Predictive Algorithms for Critical Rehabilitation Outcomes

Overview

An increasing amount of evidence from evidence-based medicine indicates that early rehabilitation intervention for patients receiving mechanical ventilation is safe and feasible, and can promote functional recovery and reduce hospital stay. However, the conscious state, respiratory function, and daily living activities of these patients after being discharged from the ICU vary greatly, and some patients do not show obvious benefits. How to identify which patients may have benefit from early rehabilitation is a key issue that needs to be addressed in critical care rehabilitation. This study aims to investigate the clinical data related to the disease of the ICU survivors who received mechanical ventilation as the research object, by collecting their clinical data when receiving early rehabilitation intervention, and constructing a clinical prediction model for the efficacy of early rehabilitation intervention in the ICU through the selection of optimal regression equation or machine learning algorithm. The application of this model can effectively determine whether ICU inpatients need early rehabilitation intervention, thereby reducing complication rates and improving their quality of life.

An increasing amount of evidence from evidence-based medicine suggests that early rehabilitation intervention (including early active and passive exercises, position management, pulmonary rehabilitation, etc.) for mechanical ventilation patients is safe and feasible, and can promote certain degree of functional recovery and reduce the length of stay in the intensive care unit (ICU). However, the differences in consciousness state, muscle strength, respiratory function, and activity of daily living (ADL) among patients who are discharged from the ICU after condition stabilization are very large, even some patients did not obtain obvious benefits. Therefore, how to identify which patients may have better benefit from early rehabilitation intervention is a key issue that needs to be focused on in ICU. This study used "Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD)" as the guideline. Survivors undergoing mechanical ventilation in the ICU were recruited as the participants, whether patients gained progress in ADL function at different time points after receiving early rehabilitation intervention in the ICU was used as the outcome which is a time-to-event indicator. Demographic data, clinical diagnostic data and disease intervention data of the subjects were collected as alternative predictors. Variable transformation and variable screening were used to find predictors that could predict the outcome. The process of constructing clinical predictive models is completed by fitting models through regression equations and machine learning algorithms, internal validation, external validation, and clinical value assessment. The model with the best prediction efficiency is selected based on the differentiation and calibration of different models after validation. This model will be presented with a nomogram or a web app. The application of this clinical predictive model will identify whether and when this patient can received better recovery on ADL after receiving early rehabilitation intervention, so as to further optimize the timing of early intervention in rehabilitation and improve his survival quality.