Inspiration
Just coming out of the Covid pandemic we are all aware how important ICUs are at the time of emergency. We must be well prepared with the resource allocation and clinical decision making logistics for the available ICUs in any hospital. So a prediction system about how long a patient will need to use the ICU will be very helpful to do the necessary planning ahead of time to take efficient care to critically ill patients keeping the cost management under control
What it does
The ICU-LOS Crystal Ball takes inputs like the initial diagnosis, gender, ethnicity, and what procedures are being applied on the patients just admitted to the hospital. Based on these the system will predict an estimated length of stay for the newly admitted patient who has been transferred to the ICU.
How we built it
First we studied the database that had all the information about the different tables loaded with patients records. Then we identified which factors can be important as decision makers for the icu stay problem. After identifying the required tables we had to find out a way to import the info into the azure workspace so that we can use the machine learning principles on the data. After the data was set up we applied the machine learning techniques to design and develop the system.
Challenges we ran into
The first challenge was to get the data from the source provided to us. Secondly how to import the data into azure workspace.
Accomplishments that we're proud of
Inspite of limited time I could develop the prediction stay for the ICU stay.
What we learned
I learnt a new skill, azure machine learning in a very short time period. I can now think how to apply ML to automate and enhance current traditional business processes. I also learnt how to become a responsible researcher while getting access to the database
What's next for ICU-LOS Crystal Ball
The current project is just the beginning of a big and useful project. To make this work in the real world, we need to add the following features. a. To have a meeting with the relevant team to know which factors are important to determine the stay of a patient in ICU b. How to get access to the real data since I think the age of patients is also important for the system to predict about the length of stay. In the current dataset the birthdates were fictitious. c. To automate the system of data ingestion into azure ml workspace. I did manually here. d. To build an app that will provision the hospital staff to enter the necessary values in a very easy and user-friendly way and get the prediction output right away.
Log in or sign up for Devpost to join the conversation.