There is a strong need to increase the efficiency of healthcare workflows, to improve the patient experience and optimize the usage of healthcare resources. One way to achieve this goal is to use Machine Learning techniques to predict how long a certain patient will be admitted in the hospital, and to assess the risk of the patient being readmitted. This project is a collaboration between Cambio Healthcare Systems (one of Scandinavia’s leading suppliers of healthcare information systems and a growing player in the European market with just over 100,000 users across general and university hospitals, specialist units and outpatient units), Capio S:t Görans Hospital and KTH Royal Institute of Technology.
Anonymized data will be provided by Capio S:t Görans Hospital to address the challenging problem of predicting how long a hospitalized patient will stay admitted and/or whether a patient is likely to be readmitted within for example a month after being discharged. A number of variables from the anonymized patient records will be available, including age, priority level, reason for admission, vital parameters, lab info, earlier diagnoses, etc.
The project has the potential of being very important for the Swedish healthcare system, since there is a general lack of hospital staff/beds and healthcare providers are struggling to give all patients sufficient care while keeping down the costs of healthcare resources and staff. If accurate admission duration predictions can be done, hospital resources can be better utilized and actions can be better planned. Suggested discharge date will also serve as a decision support for less experienced staff. Readmission risk is important to assess since a high rate of readmissions might indicate that patients are not given sufficient care or information during their admission.
It is expected that a machine learning approach has strong potential to provide a promising data driven solution, which could be scaled up and tested in real-world healthcare scenarios. It remains to be seen whether the resulting inference model can combine satisfactory prediction accuracy with interpretability to ensure that doctors and nurses understand the basis and feel confident/comfortable about the prediction.
The project focus can be adjusted according to the student’s interest, as well as the number of students applying. The student can choose which of the above questions to focus on, and also explore other questions such as how to deploy machine learning and specific models in a run-time healthcare solution, such as Cambio COSMIC.
The student will be supervised by Cambio Healthcare Systems, who will guide to the Cambio COSMIC data structure, and also help the student to reflect over how the model could eventually be used as part of the existing healthcare information systems.
Workshops will be conducted together with Capio S:t Göran to ensure that the model is built on correct assumptions and solves relevant healthcare problems. One project goal is to be able to evaluate the model performance together with Capio S:t Göran staff.
For questions about the project, please contact marcus.petersson[a]cambio.se.
The selection and interview process is ongoing, so please send your application as soon as possible.
We look forward to receiving your application,
Click here to apply for the job
Vill du veta mer? Fråga mig!
Dela med dig av sidan