Computer-based Optimization for Scheduling Emergency Department Physicians

急診醫師排班系統電腦最佳化

莊旺川(Wang-Chuan Juang) ； 廖美珍(Mei-Chen Liao) ； 施正蓮(Cheng-Len Sy) ； 張運德(Yun-Te Chang) ； 郭耀昌(Yau-Chang Kuo) ； 張朝煜(Chao-Yu Chang) ； 萬樹人(Shue-Ren Wann) ； 張宏泰(Hong-Tai Chang)

《Journal of Emergency Medicine, Taiwan》 8卷3期 (2006/09) Pp. 99-105

摘要

目的：建立電腦化的排班系統，為南台灣一醫學中心之急診部門建構最佳化的排班制度。
方法：本研究以「基因演算法」為基礎，利用在Microsoft Excel下的最佳化分析軟體Evolver 4.0，為急診部門的醫師建立一套完善的排班制度。首先，設定排班的加權指數，並訂立兩項基本限制，一為不連班限制，另一為班與班之間要間隔至少16個小時；此外也根據每位醫師的要求設定個人限制。其次，架構排班演算法的流程，該流程包括六個部分：問題、編碼、操作、設限、搜尋、求解等。一旦確立基因演算法的流程後，電腦便開始模擬物種複製、交配與突變的過程，算出各種可能的排班方法。當變異數達到每位醫師班表總加權指數的最低點，並符合各種設限時，電腦即求得最佳化的班表。
結果：本研究以三十個工作日的班表為基準，為十名急診醫師共180個班數求得最佳化班表。排班制度為三班八小時工作制為基礎，白班需要三名醫師值勤，小夜班兩名，大夜班一名。電腦演算的結果顯示每位醫師的總大夜數約2~4班，總值班約為17~19班，而總加權值數的平均值為25.76，其變異數為0.03。
結論：本研究所建立之電腦化排班系統可以符合各種設限，不僅能滿足急診醫師的個人需求，更可確保病人照護品質。

關鍵字

輪班工作 ； 基因演算法 ； 排班 ； 急診醫師

並列摘要

Objective: To organize shift work programs by creating computer-based scheduling strategies for Emergency Department physicians in one medical center in southern Taiwan.
Method: Based on genetic algorithm, the scheduling model we utilized is specified within Microsoft Excel and the optimization solved using Evolver 4.0. We defined a weighted value of shifts (WVS) and set basic constraints of discontinuous shifts, a minimum of 16 hours off-duty between two shifts and personal constraints to cater for individual requirements. Then, a 6-step procedure of algorithm was put forward, including identifying problems, encoding problems, genetic operation, setting constraints, genetic search and solving problems. Next, the computer simulates genetic search and evaluates possible solutions. The operation stops when an optimized result is achieved fulfilling the specified constraints and minimizing the variance of the sum of WVS for each physician.
Result: Taking a schedule of 30 working days as an example, we scheduled a total of 180 shifts for 10 physicians. The system is on a three 8-hour shift basis, of which the day shift requires 3 doctors on duty, 2 on evening shift and 1 on night shift. The result of optimization shows that each physician would work 2 to 4 night shifts and 17 to 19 shifts in total. The mean WVS of each physician was 25.76, while the variance of WVS was only 0.03.
Conclusion: The implementation of this computer-based scheduling system can fulfill the specified constraints, which may ensure patient care while satisfying physicians' individual needs.

並列關鍵字

shift work ； genetic algorithm ； scheduling ； emergency physician