The administration of sedation in Intensive Care Units (ICU) is a vital and complex task, essential for the well-being of critically ill patients. Proper dosing is crucial, as both under-dosing and overdosing have adverse consequences for patients, including risks such as accidental extubation, ventilator disconnections, delirium, prolonged ventilation, ICU stays, and even increased mortality.

Monitoring sedation levels is essential to maintain a balance in sedative administration. Clinical scales (RASS/SAS) are currently used as the reference method, but they have limitations. These scales may be inadequate when the patient is unresponsive or when neuromuscular blockers are used. An alternative is electroencephalography (EEG)-based monitors, though they also have limitations in accurately representing the level of hypnosis.

The proposal is to develop an advanced multiparameter system called ROMANTIC, which utilizes machine learning algorithms to monitor sedation status continuously. The aim is for this system to surpass current techniques and provide a more accurate determination of sedation levels. ROMANTIC would incorporate a variety of variables, demographic, pharmacological, hemodynamic, respiratory, and EEG data, to predict sedation status in three categories: under-dosing, appropriate dosing, and overdosing.

With ROMANTIC, clinical staff (users) are expected to be able to determine sedation levels more quickly and accurately, reducing patients' wake-up times and possibly decreasing the incidence of delirium. In the long term, this could result in fewer ICU bed days, less time on mechanical ventilation, cost savings, and reduced complications.

At the end of the project, the goal is to have a prototype model or software that allows non-specialized staff (users) to quickly determine sedation levels in the ICU. This machine learning-based model targets hospitals with ICUs and mechanical ventilation to offer more efficient and cost-effective clinical care. ROMANTIC seeks to innovate in sedation monitoring, providing a more advanced and precise tool for the care of critically ill patients (beneficiaries) in the ICU.

The project is characterized by incorporating multiple clinical, demographic, and pharmacological variables, along with prediction methods based on machine learning, to classify the level of sedation of critically ill patients into three broad categories: underdosage, appropriate dosage, and overdosage. The investigators will compare the performance of this approach against the current gold standard, which is the use of validated clinical sedation scales (RASS/SAS).

The design of this project is divided into two phases. The first phase involves a prospective observational cohort study where demographic, vital signs, ventilatory, pharmacological, and EEG variables will be collected from ICU patients. The second phase involves using the data obtained to create a multiparameter model with machine learning tools that informs ICU staff about the depth of sedation or the degree of hypnosis of each patient. Data will be collected from two national intensive care units: Hospital Clinico Red UC-CHRISTUS (associated entity) and Dr. Sótero del Río Hospital (associated entity). A technical team specialized in data collection and analysis, clinical research, and algorithm development from Pontificia Universidad Católica de Chile (beneficiary entity) will use this information to create a retrospective classifier that categorizes patients into one of three groups: underdosed, appropriately dosed, and overdosed.

Upon admission, comprehensive patient data will be collected, including age, sex, cause of connection to mechanical ventilation, comorbidities, and educational level. The researchers will also record the drugs and doses used for sedation, neuromuscular relaxants, and vasopressor drugs. Sedation will be administered with Fentanyl plus Propofol or Midazolam, according to usual clinical practices at both participating centers. Data from the RASS and SAS clinical sedation rating scales will be collected every hour during two time windows, providing a detailed picture of the patient's sedation levels.

There will be two or three data recording windows per patient. Time zero will be when the patient is intubated and connected to mechanical ventilation. The first collection window will be between 18 to 60 hours after intubation, for a period of 6 hours during business hours, aiming to obtain a record of data once the chosen sedation strategy is established and has presumably stable blood levels of drugs. The second window will start once the patient's sedation is suspended and will last for at least 6 hours or until the patient awakens, defined as SAS 4 and RASS 0, with eyes open to verbal command or if the patient presents agitation and/or delirium according to CAM-ICU. The suspension time will be recorded, along with the day and time all sedatives are stopped in the continuous infusion pump. If for any reason it is necessary to sedate the patient again, a third window similar to the second will be performed.

Demographic (age, previous educational level, SOFA, Charlson, APACHE II), hemodynamic (systolic and diastolic blood pressure, heart rate, electrocardiography), ventilatory (oxygen saturation, respiratory rate, ventilatory mode, PEEP, tidal volume, plateau pressure, peak pressure), electroencephalographic (EEG signal in time domain), pain (hourly behavioral pain scale), neuromuscular blockade (hourly TOF), and pharmacological (total mass of sedative drugs used at the time of suspending infusions and plasma samples of sedative drug concentration at the beginning of both windows) data will be collected. Once patients wake up, the onset of delirium will be assessed with the CAM-ICU tool twice a day (AM/PM) by appropriately trained nursing staff.

Inclusion Criteria:

• Patients admitted to adult critical care units
• Individuals over 18 years old
• Requiring sedation for more than 24 hours
• Patients needing mechanical ventilation for more than 24 hours

Exclusion Criteria:

• Patients with neurological pathology as the cause of mechanical ventilation (including recovered cardiocirculatory arrest, fulminant liver failure, and neurocritical conditions, except subdural hematomas or subarachnoid hemorrhages).
• Pregnancy
• Presence of psychiatric or intellectual disability prior to hospitalization
• Drug dependency
• History of chronic liver damage with hepatic encephalopathy
• Second period of mechanical ventilation during hospitalization
• Early limitation of therapeutic effort
• Patients under 18 years old

Adult patients admitted to intensive care units requiring sedation and mechanical ventilation for more than 24 hours.