triggering other kind of relationships (e.g., users referral). The implementers who have been collaborating in the program for a long time are key informants who can facilitate the process of adaptation of newly incorporated professionals. Keywords: Implementation, Information exchange, Mental health, Network analysis, Psychosocial intervention, War victims © The Author(s). 2020 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. * Correspondence: [email protected] 1 Department of Social Psychology, University of Seville, Seville, Spain 3 Research Group CAVIDA, Universidad Pontificia Bolivariana, Montería, Colombia Full list of author information is available at the end of the article Ramos-Vidal et al. Implementation Science (2020) 15:69 https://doi.org/10.1186/s13012-020-01032-9 Introduction According to data of the Single Registry of Victims (SRV) launched by the Colombian government, since 1985, the armed conflict has left almost eight million victims of different criminal acts [1]. Throughout the country, eight million forced displacements, 984,408 homicides, 328,380 threatened people, 92,946 victims of terrorist acts, 34,404 abductions, and more than 10,000 anti-personnel mine victims have been reported. These numbers mean that at least 16.5% of the country’s population has suffered from the effects of the conflict. In Córdoba, where the present study was conducted, there are 400,000 victims. This number reflects that more than 20% of the local population has been a victim of war. Empirical evidence suggested that people that have suffered episodes of violence (kidnapping, torture, massacres, or forced displacement) exhibit mental health problems with a prevalence ranging between 1.5 and 32.9% [2]. According to the results derived from a systematic review [3], the main mental health problems identified in populations affected by wartime violence are post-traumatic stress disorder (9%), major depressive disorder (5%), and generalized anxiety disorder (4%). Although the Colombian Government has developed several initiatives to improve the mental health of war victims, these efforts seem insufficient due to dimensions of the problem and the low level of adjustment of the interventions regarding the population needs. The magnitude of these numbers emphasizes the need to implement effective intervention programs to impact the psychosocial well-being of the population. This research examines different types of relationships among professionals who implement a program that provides psychosocial care to victims of war in Colombia. In 2013, the Ministry of Health and Social Protection of Colombia launched a program known as the PAPSIVI (Comprehensive Program of Psychosocial Care and Health for Victims of Conflict) to meet the psychosocial care demands of the victims. In departments such as Córdoba, governance is decentralized, with the Health Development Secretariat being the entity in charge of implementing this program. The decision to explore the implementation process of this specific program was adopted due to (a) the nation-wide coverage of the program, (b) the multilevel nature of the intervention, and (c) the lack of empirical evidence regarding to the intervention effectiveness. To benefit from the PAPSIVI, users must be previously registered in the SRV. In 2017, the PAPSIVI attended to 8803 of the 9780 targeted victims for that year in Córdoba. This figure represents almost 90% coverage. The professionals implementing the program in the municipalities (in the Colombian context, a municipality is a geographical demarcation that may include small and medium-size cities as well as rural communities) are composed of psychologists (n= 26), social workers (n= 26), community advocates (n= 13), a liaison nurse, a physician, and six (n= 6) people who constitute the administrative staff, altogether providing psychosocial care to individuals, families, and the community. The interventions are provided by teams consisting of psychologists, social workers, and a community advocate who serve between one and three municipalities, depending on the number of victims. The rest of professionals (liaison nurse, physician, and administrative staff) develop other task related with the program but not centered in provide healthcare itself for this reason they are not included within intervention teams. The liaison nurse and the physician are dedicated to evaluate particular cases of users that present special needs (for example severe mental illness) that requires be referred to mental health services. The administrative staff receipts the documentation of the users and solves questions associated with the institutional requirements to participate in the program. In healthcare settings, the intervention process is usually delivered by groups of professionals who work in a collaborative fashion to improve the service quality [4]. The literature shows the importance of work teams for providing high-quality care to users and patients [5]. There are several types of work teams depending on their structure, composition, functions, and task distribution [6]. Cross-functional, self-directed, and multidisciplinary are some of the many forms teams can adopt in organizational environments [7]. A cross-functional team is a group of professionals with different background and experiences which working toward a common objective; self-directed teams are conceptualized as a set of individuals who share the responsibility to develop specific tasks and work with low level of Contributions to the literature Empirical evidence supports that implementation process needs to be analyzed from a structural perspective, by examining multiples kinds of interactions among the implementers. Seniority of implementers within the program is a key variable to understand the positions holding by the professionals in multiple networks. Longer-lasting professionals are identified as key informants by newcomers and may develop the role of promoters of program changes. Relationships that connect the professionals in the implementation process are interwitned, and the information exchange has the potential of triggering other relationships (e.g., users referral) wich are essentials for service delivery. Ramos-Vidal et al. Implementation Science (2020) 15:69 Page 2 of 13 supervision; and multidisciplinary team is a group of people, with different academic backgrounds and professional experiences, who work together for achieving a common goal [8]. In the specific context of the program examined in this work, teams operate as multidisciplinary team care (MTC) in which professionals from a range of discipline (in this case psychologists, social workers, and community facilitators) work together to deliver comprehensive care that addresses as many of the patient’s needs as possible [9]. The users of the PAPSIVI exhibit psychosocial problems, difficulties of adjustment to the community settings they inhabit, and at collective level, those contexts presents several barriers to overcome vulnerability conditions. Considering these factors, multidisciplinary team care is considered an adequate design to solve the variety of demands affecting victims attended by this initiative. Psychologists intervene at the individual level and to a lesser extent also perform family intervention. They focus on the initial diagnosis of the user and on designing an individual work plan. The intervention process consists of the following phases: (a) contacting, (b) diagnosis to design the individual action plan, (c) the development of the action plan following the PAPSIVI guidelines, and (d) a final closing session. Each intervention consists of six to eight sessions, and the number and intensity of sessions may vary, depending on each case (https://www.minsalud.gov.co/sites/rid/Lists/BibliotecaDigital/RIDE/DE/PS/Protocolo-de-atencion-integralen-salud-papsivi.pdf). Social workers intervene at the community level. They perform collective work-based therapies with small groups and mutual aid groups. Such activities are designed so that (a) the community can confer sense to the victimizing act and reduce the sense of guilty that many of them experience, (b) promote a sense of identity with the community, and (c) increase community cohesion and strengthen social capital. Advocates (a) identify victims in communities, (b) channel them to the program, and (c) assume accompanying roles for the rest of the professionals. Community advocates reside in the municipalities where they render their services and are victims themselves. This aspect facilitates access to the population and increases the ecological validity of the intervention [10]. Once the implementers are assigned to a team, and the team is assigned to attend to a community, in the first stage the community advocate is the first actor to access the community, usually a few days before the arrival of the rest of the team members. The objective of this first stage is to identify the users and maintain contact with community leaders to explain the activities the rest of implementers will carry out in the next days. Within the multidisciplinary care team, one professional (psychologist or social worker) act as team coordinator. The main functions of the coordinator is to (a) organize the visits to the users, (b) evaluate which cases requires combined attention with more than one professional and, at the end of the visit to the community (b) receipt information about incidences during the intervention process. If within the team there is a professional newly incorporated into the program, it is assigned to the community promoter who acts as a mentor during the first days working in the communities. In the Department of Córdoba, there are a number of factors that influence the coordination between the professionals implementing the program and that may affect the effectiveness indicators. First, the heterogenous composition of the teams is important to consider. Each team is composed of professionals with different technical backgrounds, which makes it necessary to pay special attention to coordination [11]. Second, the high turnover of personnel participating in the program makes the average time of participation in the teams less than a year, which may be affecting the results of the intervention [12]. Third, the geographic dispersion of teams that render attention to distant municipalities makes coordination among professionals extremely important for obtaining positive results [13]. Program success depends on the proper coordination among the implementers [14]. Recent studies highlight the importance to deeply understand the interaction patterns connecting the implementers. The relational dynamic and the networks structure that support program development determine various implementation outcomes such as acceptability, appropriateness, adoption, feasibility, and fidelity [15]. Effective coordination requires that the professionals (a) exchange relevant information regarding their services, (b) refer users to other professionals in terms of demand, and (c) participate in meetings and activities to ensure coordination [16]. The lack of coordination among service providers is related to (a) poor intervention results, (b) poor use of resources, (c) duplication of efforts, and (d) low satisfaction levels of users [17]. Collaboration, information sharing, and the referral of patients among professionals are key elements in identifying and systematizing good intervention practices [18]. Intervention programs that regulate user referral often yield better results [19]. Under this logic, user referral and information exchange optimize the implementation of programs, and from an organizational perspective, these practices benefit the institutions that coordinate service provision [20]. Previous studies have suggested that user referral reduces the cost of the service provided and improves the quality of diagnosis and treatment [21]. The adoption of evidence-based clinical practices by healthcare professionals is affected by the ties maintained by implementers. Palinkas and colleagues note Ramos-Vidal et al. Implementation Science (2020) 15:69 Page 3 of 13 that practitioners who actively exchange information and professional advice are more likely to adopt good intervention practices [22]. Individual and organizational factors influence information exchange and user referral. Individual factors describe the characteristics of professionals that encourage or inhibit interactions with other professionals. Organizational factors refer to group variables that affect the relations among professionals. Among the individual factors are the seniority of the professional in the program, the professional profile, and the empowerment level in the workplace. Some studies suggested that seniority of implementers in the program is a key factor to understand the role of professionals in the implementation process. More experienced professionals develop deep understanding of the implementation process, serves as key informants of other team members, and increase the adherence of users to the program activities [23]. Some of the organizational variables that affect the coordination and referral of users are organizational size, the level of decentralization in decision making, and geographic dispersion [14]. It is essential to analyze the structure of interactions among implementers because some relationships encourage new interactions [24]. Therefore, analysis of the network in which professionals are integrated may be considered an interesting phenomenon from the organizational perspective. Implementers request information from other colleagues with whom they have previously exchanged information or referred patients [25]. To discern how multiple relationships interact, it is necessary to know the background of the contacts that link the professionals who implement intervention programs. Doing so will make it possible to identify the relationships whose activation is a priority for applying structural changes with a positive impact on the program outcome. It is also necessary to identify the individual variables with potential to explain the role and position of the implementers in networks of information exchange and user referral. By examining the individual and relational factors that influence information exchange and user referral, it is possible to understand the governing logic of the program’s implementation to increase the effectiveness of the intervention. Objectives of the current study The general objective is to determine how individual and relational factors influence the structure of different relationships among PAPSIVI implementers in the Department of Córdoba. The specific objectives are as follows: Objective 1. To analyze the structural features of six types of networks in which PAPSIVI implementers participate: (a) the recognition network that examines the degree to which implementers are able to recognize other professionals by name; (b) the predisposition to work network which aims to know the level of affinity between the program implementers; (c) the network of submitted information requests; (d) the network of received information requests; (e) the network of submitted user referrals; and (f) the network of received user referrals. Objective 2. The impact of seniority (how long each professional has worked in the program) in the position held in the six evaluated networks is examined at the individual level. Objective 3. The dependence among the six networks is analyzed at the relational level, evaluating overlaps and the role of (a) recognition networks, (b) the predisposition to work, and (c) information exchange (received and submitted information requests) in the user referrals (submitted and received). Material and methods Participants A total of 49 of the 65 (78.4%) professionals who implemented PAPSIVI in Córdoba were interviewed. The majority were women (94.1%, n= 48). The participants included 18 psychologists (35.3%), 21 social workers (43.1%), and 10 community advocates (19.6%). The participants stayed in the program an average of 12 months (range = 1–47; SD = 10.7). However, this figure varied notably between each profile, with community advocates lasting the longest, with an average of more than two years implementing the program, compared to 6 months for psychologists and eight months for social workers. Design and procedure The study is cross-sectional and exploratory and the data presented here is part of a broader study designed to understand the variables associated to the implementation process that could affect the program effectiveness. The protocol of this study was revised to and approved by the Center for Research, Development and Innovation of the Universidad Pontificia Bolivariana. The project that supported this research was approved by the institutional review board of the University. All study participants provided written informed consent. The PAPSIVI coordinators in Córdoba were informed of the purpose of the study, and the research team signed a confidentiality agreement. The participants signed an informed consent form. The interviews were conducted in coordination meetings held periodically in the capital of Córdoba between September and December 2016 and lasted approximately 1 h. The information gathering process occurred as follows: (a) a member of the research team traveled to the monthly coordination meetings that bring together all PAPSIVI implementers; Ramos-Vidal et al. Implementation Science (2020) 15:69 Page 4 of 13 (b) then, the researcher presents the characteristics of the questionnaire indicating how it should be answered by giving concrete examples; (c) during the completion of the relational data, the researcher offers guidelines on how to respond to the socio-centric instrument. All the implementers were invited to participate; however, those that were displaced in remote rural communities for providing service during the coordination meetings do not participate in the study. The STROBE checklist http://www.equator-network.org/reporting-guidelines/ strobe/ for observational cross-sectional studies is included as additional file. Instruments The questionnaire includes socio-demographic variables and indicators related to the time that the professionals have been working in the program, the characteristics of the service provided, and the number of municipalities served. It also incorporates openand closed-ended questions to learn the potential advantages and disadvantages associated with coordination activities, information exchange, and user referral [16]. Social network analysis (SNA) helps to understand how certain relationships favor other interactions, showing the interdependence among different social systems composed of the same professionals [26,27]. Some proposals show the interdependence among multiple interactions that shape the structure of socio-health professional teams [28]. Reciprocity is important in referring users and sharing information, which means that cohesion measures such as homophily, transitivity, and reciprocity explain much of the structural variability of networks [29, 30]. This shows interdependence among the characteristics of the micro-local units that constitute the networks (e.g., dyads) and the global network structure [31,32]. The design of the network instrument was based in previous studies which suggests that when asking about specific exchanges within healthcare settings (such information exchange or patients referral), it is interesting to differentiate among when the exchange consisting in sending or receipting that petition [16,18,33]. To analyze the six relationships among the professionals, each professional was asked about the relationships he maintains with the other program implementers indicating whether he maintains or does not maintain a relationship in each of the six networks evaluated. In the cases of users referral network and information exchange networks (both received and submitted), implementers were encouraged to nominate only professionals with whom they exchanged information or referred users within the previous month. This decision was made in order to identify the most accurate structure of real interactions. The resulting output analyzed is a dichotomous matrix for each network. A sociocentric network design was chosen because it captures direct and indirect interactions and is the most advisable when the network composition is defined by formal limits; in this case, by being part of the team of implementers of the program [34]. Data processing The socio-demographic variables and the implication indicators in the program were analyzed with the statistical package SPSS® (Versión 24.0. Armonk, NY: IBM Corp). The adjacency matrices used were processed with UCINET 6.636 [35]. The structure and composition indicators of the networks were calculated with UCINET and subsequently processed in SPSS® for multivariate analysis. The visual representation of the graphs was performed with the NETDRAW application included in UCINET. Data analysis Multiple structural parameters were calculated to evaluate (a) the centrality parameters of each professional (Indegree and Outdegree), (b) the global network cohesion (density, indegree and outdegree centralization, and average degree), (c) the number of subgroups identified through cluster analysis using the optimization procedure proposed by Glover [36], (d) the dyadic properties (homophily according to category and seniority), and (e) triadic (transitivity) properties that affect the structural features of complete networks [29,30,32,37]. To measure the effect of the seniority of the professionals in the program in the position that they hold on the six networks evaluated, first, the sample was segmented into two groups according to the seniority level. The division of groups was established using the 50th percentile as the cut-off point, which is equivalent to 8 months working on the program. The first group (group 0) is composed of 23 professionals who, on average, have been working less than 8 months in the program. The second group (group 1) is composed of 26 professionals who have been in the program for more than 8 months. The outdegree and indegree in the six networks evaluated and the density of each actor’s egocentric network were calculated to discover—through ttest—the role of seniority in the position held by professionals in the six networks [38]. To examine the degree of overlap and the level of association between the different networks, several nonparametric permutation tests were performed through the Quadratic Assignment Procedure (QAP) at the dyadic level [27,39]. To analyze the overlap among the six networks, a correlation analysis was performed. To identify how different types of relationships influence user referral, two multiple regression models were performed at the dyadic level. The double Dekker semipartialling (DSP) technique was used because it is Ramos-Vidal et al. Implementation Science (2020) 15:69 Page 5 of 13 advisable when there are high levels of collinearity and self-correlation between variables [31]. In model 1, the dependent variable is the network of user referrals received, and the independent variables are the networks of user referrals (submitted), information requests (submitted), information requests (received), recognition of professionals, and predisposition to work. In model 2, the dependent variable is the network of user referrals submitted, and the independent variables are the networks of (a) user referrals (received), (b) information requests (submitted), (c) information requests (received), (d) recognition of professionals, and (e) predisposition to work. Results Description of the networks evaluated The first research objective is focused on analyzing the structural cohesion of the six networks evaluated. The analyzed networks vary according to global cohesion, number of components, and the micro-structural features at the dyadic and triadic levels. Table 1shows the parameters of the six networks evaluated. Except for the recognition network with a high density (above 50%), the rest of the networks have low cohesion, ranging from 1.4 to 8.5%. Examining the values of centralization, it is observed that the networks are decentralized in both the input and output nominations. The input nominations are less centralized than output nominations in all relationships. In the network for information requests (referred), this indicator shows that a quarter of professionals make information requests to other colleagues and that, simultaneously, such requests are received by the same number of members. In the network of information requests received, the difference between input and output centralization implies that 26.73% of professionals request information from other colleagues and that those requests are received by 11.84% of the network components. This datum reflects that the request for information is typically focused on a few professionals. Similar differences between the centralization of input and output in the user referral networks (received and submitted) are observed. This finding suggests that between 15 and 20% of professionals refer users to other professionals, whereas the professionals who assume such referrals do not exceed 7%. The mean nodal degree shows a similar trend across all networks. With the exception of the recognition network, which showed a high density, the rest have low transitivity that barely reaches 10%, which reflects low structural integration. This finding is shown in Fig. 1, in which the structure of the six analyzed networks is presented, identifying the clusters detected, and distinguishing the attributes of the actors by the color and size of the nodes. Therearesixto16cohesivesubsets.Theinformation request network has the most groups, showing a high degree of fragmentation. Fifteen isolated nodes have been deleted from the user referrals network to calculate the number of clusters using the Tabu Search procedure [36]. Homophily values range between −1 (pure homophily) and + 1 (pure heterophily). According to the professional profile, there is a moderate heterophilic tendency, except for the information request networks (received and submitted), in which a slight homophilic tendency is noted, with professionals of the same category exchanging information. It is possible to observe this tendency in Fig. 1(3 and 4), in which several of the clusters are formed of implementers of the same category. If this tendency is analyzed according to the seniority in the program, all of the networks exhibit a homophilic tendency, presenting moderate negative values. This finding indicates that seniority in the organization is a determining factor in all relationships explored. The role of seniority in individual positions The second research objective is to learn how seniority in the program affects professionals in the position that they hold in the different networks. The longest-lasting Table 1 Cohesion parameters of the six networks evaluated Type of relationship Network cohesion measures Density Ind. Cent. Out. Cent. Mean degree Trans. Number of clusters and model adjustment Homophily (professional category) Homophily (based on seniority) Recognition 51.17% 28.6% 49.8% 25.55 36.64% 7 (R 2 = .197) .240 −.141 Preference to work 7.3% 18.1% 24.4% 3.51 7.11% 10 (R 2 = .132) .163 −.132 Information request (submitted) 8.5% 23.2% 23.2% 4.08 7.04% 16 (R 2 = .199) −.091 −.154 Information request (received) 7.14% 11.84% 26.73% 12.41 6.09% 6 (R 2 = .074) −.056 −.116 User referrals (submitted) 1.74% 6.72% 19.48% 3.31 9.57% 7 c (R 2 = .138) .171 −.024 User referrals (received) 1.45% 7.03% 13.41% 2.81 5.47% 8 (R 2 = .182) .176 −.031 Ind.Cent. indegree centralization, Out.Cent. outdegree centralization, Trans. transitivity Ramos-Vidal et al. Implementation Science (2020) 15:69 Page 6 of 13 professionals within the program (group 1) presented more connections in the recognition and information request networks. In the rest, no significant differences were observed. To determine the effect of seniority on individual positioning in the two groups, a ttest was performed, discerning between the indegree value, the outdegree value, and the ego-centered density in the six networks. Table 2shows the results of the T test analysis in the networks evaluated. To calculate the network density of each professional (ego-network density), the ego-centered model was applied [38]. In this model, each actor is treated as an ego and the ego network is trait as if the rest of the network did not exist so that ties beyond alters have no effect. Hence only are considered alter-alter ties as originally is suggested in Burt’s seminal work [38]. The most senior professionals in PAPSIVI are those who (a) know more colleagues, (b) have more nominations in the predisposition to work network, (c) receive more information requests from other colleagues, and (d) receive more referred users. However, seniority is not an important factor when determining the position that professionals hold either in the network of user referrals (submitted) or in the information requests network (submitted). In analyzing the differences in egocentered density in the information exchange networks (both received and submitted), it is observed that the group with less time in the program presents comparatively denser networks than those of the longest-lasting group. In the rest of the networks, there are no significant differences between the two groups. Relational predictors of user referrals The third objective seeks to identify whether relations of dependence among the six evaluated networks exist, specifically to determine how the analyzed multiple interactions influence user referrals (submitted and received). For that end, the association between the different networks is first examined through the QAP nonparametric permutation test [27]. Table 3presents the results. The QAP shows a significant but low-intensity association [r= (.227–.283); p< .001] between the user referrals networks, both submitted and received, and the information request networks, both submitted and received. On the other hand, the user referral (received) network has similar associations, although with a slightly lower intensity range [r= (.207–.283); p< .001]. In addition, the Jaccard coefficient (JC) was calculated to Fig. 1 Graphs of the six networks. Node size represents time working in the program and node color denotes professional category (white = psychologists; red = social workers; blue = community advocates). Isolated nodes have been deleted from 1.4, 1.5, and 1.6 Ramos-Vidal et al. Implementation Science (2020) 15:69 Page 7 of 13 determine the percentage of coincident actors in the network of user referrals submitted and received, identifying an association of 17.2% (JC = .172; p< .0001). This indicator shows a low degree of overlap if is compared with the concurrence level in the information networks submitted and received, which doubles (35.3%) its value (JC = .353; p< .0001). These results indicate a low level of integration and correspondence between professionals who refer users and those who receive referrals. Finally, a nonparametric permutation test was proposed at the dyadic level using two multiple regression QAP (MRQAP) models to determine how the Table 2 Ttest to show the mean difference according to the time working in the program in the centrality parameters of the six networks Type of relationship T test indicators assuming equal variances Group Mean Mean Difference CI: 95% (Inf.) CI: 95% (Sup.) t p Outdegree (recognition) 0 38.04 −24.69 −36.42 −12.96 −4.306 .0001 1 62.73 Indegree (recognition) 0 41.48 −18.21 −25.13 −11.27 −5.35 .0001 1 59.69 Outdegree (preference to work) 0 7.69 .728 −3.34 4.81 .359 .721 1 6.97 Indegree (preference to work) 0 5.25 −3.88 −7.21 −.55 −2.345 .023 1 9.13 Outdegree (information request submitted) 0 9.32 1.557 −2.22 5.22 .855 .397 1 7.77 Indegree (information request submitted) 0 6.52 −3.73 −7.45 −.012 −2.018 .049 1 10.25 Outdegree (information request received) 0 5.34 −3.38 −7.82 1.04 −1.153 .131 1 8.73 Indegree (information request received) 0 7.06 −.146 −2.89 2.59 −.107 .915 1 7.21 Outdegree (user referrals submitted) 0 1.53 −.383 −2.52 1.16 −.359 .721 1 1.92 Indegree (user referrals submitted) 0 1.35 −.724 −1.82 .371 −1.33 .190 1 2.09 Outdegree (user referrals received) 0 2.17 1.372 −.404 3.149 1.544 .127 1 .80 Indegree (user referrals received) 0 .72 −1.358 −2.53 −.182 −2.323 .025 1 2.08 Ego-centered density (Recognition) 0 .60 .013 −.016 .043 .931 .357 1 .59 Ego-centered density (Preference to work) 0 .18 −.063 −.196 .070 −.958 .343 1 .24 Ego-centered density (Infor. request submitted) 0 .26 .142 .053 .231 3.228 .002 1 .12 Ego-centered density (Infor. request received) 0 .18 .068 .005 .130 2.188 .034 1 .11 Ego-centered density (User referrals submitted) 0 .02 −.020 −.090 .050 −.574 .569 1 .04 Ego-centered density (User referrals received) 0 .13 .039 −.055 .113 .832 .409 1 .09 Group 0 = < 8 months working in the program (n= 23); group 1 = > 8 months working in the program (n= 26) Ramos-Vidal et al. Implementation Science (2020) 15:69 Page 8 of 13 predisposition to work and information exchange (information requests submitted and received) networks influence the user referrals (submitted and received). The DSP method described above was performed [31]. Table 4shows the results of the two MRQAP models. In the first MRQAP model, in which the dependent variable is the network of user referrals received, the results show that altogether, networks that act independently are responsible for 11.7% of the dependent variance (R 2 = .119; ΔR 2 = .117; p< .00001), with the network of user referrals submitted being the variable with the highest predictive power (β= .238; p< .0005) and the other networks moderately contributing to the explained variance. However, the information requests network and the predisposition to work network also explain some variance of the dependent variable. In the second model, the network of user referrals submitted acts as a dependent variable, whereas the network of user referrals received, the information exchange networks (received and submitted), the recognition network, and the predisposition to work network act as independent variables. The model summary indicates that the independent variables altogether account for 13% of the dependent variance (R 2 = .139; ΔR 2 = .137; p< .00001). As with the first model, the network that accounts for the largest proportion of variance is the network of user referrals received (β= .232; p< .0005). It is remarkable that the information exchange networks, particularly the network of information requests received, significantly contribute to explaining the variance of the dependent variable (β= .183; p< .0005). This finding demonstrates the connection between information exchange and user referral. Discussion This research examines different types of relationships among professionals who implement a program that provides psychosocial care to victims of war in Colombia. The literature notes that the relationships among implementers have a direct effect on service quality and user satisfaction [17,19]. To analyze these interactions, several SNA techniques are used (a) to perform descriptive analyses of the relational context of the professionals implementing the program (objective 1), (b) to determine the effect of certain characteristics such as seniority in the program when referring users (objective 2), and (c) to analyze the associations among different relationships (objective 3). The cohesion measures examined yield interesting results. With the exception of the recognition network, which shows high cohesion (51.2%), as expected due to the number of implementers of the program, the rest of Table 3 QAP correlation coefficients N° Type of relationship 1 2 3 4 5 6 rprpr prprprp 1 User referrals (submitted) 2 User referrals (received) .283 .001 3 Information request (submitted) .227 .001 .218 .001 4 Information request (received) .266 .001 .202 .001 .484 .001 5 Recognition .124 .001 .104 .001 .255 .001 .231 .001 6 Preference to work .062 .016 .171 .001 .342 .001 .315 .001 .239 .001 In terms of the overlap magnitude, when the value of the Pearson rcoefficient ranges from 0.1 to 0.3, a small correlation is reflected, values between 0.3 and 0.5 indicate a moderate correlation, and values higher than 0.5 reflect a high correlation [26] Table 4 MRQAP coefficients applying the double semi-partialling procedure N° Independent networks Dependent network Model 1 User referrals (received) Model 2 User referrals (submitted) Bβp S.E B βp S.E 1 User referrals (submitted) .217 .238 .0005 .0237 –––– 2 User referrals (received) ––– – .255 .232 .0005 .0237 3 Information request (submitted) .042 .099 .0015 .0105 .048 .103 .0005 .0108 4 Information request (received) .025 .055 .0175 .0108 .093 .183 .0005 .0118 5 Recognition .002 .012 .3128 .0056 .013 .050 .0135 .0061 6 Preference to work .046 .102 .0005 .0103 −.041 −.082 .0005 .0108 Model Adjustment R 2 = .119; ΔR 2 = .117; Prob. = .00001 R 2 = .139; ΔR 2 = .137; Prob. = .00001 S.Estandard error Ramos-Vidal et al. Implementation Science (2020) 15:69 Page 9 of 13