University of Twente Student Theses


A forecasting model for the requests for MRI scans

Schaik, J. van (2015) A forecasting model for the requests for MRI scans.

[img] PDF
Abstract:Problem description: This study investigates the arrival of patients in need of an MRI at the diagnostics department of a hospital. At this moment it is unclear for the diagnostics department how many patients will arrive with an MRI request and which department they come from. Every patient arrives unannounced, which means that there is also no insight in how busy it will be the next weeks or month. Patients are sent to the diagnostics department by the various specialties, but communication and collaboration between departments is lacking. Objective of the research: The objective of the research is forecasting the number of MRI requests that will be received by the diagnostics department in the coming six weeks. Such a forecast allows scheduling of patients and a more efficient allocation of personnel. Approach: To solve this problem we develop a causal forecasting model in Excel, which uses scheduled number of consultations as input variables and predicts the number of MRIs needed in the coming six weeks. This model is partly based on the model developed by Ooms (Ooms, 2014), which is a model for forecasting the amount of OR days by the Orthopedic Department in the Medisch Spectrum Twente (MST). We extensively analyze the historical data of last year by using data on both the outpatient clinics and of the diagnostics department. We hypothesize that the amount of consultations at the outpatient clinic will influences the number of MRI requests for the diagnostics department. We perform a correlation analysis to check this hypothesis. Measurements/findings: We have used historical data (Jan-Dec 2014) of both the outpatient clinics and the diagnostics department and managed to link 7865 MRI requests (74.6%) to the consultations that led to these MRI requests. We extensively analyzed these data and found that there is indeed a strong correlation between first visit and follow up consultations and a moderate correlation between emergency and peer consultation. The specialties that contribute the most MRIs are: Neurology (27%), Orthopedics (19%), Surgery (12%), General Practitioners (7%), Neurosurgery (7%), Cardiology (4%) and Otorhinolaryngology (4%). Model: Our forecasting model predicts the number of MRI requests per week in a six week planning horizon. As input values for the model we use two different methods. The first part is a causal forecast method which uses the number of planned First Visit Consultation and Follow up Consultations per week for the departments Neurology, Orthopedics, Surgery, Neurosurgery, Cardiology and Otorhinolaryngology. For the General Practitioners and other specialties we assumed a constant amount of MRI requests per week. For the second part we use the time based methods: Exponential Smoothing and Moving Average which uses the difference between the previous demand and the forecasts of the last five weeks as input values. The outcome of the model shows the forecasted MRI requests for the next six weeks in total and per specialty. To validate the model we used the number of first visit and follow up consultations in the period of January 2014 to December 2014 (week 2 to week 47). We compared the forecasted outcomes of the model with the dates the diagnostics department scheduled the MRIs. Our forecast model has a Mean Average Percentage Error (MAPE) of 8.0% and a correlation of 0.72. Conclusions: Our forecast model can be used as a tool to forecast MRI requests and can help the diagnostics department with the allocation of personnel and scheduling of patients in order to reduce the variation in access time. Furthermore it can be used by the diagnostics department as a negotiation and communication method to other specialties to give insight in their variation and to allocate different MRI slots to specialties. Recommendations: We recommend the MST to keep track of MRI requests, from which specialty the MRI requests come from and when the MRI is requested as it is required to run our model. Using the model without these data is possible, but will decrease its forecasting power. Furthermore we recommend the diagnostics department to get access to the data of the number of consultations specialties have scheduled for the coming six weeks. These data help improve the outcome of the model. We recommend MST not to implement this model immediately, but assess its forecasting strength during a six month shadow run. After a successful implementation this research and model can be adapted for use in other applications as well, such as forecasting of CT requests.
Item Type:Essay (Master)
Faculty:BMS: Behavioural, Management and Social Sciences
Subject:44 medicine, 85 business administration, organizational science
Programme:Industrial Engineering and Management MSc (60029)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page