Lecture 25: Hospital Scheduling © J. Christopher Beck 2008 1 Outline Healthcare & Scheduling Operating room scheduling at Mt. Sinai Problem Models Results © J. Christopher Beck 2008 2 Readings Blake & Donald, Mount Sinai Hospital Uses Integer Programming to Allocate Operating Room Time, Interfaces, pp 63-87, 32(2), 2002. © J. Christopher Beck 2008 3 Healthcare & Scheduling A growth opportunity for scheduling research Historically, much less attention than Staff scheduling (nurses, manufacturing doctors, orderlies) Operating room scheduling Patient scheduling operations, clinics, … Therapy (e.g., radiation) © J. Christopher Beck 2008 4 Healthcare & Scheduling Challenges Uncertainty Large, interacting systems ER, going into labour, complications in surgery, … The law of unintended consequences Complex “people” constraints in a highstress job Many stakeholders © J. Christopher Beck 2008 5 OR Room Scheduling How to allocate OR time among different surgical specialties e.g., ophthalmology, gynecology, surgery, oral surgery, … Cyclical schedule Number and type of ORs available Assign specialties who will be given priority at different times © J. Christopher Beck 2008 6 © J. Christopher Beck 2008 7 3-Step Process Management: total number of OR hours available Nurse manager: # of template schedules # of rooms and hours of opening each day must meet total hours must be feasible with nurses’ collective agreement © J. Christopher Beck 2008 8 3-Step Process Nurse manager: using template, assign available time to departments Competing objectives: hospital wants to reduce cost fewer hours doctors want to maximize income more hours equity among surgical departments © J. Christopher Beck 2008 9 Constraints & Preferences One department per day share by assigning alternate weeks to different departments i.e., alternate Mondays to different depts Consistent schedule from week to week Min/max bounds on number of blocks assigned to each department in a given day/week © J. Christopher Beck 2008 10 Model i – operating room type j – department k – day of week xijk - # of blocks of type i assigned to department j on day k dik – duration of block i on day k © J. Christopher Beck 2008 (long, short) X (main, EOPS) 11 Model Assign xijk such that the sum of the time allocated for a department is equal to their target number of hours penalty for dept j © J. Christopher Beck 2008 target time for dept j 12 Model sj+ – amount of oversupply for dept j sj- – amount of undersupply for dept j © J. Christopher Beck 2008 13 minimize penalty allocated time ± over/under supply all available rooms are allocated bounds on number of rooms assigned to a dept in a day © J. Christopher Beck 2008 bounds on number of specific type of room assigned to a dept in a day 14 bounds on number of specific type of room assigned to a dept in a week arbitrary bound on max. under allocation © J. Christopher Beck 2008 15 Results Full production since 1997 Time to produce schedule reduced from days to 1 or 2 hours OR manager’s time reduced saving $20K/year Faster scheduling more alternatives investigated increased quality Objective measure of quality © J. Christopher Beck 2008 16 Other Points Background section provides an interesting description of how & why the Canadian healthcare system is setup economic incentives, etc. Political realities old process (p. 68) objective criteria reduces conflict © J. Christopher Beck 2008 17 What Do I Have to Know about this Paper? As this is a fairly simple, mostly nontechnical paper, you should have a detailed understanding of both the problem and the model I could give you an OR scheduling problem and ask you to give me a MIP model for it © J. Christopher Beck 2008 18