Transition of Care (TOC)
DataFest Spring 2023
DataFest Spring 2023
Transition of care can occur in two ways: inter-hospital transfer, which involves the transfer of patients from one hospital to another, and intra-hospital transfer, which involves the transfer of patients from one department or unit to another within the same hospital.
In this project, the focus is on data of inter-hospital transfer of patients who are transferred from another hospital to Keck Medicine of USC hospital. This type of transfer can present unique challenges, including differences in medical records and communication between healthcare providers. Therefore, it is important to ensure a smooth and efficient transition of care for these patients.
Improving the transition of care can have a significant impact on healthcare organizations and patients alike. By reducing costs and promoting more efficient use of resources, healthcare organizations can allocate their resources more effectively. Increased efficiency in healthcare transition can also lead to better patient outcomes and satisfaction.
Poorly coordinated care can lead to medical errors that can have serious consequences for patient health. By improving the transition of care, healthcare organizations can reduce the risk of errors and improve patient safety. Patients can also benefit from improved communication, continuity of care, and clear guidance, which can increase their confidence in their care and build trust in their healthcare providers.
In summary, improving the transition of care is critical to reducing costs, increasing efficiency, improving patient outcomes, and promoting patient satisfaction.
Imputation
On average, each row in the dataset has 10.337 out of 21 columns with null values, and 12 out of the 21 columns have over 50% of the rows with null values.
Given the small size of the original dataset and the significant number of missing values, developing a prediction model poses a particular challenge. The objective of this project is to focus on imputing missing values for the subset of rows in the dataset where the "Request Status" column indicates that the request was "Accepted". This approach is adopted because it is expected that the accepted patients would have more complete information available, and therefore, imputing missing values for this subset would lead to a more accurate imputation result.
Prediction
In this project, a total of eight different models were applied to the dataset. These models can be classified into two categories: classifier and regressor. The classifier models were used to predict the group of length of stay, while the regressor models were employed to predict the number of days in length of stay.
Firstly, patients whose LOS is very long may not have the entire cost of their admission covered by their health insurance. Secondly, a longer LOS can reduce the availability of hospital beds, potentially preventing other patients who could benefit from receiving care at Keck from being admitted. Therefore, it is important to minimize the LOS to ensure optimal use of hospital resources while also providing effective care for all patients who require medical attention.
All the data is de-identified and doesn’t contain any PII(Personally Identifiable Information).
Patient Inter-hospital Transition Request Data
21 columns x 1000 rows
Each row represents a single transition request
The unique identifier is TC_ID which is the transition request ID
Column descriptions chart:
Map navigation to the Keck Hospital of University of Southern California.