Operations improvement systems for hospital emergency department
Student thesis: Doctoral Thesis
Emergency Department (ED) is a critical component in a hospital system providing services to patients in urgent needs of medical cares. It deals with patients from serious injuries to minor ailments, and is the major channel of admitted patients to hospital wards. Because of this, ED is a very dynamic and complicate system as compared to other hospital departments. Due to the important role of ED in the entire healthcare system, its performance has been constantly measured in many different aspects. From patients' perspective, the long waiting time and length of stay (LOS) are the most common symptoms of a poorly performing ED. While from ED staff's perspective, the poor ED performance is more associated with the overlook of staff's well-being in the working place, such as fatigue alleviation, work-family balance, and creating satisfying working schedule (e.g. fulfil staff's preference on shift types and sequences). The problems mentioned above are prominent in Hong Kong. Although the ED staff have been working very hard to maintain the service quality at an acceptable level, the brutal reality of increasing patient demand and shortage of experienced ED staff have inevitably caused ED service decline. This worrying fact has drawn increasing public criticism in recent years over the ED overcrowding and services delay, and undermined the morale of the short-handed ED staff. Therefore, it is a longstanding problem of dissatisfaction among both ED patients and staff which has plagued the ED management for a long time. Unfortunately very limited research has systematically studied this problem, especially from an engineering point of view. This research is thus carried out with two main objectives. The first is to develop an ED decision support framework based on multi-criteria decision making (MCDM) methods. This framework is to assess the performance of alternative ED operational strategies for different patient demands. The performance is measured by a set of ED performance indicators and their values are obtained from an ED simulation model. The second is to develop an efficient and practical scheduling algorithm for the ED staff aiming to alleviate staff fatigue and maximize individual's preferences on shift. Under the first objective, there are four major research outcomes. The details are described below. First, the quantitative relationship between waiting time/LOS and a number of relevant factors is determined by regression analysis. Particularly, among these factors, the volume of non-emergent patients (Categories 3, 4, and 5) in different phases of an ED journey (waiting for triage, waiting for consultation, and in treatment) and arrivals of emergent patients (Categories 1 and 2) are confirmed predictive of the patient waiting time/LOS. This study provides a quantitative understanding of the potential causes to the current services delay in the partner ED, and the result is indicative of what factors should be set controllable in the ED simulation model later. Second, a forecasting model of daily patient demand is developed. The model describes the demand as a dependent variable of weekday/end, weather condition, and influenza outbreak. The result shows that non-linear modelling technique such as artificial neural network (ANN) produces better modelling accuracy than traditional linear method such as multiple linear regression (MLR). The resulting daily demand will be fed to the ED simulation model so that the user can investigate how the change of days, weather, and influenza outbreak level will affect the ED performance. Third, a discrete event simulation model is developed for the ED operations. The knowledge of ED operations is collected via staff interviews and on-site observations. Using advanced simulation software (Arena), the model is able to animate the complex ED dynamics (patient flows, patient-staff interaction, etc.). The model contains the common ED processes (registration, triage, consultation, etc.) as well as the details of medical procedures for patient treatment. The model is flexible to adapt with different ED operational strategies, and produces a wide range of simulation outcomes at the end of a simulation run. Last, an integrated MCDM method is proposed to model the decision making process of selecting optimal ED operational strategies for various levels of patient demand. This method combines analytic network process (ANP) and technique for order preference by similarity to ideal solution (TOPSIS). The first part of the method (ANP) calculates the relative weight of the ED performance indicators with interdependent influences, and the second part (TOPSIS) ranks the ED operational strategies in terms of their performance with respect to the performance indicators. The performance is actually the simulated results of ED operational strategies on the ED simulation model. The ED performance indicators (i.e. patient waiting time/LOS and ED staff utilisation) and operational strategies mentioned above are formulated combining the opinion from ED staff and the ED simulation model. The TOPSIS scores show that some ED strategies are consistently scored higher while some others are consistently lower. Between them, there are strategies whose TOPSIS scores are heavily dependent on the levels of daily patient demands. Under the second objective, there are two major research outcomes. The details are described below. First, an efficient and user-friendly scheduling algorithm for the ED staff is developed. This algorithm is heuristic-based and written by VBA so that it can operate on Excel worksheet, which is the most common desktop environment for ED staff. The algorithm not only considers the hard constraints of scheduling rules such as legal and ED regulations (e.g. maximum work hour), but also the soft constraints regarding the staff well-being and shift preference (e.g. consecutive night shifts should be avoided). The advantage of this algorithm is that it is user-friendly and costs much shorter time to generate quality schedules than the current manual scheduling method. The downside is that the resulting schedules are not optimal solutions (near-optimal). Second, a benchmark model for the same scheduling problem is developed. The model is formulated by 0-1 linear programming based on the same hard and soft constraints. The advantage of linear programming is that the optimal schedules can be eventually achieved. The downside is that the calculation time could be very long depending on the model complexity. In the comparison of schedule optimality between the proposed algorithm and linear programming, two methods generate comparable schedules if the maximum calculation time is limited to 3 hours. The major contribution of this thesis lies in three areas. First, this is a rare research focusing on the operations improvement in a Hong Kong ED with unique operational characteristics and intense working environment as compared with its western counterparts. Second, the proposed ED decision support framework based on discrete event simulation and MCDM methods is a comprehensive decision making tool which compares and ranks alternative ED operational strategies under different levels of daily patient demand. Third, the scheduling algorithm takes the well-being and shift preference of ED staff into account. The algorithm is easy to be implemented in the ED computers and flexible to modification for new scheduling rules.
- Hospitals, Administration, Emergency services