The protocol for this review was previously published as a scoping review protocol [29]. However, further development of the objectives and aims justified the conduct of a systematic rather than a scoping review. This systematic review was carried out in accordance with the Cochrane Methods Group [30] and the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines [31]. Following them, we adapted the method from the previously published protocol by (1) reformulating the research question using PICOTS rather than Population Concept and Context, (2) excluding literature reviews from the article selection process, (3) incorporating a quality assessment, and (4) completing the PRISMA 2020 checklist for systematic reviews (Checklist 1).

The search strategy was developed in collaboration with an academic librarian specializing in medical and health care management fields (FB). Databases covering health (PubMed, Embase, and CINAHL) and business (ABI/INFORM) disciplines and multidisciplinary (Academic Search Premier, Web of Science, and ScienceDirect) databases, as well as the Google Scholar search engine, were searched to identify eligible peer-reviewed articles. The last search of each database was performed on January 5, 2026.

Key search terms were informed by previous relevant reviews and are shown in Table 1, mapped to the PICOTS framework. Searches in the electronic databases were conducted from January 1, 2004, as the BPMN was released in its first version to the public in May 2004. For Google Scholar, we limited results to the first 20 results per string and filtered out citations and patents. The full search strategy for each database is provided in Multimedia Appendix 1.

The eligibility criteria aligned with the PICOTS are provided in Table 2.

The search results were imported to Covidence (Veritas Health Innovation Ltd) [32] to remove duplicates and manage the study selection process. Study selection proceeded in 2 phases, beginning with a title and abstract screening, followed by a full-text review of retained records. At each phase, a pair of reviewers consisting of 2 of the 3 coauthors (JBG, PL, and MTH) independently screened the titles and abstracts (phase 1) or reviewed full texts (phase 2) against the inclusion and exclusion criteria. Conflicts were managed by discussion to reach a consensus, or, when necessary, an additional reviewer (AC) was consulted.

A pair of reviewers consisting of 2 of the 4 coauthors (PL, JBG, MTH, and CL) independently extracted data from the included studies using a custom-developed data extraction form in Microsoft Excel which included the following data fields: “citation details” (title, authors, year of publication, and author affiliations), “study description” (study design, setting, care trajectory, aims or objectives, key variables analyzed, and extensions to BPMN used), “study results” (findings, outcomes, and study limitations), “BPMN utility” (objective for use, benefit or advantage of use, limit of use, opportunities, and alternatives or threats). JBG and PL reviewed all the data extraction tables, presented in Multimedia Appendix 2. Discrepancies were addressed through discussion between at least 2 reviewers.

Quality assessment was performed using the Mixed Methods Appraisal Tool (MMAT) of July 2020 [33], which uses five core quality criteria to assess each of the following study designs: (1) qualitative, (2) randomized controlled, (3) nonrandomized, (4) quantitative descriptive, and (5) mixed methods. Assessments were performed independently by 2 reviewers (MTH and JBG). Disagreements were addressed through discussion and, where needed, a third independent assessor (PL or AC) was consulted. A 3-level summary of quality was assigned to each study based on the number of achieved criteria, where 5=high quality, 3‐4=medium quality, and ≤2=low quality. For mixed methods studies, the lowest scoring of 3 study design sets (qualitative, quantitative, or mixed) was used to assign the summary level. Due to contention in the literature about the use of summative approaches in critical appraisal [34-37], we also provide a detailed presentation of the ratings (refer to Table 3 in the “Results” section). Studies that failed the MMAT screening questions S1 or S2 due to a lack of a clear research question were categorized as “Experience feedback” and were not formally appraised for quality.

Descriptive data items were reduced to meaningful categories as presented in Table 4. The characteristics of the included studies were analyzed descriptively, and we used multidimensional scaling in Orange (v3.38; Bioinformatics Lab) to visualize the relations within the body of literature across 3 dimensions: geography, study design, and study (health care) setting.

The text data extracted in the “Study Results” and “BPMN utility” fields underwent inductive thematic coding. This was done by examining extracts for keywords and phrases, grouping like keywords and phrases into subthemes, and further grouping those subthemes into broader themes to generate a 2-level coding tree by which extracts were coded.

Frequency analyses of both broad and subthemes were performed on the BPMN benefits, limits, opportunities, and alternatives or threats variables to determine the most prevalent themes represented in the literature in these 4 domains. We then constructed 2 composite variables of thematic frequencies, one representing limits and benefits as a continuum (benefit-limit), and the other, opportunity and threat as a continuum (opp-threat). To do this, we negatively transformed the limit and threat frequencies to represent the negative sides of the continuums, whereas benefit and opportunity were represented as positive values. We then plotted opp-threat (y-axis) against benefit-limit (x-axis) to examine how limits and benefits related to threats or opportunities at a broad thematic level. We then narratively examined the subthemes in relation to the predominant relationships between broad themes in the plot and synthesized this with a narrative analysis of the major themes from the study findings fields.

Finally, we mapped this synthesis to a proposed causal pathway describing the mechanisms by which the purpose or objective for using BPMN to model patient health care trajectories links to desired (or observed) outcomes through characteristics of BPMN implementation expressed as benefits, limits, opportunities, and threats or alternatives. The mechanisms have been classified according to the domains of the Consolidated Framework for Implementation Research (CFIR) [99-101].

Following the identification and removal of duplicates, 1177 unique records were identified. Screening against the eligibility criteria resulted in the retention of 253 studies, of which a further 192 were excluded by full-text review, resulting in 61 included studies [38-98]. This process is shown in the PRISMA 2020 flow diagram in Figure 1.

The characteristics of the selected literature are summarized in Table 4. The first published study of BPMN in modeling patient health care trajectories was in 2008 [38]. Most studies were performed in the European context and were conducted in a hospital setting, addressing a range of health care trajectories across acute and chronic care. The concentration of studies in Europe may be considered a bias for the validity of the study results.

A relational map of the selected literature in terms of geography, study design, and study (health care) setting is shown in Figure 2. These three dimensions are illustrated in the figure as follows: geography (marker colors), study design (marker symbols), and study (health care) setting (marker labels). The more tightly clustered a group of papers is, the more closely they are related to each other across these 3 dimensions. The size of the marker symbols is scaled to the goodness-of-fit of each case, where the smaller the marker symbol, the better the fit. This analysis shows the clear dominance of Europe (indicated in green) in the development of the evidence base from conceptual or theoretical roots (circle markers) through empirical (x markers), including experimental investigation (triangle markers). It is noteworthy that at each stage of this literature development, evidence was contributed from a range of health care settings (hospital, emergency departments, primary care, and integrated care), reducing the risk of limitations to knowledge development arising from setting specificity. In contrast to this European dominance in the development of the literature, this analysis also suggests that the rest of the world has proceeded with empirical and experimental research based on European experience, with only a few examples of conceptual or theoretical research having been conducted outside of Europe. Like within Europe, the research outside of Europe represents a diverse evidence base developed from hospital, primary care, integrated care, emergency department, and specialist or multidisciplinary health care clinic settings.

Table 3 presents the quality assessment using the MMAT tool of July 2020 [33]. The full scoring can be seen in Multimedia Appendix 3. Overall, the number of methodologically high-quality studies is low, which limits the scope and generalizability of our study’s results.

The primary objective of using BPMN was to formalize care processes by modeling and visualizing them. Indeed, BPMN modeling of the patient health care trajectory involved the visual representation of workflow models and activities [41,42,44,46-49,51,54,62,64,66-70,77,80,83,85], roles and systems within care pathways [45,50,55,57,70,93,98]. Further, such representation was used to validate the workflow with the participants in the field [63] and even to outsource certain processes [40].

Some studies aimed to integrate dedicated extensions in predefined dimensions to overcome shortcomings of BPMN. These extensions had different objectives, such as (1) integrating decision support via Decision Modeling Notation (DMN) rating [72,85], or in combination with Case Management Model Notation (CMMN) [69], both developed by Object Management Group; (2) modeling clinical practice guidelines (CPGs) [80]; (3) incorporating the notion of value through the Value Stream Map notation derived from lean management [46,87]; (4) knowledge definition for a rich and expressive graphical representation [59]; (5) the Hazard Analysis and Critical Control Points [75]; (6) a dynamic approach to BPMN [82]; or (7) adaptation to specific process contexts such as modeling clinical pathways [55,61,93,95,97], or operating room processes [74].

Beyond simple process formalization, patient health care trajectory modeling using BPMN is also used to analyze existing processes, visualize constraints, and simulate changes to optimize operational performance. A comprehensive analysis [70] of existing processes using BPMN shows that it enables a specific focus to be placed on the resources and activities, to incorporate them as constraints of the optimization model [45] and to define opportunities for redesigning the process [53,60,65,70,74]. More recently, BPMN has been used to provide a set of recommendations for clinical or care pathway optimization [46,87,91,92,94,95,97,98], sometimes in support of value-based health care [98]. In addition, BPMN modeling can then be used to simulate different scenarios for improvement [57,61,91,95] and provide proof of concept for possible optimization results [59]. However, using BPMN does not necessarily enable the initial objectives to be achieved and can lead to a variety of benefits, limitations in its use becoming opportunities for optimization or even giving rise to alternatives that could become threats.

A total of 26 broad themes and 59 subthemes were identified in the BPMN utility data fields. Table 5 shows the distribution of extractions across each of these broad themes for each domain of benefits, limits, opportunities, and alternatives or threats. The italic text indicates the most prominent themes, using 10% representation of column dimensions (benefits, limits, opportunities, and alternatives or threats) as an arbitrary cutoff. These are used to map the BPMN Purpose-Implementation-Outcome Model in the next section.

These frequencies were then used to plot opportunity-threat against benefit-limit to illustrate where benefits or limits might relate to opportunities or threats across all themes in a relational frequencies of themes plotted by benefit-limit against opp-threat figure (Figure 3). For this analysis, benefits (positive) and limitations (negative) were considered as representing positions along the same continuum (x-axis) as were opportunities (positive) and alternatives or threats (negative; y-axis). Data represent column percentages presented in Table 3 using negative transformed values for both limitations and alternatives or threats. Positions right of the vertical represent benefits, left of the vertical represent limitations. Positions above the horizontal represent opportunities and below the horizontal represent alternatives or threats.

To unpack the relationships between benefit-limit and opportunity-threat at the broad theme level, that is, to better understand, for example, where limits represent opportunity and where they represent threat for BPMN, we analyzed the distribution of subthemes within the dominant broad themes (Multimedia Appendix 4).

The CFIR domain III corresponds to the structural characteristics of the setting, including the existing IT infrastructure [100].

The CFIR domain V corresponds to the activities and strategies used to implement the innovation and their impacts on practices [100].

This systematic review provides an in-depth examination of the effectiveness of BPMN in modeling patient health care trajectories. Our findings suggest that BPMN has a more pronounced contribution to a managerial approach than to clinical relevance. Indeed, the use of BPMN improves process comprehension [47,53,83,86,91-98] while also creating opportunities for optimizing patient outcomes [41,61,76,81,92,94,96] and organizational performance outcomes [43,47,61,64,76,78,82,84,91-98]. However, despite these advantages, several challenges persist, including issues related to scalability [59], integration with advanced decision-making frameworks, and the complexity of modeling dynamic health care environments [61,66] or less structured processes [61,66,98].

One of the principal benefits of BPMN in health care is its ability to facilitate workflow visualization [41,42,44,46-49,51,54,62,64,66-70,77,80,83,85,92,93,98]. This capability allows health care professionals to gain a comprehensive understanding of patient health care trajectories [42,44,55,65,76,83-85,87,90-93], making it easier to identify inefficiencies [43,44,48,57,61,81,91-95,97] and areas for improvement [53,60,65,70,74,91-98]. The definition of processes ensures that workflows remain consistent across different settings [84,87], thereby reducing variation and improving the quality of care [52,78,79,81,82]. Several studies have demonstrated BPMN’s effectiveness in process optimization, highlighting its ability to streamline workflows [43], reduce redundancies [47,64,76], and enhance operational efficiency [45,50,61,94]. Moreover, when integrated with decision-support tools such as DMN, BPMN enhances clinical decision-making by enabling better adherence to guidelines and best practices [52,64,68,76,81,91]. Another important advantage is BPMN’s interoperability with existing health care IT standards, such as HL7 [41,72,92] and FHIR [81,89,92,96], which facilitates seamless integration with EHRs and other digital health systems [64,66,74,75,96]. However, it is interesting to note the lack of consideration given to the “social dimension” of transforming practices associated with any transformation project, such as those using BPMN notation, despite this dimension being widely recognized in implementation frameworks such as the CFIR that we have used.

Despite its many advantages, BPMN also has notable limitations that impact its applicability in health care settings. One of the most significant challenges is its complexity, particularly when modeling highly dynamic, multistakeholder environments, such as long-term chronic disease management [61,66]. The ability of BPMN to represent decision-making processes is limited [64,68,76], requiring extensions such as DMN and CMMN to capture complex, evolving patient pathways [60,66,68,87]. Another critical challenge is BPMN’s limited capability to integrate real-time data analytics and AI for predictive modeling, an area of growing importance in modern health care [58,78,81,97]. Additionally, BPMN requires a specialized skill set [78], meaning that health care professionals need dedicated training, a potential barrier to widespread adoption.

Despite these limitations, there are several opportunities for enhancing BPMN’s applicability in health care. One key area is the creation of domain-specific BPMN extensions [27], which can improve its ability to represent complex clinical pathways. Another promising avenue is the integration of AI and machine learning into BPMN models, allowing for more sophisticated predictive analytics and decision automation [78,81,97]. Enhancing interoperability by developing standardized data exchange formats between BPMN and health care IT systems could further improve system efficiency and adoption [98]. Additionally, BPMN could benefit from the augmentation of dynamic simulation tools [42-44,57,61,72], which would allow for a more accurate representation of complex patient trajectories. This is why, from a managerial perspective, the use and implementation of the BPMN tool have definite potential, but its implementation and results remain uncertain. Indeed, the implementation of this technical tool suffers from a lack of support for a social approach to transforming practices and understanding the complexity of such projects, which Madan [97] describes as the perceived lack of consideration for the human dimension in the transformation of processes. That is why we recommend considering the social dimension necessary for any organizational change by integrating the “individual domain,” the IV domain that exists in the CFIR [100] and which is absent from the studies analyzed. From this perspective, a recent framework for implementing care pathways proposes integrating BPMN as a reflective tool to support the optimization and transformation of organizational and clinical practices, rather than as a tool for standardizing clinical processes [14].

While BPMN is a widely adopted modeling approach, alternative methodologies offer complementary or competing advantages. Petri Nets, for instance, provide a powerful framework for discrete-event simulation and dynamic process modeling, which may be more suitable for certain applications [38,39]. Unified Modeling Language offers a robust structural representation [53,72,77,98] but lacks BPMN’s process-focused approach [39,48]. Similarly, Business Process Execution Language is more suited for automated workflows but does not provide BPMN’s visual representation capabilities [46]. A hybrid approach, integrating BPMN with AI-driven decision support systems and big data analytics, could help bridge some of these gaps and enhance BPMN’s clinical applicability.

From a managerial perspective, BPMN offers significant potential in health care service planning [50,73,78,82,91-98], process optimization [47,53,83,85,91-98], and resource allocation [57,67,78,80,81,94,97]. By providing a standardized tool for modeling health care workflows, BPMN enables administrators and decision-makers to identify inefficiencies [43,44,48,57,61,81,91-95,97] and implement data-driven improvements [42,44,52,56,58,73,76,79,82,94,96]. However, managers must also carefully weigh the trade-offs between BPMN’s benefits and its limitations, particularly in highly complex health care settings, and ensure that the project is properly prepared, as implementing BPMN requires a significant investment in human resources and skills. The BPMN purpose-implementation-outcome model proposed in this review offers a structured framework to assess and refine BPMN transformation projects, ensuring that they align with clinical and organizational objectives.

Several areas for future research could enhance our understanding of BPMN’s role in health care. Comparative studies directly evaluating BPMN against alternative modeling approaches could provide deeper insights into its strengths and weaknesses. More extensive, real-world implementation trials are needed to assess BPMN’s impact on patient outcomes and operational efficiency at scale. Additionally, further research into the integration of AI and process mining techniques could improve BPMN’s capabilities in predictive analytics and decision automation. Finally, investigating user-centric design approaches could help make BPMN more accessible and intuitive for health care professionals, facilitating broader adoption.

While this systematic review provides valuable insights, limitations should be acknowledged. The majority of the studies included are methodologically weak (75.4%), consisting of exploratory or pilot descriptive studies. Given the methodological weaknesses of existing studies and the limited number of large-scale BPMN implementation projects, the validity of the results presented remains questionable. Indeed, most studies focused on prototypes [42,46,49,51,70,73,81,88], small-scale descriptive projects [43,45,50,53,57,62,63,77,80,82-85,91-98], or theoretical explorations [41,44,55-57,59,66,68,69], underscoring the need for further real-world research. Furthermore, the concentration of studies included in European countries (68.9%) is a key factor limiting the generalizability of our results.

This systematic review highlights BPMN’s potential as a valuable tool for modeling patient health care trajectories in a managerial approach to transforming practices. Its ability to visualize and optimize processes makes it a promising tool for improving clinical and operational efficiency. However, trade-offs between benefits and limits of BPMN characterize its implementation in patient health care trajectories, giving rise to opportunities for the development and integration of new tools and extensions to handle complexity and real-time data integration and to optimize outcomes. However, it is important to note that the methodological weaknesses of the studies and the lack of large-scale research projects mean that these results cannot be generalized. Future advancements, including the development of more sophisticated BPMN extensions, integration with AI, and improved interoperability with health care IT systems, will be crucial in realizing BPMN’s full potential. We propose a framework linking purpose to outcomes through richly characterized implementation domains, which could help managers to better specify their BPMN transformation projects and facilitate evaluation of their effectiveness. Future research could address the compatibility of systems in the hospital environment and emphasize the importance of considering the social dimension inherent in any change in professional and organizational practices.