Beyond One-Size-Fits-All: Precision Mechanical Ventilation in ARDS

Comprehensive Summary

Acute Respiratory Distress Syndrome (ARDS) is a disease that has a lot of variability to presentation. Traditionally, ARDS is managed through mechanical ventilation and low tidal volume, however, these strategies do not address the underlying causes of inflammation, lung injury, and overall recovery. This is due to the complex nature of the disease involving a broad range of complex biological, anatomical, and mechanical mechanisms. As a result of this, many variables exist in deciphering the best treatment plan available. In this narrative review study, the authors aim to investigate numerous factors such as lung size, driving pressure, and mechanical power among others to look at current treatment plans and determine if heterogeneity can be predicted to make more tailored treatment plans for individual patients. The first step in this is viewing ARDS as a spectrum rather than a disease. This involves many different subphenotypes for ARDS, including mechanical subphenotypes, where the mechanical stress of the lung is measured and categorized based on existing data. Next, there are different biological subphenotypes characterized by inflammatory markers which lead to various outcomes such as shock and mortality. Additionally, there are radiological subphenotypes which identify specific lung patterns, and finally, there are etiological subphenotypes, which cateogrize the root cause of the problem, either the lungs or extrapulmonary. The idea is that the specific disease of the individual lies in the complex web of these subphenotypes, and a correlated treatment plans lies with it. One key treatment of ARDS is a low tidal volume, however, there exists errors in determining the recommended tidal volume through the use of calculators as they rely on factors such as predicted body weight compared to lung function. These limitations call for a shift towards more “energy-based” frameworks, which look at the existing pulmonary function and base it off that. The final goal of this study was to explore the future of the treatment of ARDS, especially by using a potential machine learning model to look for pattern recognition. Through the use of data from pulmonary volumes, flow, and patient effort, the goal would be to have a machine model predict the best treatment. This can prove important in determining when to use positive end-expiratory pressure and recruitment, which when used incorrectly can lead to further lung damage. This machine learning use alongside a clinician can lead to better overall outcomes for ARDS, including increased survival rates and recovery rates. This termed “precision ventilation” can lead to an overall increase in the treatment of ARDS through the use of complex pattern recognition of ARDS that treats it is as a spectrum, not a disease.

Outcomes and Implications

This article explains the possible future of the treatment of Acute Respiratory Distress Syndrome and states that the incorporation of a machine learning model to support a clinician can lead to better overall outcomes of ARDS. The complex heterogeneity of ARDS makes it difficult to come up with one individual treatment, however, with the use of pattern recognition machine learning models, treatment plans can become tailored to the patient. This machine learning model can also be easily incorporated in hospital systems, using existing data to produce a recommended treatment plan. This can also bypass the less accurate current calculators that do not account for current lung function or patient specific problems, leading to better outcomes and survival rates. Additionally, there is the potential to develop another machine learning model that can predict potential patients to are at high risk of developing ARDS based on previous patients and their patterns and data. This risk could then be interpreted by healthcare teams to put the patient on high alert and initiate preventative measures or delay high risk surgeries if the patient has one scheduled. Ultimately, the incorporation of machine learning models could proactively reduce ARDS cases overall, alleviating the burden of ARDS on the patients and hospitals.