Martin Lee David Gunn MBChB, Presenter: Nothing to Disclose

Yasin Gocgun MSC, Abstract Co-Author: Nothing to Disclose

Brian Bresnahan PhD, Abstract Co-Author: Nothing to Disclose

Norman Joseph Beauchamp MD, MHS, Abstract Co-Author: Research grant, Koninklijke Philips Electronics NV

Archis Ghate PhD, Abstract Co-Author: Nothing to Disclose

To evaluate radiology scheduling policies using computer simulations and operations research tools to identify potential efficiency and quality improvements.

At a busy level 1 trauma center which serves inpatients, outpatients and emergency department (ED) patients, times of CT request and patient arrival time were collected over a three month period. Patients were categorized as inpatients, non-critical ED, critical ED, and outpatients. A two CT scanner scenario was used for the model. CMS reimbursement rates for the seven commonest, outpatient and discharged ED patients were obtained. Five heuristic scheduling policies were developed for CT patient scheduling, and an optimal policy was devised using Markov Decision Process (MDP) rule building. These were developed for both one and two CT scanner scenarios. Thirty-two scenarios were analyzed by varying revenue, cost and arrival probability and these were compared to the optimal scenario using sensitivity analysis. Outcome measures included daily CT revenue-penalty (or revenue gap), number of patients scanned, average patient waiting time.

The average revenue gap between the optimal and the optimal policy and the heuristic decision rules was about 5%, with a maximum value of 10%. Using the optimal policy could result in average gains of $230 / day in a single scanner scenario, and $150 / day with a two CT scanner scenario. The affect of using an MDP derived decision process has the greatest impact on revenue, waiting time, and number of patients served when resources are scarce (ie. one CT scanner). Average daily revenue gap is more sensitive to inpatient penalty cost and incremental revenue when a single scanner is used, and IP revenue and outpatient arrival probability in a two CT scanner scenario. Optimal policies schedule inpatients at the end of the day. Scheduling rules that prioritize one type of patient generally perform better than ad-hoc policies.

Scheduling policies devised by the MDP process could be implemented to optimize utilization of current CT resources, determine how to use resources if a CT scanner goes down, or to evaluate the revenue and quality benefits of purchasing more equipment.

Scheduling policies devised by the MDP process may improve radiology operational efficiency, quality, and help plan for future purchases or outages.

Gunn, M, Gocgun, Y, Bresnahan, B, Beauchamp, N, Ghate, A, An Operations Research Approach to Multicategory Patient Scheduling for Computed Tomography.  Radiological Society of North America 2009 Scientific Assembly and Annual Meeting, November 29 - December 4, 2009 ,Chicago IL. http://archive.rsna.org/2009/8002411.html