From Reactive Ventilarion to Predictive, Personalized Care.
Mechanical Ventilator
What If We Could Predict the Lung? Digital Twins in Mechanical Ventilation
Mechanical ventilation is built around continuous reassessment. Ventilator settings are adjusted, the patient’s response is observed and the next decision is made based on what happens at the bedside. This approach remains central to respiratory care—but as our ability to collect and interpret physiological data grows, a different question is emerging:
What if we could predict how a patient’s lungs might respond before changing the ventilator settings?
This is the idea behind Digital Twins in mechanical ventilation.
A Digital Twin of the Lung
A Digital Twin is a computational representation of a real world system that is informed by data from its physical counterpart. In mechanical ventilation, this concept can be used to create a patient specific computational model that represents aspects of an individual’s respiratory physiology.
Depending on the model, information such as ventilator pressures and flows, respiratory mechanics, blood gases, imaging, and other physiological measurements can be incorporated to represent the patient’s current respiratory state.
The important distinction is that the model is not simply describing what is happening now. Its potential lies in using that information to simulate possible changes and estimate how the patient’s respiratory system could respond.
Model → Simulate → Predict → Personalize
In other words, the goal moves beyond describing the present toward exploring possible next steps.
From Observation to Prediction
Traditionally, the process is familiar:
Adjust → Observe → React
A ventilator setting is changed, the patient’s response is monitored, and the clinical team decides whether another adjustment is needed.
A Digital Twin introduces another layer:
A clinician could potentially explore different ventilation scenarios within a patient-specific model and examine their predicted physiological effects before applying a change to the actual patient.
This concept is particularly relevant because respiratory physiology varies considerably between patients—and can change within the same patient over time.
Why Personalization Matters
Mechanical ventilation is not a one size fits all intervention.
The concept is also being explored across different forms of respiratory support.
Two patients with ARDS, for example, may have very different lung mechanics, regional ventilation patterns, recruitability, and responses to airway pressure. Even within the same patient, these characteristics can evolve as the underlying disease changes.
Recent research has demonstrated the potential of patient specific computational models to capture regional lung behavior. In a 2025 validation study, models generated from individual patient CT data and ventilator waveforms were compared with bedside electrical impedance tomography (EIT). The models showed strong agreement with measured regional ventilation, suggesting that computational models may eventually provide information about lung behavior that is difficult to observe directly at the bedside.
This opens an interesting possibility: rather than asking what setting works on average, could we increasingly ask what is most appropriate for this patient’s physiology?
From Simulation to Clinical Decision Support
The concept has already been tested in clinical research.
The DeVENT randomized clinical trial evaluated a physiological model/digital-twin-based decision support system in mechanically ventilated patients with ARDS. The system provided recommendations intended to support individualized ventilator management. The trial demonstrated the feasibility of evaluating this approach clinically, although it did not show a significant difference in its primary outcome of average driving pressure.
That distinction matters.
Digital Twins are promising, but current evidence does not support presenting them as a replacement for conventional clinical management. Instead, they represent an emerging form of decision support that is still being evaluated.
Beyond Invasive Ventilation
In acute hypoxemic respiratory failure, researchers have created Digital Twins from individual patient data to investigate why some patients succeed or fail on non-invasive ventilation (NIV). The models allowed researchers to examine variables such as respiratory effort, lung stress, strain, compliance, and mechanical power in ways that are difficult to measure comprehensively at the bedside.
More recent work has also explored how Digital Twins could be used to investigate different ventilation strategies and their potential relationship with ventilator-induced lung injury (VILI). One 2025 study used Digital Twins generated from data from 98 patients with ARDS to compare APRV and pressure-controlled ventilation across several VILI-related measures. Importantly, the authors concluded that the findings require further evaluation in randomized clinical trials.
This is where the technology becomes particularly interesting: the Digital Twin can function as a virtual testing environment.
A Virtual Test Before the Real Intervention
Imagine being able to explore several possible ventilator strategies using a patient-specific model:
What might happen if PEEP is increased?
How could regional ventilation change?
Could lung stress or mechanical power increase?
Would the predicted response justify the change?
The answers would still require clinical interpretation. But instead of relying entirely on trial and response at the bedside, clinicians could potentially have another source of patient specific information before making the intervention.
Research into virtual patient models has already demonstrated the ability to predict certain mechanical responses to changes in PEEP, including the potential to estimate lung distension before changing ventilator settings.
The Technology Has Limits
Despite its potential, a Digital Twin is still a model not the patient itself.
Its predictions depend on the quality of the data used to build and update it, the assumptions incorporated into the model, and how accurately the model represents the patient’s physiology.
Patient physiology is dynamic. Disease progression, spontaneous breathing, sedation, positioning, fluid status, and many other factors can change respiratory mechanics over time.
For that reason, important challenges remain around model accuracy, uncertainty, validation, data quality, interoperability, and integration into clinical workflows.
And perhaps the most important principle is simple:
A prediction is not a certainty.
A Digital Twin may help clinicians explore what could happen, but it cannot guarantee what will happen.
What Could the Future Look Like?
The long-term vision is not necessarily a ventilator that makes decisions on its own.
It is a more connected system in which patient data continuously updates a physiological model, the model evaluates potential interventions, and the clinical team uses those predictions alongside bedside assessment and clinical judgment.
In that future, mechanical ventilation could become increasingly individualized not because one algorithm has found a universal “best” setting, but because technology can help us understand how this patient’s lungs may respond.
The shift is subtle, but important:
From reacting to physiological changes
to anticipating possible responses.
From treating the ventilated patient as a set of numbers
to building a dynamic representation of their individual physiology.
Digital Twins are not yet the standard of care. But they offer a glimpse of where mechanical ventilation may be heading: toward a future where we can test, predict, and personalize before we intervene.
References
1. Patel B, et al. A randomized control trial evaluating the advice of a physiological-model/digital twin-based decision support system on mechanical ventilation in patients with acute respiratory distress syndrome. Frontiers in Medicine. 2024;11:1473629.
2. Rixner M, et al. Patient-specific prediction of regional lung mechanics in patients with ARDS with physics-based models: a validation study. Journal of Applied Physiology. 2025;139(4):1029–1049.
3. Weaver L, et al. Digital Twins of Acute Hypoxemic Respiratory Failure Patients Suggest a Mechanistic Basis for Success and Failure of Noninvasive Ventilation. Critical Care Medicine. 2024.
4. Joy W, et al. Digital Twins to Evaluate the Risk of Ventilator-Induced Lung Injury During Airway Pressure Release Ventilation Compared With Pressure-Controlled Ventilation. Critical Care Medicine. 2025;53(12):e2573–e2582.
5. Pulmonary response prediction through personalized basis functions in a virtual patient model. Journal of Critical Care / PubMed-indexed research. 2024.
The Ventilator–Sedation Connection
Critical Care
ICU delirium is one of the most common complications in critically ill patients. It is an acute disturbance in attention and awareness that can fluctuate throughout the day. It is particularly common in patients receiving mechanical ventilation and is associated with longer ICU stays, prolonged mechanical ventilation, and poorer outcomes.
For Respiratory Therapists, delirium is especially relevant because mechanical ventilation, sedation, and delirium are closely connected.
Delirium is usually multifactorial. Common contributors include critical illness, inflammation, hypoxemia, metabolic disturbances, sedative medications, pain, sleep disruption, immobility, infection, and the ICU environment.
It can be hyperactive, with agitation and restlessness, or hypoactive, where the patient appears unusually quiet, sleepy, or withdrawn. Hypoactive delirium can easily be missed.
Regular assessment with validated tools such as CAM-ICU, together with sedation assessment using RASS, is therefore important.
One of the most interesting aspects of ICU delirium is its relationship with mechanical ventilation.
Respiratory failure → Intubation → Sedation → Immobility & sleep disruption → Delirium → Agitation/ventilator asynchrony → More sedation → Prolonged ventilation
This cycle can delay recovery and make ventilator liberation more difficult.
For RTs, this means that delirium should not be viewed as a completely separate problem. A patient’s mental status can directly affect their ability to participate in an SBT, follow commands, protect their airway, and progress toward extubation.
RTs play an important role in breaking this cycle.
1. Sedation & Ventilator Liberation
When clinically appropriate, Spontaneous Awakening Trials (SATs) and Spontaneous Breathing Trials (SBTs) should be coordinated as part of the patient’s liberation strategy.
The goal is not simply to keep the patient comfortable and synchronized with the ventilator. It is to use the lightest effective sedation and assess readiness for liberation as early as possible.
The 2025 PADIS focused update suggests dexmedetomidine over propofol when light sedation and/or reduction of delirium are high priorities in mechanically ventilated adults.
2. Recognizing Changes in Mental Status
During respiratory care, a sudden change in behavior should not automatically be considered “agitation” or “non-compliance.”
Ask:
Early recognition and communication with the ICU team can prevent unnecessary escalation of sedation.
3. Supporting the ABCDEF Approach
Delirium management is part of the ICU Liberation Bundle (ABCDEF):
A – Assess, prevent, and manage pain
B – Both SATs and SBTs
C – Choice of analgesia and sedation
D – Delirium: assess, prevent, and manage
E – Early mobility and exercise
F – Family engagement
For RTs, B is particularly relevant, but respiratory care also contributes to sedation management, delirium recognition, and overall ventilator liberation.
The focus is increasingly shifting from simply treating delirium after it develops toward prevention and early prediction.
Recent research has explored prediction models for sedative-associated delirium in mechanically ventilated patients, aiming to identify high-risk patients earlier.
There is also growing interest in individualized sedation strategies and non-pharmacological interventions such as early mobility and sleep-focused care.
Importantly, the latest PADIS update does not support routine use of antipsychotics simply to shorten delirium duration. This reinforces the importance of addressing modifiable risk factors rather than relying on medication alone.
As RTs, recognizing delirium, supporting light sedation strategies, participating in SAT/SBT, and promoting early ventilator liberation can all contribute to breaking the cycle.
The goal is not just to get the patient off the ventilator — it is to get them off the ventilator awake, oriented, and ready to recover.