Can artificial intelligence using machine algorithm help in preventing fall in blood pressure in seriously ill patients
- Registration Number
- CTRI/2023/06/053973
- Brief Summary
Study will be conducted after the approval from the institutional ethical committee and will be registered with clinical trials registry of India (CTRI**)****.** Patients being managed with advanced haemodynamic monitoring and other inclusion criteria will be recruited for the study. A written and informed voluntary consent for inclusion in the study will be obtained from the patient or next of kin, after carefully explaining the procedure, its benefits and risks in vernacular. A detailed history and physical examination will be carried out.
Patient characteristics including age, gender, body weight and height will be noted. A note shall also be made of the diagnosis and vasoactive drugs being administered if any. The need for advanced haemodynamic monitoring by FloTrac sensor (without machine learning algorithm- Group I) or HPI (Acumen IQ sensor; machine learning algorithm- Group II) shall be decided by the consultant in charge of ICU, and only the events of hypotensive episodes, haemodynamic parameters and treatment instituted will be noted without any active intervention as part of the study.
The sensor is connected to the radial artery catheter of all the patients and placed at the level of right atrium before pressure equilibration (zeroing). The adequacy of the dynamic response of the pressure sensor is tested with the fast-flush test. In case of overdamping, the arterial line is tested for clots by flushing the system, and in case of intractable overdamping, a new line is placed to the other side as standard protocol. The sensor measures arterial blood pressure and derives advanced hemodynamic parameters from the arterial waveform every 20secs, advanced haemodynamic parameters such as SVV, PPV, Eadyn, CO, CI, SV, SVI, SVR and SVRI is available with FloTrac sensor and HPI with Acumen IQ sensor and displayed on the HemoSphere monitor.
Patients connected to FloTrac sensor receive treatment once hypotension (MAP<65 mmHg) has occurred and it is treated on the basis of the decision of the treating clinician based on the haemodynamic parameters.
In patients monitored by Acumen IQ sensor, if HPI is ≥85, despite MAP ≥65 mmHg, treatment to prevent hypotension is performed at the discretion of the treating clinician based on parameters of the HemoSphere monitor (SVV, PPV, Eadyn, CI, SVI or SVRI) viz, fluids, vasopressors, inotropes or a combination of these if required.
A note shall be made of the treatment instituted by the clinician in response to hypotension in the control group or HPI ≥85. Treatment performed shall be entered in both the groups. Furthermore, at the time of a hypotensive event, other parameters like PPV, SVV, Eadyn, SV, SVI, CO, CI, SVR and SVRI shall also be noted. An increase in PPV, SVV and Eadyn indicates hypovolemia and a need of fluids, a decrease in SVR and SVRI denotes decrease in resistance and hence need for vasopressors whereas, a decrease in SV, CO or CI indicates decreased cardiac contractility and need for inotropes or a combination of these, as a possible aetiology and management of hypotension.
The haemodynamic measurements will be recorded in the monitor for maximum of 48 hours after recruitment in the study or till the HemoSphere monitor is disconnected (whichever is earlier). Data will then be downloaded and number of hypotensive episodes along with its duration and time to MAP <65 mmHg after HPI ≥ 85 will be noted in both the groups. Morbidity will be defined in terms of end organ damage: AKI stage I or worsening as defined by KDIGO criteria (Annexure Ⅳ) within the first three days after the first episode of hypotensionor an increase in plasma neutrophil gelatinase-associated lipocalin (pNGAL) levels of ≥ 100 (ng/ml) and myocardial dysfunction indicated by an increase in Troponin T levels of ≥ 0.003 (ng/ml) obtained every day for three days shall also be noted. Hypotension shall be defined as mean arterial pressure (MAP) of less than 65 mmHg.
Similar to the analysis by Wijenberge et al, to assess the performance of the algorithm a forward analysis, with a HPI threshold value of 85 will be used. If, within a 1 min window, the HPI is ≥85, we will consider this as an alarm. From the start of this alarm, hypotension (<65 mmHg) will be checked for over the next 20 min window of MAP. That means that any HPI value is ≥85 for more than 1 min and a MAP <65 mmHg will be considered a positive prediction. The time between the alarm and the onset of hypotension will be noted. If the HPI does not rise above 85, we will consider this as negative prediction. A true positive, false positive, true negative or false negative prediction will be counted as every 20 min timeframe.
To make sure that each alarm and each hypotensive event will be counted only once in the analysis, the window will be shifted forward 20 min in time following a true positive, false positive or a false negative detection. The data with an increase in MAP of ≥ 5 mmHg within 20 s or an increase in MAP of ≥ 8 mmHg within 2 min, starting from a baseline MAP < 70 mmHg will most likely be due to clinical treatment, hence such measurements will not be taken.
If the hypotension persists for more than 20 minutes despite fluid resuscitation and increase in vasopressors (noradrenaline >20µg/min or two vasopressors), it would be labelled as refractory and further HPI values shall not be noted till the MAP increases to >65 mmHg. However, it would be included for the final analysis.
- Detailed Description
Not available
Recruitment & Eligibility
- Status
- Not Yet Recruiting
- Sex
- All
- Target Recruitment
- 54
1.More than 18years old 2.Critically ill patients admitted to intensive care with an arterial catheter in the radial artery for advanced hemodynamic monitoring.
1.Patients having sustained hypotension despite receiving maximum drug support 2.Patients admitted to intensive care with hypertensive emergencies.
Study & Design
- Study Type
- Observational
- Study Design
- Not specified
- Primary Outcome Measures
Name Time Method Comparison between median (IQR) number of hypotensive events in critically ill patients being managed with or without machine learning algorithm 48 hours
- Secondary Outcome Measures
Name Time Method Sensitivity, specificity, positive predictive value, negative predictive value of HPI & HPI threshold predicting hypotension using AUROC 48 hours Mean/ median time to MAP less than 65 mmHg after HPI ≥ 85 48 hours Comparison between groups of average duration (min) of each hypotensive event 48 hours Comparison between groups of percentage agreement between decision taken by the clinician to that suggested by the machine learning algorithm 48 hours Comparison between groups of proportion of patients with morbidity within 3 days of first event of hypotension (number of patients with acute kidney injury, increased troponin levels) 72 hours
Trial Locations
- Locations (1)
Lady Hardinge Medical College and Associated Hospitals
🇮🇳Central, DELHI, India
Lady Hardinge Medical College and Associated Hospitals🇮🇳Central, DELHI, IndiaDr Yugansh GuptaPrincipal investigator8860697084yug.sagittarius@gmail.com