Abstract
Overcrowding at EDs is a world known problem which negatively affects the quality of medical care. It is evident, for example, by long patients’ waiting times. EDs frequently use average patients’ length of stay (LOS) as a performance indicator. Long LOS is commonly correlated with overcrowding. In this research, a prototype of an electronic online digital dashboard, termed operational BI, was developed following the Design Science Research methodology. The system is targeted to be used by the ED staff. Simulation was used to assess if such a system, which displays ED’s critical data in a dashboard visualisation, can decrease LOS, thereby ease EDs’ overcrowding. Six scenarios were simulated, depicting various use patterns. The results show a potential decrease of 34–44% in average LOS, depending on the use pattern. These are promising results that call for further, real-world examination of the system.
1. Introduction and background
Emergency departments (EDs) are the entry point to hospitals. In recent years, the number of ED referrals is increasing beyond the population growth rate, a major challenge for public health since the late 1990s (Hwang & Concato, Citation2004). During the years 1997 – 2007 the growth in ED referrals was twice as high as could be explained by the population growth (Tang, Stein, Hsia, Maselli, & Gonzales, Citation2010). Consequently, many EDs are over-crowded. Overcrowding can be explained by population ageing, clinical staff shortage, bed shortage and ease of ED referrals (Carter, Pouch, & Larson, Citation2014). Recent studies, however, pointed at the overcrowded hospital as a main cause, since the inability to transfer ED patients to inpatient beds, also known as patients boarding, results in an over-crowded ED (Moskop, Sklar, Geiderman, Schears, & Bookman, Citation2009a). ED crowdedness is often measured by patients’ waiting times, patients’ average length of stay (LOS), bed occupancy, waiting rooms occupancy, patients hospitalisation in corridors, and patients leaving without being seen by a doctor (LWBS) (Hwang & Concato, Citation2004). ED overcrowding is believed to negatively affect the quality of medical care since over-stressed medical staff cannot provide the best care. This can result in inaccurate diagnoses and even unnecessary deaths (Carter et al., Citation2014, Richardson, Citation2006).
Efforts to ease EDs’ overcrowding include investments in medical infrastructures and staff on the one hand, and process improvements on the other hand (Lo et al., Citation2014). While the former is costly and takes a long time to be effective, the latter entails more immediate results, yet is quite challenging to implement (Moskop, Sklar, Geiderman, Schears, & Bookman, Citation2009b). The use of electronic medical records (EMR) systems is considered one of the means to improve medical processes, yet empirical findings regarding their effectiveness are scant in the ED context, and inconclusive (Handel & Hackman, Citation2010).
Israeli EDs have recently implemented the Chameleon® EMR as a mean to improve medical processes. The intention is to extend the system further, to include operational Business Intelligent (BI) capabilities for assisting on-going EDs’ administration. While the more traditional BI aims at supporting high-level managerial decisions and relies on historical data, Operational BI (also termed often as ‘Real-Time BI’), aims at supporting ongoing ‘here and now’ decisions and, accordingly, uses current data. Operational BI is particularly effective for managing on-going complex and hectic operational scenarios (Sandu, Citation2008; Search Business Analytics, Citation2014). As a concept that ‘marries’ data analysis and visualisation, and real-time processes, the operational BI is considered a novel and a somewhat revolutionary approach (Eckerson, Citation2007), allowing immediate process improvement and fast reaction to changing conditions. Operational BI is commonly operationalized as digital dashboards, dynamically displaying recently-collected data. While digital dashboards displaying static data have long been part of traditional BI implementations (Klipfolio, Citation2014), dashboards that are based on real-time data toward supporting ongoing operations are quite novel in the healthcare industry in general, and in EDs in particular (http://www.d2ihc.com/home/#welcome, accessed February 23, 2016).
Thus, following the principles of design science research (DSR), this study aims at developing and examining the feasibility and the potential contribution of using operational BI systems at EDs. The proposed prototype is designed as a flexible, customizable solution, where the displayed key performance indicators (KPIs) can be adjusted to meet specific requirements of each ED, and to changes in the environment. The system displays dynamic EMR-generated, real-time ED data on digital dashboards, and includes drill-down functionality, aimed at directing staff and administration’s decisions. It is hypothesised that the ability to observe the overall ED status with highlighted alerts and exceptions can decrease LOS. LOS is commonly measured by EDs as an indicator for effectiveness and efficiency, and used as a proxy for good service (Casalino et al., Citation2014). Thus, the prototype evaluation will focus on examining the following research question:
| (1) | Can optimal use of real-time BI by ED staff members shorten the length of stay? | ||||
The prototype has been developed based on the requirements of a large and busy Israeli ED, using real data provided by the hospital’s IT department. It has been demonstrated to the ED staff for comments and amended accordingly, and then examined on a simulated ED, developed based on the ED’s real data, to assess its effect on LOS.
The rest of the paper is constructed according to the six stages recommended by Peffers, Tuunanen, Rothenberger, and Chatterjee (Citation2007) for DSR papers, described next.
2. Methods
The study used the DSR methodology (DSRM) (Hevner, Citation2007; March & Storey, Citation2008; Peffers et al., Citation2007). DSRM requires identification of the problem content and context or the requirements of the relevant stakeholders, organisations and technologies, and then developing a novel solution that effectively addresses these requirements or problems (Gregor & Hevner, Citation2013). The final stage includes a thorough evaluation of how well the developed artefact solves the identified problem (Hevner, March, Park, & Ram, Citation2004). This study applies the framework developed by Peffers et al. (Citation2007) (Figure ), which guides the execution of an effective DSR. It consists of six activities: 1) problem identification and motivation; 2) define the objectives for a solution; 3) solution design and development; 4) solution demonstration; 5) solution evaluation; and 6) solution communication.
Figure 1. Design science research methodology (Peffers et al., Citation2007).
To carry out the first and second activities we conducted in-depth literature and state-of-the-art reviews. Next, we observed and documented the processes at three EDs, taking process measurements based on Work Research methods. Finally, semi-structured interviews with the ED staff were conducted. The interviews were analysed using the ‘unit of meaning’ method suggested by Miles and Huberman (Citation1994), aimed at extracting the system’s requirements. The screens design was based on User Experience (UX) principles (Hartson & Pyla, Citation2012), and approved by an expert. The initial prototype version (V1.0) was developed using QlikView on demo data that adhered to the data structure at the Beilinson Hospital ED, which was the primary stakeholder. The rest of the paper is reported according to the six DSRM steps.
3. Solution
The following six chapters describe the design, development and evaluation of the digital dashboard:
3.1. Problem identification and motivation
Currently, ED doctors and administrators lack a comprehensive view of the state of the ED concerning long waits. Furthermore, there is no view of hospitalisation departments’ occupancy to guide hospitalisation decisions by the ED staff. This in turn causes unnecessary boarding of released patients at the ED, indicated by prior research as one of the crowdedness reasons.
3.2. Defining solution objectives
The solution objective was to provide the ED staff with online information concerning potential bottlenecks in the patients’ streaming processes, which cause avoidable waiting. The information should be quickly and easily comprehended, accurate, timely and useful, provided in a push rather than pull mode.
3.3. Artefact design and development
The design of the artefact followed the principles of User Experience (UX) (Hartson & Pyla, Citation2012), and further approved by an expert. Figures and present the visualisations described next.
Figure 2. Dashboard main screen.
Figure 3. Departments’ occupancy screen.
Three general parameters are displayed on the upper right side of the main screen (figure ): number of patients not yet assigned a doctor or a nurse, number of assigned patients with a long wait for a doctor by doctor’s name, and number of those waiting for a nurse by nurse name.
Patients are prioritised by severity degree, and exceptionally long waiting times are indicated by colours. For both roles there is an indication how many patients are waiting longer than the ED maximum wait allowed for each procedure (customizable).
3.3.1. Alerts
Three flashlights on the upper left side of the screen indicate exceptions: exceptional blood test results, exceptional radiology results and allergies. A RED flashlight means an exceptional indicator for at least one patient.
3.3.2. Overcrowding
Displayed on the lower centre and right side of the screen, this display includes a gauge showing current occupancy rate compared to the ED’s maximum occupancy, as well as the current number of patients at the ED, average length of stay in hours, maximum length of stay, and number of patients with an exceptional length of stay. Another graph displays the last 24 h with the hourly number of patients at the ED, and how close it is to the maximum occupancy, as well as hourly arrival and discharge rates. There is also a patient list with a drill-down option.
3.3.3. Other indicators
Other indicators include pie charts of blood and radiology tests, as well as consultations yet to be handled divided into three statuses, and the number of discharged patients by destination (home or hospitalisation). Additional screens containing other indicators, as well as expansions of the ones appearing on the main screen, have also been developed.
Users can filter the display by various parameters (i.e. a doctor can filter the display to only show her patients) and drill-down to the single patient level in order to identify sources of exceptions.
An important indicator included in a separate screen displays the occupancy of other departments (Figure ). These indicators were included since data from Israeli hospitals indicate that about 50% of the times, destination departments claim full occupancy although there are available beds. It was therefore assumed that changing the current process from request to move a patient into a department to automatic assignments based on displayed occupancy, will decrease the number of boarded patients, a situation that significantly contributes to an ED crowdedness. This change, however, is politically sensitive and not straight forward to implement.
3.4. Solution demonstration
The V1.0 prototype was demonstrated to the Beilinson ED staff for feedback, to validate that the artefact is indeed perceived as a solution to the problem (finding a context). Four doctors, one nurse and one ED administrator were asked to grade each one of the fifteen dashboard components that were presented as either very useful, somewhat useful or not useful. Of the 15 components, components 4 and 5 were rated as not useful hence replaced. Components 2 and 3 were rated as somewhat useful hence repositioned in an inner screen. All other items were rated as useful by most of the participants. The participants were also asked to evaluate the overall perceived usefulness of the system and the perceived ease of use. The overall usefulness was generally perceived as high, and the perceived ease of use – as medium.
Two main screens of the amended dashboard prototype (V2.0) are presented in Figures and . A click on an item opens it in a new screen allowing a drill-down functionality to the level of the patients causing the specific warning to display. For example, clicking on an indication of an abnormal blood test will display the health records of the patients with such results, including bed number.
The system can be used on a laptop, desktop or windows-based tablet computers thereby available to the ED staff from all workstations. In an optimal use, a large screen will be displayed at the ED manager office for general administration of the ED and quick response to situations that affect crowdedness.
3.5. Solution evaluation
A field examination at the ED was impossible because the product was still a prototype, therefore the evaluation phase was carried out using a simulated ED, based on the real-world data extracted from the Chameleon EMR at the Beilinson ED. 31,968 records were extracted describing ED patients admitted between 1.5.2014 and 23.7.2015. The results are presented following the description of the simulation and its assumptions.
3.5.1. The simulation
Use of a simulation allows comparing the impact of various scenarios and interventions on dependent variables, hence is adequate for the current study. Prior research has used simulation for similar studies, including at EDs (Borgman, Mes, Vliegen, & Hans, Citation2015; Diefenbach & Kozan, Citation2011; Jerbi & Kamoun, Citation2009).
The first activity was to clean the data by culling records where many fields included incorrect (non-rational or outlier) or missing data and to add calculated fields of waiting times between activities based on start and end times that were available as part of the record. The final database included 16,286 records (patients). Next, the simulation was developed using the Arena software, where the dependent variable was the average patients’ LOS. A Monte Carlo simulation was used, that utilises the distribution of the simulation independent variables and principles of Queues Theory to generate random appearances of entities (patients). Thus, the simulation’s input is the distributions (calculated by the simulation based on the real data provided) and the outputs are the results of the dependent variable(s) (Mahadevan, Citation1997). The five main processes enacted by the simulation were: 1) random retrieval of entities and their characteristics based on the real-data distributions, 2) patient reception at the ED including assignment of a triage index and bed assignment, 3) execution of medical processes based on the patient characteristics, including entity splits to allow execution of parallel processes where possible, 4) medical treatment and tests execution, including: nurse treatment, blood tests, imaging and consultancy, and 5) decision (hospitalisation or discharge). The simulation was then run on current state scenario (without any intervention), and the average and variance of LOS for hospitalised and discharged patients (separately) was compared to the variables calculated from the real data, using z-test. This stage is called ‘validation’ and is aimed to ensure that the simulation closely resembles the real world in terms of the focal variables (Altiok & Melamed, Citation2010). A p-value greater than 0.05 is required to indicate a valid simulation before proceeding to the next stage.
In the next and final stage the simulation was run under assumptions of two types of the BI system use: ideal and realistic uses. In an ideal use, whenever an indicator turns ‘yellow’ the patient is treated immediately. In a realistic use, this happens only when an indicator is turned ‘red’. Likewise, three types of BI systems were simulated: 1) the system includes ED operational indicators, without hospital departments’ occupancy (termed hereafter occupancy transparency), 2) the system only includes departments’ occupancy without ED indicators, and 3) the system includes both ED indicators and departments’ occupancy. These types were selected to allow flexibility in use pattern.
Further details about the simulation, its structure and underlying assumptions as well as detailed intermediate results, can be obtained from the authors.
4. Results
The simulation ran under six different interventions: 0) no intervention, namely no BI system is used (current state), 1) Only the alerts of the BI system are used (without departments’ occupancy) in an ideal use pattern, 2) Only the alerts of the BI system are used (without departments’ occupancy) in a realistic use pattern, 3) Only the occupancy transparency part of the BI system is used (without the alerts), 4) options 1+3, and 5) options 2+3. Average LOS was measured for each intervention for all patients, for hospitalised patients and for discharged patients (Table ).
Table 1. Simulation results.
As depicted in Table , using only the alerts of the operational BI system (options 1 and 2) can potentially reduce the average LOS by ~30% (for realistic use pattern), with more significant effect on the LOS of discharged patients, as expected. Using only the occupancy transparency feature (option 3) has a minor effect of ~8% reduction, yet using both features even under the realistic use assumption can potentially reduce average LOS by ~34%.
4.1. Solution communication
The prototype and the simulation results were presented in a local academic conference in Israel, and to the Beilinson ED staff, who indicated that the promising results call for a real field examination. It was also delivered to Elad Health Systems, the company in charge of the Chameleon EMR system, to secure their collaboration in advancing the system from a prototype to a mature product that can be tested in a hospital’s ED. The intention is to further test the system by real doctors at an ED, and then, pending positive results, present the results to hospital and health policy decision makers, in order to advance the organisational changes required for the implementation of such a system.
The study will also be presented at academic conferences and published in academic and professional journals.
5. Summary and conclusions
This study addresses the global problem of overcrowded EDs using the design science research methodology (DSRM). DSR is about designing an innovative artefact to a real-world problem and examining its contribution to the problem solution. The artefact suggested here is an operational BI system that displays alerts and exceptional situations at the ED, as well as the occupancy of the hospital’s departments. As far as could be found in the literature, such systems are relatively novel in the healthcare industry, particularly in EDs. The proposed system retrieves real-time data from the ED’s EMR system and displays indicators that affect the patients’ waiting times in an easy-to-comprehend visualisation.
Results of a simulation used for the system’s evaluation indicate that it has a potential to decrease the average LOS by up to 44%, depending on the system’s functionality and use pattern.
5.1. Limitations
The main limitation of this study is the inability to test and evaluate the effect of the system’s use on the average LOS in a real ED. This, however, is planned for future research. Another limitation is the locality of the solution design and data. Further research in additional EDs is needed to strengthen the external validity of the results.
5.2. Expected contribution
Hospitals in general, and EDs in particular lag behind other industries in harnessing information technology (IT) for process improvement (Jha et al., Citation2009). This study, therefore, is a step towards showing policy and decision makers in healthcare the potential benefits of IT use in EDs. The results, if supported by additional research, can drive IT implementation and use at EDs in spite of barriers and challenges reported in the literature (Menachemi, Brooks, Schwalenstocker, & Simpson, Citation2009). It also hints at potential positive outcomes of using such a system that can be quantified, as opposed to other studies that yielded mixed results concerning benefits of using healthcare IT in general, and EMRs in particular (Furukawa, Citation2011; Gagnon et al., Citation2012; Johnson et al., Citation2011; Kellermann & Jones, Citation2013). It will thus be quite straightforward to calculate the ROI of such systems if field evaluations of the system yield similar results.
On the theoretical aspect, this study contributes to the DSR field by demonstrating a step-by-step use of Peffers et al. (Citation2007)’s framework. The use of a Monte Carlo simulation is by itself a methodological contribution since it is not often used as an evaluation tool in IT research. In addition, this study paves the way to future research questions. First, the system should be evaluated in a real ED to validate the simulation result. This, however, is challenging because it entails significant process changes and depends on the staff acceptance. Thus future research can use the various IT acceptance models to study the system’s adoption. Furthermore, researchers in the Business Value of IT (BVIT) domain (Melville, Kraemer, & Gurbaxani, Citation2004; Ramirez, Melville, & Lawler, Citation2010) can evaluate and investigate the business value of this system, which seems quite significant at least based on the simulation results. In Human-Computer Interaction (HCI) domain, researchers can study the system adoption and acceptance from the user interface aspect, testing various types of visualisations.
Many research questions can apply to the healthcare IT research domain, that relate to ED process changes and improvements, and their impact on patients’ health outcomes, and on staff performance.
6. Conclusions
This study lends an initial support to the feasibility of using an operational BI system in a hospital’s ED, based on a simulation. It is hoped that a production system can be evaluated in a real ED in a future research.
Disclosure statement
No potential conflict of interest was reported by the authors.
References
- Altiok, T., & Melamed, B. (2010). Simulation modeling and analysis with arena. Burlington, MA: Academic Press, Elsevier.
- Borgman, N. J., Mes, M. R. K., Vliegen, I. M. H., & Hans, E. W. (2015). Improving the design and operation of an integrated emergency post via simulation. J of Sim, 9, 99–110.10.1057/jos.2014.5
- Carter, E. J., Pouch, S. M., & Larson, E. L. (2014). The relationship between emergency department crowding and patient outcomes: a systematic review. Journal of Nursing Scholarship, 46, 106–115.10.1111/jnu.12055
- Casalino, E., Wargon, M., Peroziello, A., Choquet, C., Leroy, C., Beaune, S., Pereira, L., Bernard, J., & Buzzi, J. -C. (2014). Predictive factors for longer length of stay in an emergency department: a prospective multicentre study evaluating the impact of age, patient’s clinical acuity and complexity, and care pathways. Emergency Medicine Journal, 31, 361–368.10.1136/emermed-2012-202155
- Diefenbach, M., & Kozan, E. (2011). Effects of bed configurations at a hospital emergency department. J of Sim, 5, 44–57.10.1057/jos.2010.1
- Eckerson, W. W. (2007). Best practices in operational BI. Business Intelligence Journal, 12, 7–9.
- Furukawa, M. F. (2011). Electronic medical records and the efficiency of hospital emergency departments. Medical Care Research and Review, 68, 75–95.10.1177/1077558710372108
- Gagnon, M. -P., Desmartis, M., Labrecque, M., Car, J., Pagliari, C., Pluye, P., Fremont, P., Gagnon, J., Tremblay, N., & Legare, F. (2012). Systematic review of factors influencing the adoption of information and communication technologies by healthcare professionals. Journal of Medical Systems, 36, 241–277.10.1007/s10916-010-9473-4
- Gregor, S., & Hevner, A. R. (2013). Positioning and presenting design science research for maximum impact. MIS quarterly, 37, 337–355.
- Handel, D. A., & Hackman, J. L. (2010). Implementing electronic health records in the emergency department. The Journal of Emergency Medicine, 38, 257–263.10.1016/j.jemermed.2008.01.020
- Hartson, R., & Pyla, P. S. (2012). The UX book: process and guidelines for ensuring a quality user experience. Waltham, MA: Elsevier.
- Hevner, A. R., March, S. T., Park, J., & Ram, S. (2004). Design science in information systems research. MIS Quarterly, 28(1), 75–105.
- Hevner, A. R. 2007. A three cycle view of design science research. Scandinavian journal of information systems, 19, article 4. http://aisel.aisnet.org/cgi/viewcontent.cgi?article=1017&context=sjis
- Hwang, U., & Concato, J. (2004). Care in the emergency department: How crowded is overcrowded? Academic Emergency Medicine, 11, 1097–1101.10.1111/acem.2004.11.issue-10
- Jerbi, B., & Kamoun, H. (2009). Using simulation and goal programming to reschedule emergency department doctors/’ shifts: case of a Tunisian hospital. J of Sim, 3, 211–219.10.1057/jos.2009.6
- Jha, A. K., DesRoches, C. M., Campbell, E. G., Donelan, K., Rao, S. R., Ferris, T. G., Shields, A., Rosenbaum, S., & Blumenthal, D. (2009). Use of electronic health records in U.S. hospitals. New England Journal of Medicine, 360, 1628–1638.10.1056/NEJMsa0900592
- Johnson, K. B., Unertl, K. M., Chen, Q., Lorenzi, N. M., Nian, H., Bailey, J., & Frisse, M. (2011). Health information exchange usage in emergency departments and clinics: the who, what, and why. Journal of the American Medical Informatics Association, 18, 690–697.10.1136/amiajnl-2011-000308
- Kellermann, A. L., & Jones, S. S. (2013). What it will take to achieve the as-yet-unfulfilled promises of health information technology. Health Affairs, 32, 63–68.10.1377/hlthaff.2012.0693
- Klipfolio. (2014). Retrieved February 23, 2016, from http://www.klipfolio.com/resources/articles/operational-analytical-bi-dashboards
- Lo, S. M., Choi, K. T. Y., Wong, E. M. L., Lee, L. L. Y., Yeung, R. S. D., Chan, J. T. S., & Chair, S. Y. (2014). Effectiveness of emergency medicine wards in reducing length of stay and overcrowding in emergency departments. International Emergency Nursing, 22, 116–120.10.1016/j.ienj.2013.08.003
- Mahadevan, S. 1997. Monte Carlo simulation. In T. R. Cruse (Ed.). Reliability-based mechanical design (pp. 123-146). New York & Basel: Marcel Dekker.
- March, S. T., & Storey, V. C. 2008. Design science in the information systems discipline: an introduction to the special issue on design science research. MIS quarterly, 32, 725-730.
- Melville, N., Kraemer, K., & Gurbaxani, V. (2004). Information technology and organizational performance: an integrative model of it business value. MIS Quarterly, 28, 283–322.
- Menachemi, N., Brooks, R. G., Schwalenstocker, E., & Simpson, L. 2009. Use of health information technology by children’s hospitals in the United States. Pediatrics (123:Supplement_2), pp. S80-84.10.1542/peds.2008-1755F
- Miles, M. B., & Huberman, A. M. 1994. Qualitative data analysis: An expanded sourcebook. 1994. Beverly Hills: Sage Publications.
- Moskop, J. C., Sklar, D. P., Geiderman, J. M., Schears, R. M., & Bookman, K. J. (2009a). Emergency department crowding, part 1 – concept, causes, and moral conse-quences. Annals of Emergency Medicine, 53, 605–611.10.1016/j.annemergmed.2008.09.019
- Moskop, J. C., Sklar, D. P., Geiderman, J. M., Schears, R. M., & Bookman, K. J. (2009b). Emergency department crowding, part 2 – barriers to reform and strategies to overcome them. Annals of Emergency Medicine, 53, 612–617.10.1016/j.annemergmed.2008.09.024
- Peffers, K., Tuunanen, T., Rothenberger, M. A., & Chatterjee, S. (2007). A design science research methodology for information systems research. Journal of Management Infor-mation Systems, 24, 45–77.10.2753/MIS0742-1222240302
- Ramirez, R., Melville, N., & Lawler, E. (2010). Information technology infrastructure, organizational process redesign, and business value: an empirical analysis. Decision Support Systems, 49, 417–429.10.1016/j.dss.2010.05.003
- Richardson, D. B. (2006). Increase in patient mortality at 10 days associated with emergency department overcrowding. Medical Journal of Australia, 184, 213–216.
- Sandu, D. (2008). Operational and real-time business intelligence. Revista Informatica Economică, 3, 33–36.
- Search Business Analytics. (2014). Retrieved December 31, 2014, from http://searchbusinessanalytics.techtarget.com/definition/operational-business-intelligence
- Tang, N., Stein, J., Hsia, R. Y., Maselli, J. H., & Gonzales, R. (2010). Trends and charac-teristics of US emergency department visits, 1997-2007. JAMA: The Journal of the Amer-ican Medical Association, 304, 664-670.