1
The relationship between the planning of specialist-time
and patient flow: a supply chain perspective.
June 2013
University of Groningen – Faculty of Economics and Business
Monique Bakker
Research Master in Economics and Business Profile Operations Management & Operations Research
Student number: 1681508 Wielewaalplein 20 9713 BR, Groningen Phone: +31 (0) 627 867 068 E-mail: m.bakker.30@student.rug.nl
Supervisors/assessors
Dr. J.T. Van der Vaart2
Abstract
The literature does not offer an integral view on the allocation and scheduling of specialist-time – with variable availability – and how it affects the patient flow. Many “why” and “how” questions accompany this subject, e.g. how problematic is this type of variability? What is the effect at the level of a single process step? What is the role of the linkage between process steps in the supply chain? We conducted an explanatory in-depth case study within the surgical department of a 183-bed hospital in the Netherlands. A case study allows an in-depth analysis of a contemporary phenomenon within its real-life context. The sheer scope and complexity of our integrated view, which the literature lacks, restricts our resources to the full investigation of one single case. The rationale for the case we have selected is not only convenience sampling, but it is also our conviction that it is a representative, typical case, as the activities of the care process are typical of many other hospitals. Furthermore, the hospital under study was awarded best performing general hospital in the Netherlands by the “Dr. Yep” in 2011, an annual guide that compares hospitals on a wide range of topics.
Detailed patient flow data and the schedules of the specialists were analyzed in order to explore how variability in specialist-time and allocation of their time affects overall flow performance of the internal supply chain. To fully understand the planning considerations, priorities and rules, semi-structured interviews were conducted with all specialists, the department’s outpatient coordinator and the department’s head of planning. Findings support our proposition that specialist-time as key resource binds two supply chain activities together with capacity management and that variability in the availability of this resource negatively affects patient flow. In line with literature, variability in specialist-time availability has a negative impact on patient flow and service offered to patients. We discuss possibilities for hospitals to reduce artificial variability in the internal supply chain and to better cope with remaining variability.
The value of this study lies in its integrative perspective, since little is known about effective planning of all activities of specialists along the entire internal supply chain, given the variability in available specialist-time. Optimizing either outpatient scheduling or operating room scheduling separately is no longer sufficient in an attempt to overcome inefficiencies of the entire internal supply chain. An integral approach to the allocation of specialist-time provides opportunities for hospitals to improve the patient flow in their internal supply chains.
3
Acknowledgements
4
Table of Content
Abstract ... 2 Acknowledgements ... 3 Table of Content ... 4 Introduction ... 5 Theoretical Background ... 6 Capacity management ... 7Swift and Even Flow ... 8
Research method ... 9
Data collection ... 9
Analysis ... 9
Validation ... 10
Results ... 11
The surgical specialism ... 11
Variability in specialist-time availability ... 11
Consequences for the patient flow ... 13
Overcrowding at the OC ... 14
Re-scheduled OC appointments ... 16
OR access times ... 17
Discussion and Conclusions ... 18
Theoretical implications ... 18
Managerial implications ... 19
Limitations ... 20
References ... 21
Appendix I : Template week schedule of Surgical department ... 24
Appendix II : OR Master schedule of 2011 ... 25
Appendix III: Logic model - Time sequence of planning activities ... 26
5
Introduction
One thing that is certain in health care is that hospital patients must be patient. It is not uncommon for a patient to wait up to 8 weeks or longer for any relatively simple surgical procedure, or over 2 weeks for a 5-minute appointment. There can be many reasons for long process lead times or access times, such as a mismatch of supply and demand, but there is one important factor that is known to inhibit any multi-stage process flow performance: variability. Variability, which can come in different shapes and sizes, is often an important factor causing inefficiencies in a supply chain. (Fredendall, 2009; Schmenner and Swink, 1998).
With the theory of Swift Even Flow as a theoretical lens (Schmenner, 2001; Schmenner and Swink, 1998), we closely examine the impact of variability in the availability of specialist-time on patient flow performance. By means of an in-depth case study we aim to uncover exactly how this type of variability affects patient flow, given that the specialist is the key resource involved in multiple activities of the care process, e.g. diagnosis at the outpatient clinic and surgery in the operation room. Patient flow performance is defined as the swiftness and evenness with which patients progress through the distinct supply chain steps of outpatient consultations, surgery, and follow-up consultations. The type of variability that we focus on is the variation in the availability of the key resource: specialist-time. Capacity management is the label that we allot to the collection of planning and scheduling rules, procedures, and hierarchy of decisions that determine how specialist-time is managed and allocated. Of particular interest is the integration of these planning efforts of the outpatient clinic (OC) on one hand, and the operating room (OR) on the other, since these are two activities of the same internal supply chain that both require specialist-time.
Literature on outpatient appointment scheduling (e.g. Cayrli and Veral, 2003) and on operating room scheduling (e.g. Cardoen et al. 2010) is abundant. However, very few authors take an integrative, supply chain approach towards planning and scheduling of outpatient clinic sessions and surgery, the two activities that are concerned with appointment scheduling and operating room scheduling, respectively. With the specialist as the key resource and these two main activities as process steps of the same internal supply chain, these activities are bound together with the management and allocation of the specialist-time resource. With regard to capacity management they can thus not be treated as separate.
6
The value of this study lies in its integrative perspective, since little is known about effective planning of the main supply chain activities of the key resource – the specialist – while the availability of this resource is highly variable. Optimizing either outpatient scheduling or operating room scheduling separately is no longer sufficient in an attempt to overcome inefficiencies of the internal supply chain as a whole. An integral approach to the management and allocation of specialist-time provides opportunities for hospitals to improve their patient flow performance.
Theoretical Background
Distinct departments involved in the different steps of a care process traditionally focus on their internal processes and costs. In general they are not naturally inclined to coordinate their activities (Drupsteen et al., 2013). Most hospital departments plan their activities independently (Lega and DePietro, 2005), even when the departments are part of the same internal supply chain. Nevertheless, from a patient flow perspective, it can prove profitable not to treat the (planning of these) activities as independent from one another.
Relationship between the OR planning and the OC planning
The operating room (OR) and the outpatient clinic (OC) act as communicating vessels in the sense that allocating specialist time capacity to the former automatically reduces the capacity that is left for the latter, and vice versa. This is why, when capacity is managed, the OR and the OC cannot be treated as independent of one another. In the healthcare operations literature also, the focus is often on optimization of a single, confined hospital process, rather than planning for the supply chain as a whole. Most studies analyse either demand (e.g. outpatient appointment scheduling) or capacity (e.g. operating room planning), without exploring how these aspects might jointly affect patient flow. When looking at the capacity of the key resource, the specialist, we find that over the past 60 years scholars have addressed the capacity planning of appointment scheduling at outpatient clinics (e.g. Gupta and Denton, 2008; see Cayirli and Veral, 2003 for an overview), and of the OR (Devi, Rao and Sangeetha, 2012; Cardoen et al., 2010), but these and other papers only focus on a fraction of the capacity of the specialist. Papers that take into account the total capacity of specialists, and that concentrate on scheduling all the activities of specialists, are scarce and hard to find (Winands et al., 2005). A supply chain perspective in this area of research is lacking in the literature. Besides the necessity of a supply chain perspective, the variability in the availability of the key resource specialist-time, and how this relates to the flow of patients through the internal supply chain, has not been researched so far.
Variability in the availability of specialist-time
7
availability of specialists. The types of variability that have been studied most are generally related to process-times (e.g. Larsson, 2012; Elkhuizen et al., 2007; Ho and Lau, 1992; Rising
et al., 1973; White and Pike, 1964; Bailey, 1952) or to physical (e.g. bed) capacity (Kusters
and Groot, 1996; Beliën and Deulemeester, 2007). Other types of variability that were considered at an OC are variability in (number of) arrivals, in registration time and in departure from registration to waiting room (Chand et al., 2009). Liu and Liu (1998) consider random lateness for each physician, resulting in a random, variable number of physicians available during the early period of the clinic. This latter type of variability draws near to what we will investigate in this paper, but rather than looking at specialist availability at (part of) a single process step (the OC), we consider specialist availability in two key process steps of the internal supply chain.
Artificial variability is a type of variability that is created because schedules create peaks and troughs. As opposed to natural demand variability, it is often the result of scheduling and allocating limited hospital resources. Artificial variability should be eliminated, after which remaining natural variability should be managed optimally (Litvak et al., 2005; Litvak and Long, 2000). McManus et al. (2003), found that variability in time spent on scheduled procedures exceeded that of emergencies at their case study hospital, as the elective procedures were artificially scheduled according to highly variable schedules. Litvak (2005, p. 99) too states that, if adjusted for patient volume, the number of scheduled surgical admissions per time unit often vary more than admissions through the emergency department. It is clear from literature that a large part of variability in hospitals can and should be eliminated. In addition to the types of variability that have been studied meritoriously before, there is a need for a deeper investigation of variability in specialist-time availability, with the specialist being involved in multiple steps of the supply chain.
Capacity management
8
Swift and Even Flow
The greater the random variability, either demanded of the process or inherent in the process itself or in the items processed, the less productive the process is (Schmenner and Swink, 1998). Productivity here is output per unit of input. This is one of the laws of Schmenner and Swink’s Theory of Swift, Even Flow, derived from queuing theory. According to this theory, variability will in some way cause patients to flow less swiftly and evenly through the process. It thus induces long patient lead times and many patients waiting somewhere in the health care process. Especially variations in the number of patients that enter the system will keep causing variations in health care demand at process steps downstream. A part of this variation is natural, which hospitals try to smoothen by giving patients appointments according to certain schedules (Cayrli and Veral, 2003). But opposed to that, hospitals also introduce artificial variability in their internal supply chains. This variability is the result of irregular schedules or irregular availability of resources. Prioritizing certain supply chain activities at the expense of others, when capacity availability is low, is also a way of introducing artificial variability.
Based on what is known in the literature about variability in relation to capacity management and patient flow, we presume with a fair amount of certainty that self-introduced artificial variability in the availability of specialist-time negatively affects patient flow. We aim to uncover how variability in the availability of specialist-time affects patient flow. We are interested in how exactly a disproportionate allocation of specialist-time to two different supply chain activities causes additional variability at the level of each of those activities. How are patients’ access times affected when not only the overall availability of specialist-time, but also the amount of specialist-time allocated to each of the supply chain activities, is variable? How severe are the effects? What effects are there other than the expected increase in access times? What causes this variability, i.e. what are the planning considerations behind it, to what extent is it artificial and how much of it can be eliminated at what cost? We therefore direct our research at the following proposition:
Since specialist-time is the key resource in multiple supply chain activities, it inextricably binds those activities together in a triadic relationship with capacity management. We argue that the variability in the availability of specialist-time is of such a degree that it becomes problematic. Based on theory, we can assume that variability in the availability of this key resource in general, and at the level of specific supply chain activities, negatively affects patient flow performance. What we aim to uncover is the extent of this problem, how it arises, how exactly it affects patient flow performance at the level of distinct process steps.
Figure 1 – Triadic relationship between scheduling OR sessions, OC sessions, and Capacity management
Operating Room sessions
Capacity management
9
Research method
This study is an explanatory case study investigation of how variability in the available specialist-time and the allocation of their time to the different supply chain activities affect patient flow performance. The subject of our study is the surgical department of a 183-bed non-academic community hospital in the Netherlands. A single case study allowed in-depth analyses and understanding of capacity management in relation to patient flow, and the role of variability in the availability of specialist-time. The main limitation of our research design is that its generalizability to other specialties or hospitals might be restricted. The value of this research instead lies in its capacity to provide in-depth insights and rich detail in complex and dynamic processes, which are difficult to reveal by means of more cross-sectional research designs (Hall, 2006). Case studies provide an excellent vehicle for developing understanding (Meredith, 1998), which may have particular significance in a field where the subject matter is particularly complex (Stuart et al., 2002). Litvak and Long (2000) stress the need for very practical and healthcare specific research in collaboration with healthcare providers, and since we aim to answer “why” and “how” questions, an in-depth case study is the preferred method (Yin, 2003; Eisenhardt, 1989). Furthermore, our case study hospital is one of the best-performing hospitals in the Netherlands according to a study by the “Financieel Dagblad” (Financial Daily newspaper), September 20, 2011. It was awarded best performing general hospital in the Netherlands according to an annual guide (censored) that compares hospitals on patient satisfaction, financial performance, staff absenteeism, facility service, website quality, the general Healthcare Inspection indicators and hospital transparency. We therefore have reason to believe that our case study hospital is not an exceptional, or particularly bad performing hospital, which could indicate the use of a non-representative case.
Data collection
Detailed quantitative patient flow data and the schedules of the specialists were collected in order to investigate how the variability in available specialist-time and the allocation of their time affects the overall flow performance of the internal supply chain. Patient flow data is recorded constantly by the department staff for administrative purposes, through their medical enterprise software Chipsoft. Of interest to us is the registration of all patient visits to the outpatient clinic: date, time, type of visit, duration, and rescheduled appointments. We analysed the data on 15,700 patients who visited the OC in 2011. Data on patient surgeries were collected in a similar fashion from Chipsoft for the same period. All weekly schedules of the specialists of 2011 were used, containing all the specialists’ day-to-day activities. The Masterplan of 2011, which determines the weekly allocation of operating rooms to the surgical department and to other specialties, was also used in order to analyse the variability in the planned specialist-time allocation to the two main activities.
To fully understand the planning considerations, priorities, and rules, semi-structured interviews were conducted with all specialists, the department’s outpatient coordinator and the department’s head of planning. Observations were made and documented (i.e. tag-along with the head of planning, and with 4 of the 6 specialists). In addition, during the eight months of research much time was spent on-site and bi-weekly meetings with management helped develop a thorough understanding of all processes, rules, procedures and planning hierarchies. A combination of qualitative and quantitative data from multiple sources facilitates validation of data, which is known as data triangulation (Eisenhardt, 1989; Yin, 2003).
10
Our data analysis can be divided into four parts that were carried out in an iterative fashion. New information repeatedly inclined the researchers to go back to the stage of data collection and congregate earlier findings and sharpen earlier analyses. The first part of the analysis entailed the mapping of the internal supply chain processes of the surgical department. This step augmented our understanding of the supply chain activities and any resources involved, as well as planning efforts relevant to our research goal. The second part consisted of the analysis of all weekly schedules from 2011 of the specialists, in order to determine the degree of variability in overall specialist-time availability on a weekly basis. The third part consisted of the analysis of variability in the amount of specialist-time that was allocated to either outpatient clinic sessions, or operating sessions. The fourth and final part of the analysis entailed the assessment of patient flow performance, in which quantitative data on about 15,700 patient appointments in 2011 were analysed, from which we derived access times to the OC and OR. OC and OR access times are used as proxy for patient lead-time, as an indicator of patient flow performance. The lead-time of a (partially) diagnostic process is by nature difficult to assess, because of high uncertainty with regard to routing.
Another indicator of patient flow performance (i.e., smoothness and evenness of patient flow) is the number of rescheduled appointments on the hospital’s initiative, which was also calculated from this data. We used a sample from all rescheduled appointments in order to determine what fraction was rescheduled on the hospital’s initiative, since the cause of each rescheduled appointment is not registered (as is at e.g. the radiology department). With the following formula (e.g. Israel, 1992) we determined a sufficient sample size.
𝑛 = (N(zs/e)2)/(N-1+(zs/e)2)
With a population size of N = 6927, we applied a generally accepted confidence level of 95% meaning z = 1.96. We chose the most conservative estimated proportion of p = 0.5 which yields s = 0.25 (i.e. we do not have prior knowledge of the expected portion of rescheduled appointments on hospital initiative). We applied several different margins of error and settled for e = 3% since this yielded a required sample size of 925 rescheduled appointments, which was feasible given our time and resource constraints.
Validation
11
Construct validity was fortified through the use of multiple sources of evidence during the data collection phase, and having key informants review the draft case study report. During our analysis, a chain of evidence was established and maintained, which reassures the ability of an external observer to trace the steps of our inference (Stuart et al., 2002; Yin, 2003). External validity remains a major barrier in doing case studies. However, the health care process we investigate is not unique in its characteristics. Other surgical specialisms, as well as other hospitals have processes very similar to the one under study. Our aim was to generalize a particular set of results to a broader theory: the theory of Swift Even Flow, rather than to other case studies. No set of cases, no matter how large, is likely to deal satisfactorily with the complaint that it is difficult to generalize from one case to another (Yin, 2003). The reliability of our findings is enhanced through the use of a case study protocol, documenting the research procedures, and through data triangulation (Eisenhardt, 1989). Yin (2003) suggests that one should conduct case study research as if someone were always looking over your shoulder. In our case, this was literally done with the co-author continuously evaluating the work of the main researcher.
Results
Earlier, we proposed that specialist-time as key resource in multiple supply chain activities inextricably binds those activities together with capacity management. By means of closely examining all specialists’ weekly schedules, we have determined how much specialist-time variability the process under study contains in total, and at the level of individual supply chain activities. Furthermore, we have assessed access times at the OC and OR process steps. In addition to that, we have uncovered effects other than that on access times, which can be related to variability in specialist-time availability to a certain extent. These are: overcrowding at the OC and rescheduling of OC appointments.
The surgical specialism
The hospital has a contract with a partnership of six specialists. Each year about 20,000 visits are paid to the OC, and the specialists carry out about 6,500 surgeries. Patients enter the system with a first visit to the outpatient clinic. Next, they might have one or more repeat visits, possibly followed by surgery. After surgery, patients may visit the outpatient clinic for one or more post-surgery check-ups (monthly, quarterly, twice a year, yearly or once every two years). Over two-third of all patients are classified as general surgery patients, they can be served by any of the six specialists. In addition to general surgery, the specialists each have their own sub-specialism, which means that not every not-general patient can consult any specialist. There are three oncology specialists. One of them only does small surgeries and another is specialised in head/neck surgery. The third oncology specialist can serve all types of oncology patients. The fourth specialist does all children’s surgery in addition to head/neck surgery (except oncology). The two remaining specialists are specialised in vascular disease. At the beginning of each year a so-called Masterplan schedule is drawn up which allots operating time to all specialisms, of which the surgical specialism is one. After the master schedule is determined, remaining specialist-time can be allotted to the OC and other patient-related activities. A standard schedule is then drawn up for the other activities. The results indicate that due to all sorts of changes (e.g. due to specialist availability), the template of the standard schedule is hardly recognizable.
12
Our results show that the amount of time spent on distinct supply chain processes is subject to high degrees of variability, indicating that the patient flow is likely to suffer from this variability. Figure 2 depicts the number of half-day sessions, i.e. specialist-time, that was available, and that was spent on either surgery or on OC sessions for each week in 2011. In addition, the dotted line depicts the availability of the OR to the surgical specialism after the operating rooms have been allocated to e.g. gynaecology, urology, and orthopaedics.
The first graph in figure 2 depicts the weekly availability of specialist-time, revealing that the availability of this resource in one week can be as much as twice the amount it was in the preceding week. These extreme variations, the largest troughs, occur during the holiday periods: Spring, May, and Summer holidays (weeks 8, 18, and 30 – 32 respectively), and Autumn and Christmas holidays (week 42 and 52 respectively). During these weeks, the hospital runs a special so-called reduction schedule with fewer operating rooms in use. The main argument for this is that the OR’s support staff should take holidays during a concentrated time period, so that the support teams are either fully absent, or fully complete, and never incomplete. Similar arguments apply to the construction industry, where all construction workers are forced to take their holiday during three designated weeks in the summer. This is planned variability, which is accepted and anticipated. Nevertheless, it impedes a swift, even flow of patients. There are additional variations as a result of specialists taking time off, attending conferences or training outside the reduction-schedule periods. At all times, up to two specialists are allowed to be on leave simultaneously, which causes one-third reduction in overall specialist-time capacity, but this is not always strictly adhered to. Occasionally their holidays overlap, such that there are a couple of days when fewer than four specialists remain available. It occurred 9 times in 2011 that half of the specialists were absent simultaneously for at least one day. In addition to that, there are variations in the weekly
Figure 2 – Variability in specialist-time availability in 2011
0 5 10 15 20 25 30 35 40 45 50 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 Num be r of half -‐day se ssions Week
13
schedules without any particular reason. What remains of the weekly schedules after all holidays and other causes of variations have been incorporated, is a pattern that makes it highly unlikely to spread patients evenly by means of appointment scheduling, in order to maintain a swift, even flow of patients.
Figure 2 also shows that the number of OR and OC sessions is far from constant from week to week. Inevitably, they follow more or less the same pattern as the total available specialist-time, but the ratio of OR to OC sessions seems distorted. According to the general weekly schedules, and average number of OR and OC sessions in 2011, the number of OR and OC sessions should each be around 16 per week. In figure 2 however, it shows that in 21 of the 52 weeks, the number of OR sessions exceeded the number of OC sessions, while the number of OC sessions exceeded the number of OR sessions in only 12 of the 52 weeks. In the remaining 9 weeks the number of OR and OC sessions was (almost) equal. Note that this comparison does not take the size of the ‘gap’ between the number of OR and OC sessions into consideration. However, it is curious to note that when OR exceeds OC, the number of OR sessions is generally higher than the average of 16 sessions. When it is the other way around, with the OC exceeding the OR, the number of OC sessions is generally under or close to 16 sessions, only reaching an amount of 20 sessions once, in week 43. The point made here is that more often than not, the number of OR sessions exceeds the number of OR sessions while technically they should be equal. In the weeks when OC does exceed OR, the absolute number of sessions is likely to be insufficient to recover from the backlog.
The weekly OR availability is determined by the Masterplan schedule drawn up at the beginning of the year. This Master schedule is the same for every week, and allocates the operating rooms to each of the different specialties, or departments. Only the reduction-schedule weeks are an exception, when a separate Master reduction-schedule applies. Despite of the Master schedule being the same every week, we see variations in the OR availability. These variations are due to national holidays, absence of anaesthesiologists, or quality inspection visitations. From this analysis, we infer that there is ample variability in the specialist-time capacity made available to the OR to presume that the patient flow is negatively affected by it. Of course there is natural variability in patient demand, but the weekly schedules are not drawn up in an attempt to match natural demand variations. The weekly schedules that determine the number of OC and OR sessions are made six weeks or longer in advance, at a time when demand is not yet known. This means that the variability in the number of OC sessions scheduled per week does not result from variability in patient demand. Demand variability thus does not influence the changes made to the standard weekly schedules, which make the standard pattern become unrecognizable and the resulting schedules variable. Only occasionally, when lead times are already escalating, one additional OR or OC session would be scheduled in a week, but this evidently causes only a small variation (for the OR, this happened once at the end of 2011 and for the OC this happened once when the number of OC sessions was relatively low).
Consequences for the patient flow
14
with the number of First Patient Visits, which is practically the initial processes step in the supply chain (after requesting the first visit to the OC)1. This means that from week to week, a variable number of patients is being sent into the process, causing variable demand downstream, at the OR. The OR-time that is demanded each week – which is the number of surgery-hours demanded at the time the patient is added to the surgical wait list – also correlates positively with the number of OC sessions, with r = 0.52, p-value < 0.000. The OR-time that is demanded each week does not result from the number of OR sessions in that week, since the OR-time that is demanded in a particular week is measured at the moment when the patient is added to the surgery wait list, and not at the moment when the actual surgery will take place. Although correlation does not imply causation, we can see that there is a relationship between the number of OC sessions and a) the number of First Patient Visits, and b) OR-time demand.
Given the considerable variability in specialist-time availability, we analysed OC and OR access times to determine whether the patient flow encounters difficulties that can be related to those variations. For the OC, we focus on patients that enter the process for the first time, i.e. First Patient Visits, because these are the only appointments that have to be scheduled as soon as possible, for the patient’s convenience and to adhere to norms set by government: 80% within 3 weeks, all within 4 weeks. First Visit access times give us an honest and valid measure of access times. Repeat visits often have to have a predefined wait time, in order to assess whether a certain treatment is taking effect. Besides, since we have no data on the predefined wait times for repeat visits, and since other departments (e.g. radiology) may play a delaying role, we only focus on First Patient Visits.
Our data reveals that at any given point in time, on average, there will be 142 patients that have been or will be waiting for three days or longer, between the process steps of requesting a First Patient Visit appointment, and having that appointment which is their OC access time. This number of patients varies between 86 and 179, with a standard deviation of 17. On average 13 patients will be waiting for more than 4 weeks, which is only 3,5% of all new patients. This figure does not seem devastating for the OC. According to the interviews and observations, the OC access times for First Visits never escalate, because new and urgent patients have to get an appointment within the norm times set by government. This however has severe consequences for the daily workload at the OC, which is elaborated on in the following paragraph.
Overcrowding at the OC
Serving the patients well within the official access time norm has put the OC under high pressure. Patients with a suspicion of cancer or with severe pain have to be scheduled within 5 workdays. All specialists and the planning coordinator in the interviews agree that the OC sessions are too full, because urgent patients are squeezed into full sessions so the hospital can adhere to the access time norm. Especially in a week when relatively few sessions are scheduled, the sessions become overcrowded. The formal rule is that patients may not be double-booked, which means that two patients cannot be scheduled on a single time slot, but
1 This seems logical, but if the number of patients entering the system were purposefully kept more constant, the
15
the nurses that schedule them often have no choice. The specialists all claim that more often than not, their OC sessions are delayed by 20 minutes or more because there are too many patients and too little time2. However, they assured us that there is no structural capacity shortage. Not all OC sessions are excessively fully booked, which indicates that it is not simply a matter of too little capacity. Rather, the available capacity for OC sessions (as well as for OR sessions and in total) is variable.As one specialist stated in the interview (translated):
“The sessions are very full, there are many more patients scheduled in one session than the number of available time slots. This in combination with our additional tasks during a session – emergency patients, phone calls from colleagues, being paged – makes it impossible to serve all patients in time.”
Generally, hospitals use appointment systems in order to distribute demand evenly (Litvak, 2002). This means that appointment systems should be designed such that natural demand variability is transformed into known, even arrivals at appointed dates and times. Similarly, schedules are used so the different activities are all performed with certain regularity. If all goes according to a regular schedule, then for instance an average of 400 patients pass through the OC every week, with little deviation. If the schedule is not regular, with only 8 OC sessions one week and 19 the next (see figure 2), then evidently the number of patients that pass through the OC will vary too, from e.g. 200 one week and almost 500 the next. At the hospital under study, an average of 16 OC sessions per week is needed in order to meet demand. If in one week 13 sessions are scheduled, and in the next week 19, there is no real problem, but if only 8 sessions are scheduled in one week, the next week 24 would be required in order to compensate for the ‘dip’. Such great variations are not desirable because, as is known from queuing theory, it is associated with longer throughput times and many patients ‘buffered’ somewhere in the system. In addition, at the case study hospital physical capacity allows a maximum of 20 sessions in one week, which can only be attained if all specialists are available. So, if in one week only 8 OC sessions are scheduled, not all of the capacity loss can be compensated in the following week. Many patients thus incur additional access waiting time indicating that the patient flow suffers from schedule irregularities, caused by variable specialist availability.
Initially, no clear relationship could be uncovered analysing the correlation between weekly average OC access times and specialist-time availability in the same week. The weekly access times are measured at the moment of appointment request. However, comparing average weekly access times with specialist-time availability one week later (which is the most common wait time) results in a correlation of -0.25 significant at the 10% error level, with a
p-value of 0.074. This indicates that access times tend to be higher if patients request
appointments in week x while specialist availability is low in week x+1. Although a correlation does not imply causation, we can infer that there is a relationship between specialist-time variability and access times. We argue that this relationship would be stronger, perhaps with a higher confidence level, had the OC never conceded to overbooking their sessions.
Inevitably, demand variability also affects access times to some extent. We have assessed
2 Rather than access times, wait times in the waiting room escalate, which is also an indicator of patient flow
16
demand variability per time unit of one week since almost half (46%) of the new patients (First Patient Visit) require access times that do not exceed one week. Recall that this concerns the most urgent patients (oncology, severe pain). Furthermore, a one-week period was convenient for assessing specialist-time variability, since they use weekly patterned schedules. Maintaining this unit of time is thus not unnatural. The weekly average First Patient Visit demand is 129 appointment requests, with a standard deviation of 29, thus a coefficient of variation (CV) of 0.22. According to Hopp and Spearman (2008, p. 269), and Waller et al. (1999), this classifies as low variability3. First Visit demand in week x correlates positively with the number of patients waiting in week x (correlation of 0.34 with a p-value of 0.016), but not with wait times in week x+1 (p-value of 0.623). Nevertheless, had the OC schedules been more regular, support staff would have had a better opportunity of spreading demand. As the department’s head of planning stated in the interview:
“We really should not have fewer than 16 normal OC sessions a week [indicating that non-normal sessions means shortened sessions], preferably 18 or 19 sessions, at least. As soon as we are short on sessions, we have trouble giving appointments in due time. [...] and it is hard to catch up, because even if a third surgeon [specialist] is available, we only have two treatment rooms that we can use simultaneously4.”
Furthermore, as discussed earlier, the number of OC sessions is not adjusted according to demand at that point in time. In other words, even if demand is high in a particular period, the number of OC sessions is not increased (rather the sessions become overfull). So we can exclude variable demand resulting in a variable number of OC sessions, in turn negatively affecting OC access times.
Re-scheduled OC appointments
The analysis of the patient flow furthermore revealed that as much as 7,8% of all patient-specialist appointments at the OC is rescheduled at least one day in advance, due to patient-specialists becoming unavailable5. This is another indication that the patient flow is hindered. Regardless of the amount of specialist-time available, by hook or by crook specialists and support staff make sure that all urgent patients (e.g. oncology) are always served in due time, at the OC as well as at the OR stage. This affects the workload at the OC in two ways. Firstly, OC sessions become excessively full because urgent patients are added to already full sessions. Secondly,
surgery on urgent patients6 happens at the expense of OC sessions (recall from figure 2 that the number of OR sessions exceeds the number of OC sessions more often than the other way around, while they should be equal). The specialist has to do surgery on an urgent patient who does not fit into one of the regular OR sessions, thus (part of) his OC session is cancelled. According to the specialists, this particular reason for them becoming unavailable to the OC accounts for most of the 7,8% of rescheduled appointments. Unfortunately, we could not re-trace the exact cause of each rescheduled appointment or cancelled OC session. What we do
3 It must be noted that Hopp and Spearman’s classification is with reference to process times, not weekly (health
care) demand patterns. Waller et al. specified a range of demand variability for modelling purposes.
4 Further inquiry revealed that an additional treatment room would not be profitable, since it only rarely occurs
that a third specialist is available (about 10 half-day times a year according to the interviewee).
5 With the margin of error set at 3% (see Analysis), we can state with 95% confidence that this percentage lies
somewhere between 4,8% and 11,8%.
6 Note that urgent surgery (e.g. oncology) is not the same as emergency surgery, for the latter by definition
17
know is that cancelling OC sessions introduces additional variability in the (weekly) number of sessions.
Generally, 10% of each OR session duration is reserved for emergency and urgent surgical procedures. Logically, 10% of the durations of 16 OR sessions (which is the average of the initial Master schedule) evenly distributed over the week entails more slack to cope with urgent patients, than 10% of only 8 OR sessions. Therefore, if the number of OR sessions were more constant, the need to cancel an OC session to make room for urgent surgery would occur less often. The specialists as well as hospital management had not been aware that specialist-time availability was subject to variability of this magnitude, until this research project. Therefore, reserving a pre-set percentage of OR-time rather than an absolute number of hours/minutes had not raised any doubts thus far, but might prove to be viable as long as this type of variability is not reduced to a minimum.
Check-up appointments at the OC are made 8 weeks up to 2 years in advance. This concerns about 38 patients per week who are given an appointment according to the standard weekly schedule. Problems occur when the final schedule no longer resembles the template. Therefore, we must note that if patients are given their check-up appointment less far in advance, once the final schedule is fairly certain, this should also result in less rescheduling.
OR access times
Since 10% of OR-time is reserved for urgent and emergency surgeries, it implies that all elective surgical procedures must fit into the 90% of OR time that is left, which is a carefully calculated percentage, based on demand and quota agreements. As is the case at the OC, emergency patients will always be served immediately, and urgent patients within 3 to 5 days. At the OR this means that these two groups of patients are served at the expense of elective patients, when the 10% reserved time is insufficient. The planners try not to cancel scheduled elective surgeries when the patient is already notified of his/her surgery date. They would rather postpone them by a few hours, which is sometimes an option. This is not desirable, but emergency patients are in a life-threatening situation and need to be served immediately. Urgent patients do not require emergency surgery, but as was mentioned, can generally wait no longer than 3 to 5 days. These patients therefore often cause elective patients to incur additional wait time, when the 10% reserved OR time is insufficient. We analysed the access times of two exemplary groups of elective patients with carpal tunnel syndrome (CTS), or inguinal hernia. These two groups of patients make up almost 30% of all electives. The so-called ‘Treek’ norms set by the Dutch government state a maximum wait time of 6-7 weeks for clinical treatment, but this is often not achieved. No less than 24% of all CTS patients, and 28% of inguinal hernia patients7 exceeded a 7-week wait period for surgery in 20118. The access times of all elective patients varied between 3 and 330 days, excluding the patients that are put on the waiting list with an urgency indication of ‘on patient’s request’ or ‘on specialist’s request’.
Variability in specialist-time available to the OR has a knock-on effect on downstream
7 It is difficult to state such percentages and access times for other patient groups since they may require other
process steps, such as radiology, adding wait time for reasons other than the variability in specialist-time availability.
8 All specialists agree that it is not demand exceeding capacity that causes long wait times, but that it must be
18
processes in the same way it does at the OC: only 8 OR sessions in one week causes patients – electives – to incur additional access time, while more than twice as many OR sessions the following week (e.g. weeks 18 and 19 in figure 2) may cause congestion further downstream, at the post-surgery clinic (which is outside the scope of this study, since the specialist is not involved). Note that the process times of surgeries are far more variable than that of OC visits. An OR session can be filled with 2 long procedures, or with 10 short ones, so when in one week specialist-time available to the OR is 50%, it does not necessarily mean that 50% of the average number of patients will progress through this process step in that particular week, it could be more. However, in case of low specialist-time, urgent and emergency procedures will have priority, while these are on average long procedures, with µ = 41 and σ = 44 minutes. The elective procedures, CTS and inguinal hernia, have an average process time of µ = 7 and σ = 6 minutes, and µ = 26 and σ = 13 minutes respectively. Therefore, specialists cannot simply choose to keep the number of patients that pass through the OR process step as constant as possible; the urgent patients have priority.
The problem of variability at the OR is intensified because, once patients ‘accumulate’ due to low specialist-time availability, this backlog is not easily made up for. When more than sufficient specialists are available and they want to gain on their list of electives, the number of ORs available to them is restricted. This is experienced first-hand by the planners, who claim that especially the simple, elective, short procedures keep being pushed forward, and that it is difficult to explain to patients that they have to wait two months for a 10-minute procedure.
Discussion and Conclusions
According to Cardoen et al. (2010) healthcare operations literature is dominated by studies that focus on the optimization of single steps in care processes, with fixed or independent capacity (see Cayirli and Veral, 2003 for an overview). Our aim was to provide insight in the relationship between variability in the availability of the key resource (capacity) and patient flow performance, looking at more than one single process step. This research thus sheds light on a particular, important type of artificial variability in specialist-time availability, where it sources from, and how it affects patient flow performance, from a supply chain perspective. In addition we have shown how variability in availability of specialist-time can affect the workload at the OC (overcrowding) since all urgent patients will always be served within a maximum of 5 days, regardless of the number of OC sessions and the fullness of those sessions. Furthermore, rescheduled OC appointments (on the hospital’s initiative) were analysed in light of a variable number of OC sessions each week, while having to cope with on-time serving of urgent patients.
Theoretical implications
19
variability that should be eliminated, but there is more to it.
With the specialist being the key resource in multiple internal supply chain activities, it is surprising that this type of variability had not yet been subject to close examination in the health care operations literature, let alone from a supply chain perspective. The allocation of specialist-time to OR directly influences the available time that can be allocated to OC, and vice versa; they are inextricably bound together with capacity management. This is a remarkable characteristic that only becomes eminent when we apply a supply chain perspective. From White et al. (2011) we know that in outpatient healthcare clinics, capacity, patient flow, and scheduling are rarely managed in an integrated fashion, and that integration improves clinic performance. We add to this knowledge by including multiple supply chain activities that involve a key resource, i.e. the specialist.
Our results are in line with Litvak and Long (2000), in that eliminating artificial variability in healthcare systems has much potential for reducing waste, in this case excessive access times and rescheduling. Our study revealed considerable variability and clear signs of patients not moving swiftly and evenly through the process. We have attempted to explicate the relationship between specialist-time variability and patient flow performance, but we must yield to the fact that it remains difficult to establish a causal relationship that excludes all other possible causes of high access times, such as demand variability. We know from theory that variability – especially artificial variability – is detrimental to the process and should be eliminated. Variability in specialist-time availability is such an artificial variability, of which we have aimed to expose its relationship with patient flow performance.
Managerial implications
As a result of our investigation we conclude that by means of capacity management, efforts should be directed towards reducing variations in total specialist-time, as well as per supply chain activity. This will allow patients to flow more swiftly and evenly through the process, which is in line with, and perhaps the cornerstone of the theory of Swift, Even Flow. Knowing that not all variability can be eliminated (or at what cost), it should be kept within acceptable boundaries by means of integrative planning and allocation of specialist-time to the different supply chain activities. In this way, patient flow performance is less likely to suffer from variability in the availability of specialist-time. In terms of capacity management, we thus argue that allocating specialist-time to the main supply chain activities in an integrative fashion helps further reduce variability at the level of each of those activities. This means that the OR should not have priority at the expense of the OC if that causes additional variability. We analysed access times of patients’ first visits to the OC, and access times of the OR for elective patients, as indicators of patient flow performance.
20
capacity. Surely check-up patients do not affect the workload downstream. Many hospitals in the Netherlands schedule check-ups up to 1,5 or 2 years in advance. These appointments can easily take place a few weeks earlier or later with no medical consequences. Note that these appointments must then be made less far in advance, once the availability of specialist-time is more or less known.
Limitations
The complex nature of hospitals and their care processes makes it difficult to control for all the relevant variables that might influence performance (Drupsteen et al., 2013). Data triangulation and maintaining a chain of evidence has helped us link the variables in order to oppugn this problem, but one must keep in mind that findings are supportive of our proposition rather than decisive.
21
References
Bailey, N. T. (1952), “A study of queues and appointment systems in hospital out-patient departments, with special reference to waiting-times”, Journal of the Royal Statistical Society. Series B (Methodological), Vol. 14, No. 2, pp. 185-199.
Beliën, J. and Demeulemeester, E. (2007), "Building cyclic master surgery schedules with leveled resulting bed occupancy", European Journal of Operational Research, Vol. 176, No. 2, pp. 1185-1204.
Cardoen B., Deulemeester E., Beliën E. (2010), “Operating room planning and scheduling: a literature review”, European Journal of Operational Research, Vol. 201, No. 3, pp. 921-932.
Cayirli T., Veral E. (2003), “Outpatient scheduling in healthcare: a review of literature”, Production and
Operations Management, Vol. 12(4) pp. 519-549.
Chand S., Moskowitz H., Norris J.B., Shade S., and Willis D.R. (2009), “Improving patient flow at an outpatient clinic: study of sources of variability and improvement factors”, Health Care Management
Science Vol. 12, No. 3, pp. 325-340.
Drupsteen, J., Van der Vaart, T., and Van Donk, D. (forthcoming). “Integrative practices in hospitals and their impact on patient flow”, International Journal of Operations and Production Management.
Eisenhardt, K. M. (1989), “Building theories from case study research”, The Academy of Management
Review, Vol. 14, No. 4, pp. 532-550.
Elkhuizen, S. G., van Sambeek, J. R., Hans, E. W., Krabbendam, K. J., & Bakker, P. J. (2007), “Applying the variety reduction principle to management of ancillary services”, Health care management review, Vol. 32, No. 1, pp. 37-45.
Fredendall, L.D., Craig, J.B., Fowler, P.J. and Damali, U. (2009), “Barriers to swift, even flow in the internal supply chain of perioperative surgical services department: a case study”, Decision Sciences, Vol. 40, No. 20, pp. 327-349.
Gupta, D. and Denton, B. (2008), “Appointment scheduling in health care: challenges and opportunities”, IIE
Transactions, Vol. 40, No. 9, pp. 800-819.
Hall, P. A. (2006), “Systematic process analysis: when and how to use it:, European Management Review, Vol. 3, No. 1, pp. 24–31.
Ho, C. J., and Lau, H. S. (1992), “Minimizing total cost in scheduling outpatient appointments, Management
Science, Vol. 38, No. 12, pp. 1750-1764.
Israel, G. D. (1992), Determining sample size, University of Florida Cooperative Extension Service, Institute of Food and Agriculture Sciences, EDIS.
Kusters, R. J., and Groot, P. (1996), “Modelling resource availability in general hospitals design and
implementation of a decision support model”, European Journal of Operational Research, Vol. 88, No. 3, pp. 428-445.
Larsson, A. (2012), “The accuracy of surgery time estimations, Production Planning & Control: The
Management of Operations, DOI: 10.1080/09537287.2012.666897.
Litvak, E., Buerhaus, P. I., Davidoff, F., Long, M. C., McManus, M. L., and Berwick, D. M. (2005), “Managing unnecessary variability in patient demand to reduce nursing stress and improve patient safety”, Joint Commission Journal on Quality and Patient Safety, Vol. 31, No.6, pp. 330-338.
Litvak, E., and Long, M. C. (2000), “Cost and quality under managed care: irreconcilable differences”, The
American Journal of Management Care, Vol. 6, No. 3, pp. 305-312.
Liu L, Liu X. (1998), “Block appointment systems for outpatient clinics with multiple doctors”, Journal of the Operational Research Society, Vol. 49, No. 12, pp. 1254–1259.
Meredith, J. (1998), “Building operations management theory through case and field research”, Journal of
Operations Management, Vol. 16, No. 4, pp. 441-454.
Rising, E. J., Baron, R., and Averill, B. (1973), “A systems analysis of a university-health-service outpatient clinic”, Operations Research, Vol. 21, No. 5, pp. 1030-1047.
Schmenner, R. (2001), “Looking ahead by looking back: swift, even flow in the history of manufacturing”,
Production and Operations Management, Vol. 10, No. 1, pp. 87-96.
Schmenner, R., and Swink, M. (1998), “On theory in operations management”, Journal of Operations
Management , Vol. 17, No. 1, pp. 97-113.
Swisher, J. R., and Jacobson, S. H. (2002), “Evaluating the design of a family practice healthcare clinic using discrete-event simulation”, Health Care Management Science, Vol. 5, No. 2, pp. 75-88.
Vasilakis, A., Sobolev, B. G., Kuramoto, L., Levy, A.R. (2007), “A simulation study of scheduling clinic appointments in surgical care: individual surgeon versus pooled lists”, The Journal of the Operational
Research Society, Vol. 58, No. 2, pp. 202-211.
22
chain", Journal of Business Logistics, Vol. 20, No. 1, pp. 183-203.
White, D., Froehle, C., and Klassen, K. (2011), “The effect of integrated scheduling and capacity policies on clinical efficiency”, Production and Operations Management, Vol. 20, No. 3, pp. 442-455.
White, M. B., and Pike, M. C. (1964), “Appointment systems in out-patients' clinics and the effect of patients' unpunctuality”, Medical Care, Vol. 2, No. 3, pp. 133-145.
Winands, E., De Kreuk, A., and Vissers, J. (2005), “Master scheduling of medical specialists”, in Vissers, J. and Beech, R. (Ed.), Health Operations Management: Patient flow logistics in health care, Routledge London.
Wright, M. B. (1987), “The application of a surgical bed simulation model”, European Journal of
Operational Research, Vol. 32, No. 1, pp. 26-32.
Yin, R. K. (2003), Case Study Research: Design and Methods, Sage Publications, London.
