Inspiration
The length of stay in the ICU is an important metric for healthcare professional as it can impact patient outcomes and resource utilisation. In the United States, the average length of stay in the ICU is 3.3 days, with a median of 2.2 days. However, this can vary widely depending on the patient's condition, with some patients staying in the ICU for several weeks or even months. Also ,a longer stay can increase the risk of contracting a hospital acquired condition like a staph infection as well.In light of the current global health crisis, I strongly believe that equipping our healthcare professionals with AI functionalities will have a huge impact.
What it does
I have developed a model that predicts the length of stay for a patient admitted to the ICU, based on demographic, services, vitals and lab data, that is available within 24 hours of admission. Predicting the patients' length of stay can empower medical professionals to optimize patient care and resource allocation. This eventually leads to better outcomes, further enhancing the credibility of the hospital.
How we built it
The MIMIC-III dataset was extensively explored to understand the various aspects of an ICU stay that can affect the patients length of stay. I performed detailed research using medical reference literatures and speaking to medical experts to understand the data from a medical domain perspective. Further , the data was transformed using python to form our final dataset. Our model was built using Azure AutoML, which allowed us to leverage a powerful cloud-based platform to automate the model training process.
Challenges we ran into
One of the significant challenges I encountered was the need for extensive data pre-processing. I had to clean, transform, and format the MIMIC-III Dataset to make it suitable for training the model. This process required significant effort and time to ensure the data was of high quality.
Accomplishments that we're proud of
This being my first data science project, I am proud of the end-to-end visualization of the project, which showcases the model's capabilities and insights that can be gained from the predictions. Extensive data preprocessing and analysis has helped to hone my python skills as well.
What we learned
Python,Exploratory Data Analysis, Microsoft Azure AutoML,Machine Learning
What's next for J&J PROBLEM STATEMENT - ICU LENGTH OF STAY PREDICTION
Our next steps will be to improve the accuracy and performance of the model by incorporating additional features and refining the training process. We also plan to explore the use of the model for other applications, such as predicting patient outcomes and identifying patients at high risk of adverse events. We believe that our model can be a valuable tool for medical professionals to improve patient care and outcomes. Also, an application can be designed using Power Apps to deploy the model and deliver it to the key stakeholders.
Log in or sign up for Devpost to join the conversation.