Predicting changes in quality of life of ICU survivors

Clinical problem

Annually, over 85,000 patients are admitted to Dutch Intensive Care Units (ICUs), often in life-threatening circumstances. Due to advances in critical care medicine, more patients survive their critical illness. However, the survival of a serious illness does only rarely pass without consequences. It is estimated that 25% to 75% of ICU survivors experience physical problems (e.g. pain, shortness of breath, reduced muscle strength), psychological complaints (e.g. anxiety/depression), cognitive problems (e.g. memory-related) and/or problems related to daily functioning. These issues often negatively influence the quality of life (QoL) and the financial and social situation of former ICU patients.

Where the emphasis of ICU healthcare professionals initially was laid on the prevention of a patient's death, the challenge now also lies in studying what the survival of a serious illness means for patients in the long term, and including these adverse consequences in the decision-making process about treatment in the ICU. At the moment, medical decisions about ICU treatment and predictions concerning the post-ICU period and QoL are often made based on the experience and intuition of physicians.

Solution

In 2016, the MONITOR-IC study (www.monitor-ic.nl) was set up with the aim of including patient-reported outcome measures in clinical decisions. The study should give insight into the long-term outcomes on the QoL of ICU survivors by monitoring them during a five-year follow-up period. At the moment, the study’s prediction model after a one-year follow-up period has shown that an explained variance of 0.52 can be reached by using traditional statistical analysis methods.

Using the patient-reported outcomes from the one-year follow-up of the MONITOR-IC project in addition to data from the Electronic Health Records (EHR), this research project will focus on the development of a prediction model for changes in QoL using machine learning methods. The hypothesis is that including EHR data and using AI techniques will improve upon the current model’s performance.

Tasks

  • Deciding on which parameters in the EHR data to include and doing the pre-processing.
  • Developing and evaluating a prediction model for QoL using machines learning techniques.
  • Conducting analyses of the important factors influencing the QoL.

Innovation

This project is part of the extension of an ongoing study focused on the integration of patient-reported outcomes in the clinical decision-making process. Worldwide, using patient-reported outcomes during ICU treatment is new in Intensive Care Medicine. In 2020, the use of prediction models in daily clinical practice will be pilot tested in the ICU of the Radboudumc and in 2021 this will be extended to the regional hospitals participating in the MONITOR-IC study. The findings of this project will be used to optimize the traditional prediction models that have been/will be tested before and during this project. This project, therefore, participates in supporting ICU employees with data-driven information in order for them to make substantiated decisions and provide better counseling based on these decisions. Patients and family profit from this technological advance by gaining an understanding of possible consequences following ICU admission and treatment.

People

Manon de Jonge

Manon de Jonge

Master student

Artificial Intelligence, Radboud University

Marieke Zegers

Marieke Zegers

Senior researcher

Intensive Care, Radboudumc

Mark van den Boogaard

Mark van den Boogaard

Assistant professor

Intensive Care, Radboudumc

Ruud van Kaam

Ruud van Kaam

PhD student

Intensive Care Medicine, Radboudumc

Luca Ambrogioni

Luca Ambrogioni

Assistant professor

Artificial Intelligence, Radboud University