IMPROVING THE OUTFLOW OF PATIENTS
AT THE EMERGENCY DEPARTMENT
Using the admission predictions of physicians and nurses
MSc. Technology and Operations Management
&
MSc. Supply Chain Management
University of Groningen
June 2019
A.J. Komdeur
a.j.komdeur@student.rug.nl
S2667606
Supervisors Dr. M.J. Land Prof. dr. J.T. van der VaartInformants MCL:
Drs. J.M.G Theunissen, SEH arts KNMG, MD Drs. R.S. Stolmeijer, SEH arts KNMG, MD Dr. H. Lameijer, SEH arts KNMG, MD PhD
Informant UMCG:
2
Abstract
Emergency departments (EDs) are experiencing crowding issues, which inhibits them to provide timely care to arriving patients. The main cause of ED crowding is the troubled outflow of ED patients to the inpatient department. This results in waiting time for the patient while unnecessary occupying a scarce ED bed. A proposed solution is to start the admittance process earlier, in parallel with the treatment at the ED, when an admission is predicted. Therefore, this paper examines the effect of an earlier start of the admittance process, based on the admission predictions of physicians and nurses, on the outflow of patients. To examine this effect, a single-case study is conducted at a hospital in the Netherlands. Here, the admission predictions of the physicians and nurses after one hour of treatment are analysed and compared. The results show that physicians are significantly more accurate in predicting admissions. Additionally, the results show that admissions are better predicted after one hour of treatment than during triage. This implies a trade-off between how far in advance the admission prediction is shared and how accurate the prediction is. Further, the results show that merely initiating an earlier start does not guarantee a reduction in waiting time. This study discovered multiple factors that can potentially cause a delay in the admittance process beyond the time slack created by the earlier start. Additionally, this study shows that due to particular circumstances an earlier start initiated at the ED, does not guarantee an earlier start of the activities at the inpatient department.
3
Contents
1. Introduction... 4 2. Theoretical Background... 6 2.1 Emergency departments ... 6 2.2 Crowding problems ... 6 2.3 Causes of crowding ... 6 2.4 Proposed solutions ... 72.4.1 Admittance process improvement ... 7
2.4.2 Use of prediction models ... 8
3. Methodology ... 9
3.1 Single-case study design ... 9
3.2 Hospital case description... 9
3.3 Data collection ... 9
3.4 Data analysis ... 11
4. Results ... 14
4.1 Phase 1: the admission predictions of ED physicians and nurses ... 14
4.2 Phase 2: identifying activities of the admittance process and causes of delay ... 16
4.3 Phase 3: effect of starting the admittance process earlier ... 20
5. Discussion ... 27
5.1 Admission predictions ... 27
5.2 The admittance process ... 27
5.3 Earlier start of the admittance process ... 29
6. Conclusion ... 30
6.1 Conclusion ... 30
6.2 Practical implications ... 30
6.3 Limitations and future research ... 30
References ... 32
Appendix I – Admittance process with admission department ... 36
Appendix II – Admittance process with evening-night coordinator ... 37
Appendix III – Interview protocol inpatient departments phase 2 ... 38
Appendix IV – Admittance process during intervention ... 39
Appendix V – Interview protocol phase 3 ... 40
Appendix VI– Admission predictions per specialism ... 42
Appendix VII – Complete analysis admittance process ... 43
4
1. Introduction
Emergency departments (EDs) provide treatment to the seriously ill and injured patients (Asplin et al., 2003). The effectiveness of the treatment at the ED relies on time-sensitive and rapid interventions (Hirshon et al, 2013). However, EDs have been struggling to provide timely care to arriving patients due to crowding problems (Anneveld et al., 2013). In many studies it has been acknowledged that outflow obstruction is the main cause of crowding (Rabin et al., 2012). This results in the inability to timely transfer admitted patients to an inpatient bed (Handel et al., 2010; Forster et al., 2003). To enable the ED in providing timely and effective care, the crowding issue has to be resolved. Therefore, this study focusses on improving the outflow of patients from the ED to the inpatient departments, which in turn should lead to a reduction of the crowding issue and enables EDs in providing timely care. Rabin et al. (2012) state that crowding is not merely an ED-based problem. Rather it is caused by the dysfunction in the interrelated parts of the whole hospital, considering that a hospital is a system where patients flow through multiple processes and departments. Drupsteen, van der Vaart and van Donk (2013) define patient flow performance as the speed at which patients move through the required processes and departments in the hospital. When patient flows are not managed properly, flow problems will emerge resulting in queues and delays (Villa, Prenestini and Giusepi, 2014). The flow problems at EDs have resulted in hampered outflow and delays in transferring admitted patients. Consequently, solutions must focus on improving these patient flows and specific on improving the outflow process by enabling efficient transfers to the inpatient departments (Abraham and Reddy, 2010).
Starting the admittance process of ED patients earlier may enhance patient flows (Peck et al., 2012) and by that improve the outflow of admitted patients. The admittance process consists of all activities necessary before a patient can be admitted to an inpatient department. These activities include requesting an inpatient bed, preparations at the inpatient department and transferring the patient (Qui et al., 2015). To enable an earlier start of the admittance process, it first has to be decided whether a patient has to be admitted and to which inpatient department. Multiple studies examined the accuracy of admission predictions made during triage (Qui et al., 2015). The results of these studies show that discharges can be predicted accurately, while accurately predicting admissions is difficult. A limitation is that these studies exclusively focus on admission predictions of triage nurses, whereas physicians are potentially more accurate in predicting admissions (Kosowsky et al., 2001). Physicians interpret the same information differently (Brillman et al., 1996), which could result in different and better predictions.
Despite the attention for admission predictions, the accuracy of the admission predictions of ED physicians and nurses has never been compared. Further, the effect of an earlier start of the admittance process on the outflow of patients has never been empirically tested. Therefore, the main research question is: How does an earlier start of the admittance process, using admission predictions, influence the outflow of patients at the ED?
6
2. Theoretical Background
2.1 Emergency departments
Asplin et al. (2003) developed a framework to illustrate patient flows through the ED, distinguishing input, throughput and output. Input refers to the arrival of patients. Characteristics of the input are the total volume of arriving patients and the severity of sickness (Mackway-Jones, Marsden and Windle, 2013). After arrival the throughput phase begins, which is the treatment and diagnosis of the patient. EDs are often divided into several areas like a waiting room, a triage area and treatment rooms. At triage the patients are prioritised according to the severity of their conditions. Depending on their triage level, patients have to wait or are treated immediately (Mackway-Jones et al., 2013). When the treatment and diagnosis at the ED are completed, the output phase starts. Here patients can be hospitalised, referred to an outpatient clinic or discharged to go home (Kao, Yang and Lin, 2015).
2.2 Crowding problems
The ageing population caused an increase in the demand for care. Not just volume has grown, but older patients generally need more and complex care (Morris et al., 2012). Unfortunately, these patients are also the ones likely to spend disproportionate time at the ED (Boyle et al., 2012). The increase in volume together with the increase in complexity of the provided care has led to crowding at the ED (Anneveld et al., 2013). ED crowding is a major issue for EDs since it compromises the care provided to patients (Van der Linden, 2015). According to Anneveld et al. (2013) crowding appears when a mismatch between the supply of resources and the demand for timely care is present, causing long waiting times and delays in critical treatment.
Van der Linden (2015) provides an extensive overview of the consequences of crowding discussed in literature. The main negative consequences are increased length of stay (Higginson, 2012), higher complication and mortality rates (Schull et al., 2004), and patient and staff dissatisfaction (Derlet and Richards, 2000). Eckstein and Chan (2004) add that the negative consequences of crowding reach beyond a particular ED. Ambulance crews are not allowed to transport a patient to the nearest ED when suffering from crowding, reducing their capacity to respond to new emergency calls.
2.3 Causes of crowding
The causes of crowding can be analysed using the three phases of patient flow: input, throughput and output (Boyle et al., 2012). As mentioned, input factors contributing to crowding are the increase in volume and complexity of care due to the aging population (Morris et al., 2012). Throughput factors relate to the treatment and diagnosis at the ED. Throughput factors that can hamper patient flow and contribute to crowding are poor ED design (Hwang et al., 2006) and delays in diagnostic imaging and laboratory results (Boyle et al., 2012).
7
2.4 Proposed solutions
Limited literature is available on effective solutions to reduce crowding (Boyle et al., 2012). Solutions can aim at solving either input, throughput or output related causes of crowding. As the output factors are acknowledged to be the main cause of crowding, it is obvious that solutions should aim at improving these factors. However, of the proposed solutions in literature, only few are focused on improving the output of the ED (Morris et al., 2012). These solutions are either focused on increasing available resources in the hospital (Hoot and Aronsky, 2008) or on improving the outflow process of admitted patient from the ED to the inpatient department (Khare et al., 2009).
An Acute Medical Unit (AMU) is a solution often proposed to increase available resources (Scott, Vaughan and Bell, 2009). An AMU is designed and staffed to receive acute patient from the ED and provide treatment for 24 to 72 hours prior to discharge or admission. Another solution is to increase the capacity of inpatient beds available for ED patients (Derlet and Richards, 2000). Increasing the available inpatient beds can be achieved by adding more beds, improving the discharge process at the inpatient departments or by cancelling elective care (Morris et al., 2012). However, increasing available resources is costly and hospitals are already under financial pressure (Van der Linden 2015) and cancelling elective care had negative consequences for the wider hospital (Lane, Monefeldt and Rosenhead, 2000).
In contrast, Khare et al. (2009) analysed the effect of an improved outflow process for admitted patients by ensuring rapid transfers to an inpatient bed. Their results show a decrease in both the average length of stay of patients and the number of boarders at the ED, which in turn should reduce crowding at the ED. Thus, improving the outflow process of admitted patients to the inpatient departments is likely to lessen ED crowding.
2.4.1 Admittance process improvement
The admittance process encompasses all activities needed to admit a patient to an inpatient department, like requesting an inpatient bed, preparations at the inpatient department and transferring the patient. Currently, the admittance process is generally initiated after the treatment at the ED is finished and the patient is allowed to leave the ED (Qiu et al., 2015). This approach causes a delay in the bed request and resulting preparations at the inpatient department (Peck et al., 2012), which in turn leads to outflow problems and boarders at the ED. A quantitative analysis of Witvoet (2018) also revealed that a substantial delay between the bed request at the end of treatment and the moment a patient leaves the ED is present.
8
Figure 2.1: Early prediction sharing (Peck et al., 2012)
2.4.2 Use of prediction models
An earlier start of the admittance process is only beneficial, if the admission of a patient is accurately predicted. Multiple studies have investigated the accuracy of admission predictions made during triage (Qui et al., 2015). These predictions are either based on prediction models or on predictions of triage nurses (Beardsell and Robinson, 2011). Cameron et al. (2017) advise to always use a prediction model because triage nurses are often uncertain about their prediction. In contrast, Uniken Venema (2018) concludes that triage nurses are more accurate than prediction models, indicating to use their predictions. According to Vlodaver et al. (2019) both prediction models and triage nurses are unable to predict admissions with sufficient accuracy. Therefore, they examined the accuracy of the admission predictions of physicians during triage, without seeing the patient. Unfortunately, the findings of Vlodaver et al. (2019) regarding the admission predictions of physicians are also disappointing. They argue that this is because the physician did not see the patient.
9
3. Methodology
The goal of this study is to investigate the effect of an earlier start of the admittance process on the outflow of patients. Since the earlier start is based on admission predictions, first the accuracy of the admission predictions is analysed. Additionally, to develop a good understanding of the admittance process, the admittance process is analysed extensively. Therefore, this research is divided into three phases:
Phase 1: Analysing the accurateness of the admission predictions of physicians and nurses at the ED. Phase 2: Identifying activities of the admittance process and causes of delay within this process. Phase 3: Analysing the effect of an earlier start of the admittance process on the outflow of patients. To answer the research question a single-case study at a hospital in the Netherlands is conducted.
3.1 Single-case study design
Case studies enable the researcher to study a phenomenon in a real-life setting (Meredith, 1998) and to get a good understanding of the complexity and nature of the investigated phenomenon (Karlsson, 2009). Since hospitals are complex systems and the effect of an earlier start needs to be empirically tested, conducting a case study seems a logical choice. A case study enables that the accurateness of the admission predictions is assessed based on real-life admissions and the effect of starting the admittance process earlier is studied in the same context. Further, the explorative nature of a case study is helpful (Meredith, 1998), since not much is known about the effect of starting the admittance process earlier. Despite the limitation in terms of generalizability, this study focusses on a single case. According to Karlsson (2009) the fewer cases studied, the better the opportunity to get in-depth insights.
3.2 Hospital case description
This study is conducted at the ED of the Medisch Centrum Leeuwarden (MCL). This ED is chosen because Witvoet (2018) showed that a large delay is present between the admission decision and the moment that the patients leaves the ED. Besides, MCL is planning to reduce the number of inpatient beds to achieve financial savings. Therefore, the utilization of beds, the flow of patients and the coordination between departments become even more important.
The process of patients at the ED of MCL is similar to the process discussed in section 2.1. Arriving patients first go through triage. Here the urgency of the patient is determined and some initial diagnostic tests are performed. The urgency of the patient determines whether a patient must wait or is immediately treated. When treatment at the ED is finished, the physicians decides whether the patient has to be admitted. If so, the physician initiates the admittance process by sending a bed request to the admission department. When a bed has been assigned, the ED nurse contacts the inpatient department to inform them that the patient can be picked up. The progress of a patient is tracked with the patient information system EPIC. The admittance process is presented in Appendices I and II and is discussed in more detail in section 4.2.
3.3 Data collection
Phase 1: Analysing the accurateness of the admission predictions
10 monitored the progress of each patient in EPIC. Within the first hour of treatment, the researcher approached both the physician and nurse to answer the following questions:
1. Do you think that this patient has to be admitted? If yes,
2. For which specialism does the patient has to be admitted?
3. Please indicate the time when the treatment of the patient is most likely to be finished. After the sampling period, data of each patient that visited the ED during the sampling period is extracted from EPIC. This data describes the real admission decision, the specialism and the time that treatment was finished. The collected data and the data extracted from EPIC are linked based on patient number and date to enable comparison.
Phase 2: Identifying activities of admittance process and causes of delay
To analyse the admittance process extensively, observations at the ED and interviews with the managers of four inpatient departments and with an employee of the admission department are conducted. The inpatient departments are trauma surgery, internal medicine, lung diseases and neurology. These departments were chosen because they are responsible for the majority of patients at the ED. During the observations and interviews the first aim was to identify the activities performed at the ED, admission department and inpatient department in relation to the admittance process. Furthermore, the focus was on identifying the factors that potentially cause delays within the admittance process, resulting in additional waiting time. Each interview is recorded and transcribed. The interview protocol is given in Appendix III.
Phase 3: Analysing the effect of an earlier start of the admittance process
An intervention at the ED is implemented to analyse the effect of starting the admittance process earlier. During the intervention period of 9 days, the researcher approaches the physician one hour after the patient arrived at the ED and asks whether the physicians predicts that the patient has to be admitted. If so, the researcher contacts the ED nurse and together they initiate the admittance process of that patient by generating a pre-order. The pre-order indicates the preferred inpatient department and provides additional information regarding the patient. The admission department receives this pre-order and unofficially assigns a bed. Thereafter, the admission department informs the inpatient department about the admission, enabling them to perform their activities based on the provided information in parallel with the treatment at the ED. After treatment is completed, the official bed request is generated by the physician and the bed is officially assigned. Finally, the ED nurse contacts the inpatient department to announce that the patient can be picked up (Appendix IV).
Quantitative data about the throughput times of patients is extracted from EPIC to analyse the effect of the intervention. This data is extracted for the intervention period and for a control period of four months, January till April 2019. The quantitative data relates to:
- Arrival time at the ED
- Time request for inpatient bed - Time patient left the ED
11 During the control period the ‘time request for inpatient bed’ equals the time that the admittance process is initiated. However, during the intervention the admittance process is initiated by the pre-order. Therefore, during the intervention period additional data is collected regarding the time that the pre-order is generated.
Additionally, qualitative data is collected to support the findings from the quantitative analyses. Data is collected by conducting an interview with two employees of the admission department and with several inpatient departments (Appendix V). For Neurology and Internal medicine, the managers are interviewed. The interviews with Lung diseases and General surgery are conducted with a coordinating nurse and for trauma surgery both the manager and a coordinating nurse are interviewed simultaneously. The aim of these interviews is to evaluate the effect of the intervention on the admittance process and the related waiting time and discuss the causes of delay identified in phase 2. The insights from the interviews are used to explain findings from the quantitative analyses.
3.4 Data analysis
Phase 1: Analysing the accurateness of the admission predictions
In this phase, the accurateness of the admissions predictions of the physicians and nurses is assessed to determine whether these predictions are useful to initiate an earlier start of the admittance process. The admission predictions are compared to the real admission decision to check whether they are correct. Then, the accurateness is calculated together with the sensitivity, specificity, positive predictive value and negative predicted value based on the following definitions of Vlodaver et al. (2019) and the confusion matrix shown in Table 3.1:
1) Accurateness Proportion of correctly predicted predictions 2) Sensitivity Proportion of admissions predicted as admissions 3) Specificity Proportion of discharges predicted as discharges
4) Positive predictive value (PPV) Proportion of admission predicted resulting in admission 5) Negative predictive value (NPV) Proportion of discharge predictions resulting in discharge
12 A 95% confidence interval is calculated for the accurateness of the physicians and nurses, considering the sample size, with the following formula:
p ±Z∝/2 * √
(p)(1-p) n Where p = the accurateness during sampling period
Zα/2 = the confidence level (95% = 1,96)
n = sample size
After assessing the performance of the physicians and nurses separately, it is analysed whether there is a significant difference in performance between the two groups. The McNemar statistical test is used because a comparison is made between two paired groups and the data is binary, i.e. correct or incorrect prediction. The groups are paired because the prediction of a physician and a nurse are made for the same patient. Further, it is analysed whether the physicians and nurses are able to predict the right specialism. Additionally, the incorrect predictions are analysed based on their specialism to determine whether the accurateness differs per specialism.
Phase 2: Identifying activities of the admittance process and causes of delay
The goal of this phase is to get insight in the activities of the admittance process, the places where waiting time emerges and the factors causing a delay. First, the transcripts of the five interviews are analysed to identify the activities of the admittance process and their sequence. Based on these insights, process flow diagrams are created for the admittance process during the day and during the evening/night (Appendices I and II). With the help of the process flow diagrams, the places where waiting time emerges are identified and the total waiting time is divided into waiting time components per department. Further, the points were one waiting time component ends and another waiting time component begins are identified together with possible points of overlap between multiple components. Eventually, interviews are systematically analysed to identify factors that influence these waiting time components and can potentially cause delays in the admittance process. These potential delays are then analysed together to identify higher level categories.
Phase 3: Analysing the effect of an earlier start of the admittance process
In the third phase, the goal is to analyse the effect of an earlier start of the admittance process initiated during the intervention. The analyses of this phase compare the performance of the intervention period with the control period. First, the data is prepared for the analyses. This implies that discharged patients are excluded from the data, as the intervention is only directed towards admitted patients. Further, specialisms cardiology, psychiatry and gynaecology are excluded as these specialisms arrange their own beds instead of consulting the admission department. The study populations are further explained in section 4.3.
13 For both periods a waiting time distribution is generated to determine the frequency of each waiting time interval. These waiting time distributions are compared to analyse whether a difference in waiting time is noticed. Additionally, the difference in waiting time between the two periods is statistically tested. First, the data is checked for normality with the Shapiro-Wilk test. As it is concluded that the data is not normally distributed (p < .05), the Mann-Whitney U test is used to test whether there is a significant difference in waiting time between the periods. Finally, the waiting time related to boarding is analysed per specialism to identify differences per specialism.
14
4. Results
This chapter presents the results from the case study. First, the admission predictions are evaluated. Then, the activities of the admittance process and the possible causes of delay within this process are discussed. Lastly, the results from the intervention, where an earlier start was initiated at the ED, are analysed.
4.1 Phase 1: the admission predictions of ED physicians and nurses
324 patients visited the ED during the sampling period. From these patients, 265 patients are included and 59 patients are excluded. Reasons for exclusion are that the patient did not give permission for inclusion or that the patient already left the ED within one hour. Both the physicians and nurses made 264 predictions. Thus, there are two patients for whom only the physician or only the nurse made a prediction. The reason for a missing prediction is that the physician or nurse was not available to make a prediction within one hour. To analyse the differences between physicians and nurses 263 admission predictions are use (Figure 4.1). From this sample, 143 patients were admitted resulting in an admission rate of 54%, which is slightly higher than the overall admission rate. This is expected, as patient with minor injuries that left the ED within one hour are excluded from the sample.
Figure 4.1: Study population
Accurateness of the admission predictions
Figures 4.2 and 4.3 show the number of correctly and incorrectly predicted admissions and discharges for the physicians and nurses. The results show that the physicians performed better in correctly predicting admissions and discharges. Further, the nurses have considerably more false negatives, where a discharge is predicted but the actual outcome is an admission.
15
Figure 4.3: Overview admission predictions nurses
The performance of both groups, regarding their admission predictions, is assessed based on five performance measurement (Table 4.1). Logically, the accurateness is higher for the physicians compared to the nurses. The accurateness refers to the overall proportion of correctly predicted predictions and above it was already discussed that the correctly predicted predictions are higher for the physicians. The accurateness of the physicians is 89,02% and the accurateness of the nurse is 83,71% for this sample. Considering the sample size of 264 predictions, it can be concluded that with a confidence level of 95% the accurateness of the physicians will be between 85,2% and 92,8% and the accurateness of the nurses will be between 79,3% and 88,2%. Further, especially on sensitivity and negative prediction value the physicians perform better. Both these performance measurements are calculated using the number of false negatives of which nurses have considerably more.
Table 4.1: Performance of physicians and nurse admission predictions
When calculating the five performance measurements, purely the fact whether an admission or discharge was predicted correctly is considered. Of the 130 correctly predicted admissions, the physician incorrectly predicted the specialism 12 times. The nurses incorrectly predicted the specialism 14 times of the 120 correctly predicted admissions. In these cases, a bed at the wrong inpatient department would have been reserved.
Differences between physicians and nurses
For 263 patients both the physician and the nurse made an admission prediction. Therefore, these predictions are used to compare the performance of the groups. For 34 patients the physician and nurse made a different prediction. For 24 patients the physician was correct with 15 correctly predicted admissions and 9 correctly predicted discharges. For 10 patients the nurse was correct with 5 correctly predicted admissions and 5 correctly predicted discharges (Figure 4.4). The difference between the correctly predicted admissions of both groups is statistically significant (p = .026), which means that physicians are significantly better in predicting whether a patient is going to be admitted.
There are 19 patients for whom both the physician and the nurse made a wrong prediction. All the wrong predictions are analysed in more detail related to the specialism involved. These results are discussed further in Appendix VI.
16
Figure 4.4: Different admission predictions
4.2 Phase 2: identifying activities of the admittance process and causes of delay
This section elaborates on the findings from the observations and interviews related to the admittance process. The admittance process is analysed on three levels. First, the admittance process is analysed on the level of the departments and their activities. Thereafter, the admittance process is analysed on the level of the waiting time related to boarding to determine the components of waiting time and identify the boundaries between the components. Finally, the admittance process is analysed on the level of potential delays. Here, factors are identified that can cause a delay and by that additional waiting time in the admittance process (Figure 4.5). A detailed discussion of the results is given in Appendix VII. Below the most relevant results are discussed.
Activities of the admittance process
The admittance process encompasses all activities that must be completed before a patient can be admitted. There are three departments involved in the admittance process: the emergency department, the inpatient department and the admission department. A detailed flow diagram of the admittance process is shown in Appendix I. Below the most relevant activities are discussed in relation to Figure 4.5.
17
Figure 4.5: The admittance process and possible causes of delay
Waiting time components
The waiting time related to boarding begins at the moment that treatment at the ED is completed and the physician decides that the patient has to be admitted and ends at the moment that all activities of the admittance process are completed and the patient had left the ED. This implies that the total time required to perform all the activities of the admittance process can be classified as the total waiting time related to boarding. The results show that this waiting time emerges because of the late initiation of the admittance process and because of the late information sharing between the departments. The results indicate that the total waiting time related to boarding can be divided into several components related to the departments (Figure 4.5). Several moments are identified that indicate that waiting time created at one department ends and waiting time starts emerging at another department. Component 1: Waiting time starts emerging at the ED after the admission decision and ends after the physician has generated the bed request.
Component 2: Waiting time starts emerging at the admission department after receiving the bed request and ends after a bed has been assigned and the other departments are informed.
Component 3: Waiting time starts emerging after being informed that a bed is reserved and ends after the ED nurse has contacted the inpatient department to provide the required information.
Component 4: Waiting time starts emerging after being informed about the admitted patient and ends after the inpatient department has picked up the patient and the patient left the ED.
18 contacts them quickly, it is likely that the inpatient department did not finish their initial preparations yet. In that case, all the waiting time related to component 3 overlaps with waiting time component 4 and no additional waiting time is created. However, when it takes longer for the ED nurse to contact them, it is likely that they did finish their initial preparations and have to wait before starting the final preparations. Then, additional waiting time is created by the ED nurse due to the late information sharing.
Causes of delay
The results show that the admittance process in itself causes waiting time related to boarding due to the late initiation. Additionally, based on the interviews, several factors potentially causing a delay within the admittance process are identified. These factors can cause a delay in one or multiple waiting time components (Figure 4.5), increasing the total waiting time. The identified causes of delay are: capacity constraints related to the work force, capacity constraints related to the beds and room, and receiving multiple requests simultaneously. Below the identified causes of delay and their influence on the waiting time components are discussed.
1) Work force related capacity constraints
The activities of the ED and of the inpatient department can be delayed by a constraint in the available capacity of the physicians or nurses, increasing the waiting time in components 1, 3 and 4. The physicians and nurses at the ED treat multiple patients simultaneously. Thus, the physician and/or nurse can be occupied treating another (more urgent) patient. In case of the physician, this indicates that the bed requests of other patients get a lower priority, resulting in a delay in generating the bed requests and an increase in waiting time component 1. In case of the nurse, the contact with the inpatient department about another patient is delayed, increasing the additional waiting time in component 3. When the ED nurse is having a break, the call is regularly not realized until after the break. Occasionally, another ED nurse calls the inpatient department, but this does not happen consistently. Thus, the time available of the physician and nurse at the ED influences the duration of waiting time component 1 and 3.
The physician at the ED needs permission from the receiving specialist to admit a patient. It can occur that the receiving specialist is not immediately available to discuss and approve the admission, resulting in a delay in generating the bed request and increases waiting time in component 1. Constraints in the available time of the inpatient department nurse can cause a delay as well. A nurse must have time available to execute the activities of the admittance process. When a nurse is occupied with other activities, completing the activities of the admittance process is delayed and waiting time in component 4 increases. Additionally, it can occur that there are a limited number of nurses present at the department. In that case, a nurse cannot immediately leave the inpatient department to pick up the patient from the ED. Thus, if there is not immediately time available to perform the activities or there is no nurse available to pick up the patient, a delay is caused at the inpatient department and waiting time in component 4 increases.
2) Room and bed related capacity constraints
19 departments to determine the most appropriate location, resulting in an increase in waiting time component 2. When the physician insists that the patient is assigned to the preferred inpatient department, several additional activities, like moving patients to another inpatient department or arranging earlier discharges, have to be performed to make a bed available. Then, a delay emerges at the inpatient department and increases waiting time in component 4. This is similar for a request for a single room, as the number of single rooms is limited and these are generally fully occupied. It takes the admission department more time to satisfy requests for single rooms, as they first have to coordinate with the inpatient department about the best solution. Usually, a single room is then made available by moving other patients. To summarize, if there are capacity constraint related to the available bed and/or rooms, delays arise at the admission department and the inpatient department increasing the waiting time in component 2 and 4.
3) Multiple requests simultaneously
Another factor identified causing a delay at the admission department and the inpatient departments, is when the admittance process is initiated for multiple patients simultaneously. In that case, the admission department receives multiple bed requests simultaneously. These bed requests are either for the same or for different inpatient departments. Simultaneous bed requests for different inpatient departments are generally caused by the coordinating ED physician. When the ED is getting crowded, the coordinating ED physician encourages other physicians to finish treatment and initiate the admittance process. This leads to multiple bed requests at once. If the bed requests are for the same inpatient department, this is generally caused by an unexperienced physician at the ED. These physicians first have to discuss each patient with their supervisor before initiating the admittance process. Generally, the supervisory specialist visits the ED when multiple patients can be discussed. Afterwards, the admittance process is initiated for these patients simultaneously. Consequently, the admission department receives multiple requests simultaneous, while they can process one at a time. Thus, automatically the other requests are delayed until the ones before them are processed, resulting in additional waiting time in component 2.
20
4.3 Phase 3: effect of starting the admittance process earlier
This section presents the results of the intervention that initiated an earlier start of the admittance process. First, the various study groups are explained. Then, some general analyses for the control and intervention period are conducted to identify differences that could affect the results. Finally, more in-depth analyses regarding the length of stay and the waiting time related to boarding are performed. Study groups
3611 patients visited the ED during the control period, from whom 1073 are excluded because of their specialism. To analyse the waiting time related to boarding, purely admitted patients are relevant, thus discharged patients are excluded. Further, 8 patients are excluded due to missing data. Thus, 1527 patients are included in the control period (Figure 4.6). During the intervention period 400 patients visited the ED, of which 246 are included. 27 patients are excluded because of their specialism. The other patients are either excluded because they did not give permission for inclusion or left within one hour. Of the included patients, 99 patients were admitted and for 61 patients an earlier start was initiated. However, 4 of these patients were eventually discharged and for 7 patients the specialism was predicted incorrectly (Figure 4.7). Thus, for 50 patients an earlier start is initiated correctly and for 49 admitted patients an earlier start is not initiated or not initiated correctly. Reasons for not initiating an earlier start are when the physician did not predict an admission or was cautious with prediction of an admission, meaning that the physician was not confident enough to initiate an earlier start. This resulted in selecting the safer option, predicting a discharge, while stating that they will know more after test results are available.
Some general characteristics of each period are shown in Appendix VIII. The number of arriving patients and the admittance rate are slightly higher during the control period. This is mainly caused by Lung diseases, which have a considerable higher admittance rate during the winter period.
21 Average length of stay and patient waiting time
For 50 patients an earlier start is initiated correctly. However, the admission department stated “a pre-order that initiates an earlier start should arrive at least 20 or 30 minutes before the real bed request. Otherwise, there is not sufficient time in between to prepare and a short time frame causes confusion”. With confusion the admission department means that it is difficult to maintain a clear overview when the pre-orders and real bed requests arrive shortly after each other. Then, they are still processing the pre-order when the real bed request arrives. Therefore, an additional group is distinguished in the analyses. This group includes all patients for whom the time between the pre-order and the bed request exceeds 20 minutes. Figure 4.8 shows the percentages of pre-orders placed per time interval. Half of the pre-orders are placed less than 20 minutes before the bed request. This is primarily caused by the cautiousness of the physicians in making a prediction. Whenever physicians were not confident enough to initiate an earlier start, they usually waited until they were certain about the admission and then initiated an earlier start. However, at that moment the actual bed request is generated shortly after.
Figure 4.8: Overview time between pre-order and bed request
Table 4.2 shows the average length of stay (LOS), which is the time between arrival and the moment the patient left the ED, and the average waiting time related to boarding (WT), which is the time between the bed request and the moment the patient left the ED. There are no considerable differences in the average LOS, except for the group ‘pre-order > 20 minutes’. That group has the longest average LOS. An explanation is that for patients with a longer average LOS, it is more likely that there is sufficient time between the pre-order and the actual bed request.
The average WT for patients with an earlier start and for patients with ‘pre-order > 20 minutes’ are slightly shorter than the average WT of the control group, but no significant differences are found (p = .59, p = .27, respectively). Control period (1527) Intervention period (99) Earlier start (50) No earlier start (49) Pre-order > 20 minutes (25) Average LOS 3:42 3:33 3:37 3:30 4:11 Standard dev. LOS 1:15 1:09 1:09 1:09 1:10 Average WT 1:01 1:01 0:59 1:03 0:58 Standard dev. WT 0:30 0:31 0:29 0:34 0:32
22 Figure 4.9 shows the distribution of waiting time related to boarding for the control period and the patients with an earlier start. A minor shift to the left can be noticed when comparing the intervention period to the control period. However, the distribution of the intervention period is very irregular due to the small sample size, thus no conclusions can be drawn.
Figure 4.9: Waiting time distributions
Specialisms with an earlier start
In table 4.3 the patients with an earlier start, the patients without an earlier start and the patients with a ‘pre-order > 20 minutes’ are categorized per specialism. Clearly, earlier starts are mainly initiated for patients from four specialisms: Surgery, Internal medicine, Lung diseases and Neurology. It is difficult to measure and discuss the effect of an earlier start when the total amount of patients with an earlier start is limited. Therefore, these four specialisms are the focus of the coming analyses.
Specialism Earlier start No earlier start Pre-order > 20 min
Surgery 13 18 4 Internal medicine 14 9 10 Lung diseases 8 3 4 Neurology 7 7 3 Gastroenterology 2 3 0 Orthopaedics 1 2 1 Geriatrics 1 0 1 Paediatrics 1 1 1 Oral surgery 1 1 1 Urology 1 2 0 Rheumatology 1 1 0 Oncology 0 1 0 Plastic surgery 0 1 0
Table 4.3: Number of earlier starts per specialism
0% 2% 4% 6% 8% 10% 12% 14% 16% Perc en ta ge o f to ta l
Waiting time related to boarding
Distribution waiting time
23 Waiting time related to boarding per specialism
During the interviews with the inpatient departments, the effect of the intervention is discussed. Both Surgery and Internal medicine noticed a difference during the intervention. They state that they observed a difference in the admittance process compared to before and additionally observed a difference between patients with an earlier start and patients without an earlier start. One inpatient department elaborates by “The earlier start is incredibly practical as it provides additional information earlier and we need this information to steer our activities. Due to this earlier information sharing we can already decide where to place the patient and perform some preparing activities. Additionally, I can already inform the assigned nurse about the incoming patient and we can make room in the schedule of the nurse. It just helps with using your time more effective”. In contrast, Neurology and Lung diseases observed no difference in the admittance process during the intervention. A possible reason for this difference in notion is that less patient, with an earlier start are included for Neurology and Lung diseases (Table 4.3).
Table 4.4 shows the average waiting time related to boarding (WT) for each group. Due to the small sample size and large standard deviation, no conclusions can be drawn. The results are coherent with the observations of the inpatient departments. For Surgery and Internal medicine, the average WT is slightly shorter for patients with an earlier start compared to the control group and patients without an earlier start. For Neurology and Lung diseases the average WT is longer for patients with an earlier start compared to the control group and patients without an earlier start. Thus, the average WT is slightly improved for the specialisms that observed a difference in the admittance process and the average WT did not improve for the specialisms that did not observe a difference.
Control period Intervention period
Earlier start No earlier start
Pre-order > 20 minutes Specialism Average SD Average SD Average SD Average SD Average SD Surgery 1:05 0:34 1:02 0:29 0:59 0:22 1:05 0:33 0:46 0:12 Internal medicine 1:04 0:30 0:57 0:37 0:55 0:34 1:01 0:44 0:53 0:36 Lung diseases 0:59 0:26 1:03 0:20 1:05 0:23 0:56 0:15 1:03 0:22 Neurology 0:55 0:25 0:54 0:20 1:01 0:25 0:46 0:13 0:52 0:17
Table 4.4: Average waiting time per specialism
24 Additional findings related to waiting time
During the interviews, several factors influencing the admittance process during the intervention are discussed. These factors could give explanations for the limited effect of the intervention on the average waiting time of patients.
1) Single rooms. The admission department stated “During the intervention there were relatively more requests for single and two-person rooms than usual. In our opinion, this was not always needed. But since the ED had to make a choice and these options were given, it was easier to choose a single of two-person room than before”. Thus, according to the admission department there was a greater tendency to request a single or two-person room during the intervention. As stated in section 4.2, this tendency could lead to an increase in both waiting time components 2 and 4 because of the related capacity constraints. Table 4.5 shows the frequency that a particular room type is requested and the average WT per room type. These results indeed show that for patients with a single or two-person room the average WT is longer compared to a four-person room. However, these patients generally require a single room because of certain aspects in their condition or background. These aspects could also be the reason for a longer average WT instead of the room type.
Table 4.5: Waiting time per room type
2) Multiple requests simultaneously. Also during the intervention multiple requests to pick up patients arrived simultaneously. The intervention mainly focused on initiating the admittance process earlier. Therefore, it is not surprising that the delay caused by receiving multiple requests simultaneously still exists. An inpatient department stated “We can prepare everything for each patient that was announced in an earlier stage, which is an advantage. However, even with a pre-order it is still impossible to pick up multiple patients simultaneously”. Thus, the earlier start of the admittance process does enable the inpatient department in their preparations, reducing waiting time component 4. However, when requests to pick up patients still arrive simultaneous, the delay that increases waiting time component 4 still occurs. In contrast, the admission department stated “The pre-orders generally did not arrive simultaneously, which allowed us to assign most of the beds before a new pre-order arrived. When the actual bed requests did arrive simultaneously, we had already assigned a bed based on the pre-order. Therefore, it did not require a lot of time to assign them officially”. Thus, the additional waiting time caused by receiving multiple requests simultaneously in waiting time component 2 is reduced due to the earlier start of the admittance process.
25 patient during lunch, dinner or at the end of a shift. At these moments we have fewer nurses available, which makes it harder to pick up patients quickly”. Thus, at these moments there is a constraint in the available nurses, leading to an increase in waiting time. In contrast, another inpatient department stated “Sometimes it is difficult to anticipate because generally the moments that it is crowded at the ED, we are also occupied with our own patients and it can be difficult to make room in the schedule of the nurses. However, in my opinion it does make a difference when we are informed half an hour before than just a few minutes, because it is easier to make a nurse available half an hour before than instantly”.
Figure 4.10 shows the amount of pick up requests and the average waiting time per hour. From 11:00 till 17:00 the amount of pick up requests keeps increasing. Minor increases in the average waiting time can be noticed between 12:00 and 13:00, 15:00 and 16:00 and after 17:00, which is around lunch, the end of the day shift and dinner. However, between 11:00 and 20:00 the average waiting time varies between 60 and 70 minutes. Thus, in contrast with the statements of the inpatient department, the total average waiting time is not significantly longer at certain time moments. Perhaps it is merely the perception of the inpatient department that the admittance process takes longer at these moments, since it is harder for them to arrange capacity for the activities.
Figure 4.10: Number of requests and average waiting time per time interval
4) Accuracy of information. The inpatient departments stress the importance of accurate and complete information: “We make choices related to the activities of the admittance process based on the information provided by the ED. If this is faulty information, we either perform unnecessary activities, which take additional time or we do not prepare sufficiently which indicates that preparations have to be performed last minute”. One example is “During the intervention one patient required isolation. However, this requirement was not communicated by the ED in the pre-order and did not became apparent until the patient could already leave the ED. Then, all activities to prepare the isolation still had to be performed”.
Unnecessary preparations are also performed when an admission is communicated, but the patient is discharged. According to the admission department it has negative consequences when too much false admissions are communicated. They stated “An excessive amount of false
0:00 0:10 0:20 0:30 0:40 0:50 1:00 1:10 1:20 1:30 1:40 1:50 0 20 40 60 80 100 120 140 160 180 200 220 A ve rag e WT Fr e q u e n cy Time interval
Influence of
timemoments
27
5. Discussion
The main objective of this research is to investigate the effect of an earlier start of the admittance process on the waiting time related to boarding. To initiate an earlier start, the admission predictions of physicians and nurses at the ED have been analysed and used. Further, an explorative analysis of the admittance process has been conducted to get a good understanding of the admittance process and factors causing a delay.
5.1 Admission predictions
Admission predictions studied in literature are essentially made by triage nurses. Only Vlodaver et al. (2019) studied the admission predictions of physicians during triage. However, these studies show disappointing results with regard to predicting admissions. Additionally, the accurateness of physicians and nurses has never been compared. Therefore, this study compares the accurateness of the admission predictions of nurses and physicians to determine who performs better. The results show that the physicians are significantly more accurate in prediction admissions compared to nurses. Furthermore, a different approach is taken, since the predictions are made after one hour of treatment instead of during triage. In this study, the nurses scored an accurateness of 83,71%, with a sensitivity of 84,51%, a specificity of 82,79%, a PPV of 85,11% and a NPV of 82,11%. Compared to the outcomes of similar studies (Stover-Baker, Stahlman and Pollack, 2012; Vaghasiya et al., 2014; Alexander et al., 2016) the sensitivity and the PPV are considerably higher. This implies that during this study the nurses performed substantially better with regard to predicting admitted patients. A possible reason is that the extra hour gives the opportunity to gather extra information and perform initial tests. The physicians scored an accurateness of 89,02% with a sensitivity of 91,55%, a specificity of 86,07%, a PPV of 88,44% and a NPV of 89,74%. Compared to the results of Vlodaver et al. (2019), the physicians in this study perform better on all performance measurements except on specificity. In the study of Vlodaver et al. (2019) the physicians were accurate in predicting discharges but were inadequate in prediction admissions. In this study, the physicians accurately predicted both discharges and admissions. A possible reason for the difference is that Vlodaver et al. (2019) asked the physicians to make a prediction during triage without seeing the patient, while the predictions in this study are made after one hour of treatment and having contact with the patient.
Thus, physicians are more accurate in predicting admission than nurses and the accuracy of these predictions is higher after one hour of treatment compared to during triage. This implies a trade-off between how far in advance the admittance process is initiated and the accuracy of the prediction. The results indicate that too much faulty predictions can have a negative effect on the perception of the inpatient department and diminish the effect of an earlier start. Therefore, it is essential to consider this trade-off during the admittance process of patients.
5.2 The admittance process
28 helpful in the development of improved solutions focused on improving the outflow process and resolving the crowding issue.
The three departments involved in the admittance process are interdependent. The ED depends on the admission department to assign a bed quickly and on the inpatient department to pick up the patient rapidly. However, these two departments depend on the ED to share accurate and complete information to steer their activities. Currently, this information is not shared in an early stage, which makes it difficult for the other departments to anticipate, resulting in waiting time related to boarding. This corresponds with Drupsteen et al. (2013), who state that departments are too focused on their internal process instead of the whole patient flow, which reduces the tendency to coordinate with other departments. Uncertainty and variability are factors that inhibit coordination (Drupsteen et al., 2016). Uncertainty and variability are also observed to be an inhibiting factor in initiating an earlier start. Less experienced physicians were cautious with their prediction when there was uncertainty about the outcome. They did not want to inform the other department until they were absolutely certain. Thus, this uncertainty inhibited the opportunity to coordinate an admission earlier and reduce waiting time.
Possible causes of delay in the admittance process are the capacity constraints related to beds, rooms and personnel. Asplin et al. (2003) already suggested that the lack of inpatient beds and nurses increases waiting time related to boarding at the inpatient department. This study shows that these capacity constraints increase the waiting time at the admission department and the ED as well. It is acknowledged that limited capacity leads to waiting time in a hospital. Hospitals have to deal with variability in the inflow of patients and according to Hopp (2011) there are three types of buffers to cope with variability, namely inventory, capacity and time buffers. Inventory buffers are not possible within hospitals (Jack and Powers, 2004). Therefore, hospitals create either capacity or time buffers. Time buffers indicate waiting time for the patient and are essentially used in elective care. Since waiting time in undesirable in acute care, here mainly capacity buffers are used. Nevertheless, this study shows that capacity constraints occur during the admittance process of acute patients. Consistently, a time buffer appears and waiting time emerges.
29
5.3 Earlier start of the admittance process
In literature an earlier start of the admittance process is repeatedly mentioned as the solution for crowding issues. However, this study shows that solely communicating an admission earlier is not sufficient. The admission department and the inpatient department need additional information to steer their activities. It is crucial that this information is accurate, otherwise even more waiting time is created due to the unnecessary activities performed. Initiating an earlier start by sharing complete and accurate information moves the whole admittance process to an earlier stage. Then, the admittance process is performed in parallel with treatment at the ED, reducing the waiting time related to boarding. Additionally, this study shows that there are multiple factors that potentially cause a delay in the admittance process and therefore increase the duration of the admittance process beyond the time slack provided by the earlier start. Thus, even with an earlier start, waiting time related to boarding still emerges. Thus, to reduce the waiting time related to boarding, an earlier start should be initiated and the causes of delay should be tackled.
30
6. Conclusion
6.1 Conclusion
Although, solutions for the crowding issue are extensively researched, the crowding issue remains. One solution proposed in literature is to initiate an earlier start of the admittance process based on the predictions of triage nurses. However, the accuracy of these predictions has been disappointing and are therefore barely used in practice. This study shows that physicians are significantly more accurate in prediction admissions than nurses. Additionally, the results show that the accuracy of the admission predictions improves considerably when made after one hour instead of during triage. However, the earlier the admittance process is initiated the more time the other departments have to prepare. This implies that there is a trade-off between how far in advance the admittance process is initiated and the accuracy of the predictions. Too much incorrect predictions can have negative consequences and the time needed to execute the admittance process differs per patient and per moment. Therefore, it is difficult to suggest a fixed moment, with accurate predictions and sufficient time to prepare, to initiate the admittance process for each patient.
To reduce the waiting time related to boarding, an earlier start of the admittance process should be initiated based on the prediction of the physician. Further, it is essential that relevant information is shared along with the admission prediction to enable the other department in their preparations. Lastly, the results indicate that initiating an earlier start does not guarantee a reduction in the waiting time related to boarding. There are several factors that can cause a delay in the admittance process beyond the time slack provided by the earlier start. Further, there are situations that the earlier start initiated at the ED, does not trigger an earlier start of the activities at the inpatient department.
6.2 Practical implications
The health care sector is pressured to increase quality and patient satisfaction with less financial resources. Therefore, hospitals have to focus on using their available resources more efficiently. In this case, the average waiting time related to boarding is one hour, which is a significant share of the total length of stay. During this hour the patient is unnecessarily occupying an ED bed. Too much boarders result in crowding issues and ambulance diversions. This study creates more understanding of the admittance process and the factors influencing waiting time. When considering these insights, hospitals can improve their admittance process and reduce the waiting time related to boarding. This reduced waiting time enables the ED to efficiently use their resources and avoid ambulance diversions, while it increases patient satisfaction as well. Further, our study shows that the coordination between the involved departments should be intensified to achieve an improved admittance process and agreements must be made regarding capacity, buffering and prioritization. The intensified coordination creates more understanding between the departments, which reduces frustration and improves employee satisfaction.
6.3 Limitations and future research
32
References
Abraham, J., & Reddy, M. C. (2010). Challenges to inter-departmental coordination of patient transfers: a workflow perspective. International Journal of medical informatics, 79(2), 112-122.
Alexander, D., Abbott, L., Zhou, Q., & Staff, I. (2016). Can triage nurses accurately predict patient dispositions in the emergency department?. Journal of Emergency Nursing, 42(6), 513-518.
Anneveld, M., Van Der Linden, C., Grootendorst, D., & Galli-Leslie, M. (2013). Measuring emergency department crowding in an inner city hospital in The Netherlands. International journal of emergency medicine, 6(1), 21.
Asplin, B. R., Magid, D. J., Rhodes, K. V., Solberg, L. I., Lurie, N., & Camargo Jr, C. A. (2003). A conceptual model of emergency department crowding. Annals of emergency medicine, 42(2), 173-180.
Beardsell, I., & Robinson, S. (2011). Can emergency department nurses performing triage predict the need for admission?. Emerg Med J, 28(11), 959-962.
Boyle, A., Beniuk, K., Higginson, I., & Atkinson, P. (2012). Emergency department crowding: time for interventions and policy evaluations. Emergency medicine international, 2012.
Brillman, J. C., Doezema, D., Tandberg, D., Sklar, D. P., Davis, K. D., Simms, S., & Skipper, B. J. (1996). Triage: limitations in predicting need for emergent care and hospital admission. Annals of emergency medicine, 27(4), 493-500.
Cameron, A., Ireland, A. J., McKay, G. A., Stark, A., & Lowe, D. J. (2017). Predicting admission at triage: are nurses better than a simple objective score?. Emerg Med J, 34(1), 2-7.
Derlet, R. W., & Richards, J. R. (2000). Overcrowding in the nation’s emergency departments: complex causes and disturbing effects. Annals of emergency medicine, 35(1), 63-68.
Drupsteen, J., van der Vaart, T., & Van Donk, D. P. (2013). Integrative practices in hospitals and their impact on patient flow. International Journal of Operations & Production Management, 33(7), 912-933.
Drupsteen, J., van der Vaart, T., & Van Donk, D. P. (2016). Operational antecedents of integrated patient planning in hospitals. International Journal of Operations & Production Management, 36(8), 879-900.
Eckstein, M., & Chan, L. S. (2004). The effect of emergency department crowding on paramedic ambulance availability. Annals of emergency medicine, 43(1), 100-105.
33
Haddad, C. J. (1996). Operationalizing the concept of concurrent engineering: a case study from the US auto industry. IEEE Transactions on Engineering Management, 43(2), 124-132.
Handel, D. A., Hilton, J. A., Ward, M. J., Rabin, E., Zwemer, Jr, F. L., & Pines, J. M. (2010). Emergency department throughput, crowding, and financial outcomes for hospitals. Academic Emergency Medicine, 17(8), 840-847.
Higginson, I. (2012). Emergency department crowding. Emerg Med J, 29(6), 437-443.
Hirshon, J. M., Risko, N., Calvello, E. J., Ramirez, S. S. D., Narayan, M., Theodosis, C., & O'Neill, J. (2013). Health systems and services: the role of acute care. Bulletin of the World Health Organization, 91, 386-388.
Hoot, N. R., & Aronsky, D. (2008). Systematic review of emergency department crowding: causes, effects, and solutions. Annals of emergency medicine, 52(2), 126-136.
Hopp, W.J. and Spearman, M.L. (2008). Factory Physics International Edition, New York, NY: McGraw-Hill.
Hopp, W. (2011). Supply chain science. Long Grove, Ill.: Waveland Press.
Howell, E., Bessman, E., Kravet, S., Ken, K., Robert, M., & Wright, S. (2008). Active Bed Management by Hospitalists and ED Throughput. Annals of Internal Medicine, 804–810.
Hwang, U., Richardson, L. D., Sonuyi, T. O., & Morrison, R. S. (2006). The effect of emergency department crowding on the management of pain in older adults with hip fracture. Journal of the American Geriatrics Society, 54(2), 270-275.
Jack, E. P., & Powers, T. L. (2004). Volume flexible strategies in health services: a research framework. Production and Operations Management, 13(3), pp. 230-244.
Kao, C. Y., Yang, J. C., & Lin, C. H. (2015). The impact of ambulance and patient diversion on crowdedness of multiple emergency departments in a region. PloS one, 10(12).
Karlsson, 2009. Researching Operations Management. 1 ed. New York: Routledge.
Khare, R. K., Powell, E. S., Reinhardt, G., & Lucenti, M. (2009). Adding more beds to the emergency department or reducing admitted patient boarding times: which has a more significant influence on emergency department congestion?. Annals of emergency medicine, 53(5), 575-585.
