Allocation of Resources in the Emergency Room
Draper Laboratory
Draper is applying its decision-making architecture to problems in healthcare. The figure below illustrates this architecture.
Figure: Information Exploitation and Planning in a Hospital Setting
We will focus at reducing uncertainty in hospital resource allocation with information exploitation and planning. Our initial goal and the corresponding sub-problems are stated below:
Optimize allocation of resources in the ER by predicting the ER load.
- Estimate individual patient’s Length Of Stay
- Incorporate seasonal, weekly and daily variations
- Estimate requirements for ER staffing
We applied some of our non-linear information exploitation techniques on a sample of patient data from BIDMC’s Emergency Department (ED). We evaluated temporal characteristics of the ED load over time (weekly and daily variations), and the patient’s length of stay in the ED. As an example result from our analysis we found that we could predict which of three classes a patient will fall in based on the initial evaluation upon entry into the ED. (Class 1 = stays less than 3 hours; class 2 = stays between 3 and 7 hours; class 3 = stays more than 7 hours.) We achieved a lift of 2.4 for class 1, 1.3 for class 2, and a lift of 1.5 for class 3. Our future goal is to incorporate the length of stay estimations together with the temporal variations into optimization model for the ER resource allocation.
