The problem of operating-room scheduling: exploring the gap between theory and practice Open Access

Planning and control in complex service systems can be overwhelmingly intricate tasks. Analytical decision-support methods hold significant promise for enhancing planning and control in such environments. However, many of these technically sophisticated solutions struggle to gain traction in practice. What can theory in operations management (OM) contribute to understanding and addressing this persistent gap?

In this paper, we focus on the planning and scheduling of operating rooms (ORs). ORs play a key role in hospitals and the broader healthcare system as centers for surgical intervention. They are typically the most expensive resources in hospitals, and their utilization is critical to the overall efficiency of hospital care (Zhu et al., 2019). Planners make planning and scheduling choices such as the sequence of patients in an OR session, the case mix or cancellations of scheduled operations. By doing so, they seek to optimize outcomes relevant to various groups of stakeholders, including the utilization of resources and staff, and the overtime of surgical sessions. If scheduling goes well, surgeons, staff and facilities have minimal idle time, the session ends in time, and no surgeries are cancelled.

Algorithmic decision support for OR scheduling is a prominent topic of study in the field of management science and operations research (MS/OR). Representative examples of this extensive and rapidly expanding literature include Denton et al. (2010), Day et al. (2012), Fügener et al. (2014), Akbarzadeh and Maenhout (2025) and Tsang et al. (2025). These works approach the task of OR planning and control by ex ante schedules determined by mathematical optimization. To do this, the goals and preferences of stakeholders are captured in an unambiguous objective function and limitations, dependencies and conditions are treated as mathematically defined constraints. Algorithms compute schedules that maximize the objective function within the given constraints, dealing with uncertainty by determining optimal capacity buffers in the schedule. Despite the great importance of efficient scheduling of ORs, however, the adoption and real-world impact of such approaches in healthcare practice seems disappointingly low (Samudra et al., 2016; Hof et al., 2017; Zhu et al., 2019).

Prima facie, the argument for decision support algorithms for OR planning is compelling: optimized schedules could make the deployment of scarce and expensive resources in the OR more efficient, and computers may outperform human planners in complex scheduling tasks. Yet, May et al. (2011) conclude: “Over the past 55 years, a variety of standard operations research models have been proposed for use in hospital scheduling, but none appear to have had widespread impact on the actual practice of surgical scheduling.” Harris and Claudio (2022), reviewing 264 papers on OR scheduling, observe that less than 5% of proposed models have been implemented in practice, and we could not find systematic evaluations of realized results and real-world impact (c.f. Lamé et al., 2022). MS/OR, with its traditional focus on mathematics and formal modeling, offers little empirical research into the validity of its core premises (Van Riet and Demeulemeester, 2015; Beliën et al., 2025). There is almost no empirical research examining whether the key assumptions underlying these algorithmic approaches align with the goals, constraints and practical realities of OR planning. Moreover, it is not evident whether ex ante scheduling is the most suitable solution pathway, nor whether alternative approaches may need to complement it. There are concerns that many studies solve oversimplified versions of problems that are in fact more complex and messy (Eldabi, 2009; Diamant et al., 2023; Beliën et al., 2025).

OM and MS/OR share substantial overlap in their domains of inquiry. Despite their common subject matter, however, theory in both fields has remained largely disconnected. While MS/OR studies healthcare operations from mathematical theory and formal modeling, OM focuses on empirical theory building, such as the Theory of Swift and Even Flow (TSEF; Devaraj et al., 2013; Johnson et al., 2020). This theory proposes that healthcare operations are improved by reducing variability (Law of Variability), better management of bottleneck resources (Law of Bottlenecks), applying PDCA-style approaches for continuous improvement (Law of Scientific Methods), improving quality (Law of Quality) and limiting the variety of tasks (Law of Factory Focus). TSEF holds that these “Five Laws of OM” (Venkataraman et al., 2018) create a swift and even flow of patients, thus improving productivity. When we examine MS/OR approaches to OR planning through this lens, we note their primary focus on optimizing OR capacity – the most constrained resource within a hospital's surgical services – thereby implicitly applying the Law of Bottlenecks (Venkataraman et al., 2018). In addition, using schedules to control the release of work into the system reduces system variability (Thurer et al., 2012) and could thus be seen as an application of the Law of Variability.

The purpose of this paper is to study MS/OR approaches for OR scheduling through the lens of OM theory and methodology. We use empirical research methods to study the alignment of MS/OR with healthcare practice and interpret the observed discrepancies from theory in OM. This leads us to propose an integrative framework, where MS/OR theory and alternative theories in OM complement each other to offer a more comprehensive framework for OR planning.

MS/OR offers a substantial and sophisticated body of prescriptive theory for OR scheduling, with little convincing support, however, that its premises are sufficiently aligned with practice. This motivates our first research question: (RQ1) To what extent is prescriptive theory in MS/OR on OR scheduling aligned with the needs and requirements of OR planning in practice? To address this question, we first make a synthesis of MS/OR theory that identifies its essential premises. This theoretical perspective captures how the OR scheduling problem is framed in the MS/OR literature. By means of a multiple-case study involving nine hospitals, we confront this theoretical perspective with how OR planning is performed in healthcare practice, which we call the practical perspective. We test to what extent the premises of MS/OR theory are aligned with the needs and requirements of practice, thus identifying several discrepancies, where MS/OR and practice are misaligned.

In the second part of the paper, we interpret the observed discrepancies from OM theory, exploring whether OM can suggest modifications in the MS/OR framework to make it better aligned with practice, and whether alternative theories in OM could complement MS/OR to resolve observed discrepancies. Our research question in this second part is exploratory: (RQ2) How can we integrate MS/OR with alternative theories in OM into a potentially better framework that overcomes the observed misalignments? Our analysis results in a set of theoretical propositions, which we present as a proposal for further research.

In the next section, we synthesize the MS/OR literature to obtain the theoretical perspective. In Section 3, we motivate and explain the design of the multiple-case study. In Section 4, we discuss our findings in the nine studied hospitals, noting discrepancies with the theoretical perspective. Section 5 analyzes and interprets the observed discrepancies from theory in OM, resulting in our propositions. The Conclusions section presents implications for theory, practice and discusses the study's limitations.

OR planning has attracted considerable attention; an initial search found around 5,000 studies from 2010 onwards (the Supplemental Materials describe our literature search in detail). We focused our literature review on the MS/OR literature, seeking to create a synthesis that identifies its key premises and solution pathways. Besides hundreds of individual contributions, MS/OR has produced many review papers, synthesizing this large body of research. After screening the abstracts and full texts, the search identified 19 key review papers from the period 2010–2025 (exclusion criteria and details are given in the supplement). Given the availability of these comprehensive review papers, we took these as the basis for constructing our theoretical perspective, using key works cited in them, such as Tsang et al. (2025), as further sources where a deeper understanding was needed.

Table Sup-I, also in the supplemental materials, presents the resulting 19 review papers. The table specifies for each review how many original studies they discuss, observations about the degree of implementation and real-world impact of MS/OR approaches and reflections about the applicability of the discussed techniques. Table Sup-I also identifies the decision horizons of OR planning covered in each review paper and how the various decision levels are characterized.

The MS/OR literature presents OR scheduling as a multi-level problem with multiple decision horizons (Van Riet and Demeulemeester, 2015; Zhu et al., 2019; Wang et al., 2021). Terminology is not used consistently, but we will refer to the levels as the strategic (long-term), tactical (mid-term) and operational (short-term) levels. Strategic planning refers to long-term capacity dimensioning and allocation decisions: what volume and mix of surgical services does the hospital want to provide for the following year(s)? This is translated into decisions for expanding or reducing the number of ORs, staff and equipment. The result is a case-mix planning, where each specialty has been allocated OR time, and the necessary staff and equipment have been budgeted. At the tactical level, the strategic allocations are translated into a Master Surgery Schedule (MSS), which assigns OR time blocks to specialties, including the baseline support staff. The MSS is a cyclical scheduling process (typically at a cycle of 2–4 weeks), which allows the strategic allocations to be adjusted based on midterm demand forecasts. The operational level breaks down into off-line planning and online control (Cardoen et al., 2010). In off-line planning, patients are booked in the allocated time blocks (position in the session and planned duration), and the required resources are made available, including staff and equipment (Guerriero and Guido, 2011; Samudra et al., 2016). Online control consists of last-minute adjustments to the schedule in response to changing conditions on the day itself. This includes the allocation of emergency patients, dealing with cancellations, and responding to congestion and session overruns (Hulshof et al., 2012; Wang et al., 2021).

Our literature review (Table Sup-I) reveals the current unclarity about the alignment of theory with practice. Even though many research papers are grounded in a real OR scheduling problem and use real data, there is almost no evidence of MS/OR being implemented successfully in hospitals, giving real and sustained efficiency benefits. This motivates our first research question RQ1. Identifying gaps between the theoretical and practical perspectives on OR management, we then seek to explore potential ways to improve the alignment, hoping to strengthen the practical relevance of MS/OR theory. In the second part of the paper, we therefore interpret observed misalignments from the perspective of OM theory, exploring how we can integrate MS/OR and OM to give a potentially stronger framework that overcomes the observed misalignments.

As a point of departure for our study, we reconstruct the problem frame that underlies approaches proposed in the MS/OR literature (theoretical perspective). We thus obtain our research model, which guides data collection in the case studies. For each of the decision levels (operational, tactical and strategic), we identify the common assumptions about goals, challenges and constraints that approaches make in terms of the following core elements (E1 through E4) of mathematical optimization problems (Kuiper et al., 2021):

First, we synthesize how MS/OR literature frames the problem at the operational level (Figure 1). This level of the OR scheduling problem gets substantially more attention than the tactical and strategic levels (Zhu et al., 2019). The setting is that OR time blocks have been allocated to specialties; that is, for specific blocks of time, each OR is reserved for a number of patients from a certain specialty. The scheduling problem here is to determine the sequence of patients within a session and allocate time for each surgery reflecting the expected duration. Patients often have been allocated to a specific date and time long before, but as they must be present in the hospital hours in advance, there is enough latitude to “reshuffle” and establish a more final sequence one to three days prior (Bandi and Gupta, 2020).

Many MS/OR approaches focus on the combinatorial challenge of finding a schedule that fits in the available ORs and with limited idle and overtime, subject to a complex combination of constraints (Zhu et al., 2019). This combinatorial aspect of OR scheduling is often seen as a variant of the bin-packing problem (Bandi and Gupta, 2020), and algorithms use various forms of mathematical programming (constraint programming, mixed-integer linear programming, etc.). The constraints include the limited availability of ORs with specialized equipment such as surgical robots, and restrictions brought about when staff is specialized and therefore cannot support all procedures or needs to be paired with certain surgeons. Other constraints involve preferences of patients and surgeons, for example, surgeons' preference for back-to-back scheduling or their preference to conduct all surgery in the same OR. In precedence constraints, scheduling is bound by dependencies on steps upstream in the patient's care path. Some papers (Freeman et al., 2018; Rachuba et al., 2024) take into account the ramifications for downstream units in determining an OR schedule, such as the recovery ward, the post-anesthesia care unit (PACU) and (post-) intensive care units ((P)ICU) or for upstream units (Bovim et al., 2025). Although most combinatorial approaches are deterministic, some incorporate uncertainty using techniques such as chance constraints (Van den Broek d’Obrenan et al., 2020), robust optimization (Maleki et al., 2023; Bandi and Gupta, 2020; Tsang et al., 2025) or stochastic optimization (Denton et al., 2010; Tsang et al., 2025).

Other MS/OR approaches focus less on combinatorial optimization, and more on optimizing how variability and uncertainty are absorbed in the schedule (for example, Denton et al., 2007; Tsai and Teng, 2014; Leeftink and Hans, 2018; Van den Broek d’Obrenan et al., 2020). The actual durations of surgeries are variable and uncertain. Surgeries may start late, and there may be emergency patients who were not scheduled but must be accommodated in the OR session at the last moment. Also, planned operations may be cancelled for various reasons: when the patient is unfit to undergo the planned surgery (medical cancellations in Figure 1), when materials or equipment are missing (logistical cancellations) or when a patient fails to appear (no-show cancellations). Scheduling algorithms focusing on variability and uncertainty typically use queuing theory to determine suitable slack in the schedule and in the allocated durations as buffers to absorb uncertainty, or they take uncertainty into account in the sequence of patients, scheduling surgeries with a high likelihood of running late at the end of the day to minimize the impact on delays for subsequent surgeries (Guda et al., 2016).

Literature considers up to a dozen outcome measures, including utilization of resources, overrun time, overrun cancellations and waiting time for patients (cf. Charlesworth and Pandit, 2020, for a review and discussion of OR performance metrics). Collectively, these outcome measures fall into two classes: measuring the consequences of overruns and congestion, and measuring the consequences of underruns and unutilized capacity. On the one hand, too tight a schedule could result in sessions overrunning the schedule, which in turn results in overtime for staff and delays for patients waiting to undergo surgery. Overruns could also result in inefficiencies due to so-called congestion effects (Berry Jaeker and Tucker, 2017), where the fact that a schedule runs late itself creates extra work or inefficiencies. For example, when a session runs late, the hospital could decide to cancel a scheduled surgery (overrun cancellations in Figure 1). This brings about inefficiencies such as the extra rescheduling and needless hospitalization and preparation of patients. On the other hand, too loose a schedule could result in underused capacity for scarce resources such as surgeons, staff and facilities. This, in turn, lowers the effective capacity of the unit (with consequences for the accessibility of care) while also lowering efficiency (with consequences for the cost of care) and productivity (with consequences for required staffing levels).

Tactical planning is typically done in a weekly, monthly or sometimes quarterly cycle. Given the strategic goals for volume and case mix, tactical planning makes adjustments to compensate for midterm demand forecasts and balances the workload across specialties (Figure 1). MS/OR approaches use mixed-integer linear programming to allocate specialties to days and ORs, considering the preferences of the surgeons, thus creating the MSS. Schedules are also optimized for balanced outflow (Choi and Wilhelm, 2014): sessions where patients need follow-up care are spread out over the week to avoid peaks in follow-up care. As an alternative to the system where blocks of OR time are assigned exclusively to a particular specialty (“block scheduling”), in an “open scheduling” strategy, ORs are used on a first-come-first-served basis, and surgeons can choose to operate in any available OR (e.g. Fei et al., 2009). Combinations of both strategies include the modified block and integrated block scheduling systems (Day et al., 2012).

The horizon for strategic decisions is several months to a year, focusing on long-term capacity planning and case-mix planning (Figure 1). This includes strategic choices, such as determining the types and the proportion of medical services the hospital will offer. Case-mix planning establishes the mix of surgical cases to optimize resource utilization and financial performance, ensuring a balanced and profitable portfolio of procedures (Ma and Demeulemeester, 2013; Koppka et al., 2018). Capacity planning and allocation (e.g. Roshanaei et al., 2017; Fügener et al., 2017) involves estimation and forecasting problems, such as how many surgeries can be done in given time, and how will demand develop? Based on these choices and estimates, hospitals make investment and allocation decisions, such as: should they expand the number of ORs or hire new surgeons and staff to meet future demand? Should they invest in specialized equipment to support high-demand surgeries such as orthopedic or cardiovascular procedures? How much OR time should they allocate to each specialty?

Figure 1 represents the theoretical perspective, that is, the structure of the OR scheduling problem as hypothesized in the MS/OR literature. To research the extent to which MS/OR accurately captures the goals and challenges in healthcare practice, we study OR planning in nine hospitals in the Netherlands. Case study research is the traditional approach to studying practices in the application domain, where the richness of detail that case studies bring compensates for the limited number of cases.

Our study is aimed at confronting key elements of MS/OR theory with practice, and then interpreting observed discrepancies. In the framework of Wellman et al. (2023), such an objective is called “Test and Explore.” Such studies employ a deductive approach to test a theory (primary research objective), followed by inductive methods to explore and deepen the understanding of findings (secondary objective). The “Test” part in our study is where we derive key premises from the extant MS/OR literature (Figure 1) and confront them with actual healthcare practice, thus identifying discrepancies between theory and practice. The “Explore” part is where we interpret these discrepancies from alternative theories in OM, arriving at a set of propositions. Figure 2 summarizes the research process.

The unit of analysis (that is, a single case) in our study is a hospital, which has several ORs (ranging from 7 to 44 in the selected hospitals), shared by multiple specialties and scheduled as a single operational unit, sometimes called the surgical suite or operating theater. Our selection of nine hospitals is partly motivated by literal replication (Voss et al., 2002), where we have included similar cases to verify if our findings and conclusions are reproduced. In addition, we made sure that the selected hospitals represent variation with respect to characteristics that we anticipated to affect OR planning, thus applying theoretical sampling (Barratt et al., 2011). In particular, our sample was designed to exhibit contrast with respect to.

1. Hospital size: small (275 beds, 7 ORs) to large (1,400 beds, 44 ORs).

2. Complexity of care. In the Dutch healthcare system, academic hospitals focus on complex care, while peripheral hospitals focus on general care and routine surgery. In between these are the advanced clinical hospitals, which offer a combination of routine and complex care.

Table 1 presents the key characteristics of the nine hospitals. Our primary data collection instrument is the structured interview. Interview questions were based on the elements E1 through E4 of the theoretical perspective. All our interviews were conducted by two or three authors, with one author leading the interview, while the other researchers took notes. Interviews, lasting between 60 and 90 minutes, were also recorded and later transcribed. In each hospital, two to four respondents were interviewed independently from each other, representing different positions and responsibilities. This ensured that for each case, multiple perspectives were captured. There were two versions of the interview questions, one aimed at respondents with responsibilities at the operational level, and the other for the tactical or strategic level. In total, we interviewed 27 respondents across nine hospitals.

Table Sup-II in the Supplemental Material presents the interview results, showing whether and how the elements E1 through E4 (on the operational, tactical and strategic levels) of the theoretical perspective were recognized and applied in the nine hospitals. We also collected evidence of work procedures, decision support tools, dashboards and other practical tools used in OR planning, which we present in Table Sup-II(e). In addition, two authors participated in a 2-h guided tour through an operating theater in operation, following the entire workflow, and with ample opportunities to discuss the process with various OR staff. This helped in building a frame of reference and a good understanding of the practical functioning of an operating theater. The interview results are the basis for identifying misalignments between MS/OR theory and healthcare practice. We discuss these findings in Section 4, thereby answering RQ1, and resulting in six key discrepancies D1 through D6, which are listed in Table 2.

Next, in the “Explore” part of the study (Section 5), we interpret these discrepancies, seeking to understand them from theory in OM. This analysis yields our proposed integrative structure for OR planning in the form of a set of propositions P1 through P6 (Table 2), where OM theory is used to either suggest modifications to MS/OR premises or as an alternative theory to complement MS/OR. This integrative framework was discussed with respondents in follow-up interviews. We did seven follow-up interviews in seven hospitals, which lasted up to 2 h. In each of them, one author discussed our emerging interpretation and proposed reconciliation of discrepancies with one of the original respondents. After each follow-up interview, the proposal was adjusted based on the feedback from the respondents until a consolidated version emerged that was confirmed by respondents in subsequent interviews. This final version of the propositions, answering RQ2, is presented and discussed in Section 5.

Section 4 in the Supplemental Materials offers a further justification of the research approach. The section contrasts our “Test and Explore” approach with alternatives and specifies various forms of triangulation and checks that we performed to strengthen the reliability of the conclusions. It also maps the chain of evidence leading from the research model and interview results to the observed discrepancies D1 through D6 and the propositions P1 through P6.

Table Sup-II provides detailed results of the case analyses, indicating the extent to which each element in Figure 1 was observed across the hospitals. Owing to its length, the table is made available in the Supplemental Materials, and we discuss the most salient findings below.

The decision parameters, outcomes, constraints and sources of variability in the theoretical perspective, as well as their organization in the strategic, tactical and operational levels were largely confirmed in all nine hospitals. This confirms that the basic framing of the OR scheduling task in MS/OR, and its organization over three decision horizons, is valid. All hospitals use block scheduling as their main tactic, reserving blocks of OR time for the exclusive use of specialties. In addition, most hospitals have one or a few ORs dedicated to emergencies. The overriding concern in all nine cases is the efficient deployment of staff rather than the optimization of financial performance, reflecting the current pressing staff shortages in the Dutch healthcare system.

Table Sup-II(e) shows whether hospitals use analytical decision support tools in their OR scheduling. All hospitals use historical estimates to determine the scheduled duration of surgeries. Some use dashboards displaying OR utilizations and other outcomes, and surgery status display boards to visualize ongoing surgeries during the day. Three hospitals had adopted an application for predicting OR outflow to downstream units.

However, even though in the nine hospitals the main structure of the decision problem was found to be consistent with the theoretical perspective in Figure 1, the actual process by which schedules come about in practice differs entirely from the conceptualization in MS/OR literature. Table Sup-II(e) reveals that we found no evidence in any of the hospitals of the application of scheduling techniques as proposed in the MS/OR literature discussed in Section 2, underscoring the motivation for this study. Below, we discuss the observed ways of working in practice and how they deviate from the theoretical perspective.

One of the findings that stood out, is that in all hospitals, planners take into consideration a myriad of wishes and constraints brought about by idiosyncratic characteristics of rooms and facilities, restrictions and personal preferences of surgeons, and preferences of patients. These are listed in Table Sup-II(d) for the operational, tactical and strategic levels. In Hospital 5, planners took into account that some surgeries require a cooler OR. In Hospital 2, patients infected by certain types of bacteria were to be scheduled at the end of the session, to avoid the required extra thorough cleaning procedure in the middle of the session. In Hospital 6, the elderly with a broken hip were not scheduled on Sundays because such surgeries have a higher risk of complications, requiring resources unavailable on Sundays. Surgeon-related constraints include special expertise, preference for a particular OR, or shifts (for example, when a surgeon has done night shifts). Examples of highly idiosyncratic constraints include surgeons' activities elsewhere, such as bringing their children to school or sports (Hospital 4) or a surgeon and anesthesiologist who could not work together in view of their personal history (Hospital 6) or a preference to do knee surgeries on left-right knees alternatingly (Hospital 8). Such surgeon preferences carry much weight in hospitals and are second in priority only after the availability of staff. Patient-related preferences are whether they can take off from work, their traveling time to the hospital, whether they can arrange daycare for the children, and many planners try to schedule children early on the day to make it easier to follow fasting requirements. Interviewees recognize how this “… Complex structure of individual ifs and buts” (Hospital 1) makes scheduling difficult. “There are no deliberate steps to simplify complexity. This would require an intervention from higher up. On the contrary, ever more exceptions and restrictions are being added,” one interviewee in Hospital 2 acknowledged.

Scheduling procedures in practice are to some extent articulated in protocols, for example, in the “Planning Bible” in Hospital 1, or the “OR Guide” in Hospital 2. However, they are primarily driven by tacit knowledge. The numerous constraints and preferences mentioned above largely reside in people's minds (Van der Ham et al., 2021, found 85% of constraints in OR planning to be tacit). We noted that heuristics derived from MS/OR theory are often negated by tacit idiosyncrasies in practice. As an example of this, we found in Hospitals 1, 7 and 9 that they schedule complex cases with much uncertainty early in the day. This goes against MS/OR heuristics (e.g. Guda et al., 2016), which prescribe that highly uncertain cases be planned at the end of the day, as this reduces the impact on subsequent surgeries. In practice, however, surgeons prefer scheduling simple and short cases at the end of the day because it is easier to do these when they are fatigued, and because long cases at the end of the day would complicate the response to overruns and emergencies.

In all hospitals, the process of scheduling is much more informal and improvised than the structure of a mathematical optimization problem (see the Observed discrepancies and idiosyncrasies in Table Sup-II(a)). Instead of goals and constraints captured in the precise and unambiguous form of an objective function, we found satisficing (Simon, 1956) as the best way to describe how planners work in practice. Planners settle for a satisfactory solution that makes everyone somewhat satisfied instead of searching for an optimal solution. One respondent described this way of working as “A dance between everyone's interests” (Hospital 1), another one as “horse-trading” (Hospital 6). In most cases, tactical planning is conducted through so-called Tactical Planning Meetings (TPMs), adopted by many hospitals in The Netherlands. In the TPMs, all relevant parties are represented, and they coordinate and negotiate until they find a solution that is satisfactory for everyone. The TPMs also handle resource allocation, prioritize critical surgeries and update schedules to reflect current needs and constraints. Respondents noted that these tactical meetings improve interdepartmental coordination. On the operational level, teams do weekly sync-up meetings to align and coordinate, but more prominently, planning and control are performed by key persons who act in a broker role (Burt, 2004). These broker-planners talk to everyone involved or affected, collecting information, resolving conflicts, balancing various interests and coordinating everything.

Moreover, actual strategies emerge over time because of experience or routine, instead of being the conclusion of a formally structured decision problem. Instead of looking forward to optimize or at least improve planning, planners look back at what they did recently and how that worked out and make ad hoc adjustments from there. They do this on all three levels (strategic, tactical and operational). For example, in Hospitals 1, 4, 6, 8 and 9, the yearly allocation of OR time to specialisms (the case-mix dimensioning) is not driven by strategic goals, demand forecasts or economic optimizations. Instead, the historical case mix is taken as a point of departure, with ad-hoc modifications driven by the lobbying of departments. One of the respondents (Hospital 8) coined the term “retrospective planning” to describe this approach. Metaphorically, they decide what clothes to wear based on what they wore yesterday and how that worked out, instead of looking at the weather forecast and optimizing for the expected conditions. Consequently, scheduling practices fluctuate around an equilibrium, where planners refrain from considering whether this equilibrium is desirable or can be improved. For example, we observed in several hospitals how emerging idle time in the OR results in surgeons doing more consults in the outpatient clinic, which in turn creates more demand for surgery, thus steering the utilization of the OR back to a normal state. Retrospective planning results in a continuation of the status quo and commonly degenerates into firefighting and unintended stability (Anand et al., 2012).

The MS/OR literature focuses on resource utilization and its trade-off with overruns and cancellations as outcomes to be optimized (as summarized in Figure 1), and we saw this reflected in metrics used for performance management in practice (Table Sup-II(c)). The norm for OR utilization was typically 85% or 90–95% (depending on whether they count changeover times as idle or utilized time), and these norms were usually met. These fairly good figures beg the question, however, how meaningful the focus on such metrics is: if utilizations are really at such high levels, then is there any room for improvement? Exploring this issue more in-depth with respondents in the follow-up interviews revealed that the realized utilizations are to some extent cosmetic: they can be artificially improved by allocating longer time to individual surgeries, at the expense of production volumes, or by prioritizing OR utilization at the expense of other services in the hospital (c.f. Charlesworth and Pandit, 2020). Also, the high utilization figures hide the considerable amount of hassle and commotion that go into creating and readjusting the OR schedule. The case interviews and follow-up interviews revealed that planners took additional goals into consideration, such as maintaining a stable patient flow. Staff shortages, drop-out and turnover are pressing concerns in the hospitals, and we observed almost universally that avoiding job dissatisfaction (work pressure, hassle and overruns) was a primary concern in OR planning.

At the heart of many scheduling techniques proposed in the MS/OR literature is the notion of variability and uncertainty. Besides the sources of variability mentioned in Figure 1, during the interviews, respondents mentioned other examples, including complications during surgery, patients' tardiness and scheduling mistakes (for example, scheduling a surgeon who happens to be on vacation). On the tactical and strategic horizons, examples include fluctuating staffing levels and absenteeism, epidemics, and an aging population with more diverse and less predictable requirements (Table Sup-II(b)). A key challenge in scheduling, as conceived in MS/OR, is dealing with such variability and uncertainty. This notion appears controversial in healthcare practice. On the one hand, hospitals accommodate uncertainty in the schedules, either by leaving slots open (“white spots,” as they refer to this method in Hospitals 1, 5 and 6) or by working with one or a few ORs dedicated to handling emergency care. On the other hand, there is a strong belief among healthcare professionals that most issues could be anticipated with the right information, and therefore, that session overruns or unanticipated idle time are avoidable. Especially in Hospitals 2, 5, 6, 8 and 9, variability and uncertainty were ignored or denied during scheduling. One respondent (Hospital 2) mentioned that 80% of surgeries deviate from their allocated slot length by more than 20 min, which has a seriously disruptive impact on how the session goes, but this observation is not taken into consideration in OR scheduling. In Hospital 2, uncertainty is acknowledged for complex surgeries, but denied for routine cases, as is reflected in the hospital's policy that for the OR blocks assigned to these specialties, overruns are officially designated as “unacceptable.” In the current situation, the belief that most issues can be anticipated seems misguided and based on an overconfidence in the controllability of the process.

Besides buffering and denial, the main response to variability and uncertainty in the nine cases was reactive. Simple responsive mechanisms are common; for example, day planners respond to emergencies and overruns by readjusting the order of scheduled operations, switching ORs or cancelling surgeries as a last resort. Such responses have an improvised rather than systematized character.

We did not find salient differences between the three classes of hospitals (academic, advanced clinical and peripheral). Instead, the findings discussed above were consistent across all nine hospitals. Below, we integrate the within-case findings into an overall conclusion about the discrepancies between the theoretical and practical perspectives.

The overarching finding presented above, and our answer to research question RQ1, is that the structure of the OR scheduling problem in practice is largely consistent with MS/OR theory (Figure 1), but that the processes through which schedules are determined and OR planning is executed are fundamentally different from the logic of optimization.

In this section, we seek to deepen our understanding of this disparity by interpreting it from theory in OM. As a starting point, we performed a process of category abstraction, starting from the idiosyncrasies and discrepancies listed in Table Sup-II that surfaced in the individual cases. Figure Sup-2 in the Supplemental Materials shows, in the form of a data structure diagram, how we abstracted from these observations six key discrepancies D1 through D6, holding across the nine hospitals, between MS/OR theory and healthcare practice. Interpreting these discrepancies from OM theory (Table 2), we then derived propositions for making MS/OR theory better aligned with the realities of practice, either by proposing modifications to some of its premises or by proposing an integration with alternative theories in OM.

The first discrepancy D1 observes that planners in practice take into account myriad and interdependent constraints and preferences. Many of these constraints and preferences are idiosyncratic. The complexity and tacitness of these nitty-gritty details of OR scheduling represent a significant challenge for experts in MS/OR techniques to define the optimization problem. They make the application of MS/OR techniques difficult, and also create a lot of work for planners, hassle in the process, and in the end reduce the available options for creating a schedule.

Many constraints are important or even critical and cannot be ignored. Other constraints, however, are nice to have if there were no consequences, but may not warrant the resulting increase in complexity. The Law of Variability in the TSEF suggests that simplifying OR scheduling by pruning unnecessary and unwarranted constraints is likely to improve OR management. For MS/OR techniques, this is an essential preliminary step to make the problem tractable and definable.

As part of this simplification step, standardization should also be considered. De Regge et al. (2019) contrast standardization of clinical practices (as many healthcare providers aim to do in their pursuit of evidence-based medicine) to standardization of operational practices, showing that the latter is essential for obtaining improved operations performance. Litvak and Long (2000) use the term artificial variability to describe the variability in processes not induced by natural variability, but by poor alignment and standardization in the hospital's operations management. In Hospitals 2, 3, 6 and 7, for example, various departments used different software for planning and administration (including some departments using pen-and-paper methods), adding complexity to the OR scheduling process. Studies such as Von Strauss und Torney et al. (2018) demonstrate the impact on efficiency of standardizing instrumentation and OR layout.

Our first proposition, therefore, is that (P1) The application of MS/OR scheduling techniques may be more effective when complemented by focused simplification and standardization initiatives. It takes dedicated effort to prune needless complexity (De Regge et al., 2019), for instance, using an approach such as lean (Dobrzykowski et al., 2016; Roemeling et al., 2017). Such efforts are also likely to meet with resistance. Johnston et al. (2019) found that surgeons resist standardized routines and centralized control. They speculate that this may be based on the surgeons' mistrust of the managerial competence in hospitals, noting that this creates a vicious circle where this mistrust itself hampers simplification and standardization, thereby undermining attempts at fixing ineffective scheduling practices, which in turn appears to justify the surgeons' mistrust.

The Law of Scientific Methods in the TSEF emphasizes the role of sound improvement methods, ranging from simple PDCA-style approaches to, indeed, the sophisticated techniques in MS/OR. We propose that (P2) The application of MS/OR scheduling techniques may be more effective when complemented by building a culture promoting continuous improvement and the use of scientific methods.

Discrepancy D2 captures our observations of dysfunctional practices being perpetuated by staff, such as the practice that we called retrospective planning, or the continued denial of uncertainty. We observed a strong fixation of staff members on immediate demands, relying on workarounds to fix issues and “keep things running,” instead of a pursuit of structural improvement. This reliance on workarounds masks underlying problems and perpetuates inefficiencies (Tucker and Edmondson, 2003; Tucker et al., 2020). Tucker et al. (2002) define second-order problem solving as focused on diagnosing and altering the underlying causes of problems to prevent recurrence. Workarounds, on the other hand, are first-order problem-solving, where immediate issues are fixed but without solving the underlying problems. Antecedents that perpetuate the focus on first-order problem solving in healthcare include structural barriers, such as hierarchies and continuous improvement not being part of job designs, and cultural aspects, such as the psychological gratification that workarounds bring to nurses demonstrating independence and stoic firefighting heroism (Tucker et al., 2002). Our proposition poses that strengthening a culture that promotes second-order problem solving and the use of sound methods to achieve it is a prerequisite for the successful adoption of MS/OR techniques. Dreyfus et al. (2020) found that standardization (as in our proposition P1) and a continuous improvement culture (our P2) reinforce each other, as a lack of standardization leads to workarounds, whereas an improvement culture promotes that lessons learned are codified into standards.

Besides their abundance, we also found constraints and preferences to be tacit and ambiguous (there is disagreement about their legitimacy and priority). This makes the OR scheduling problem underdefined: it is difficult to get a complete and precisely articulated formulation of the objectives, preferences and constraints, which may defeat optimization attempts based on an unambiguous objective function as is done in MS/OR.

Therefore, the satisficing approach observed in practice, making all stakeholders somewhat satisfied and none too unsatisfied, may be a more realistic framework than mathematical optimization. Schwartz et al. (2011) argue that satisficing is a more rational decision framework than objective-function maximization in situations where essential elements of the decision problem cannot be specified and where uncertainty is so radical that you cannot even model it meaningfully as a stochastic component in a model. Especially the tacitness and ambiguity of preferences and constraints introduce a radical uncertainty in the OR-scheduling problem. Long et al. (2023) have translated the idea of satisficing into a mathematical decision framework. The idea is to define a satisfactory (as opposed to optimal) performance threshold, and then compute a solution that maximizes the range of circumstances under which this satisfactory outcome will be achieved. To our knowledge, such satisficing frameworks have not yet been applied to the OR-scheduling problem. We propose that (P3) Incorporating a satisficing framework in MS/OR, instead of the pursuit of a uniquely optimal schedule, may make scheduling techniques better aligned with the realities of healthcare practice.

Discrepancy D4 captures our finding that instead of formal, algorithmic approaches, schedules emerge in practice through a process of negotiation and brokerage. On the one hand, we discussed in Section 5.2 that the adoption of scientific methods, including MS/OR, could improve OR management. Proposition P4, on the other hand, holds that some degree of coordination through negotiation and brokerage is unavoidable in view of the complex social and political context in hospitals.

In a seminal paper, Glouberman and Mintzberg (2001) describe how healthcare is divided into distinct but interdependent domains (cure, care, administrative control and the patient community), each with a different culture, structure, coordination mechanisms and professional logic. Various professional groups, say, doctors, managers, insurers and IT professionals, have different, often conflicting interpretive frameworks, resulting in fragmented understanding, goals and priorities (Dougherty, 1992). Fügener et al. (2017), for example, found that surgeons make irrational choices in trading off the risks of over- and underutilization of ORs, as they are unaware of the relative costs of these, and in general, are not trained in interpreting such choices in terms of their economic consequences. Managers and insurers, on the other hand, may be unaware of the concerns of high priority in the interpretive frameworks of surgeons.

Johnson et al. (2020) draw attention to this complex social and political context, claiming that “To date, research on process improvement in healthcare has tended to marginalize its multi-jurisdictional nature.” They found that regular multi-jurisdictional meetings are an effective routine for improving patient flow. Even a good schedule does not provide enough coordination (Venkataraman et al., 2018) and must be complemented by intense communication to coordinate between the different perspectives across organizational boundaries (Dreyfus et al., 2020; Stefanini et al., 2020). The TPM meetings on the tactical level, and the broker-planners on the operational level are examples of such boundary-spanning coordination mechanisms.

Discrepancy D5 (Actual goals are more complex than outcome measures suggest) was observed both in theory and practice. The MS/OR literature tends to take the goals as summarized in Figure 1 as given, focusing on resource utilization and its trade-off with overruns, and we did, in general, not find critical consideration of their meaningfulness. In practice, they monitor simple metrics such as utilization and cancellations. As discussed in Section 4, however, respondents acknowledge that the actual goal structure is more complex and ambiguous.

This ambiguity may be partly explained by the multiplicity of stakeholders as discussed above in Section 5.4. Surgeons, management and insurers will have partly conflicting views on goals and priorities, stemming from their different interpretive frameworks, backgrounds and responsibilities (Dougherty, 1992). Research in OM shows that the goal structure is also inherently more complex than reflected in current MS/OR literature, for example, when optimizing OR schedules for the outcomes in Figure 1 has negative consequences for quality outcomes, resulting in higher rates of post-surgical complications (Diamant et al., 2023).

The goal structure is further complicated by the fact that the operating theater is part of the larger hospital and healthcare system in the area: surgical processes are interdependent with up- and downstream services. These interdependencies are becoming more complex as the aging population brings about greater demand for integrated and multidisciplinary care. Also, the trend towards more personalized medicine and patient-centered care makes processes more interwoven (Beliën et al., 2025). This suggests a need for research to develop a more integrated view of OR planning, focusing on entire patient pathways within and across healthcare providers (Liu et al., 2019; Rachuba et al., 2024). Integrated planning in hospitals is impeded by a lack of standardization of work and norms (mirroring our P1), unreliable or outdated IT systems, compartmentalized performance management done at the department level, and low process visibility (Drupsteen et al., 2016). The TSEF poses that the overarching goal should be a swift and even patient flow through the system (Devaraj et al., 2013). We propose that (P5) MS/OR scheduling techniques may be more effective if they adopt a systems perspective.

Uncertainty and variability are key challenges for OR management. Stemming from its focus on ex ante scheduling, MS/OR techniques predominantly use capacity buffers to absorb uncertainty. Our D6 (Table 2) noted the discrepancy with practice, which mostly deals reactively with uncertainty and in an improvised, unsystematic manner.

Creating an even flow is a key concept in the TSEF (Devaraj et al., 2013). Jack and Powers (2004) list common ways in which even flow is achieved by responding flexibly to variability and uncertainty: using overtime/flex workers, cross-training workers, offering complementary services and improving forecasting. Responding to uncertainty and variability flexibly may be more efficient than capacity buffering, as the latter implies idle time and underutilized resources. We propose that (P6) OR management may be improved by complementing MS/OR scheduling techniques by flexibility strategies such as deliberate responsiveness, flexibility by design, uncertainty reduction and focus. We elaborate on these approaches below.

By deliberate responsiveness, we mean that uncertainty and variability are not tackled by creating slack in the schedule to absorb them, but by flexibly deploying resources and adjusting the process to accommodate them. Such a planning process can be characterized as predictive-reactive (May et al., 2011), where a baseline schedule is created ex ante, and where uncertain and unpredictable events are handled reactively during execution. We observed this way of working in the cases, but in an unsystematic, improvised manner, resulting in commotion, stress and on-the-fly solutions. We propose that there is much to gain from systematizing responsiveness as the primary approach for dealing with uncertainty. Responsiveness is then not applied as a makeshift solution, but as a deliberate, rational strategy for an environment with much uncertainty and dynamics, as is explored by Van Essen et al. (2012).

Responsiveness is leveraged by maximizing flexibility. By flexibility-by-design, we refer to the purposeful design of OR processes to be as flexible as possible and better able to adapt. For example, one respondent was promoting the idea to test a strategy where patients are initially scheduled for a particular week, but not yet for a specific day or hour. Three weeks in advance, patients are assigned to a specific day, but not yet to a specific time. The day before the surgery, the specific times are set. Instead of committing OR time long in advance (and thus reducing the flexibility to respond to later events and information), options are kept open as long as possible. A similar idea was proposed by Ferrand et al. (2014), which is also related to the set-based design approach in design engineering (Toche et al., 2020). Other flexibility-by-design approaches that we found in practice include.

1. Pooling of resources and ORs (e.g. between hospitals in an area). The earlier-mentioned open or integrated block scheduling strategies (Day et al., 2012) apply this principle by pooling demand for various specialties, deploying ORs flexibly on a first-come-first-served basis.

2. Systematize the use of flexible work hours, for example, by working with two-stage procedures: a baseline schedule based on forecasts, and last-minute adjustments using overtime or shift cancellations (Kim and Mehrotra, 2015).

3. Discern between tasks that are inflexible (must be done at a given moment) versus tasks that are flexible (can be done when a gap in the schedule emerges). Instead of framing idle time as a loss, processes could be designed such that surgeons and staff use it to do tasks that are flexible, such as administrative and educational tasks, e-consults or taking a necessary break.

4. Designing ORs and equipment for flexibility and cross-training staff for flexible deployment (Fügener et al., 2018).

A sensible preliminary step is to systematically consider whether variability and uncertainty can be reduced, thus applying the Law of Variability in the TFEF. In the context of OR scheduling, this seems underexplored both in literature and practice. A systematic consideration of the sources of uncertainty listed in Figure 1, and how they can be made less variable or less uncertain by using predictive analytics, is likely to result in approaches that are more efficient than absorbing variability in slack and buffers. Finally, the Law of Factory Focus in the TFEF (Venkataraman et al., 2018) suggests that variation can be reduced by the principle of focused factory. For example, the Netherlands government is implementing a strategy where complex surgery is concentrated in academic hospitals, channeling routine operations to peripheral hospitals. This could make surgery in the latter more predictable as variability in durations and medical cancellations decreases.

The planning and control of complex service systems seems a natural application area for analytical decision-support systems. There is a substantial and rapidly growing literature in MS/OR developing algorithmic methods for OR scheduling. These technically refined methods struggle, however, to gain traction in healthcare practice, and we find disappointingly little evidence of real-world impact. We study this gap through the lens of OM theory and methodology. We make a synthesis of the MS/OR literature on OR scheduling, which we call the theoretical perspective. Next, based on an elaborate case study in nine hospitals, we study the routines and realities of OR planning in practice. Interpreting the observed discrepancies between MS/OR theory and healthcare practice from theory in OM, we propose how scheduling methods may be better aligned with practice by modifying their premises or complementing them with alternative theories in OM.

We show that the TSEF (Devaraj et al., 2013; Venkataraman et al., 2018) is a useful framework for understanding the purpose and solution strategies of algorithmic scheduling approaches, thus connecting the technical literature in MS/OR to theory in OM. Our study shows that the problem structure assumed in MS/OR, on the strategic, tactical and operational levels, is valid. Our analysis suggests, however, that to deploy algorithmic scheduling approaches effectively in complex service organizations, there are some prerequisites and complementary principles with which MS/OR theory must be integrated.

Building on the literature on problem solving (e.g. Tucker et al., 2020), we found that a reactive workaround culture is a barrier to the deployment of sophisticated decision support methods. In line with wider theory on process innovation (Adler et al., 2003; Kim et al., 2012; Gutierrez et al., 2022), our empirical findings highlight the key role of a CI culture as a facilitator for the deployment of MS/OR techniques. Earlier research studied the role of standardization and simplification in improving operations performance (Roemeling et al., 2017; De Regge et al., 2019). Our study identifies their role as a prerequisite for the deployment of advanced analytics methods.

Our research contributes to the literature investigating coordination in complex, multi-jurisdictional organizations (Johnson et al., 2020; Stefanini et al., 2020; Dreyfus et al., 2020). We found that the complex and ambiguous goal structure in healthcare necessitates negotiation and brokerage structures to complement ex ante calculated schedules, while at the same time calling upon the MS/OR literature to adopt the framework of satisficing and a perspective of systems integration.

We share the findings in extant theory on flexibility (e.g. Jack and Powers, 2004) that uncertainty and variability should be addressed by a combination of strategies, including buffers in precalculated schedules, but also including responsive approaches. The resulting combined approach could be characterized as predictive-reactive (May et al., 2011).

Besides OM theory, we found empirical research approaches in OM to be a valuable complement to the mathematical research tradition in MS/OR, thus responding to the call of Beliën et al. (2025) to embrace interdisciplinary approaches. Critical premises about goal structure and constraints are often treated as given in the scheduling literature, without empirical research to corroborate them, resulting in misalignments with practice and sometimes oversimplified problem definitions. An integration of research in MS/OR with OM, therefore, is proposed to strengthen its real-world impact.

The propositions P1 through P6 (Table 2) offer a basis for an integrative, prescriptive theory for improving OR management. Preliminary efforts should focus on simplification and standardization of OR processes and constraints and preferences, and also on strengthening a culture promoting continuous improvement and sound methods instead of problem-fixing based on workarounds. Structures for OR planning should have a predictive-reactive design, where a baseline schedule is created ex ante, and uncertain and unpredictable events are handled reactively during execution. Scheduling should follow the framework of satisficing and should be based on a systems perspective. The primary response to uncertainty should follow the approaches of uncertainty reduction, deliberate responsiveness and flexibility by design. In view of the complex, multi-jurisdictional context, scheduling should also be complemented by structures for coordination and negotiation, such as the broker-planners and TPMs.

We discuss limitations arising from the study's context and methodological design. The setting for the case-study was the Dutch healthcare system, which is comparable in terms of institutional logic and policy to that of Canada, Denmark and the UK (Reibling et al., 2019). As in many countries, the affordability and especially staffing of the system are under pressure (De Visser et al., 2021). The Netherlands has prominent MS/OR groups with close collaborations with hospitals, making it a particularly fertile context for examining the application of MS/OR in healthcare practice. Our findings are compatible with the OM literature used to analyze them, which has no country-specific context. This suggests that the general conclusions may extend beyond the Dutch setting, but specific forms in which conclusions manifest may vary by institutional context. For example, the noted complexity of the goal structure (discrepancy D5) is likely to be universal, but differences in ownership structure (private versus government-run), the role of insurers and incentives embedded in the financing system may shape how specific priorities and goals emerge across countries.

The exploratory nature of the study's second part implies inherent limitations to generalizability. Insights provide directions for further research, not definitive claims. Table 2 suggests theoretical lenses that may guide future work aimed at testing, refining, or extending our propositions P1 through P6. In the first part of the study, we focused on the question of alignment of theory with practice. Other lenses through which the low adoption of MS/OR in practice could be studied include those of absorptive capacity, technology adoption and implementation. Even if MS/OR theory and techniques are well-aligned with the needs and realities of practice, getting them successfully implemented is a challenge in itself. Besides elaborating the integrative strategy proposed above, research identifying effective implementation pathways for OR planning techniques on the strategic, tactical and operational levels may help strengthen the real-world impact of MS/OR and contribute to closing the persistent gap between theoretical advances and their uptake in practice.

The paper has benefitted greatly from the thoughtful feedback from the Associate Editor and the editorial team.

The supplementary material for this article can be found online.