An adaptive non-linear great deluge algorithm for the patient-admission problemby Saif Kifah, Salwani Abdullah

Information Sciences


Accepted 5 October 2014

Available online 13 October 2014


Patient admission problem

Meta-heuristic ing patients’ comfort. The goal of this work is to propose a meta-heuristic approach called lity requirements , medical e the assig artment to he/she is assigned, and patient personal preferences.

It is difficult, if not impossible, to attain all of the objectives for patients in such a complicated problem. Often it is ad to generate a solution achieving only the most important objectives [18]. Therefore, many researchers have attem 0020-0255/ 2014 Elsevier Inc. All rights reserved. ⇑ Corresponding author.

E-mail addresses: (S. Kifah), (S. Abdullah).

Information Sciences 295 (2015) 573–585

Contents lists available at ScienceDirect

Information Sciences journal homepage: www.elsevier .com/locate / insIn this work, we focussed on the operational level of the patient admission problem by elevating the qua of patients’ treatments. The problem includes several different components, e.g., departments, specialism ment and patients. The arrival and discharge dates for a patient need to be known in advance to facilitat to a bed. Constraints influencing a patient’s admission include the required medical treatment, the depequipnment which visable pted toProviding healthcare to a hospital patient is an essential task for gradual health recovery. A patient requires health services in an environment that satisfies all medical necessities and personal preferences. Patient admission is therefore a complex task. This is because the many hospital resources, such as oxygen or telemetry, are either necessary or not to any given patient. Assigning a patient to a bed in the department specialised to treat his/her condition is a critical requirement. In addition, there is uncertainty in the health condition of the patient, and changes often occur that necessitate provision of various medical equipment. These challenges prompt the healthcare system to offer high quality treatment with limited resources [20].Adaptive non-linear great deluge 1. Introductionan adaptive non-linear great deluge algorithm to address patient admission problems. We investigate the effect of adaptively updating the non-linear decay rate of the water level in the great deluge algorithm during the course of optimisation. The great deluge algorithm needs only to set up the water level as its basic parameter, which makes it naturally very attractive for solving optimisation problems. The performance of the proposed approach is verified using standard benchmark datasets. Empirical results demonstrate that the proposed algorithm obtains competitive results when compared to other approaches in the scientific literature.  2014 Elsevier Inc. All rights reserved.An adaptive non-linear great deluge algorithm for the patient-admission problem

Saif Kifah ⇑, Salwani Abdullah

Data Mining and Optimisation Research Group (DMO), Centre for Artificial Intelligence Technology, Universiti Kebangsaan Malaysia, 43600 Bangi,

Selangor, Malaysia a r t i c l e i n f o

Article history:

Received 16 December 2012

Received in revised form 1 October 2014 a b s t r a c t

The patient-admission scheduling problem is a combinatorial optimisation problem that is receiving attention in health care practice. This is because hospitals are experiencing increasing pressure to optimise the usage of their resources while simultaneously increasimproving and equal, only improving, simulated annealing, and great deluge. The hyper-heuristic applied was based on the 574 S. Kifah, S. Abdullah / Information Sciences 295 (2015) 573–585claim by Burke et al. [7] that it is able to manage a wide range of domains by using a general system. Demeester et al. [13] introduced a mathematical model for the patient admission problem. They argued that the patient admission problem can be treated as a combinatorial optimisation problem. The tabu search algorithm was hybridised with a token ring approach and used as a decision support system at the operational level. An integer programming approach was also considered to apply to the patient admission problem. One hour of calculation was granted for integer programming without finding a feasible solution. In addition, a week of processing time has been tried in an attempt to find the optimal solution, but it resulted in failure. Ceschia and Schaerf [8,9] considered both simulated annealing and a tabu search to address the problem. They supported the algorithms with a multi-neighbourhood local search procedure with different combinations of neighbourhoods and were able to have effective diversification with different sets of weights for the cost components that constituted their objective function. They believed that the better performance in solution quality and processing time were attributable to a pre-processing stage and new neighbourhood combinations. Bilgin et al. [5] employed a general high-level hyper-heuristic for handling the two different yet similar optimisation domains of patient admission and nurse rostering problems; they applied a local search to serve as low level heuristic to gain generality inside the hyper-heuristic framework. Vermeulen et al. [31] present a different patient admission model compared to [4]. The model includes independent departments, resources, individual patients and their activities. Staff was treated as one of the resources in the model. The patients were assigned to two different departments with a set of resources, constraints, staff preferences and budget considerations; all departments work efficiently and consider the staffs’ preferences. They applied a multi-agent pareto appointment exchange to obtain a distributed solution for scheduling patient activity. Chien et al. [12] focused on physical therapy for rehabilitation patients. Based on a restricted sequence of assigning patients to treatment, they formulated the problem as a hybrid job-shop scheduling problem in which a genetic algorithm was employed to improve the operation efficiency by reducing the patient waiting time. Chen et al. [11] introduced a genetic algorithm to optimise a long-term admission strategy for ophthalmology departments in hospitals. They formulated a coding scheme of chromosomes to evaluate two objective functions: efficiency and fairness. Experimental results showed that the genetic algorithm strategy obtained better performance than both the first-come first-served strategy and the greedy strategy. Ceschia and Schaerf [10] introduced a new version of a patient admission procedure that includes both static and dynamic environments. The procedure considers the uncertainty of stay length and the possibility to delay the patient’s admission. Moreover, it considers the risk of overcrowding.