Reza Samavi
Software Systems
Emergency department (ED) wait times in Ontario have consistently been an issue and can be improved through the use of technology. Reducing ED overcrowding and improving the ED’s operational efficiency is an important goal of Ontario healthcare delivery.
The objective of this project is to develop a predictive resource allocation tool for the emergency department (ED) of a medium size hospital in Ontario in order to predict incoming ED events given a number of social and environmental determinants.
The system will extract the occupancy patterns of the ED in the past few years, and will investigate the relationship (and correlations) between social or environmental determinants and ED visits for that period. The tool also include an interactive "what-if" scenario analysis component for the hospital management team and healthcare policy makers in order to improve the resource allocation given the change in some of the social determinants.
- A machine learning (ML) based predictive model is suggested for the tool's engine.
- An object oriented programming approach is suggested - Python would be the best choice for the programming language as the ML module libraries are mainly in Python.
- the system should be designed with modularization in mind as the system will use multiple external libraries.
- The suggested approach for the back-end design has two parts: (1) a comparative analysis of type of ML algorithms (e.g., NN or SVM, etc.), (2) training and testing the selected ML model
- For the front-end a web-based design using active server pages is suggested.
The students will first collect the necessary data and develop a machine learning model to see how the current visits of ED are correlated to a number of social or environmental determinants including economic conditions, weather conditions, even tweeter feeds etc. Then the student turn the predictive model to a software tool that can be used by the hospital management teams and policy makers to understand the sensitivity of the ED resource utilization to the external stimulus. The dataset for the project will be either provided by the hospital if the ERB application is timely and successful or by using an open dataset or a generated synthetic dataset. The back-end of the system will be the machine learning (ML) model developed by the students after thorough investigation of different ML algorithms. The front-end of the system allows hospital management teams and policy makers to interact with the predictive model in an easy-to-use manner and investigate the outcome based on different scenarios in the outside world that is out of control of the ED team.
- Leading Requirements elicitation and analysis
- Leading identification and documentation of system specifications
- Ensuring the SDLC is improving according to the selected methodology
- Ensuring different modules and sub-modules can be integrated and synced.
- Leading integration test and ensuring software testing and quality assurance.
- Data collection (search for open datasets or generating synthetic data if necessary)
- Data cleaning and preparation
- Data quality assurance
- Data query (and procedures, triggers if necessary) development
- Design "what-if" scenarios working with Student A
- Identify feature vectors
- Develop ML model
- Train and Test ML model
- Experimental evaluation of the ML model
- Implement back-end of "what-if" scenarios working with Student B
- Design user interface (UI) module
- Implement front-end of "what-if" scenarios working with Student B
- Prototyping the system UI
- Running usability test
- Integrate with back-end module
RSA01: Emergency Department Predictive Model | Reza Samavi | Monday September 6th 2021 at 04:38 PM