Make my haemodynamic monitor GREEN: sustainable monitoring solutions.

Anaesthesiologists overwhelmingly favour pulse wave analysis techniques as their primary method to monitor cardiac output during high-risk noncardiac surgery. In patients with a radial arterial catheter in place, pulse wave analysis techniques have the advantage of instantly providing non-operator-dependent and continuous haemodynamic monitoring information. Green pulse wave analysis techniques working with any standard pressure transducer are as reliable as techniques requiring dedicated pressure transducers. They have the advantage of minimising plastic waste and related carbon dioxide emissions, and also significantly reducing hospital costs. The future integration of pulse wave analysis algorithms into multivariable bedside monitors, obviating the need for standalone haemodynamic monitors, could lead to wider use of haemodynamic monitoring solutions by further reducing their cost and carbon footprint.

The microcirculation in perioperative medicine: a narrative review.

The microcirculation describes the network of the smallest vessels in our cardiovascular system. On a microcirculatory level, oxygen delivery is determined by the flow of oxygen-carrying red blood cells in a given single capillary (capillary red blood cell flow) and the density of the capillary network in a given tissue volume (capillary vessel density). Handheld vital videomicroscopy enables visualisation of the capillary bed on the surface of organs and tissues but currently is only used for research. Measurements are generally possible on all organ surfaces but are most often performed in the sublingual area. In patients presenting for elective surgery, the sublingual microcirculation is usually intact and functional. Induction of general anaesthesia slightly decreases capillary red blood cell flow and increases capillary vessel density. During elective, even major, noncardiac surgery, the sublingual microcirculation is preserved and remains functional, presumably because elective noncardiac surgery is scheduled trauma and haemodynamic alterations are immediately treated by anaesthesiologists, usually restoring the macrocirculation before the microcirculation is substantially impaired. Additionally, surgery is regional trauma and thus likely causes regional, rather than systemic, impairment of the microcirculation. Whether or not the sublingual microcirculation is impaired after noncardiac surgery remains a subject of ongoing research. Similarly, it remains unclear if cardiac surgery, especially with cardiopulmonary bypass, impairs the sublingual microcirculation. The effects of therapeutic interventions specifically targeting the microcirculation remain to be elucidated and tested. Future research should focus on further improving microcirculation monitoring methods and investigating how regional microcirculation monitoring can inform clinical decision-making and treatment.

PeriOperative Quality Initiative (POQI) international consensus statement on perioperative arterial pressure management.

Arterial pressure monitoring and management are mainstays of haemodynamic therapy in patients having surgery. This article presents updated consensus statements and recommendations on perioperative arterial pressure management developed during the 11th POQI PeriOperative Quality Initiative (POQI) consensus conference held in London, UK, on June 4-6, 2023, which included a diverse group of international experts. Based on a modified Delphi approach, we recommend keeping intraoperative mean arterial pressure ≥60 mm Hg in at-risk patients. We further recommend increasing mean arterial pressure targets when venous or compartment pressures are elevated and treating hypotension based on presumed underlying causes. When intraoperative hypertension is treated, we recommend doing so carefully to avoid hypotension. Clinicians should consider continuous intraoperative arterial pressure monitoring as it can help reduce the severity and duration of hypotension compared to intermittent arterial pressure monitoring. Postoperative hypotension is often unrecognised and might be more important than intraoperative hypotension because it is often prolonged and untreated. Future research should focus on identifying patient-specific and organ-specific hypotension harm thresholds and optimal treatment strategies for intraoperative hypotension including choice of vasopressors. Research is also needed to guide monitoring and management strategies for recognising, preventing, and treating postoperative hypotension.

Cardiac index-guided therapy to maintain optimised postinduction cardiac index in high-risk patients having major open abdominal surgery: the multicentre randomised iPEGASUS trial.

BACKGROUND: It is unclear whether optimising intraoperative cardiac index can reduce postoperative complications. We tested the hypothesis that maintaining optimised postinduction cardiac index during and for the first 8 h after surgery reduces the incidence of a composite outcome of complications within 28 days after surgery compared with routine care in high-risk patients having elective major open abdominal surgery. METHODS: In three German and two Spanish centres, high-risk patients having elective major open abdominal surgery were randomised to cardiac index-guided therapy to maintain optimised postinduction cardiac index (cardiac index at which pulse pressure variation was <12%) during and for the first 8 h after surgery using intravenous fluids and dobutamine or to routine care. The primary outcome was the incidence of a composite outcome of moderate or severe complications within 28 days after surgery. RESULTS: We analysed 318 of 380 enrolled subjects. The composite primary outcome occurred in 84 of 152 subjects (55%) assigned to cardiac index-guided therapy and in 77 of 166 subjects (46%) assigned to routine care (odds ratio: 1.87, 95% confidence interval: 1.03-3.39, P=0.038). Per-protocol analyses confirmed the results of the primary outcome analysis. CONCLUSIONS: Maintaining optimised postinduction cardiac index during and for the first 8 h after surgery did not reduce, and possibly increased, the incidence of a composite outcome of complications within 28 days after surgery compared with routine care in high-risk patients having elective major open abdominal surgery. Clinicians should not strive to maintain optimised postinduction cardiac index during and after surgery in expectation of reducing complications. CLINICAL TRIAL REGISTRATION: NCT03021525.

The Urethral Perfusion Index During Off-Pump Coronary Artery Bypass Surgery: An Observational Study.

OBJECTIVES: The IKORUS system (Vygon, Écouen, France) allows continuous monitoring of the urethral perfusion index (uPI) using a photoplethysmographic sensor mounted near the base of the balloon of a dedicated urinary catheter. We aimed to test the hypothesis that the uPI decreases during off-pump coronary artery bypass (OPCAB) surgery and to investigate the relationship between the uPI and macrocirculatory variables. DESIGN: Prospective observational study. SETTING: University Medical Center Hamburg-Eppendorf, Hamburg, Germany. PARTICIPANTS: Twenty patients having OPCAB surgery. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: The primary endpoint was changes in the uPI during OPCAB surgery. We additionally investigated associations between the uPI and cardiac output, mean arterial pressure, heart rate, and point-of-care variables. Twenty patients with 24,137 uPI measurements were included. Overall, there was a high interindividual variability in the uPI. Compared with the preparation phase (during which the median [interquartile range] uPI was 7.7 [5.6-12.0]), the uPI decreased by 14% (95% CI 13%-15%) during the bypass grafting phase, by 35% (95% CI 34%-36%) during the cardiac positioning phase, and by 7% (95% CI 6%-9%) during hemostasis. There was no clinically important association between uPI and either cardiac output, mean arterial pressure, or heart rate. CONCLUSIONS: The uPI decreases during OPCAB surgery, specifically during the cardiac positioning phase. There was no clinically important association between uPI and either cardiac output, mean arterial pressure, or heart rate. It, therefore, remains to be determined whether intraoperative uPI decreases are clinically important, reflect alterations in intra-abdominal tissue perfusion that are not reflected by systemic macrohemodynamics, and can help clinicians guide therapeutic interventions.

Haemodynamic monitoring during noncardiac surgery: past, present, and future.

During surgery, various haemodynamic variables are monitored and optimised to maintain organ perfusion pressure and oxygen delivery - and to eventually improve outcomes. Important haemodynamic variables that provide an understanding of most pathophysiologic haemodynamic conditions during surgery include heart rate, arterial pressure, central venous pressure, pulse pressure variation/stroke volume variation, stroke volume, and cardiac output. A basic physiologic and pathophysiologic understanding of these haemodynamic variables and the corresponding monitoring methods is essential. We therefore revisit the pathophysiologic rationale for intraoperative monitoring of haemodynamic variables, describe the history, current use, and future technological developments of monitoring methods, and finally briefly summarise the evidence that haemodynamic management can improve patient-centred outcomes.

Post-anesthesia care unit hypotension in low-risk patients recovering from non-cardiac surgery: a prospective observational study.

Intraoperative hypotension is common and associated with organ injury. Hypotension can not only occur during surgery, but also thereafter. After surgery, most patients are treated in post-anesthesia care units (PACU). The incidence of PACU hypotension is largely unknown - presumably in part because arterial pressure is usually monitored intermittently in PACU patients. We therefore aimed to evaluate the incidence, duration, and severity of PACU hypotension in low-risk patients recovering from non-cardiac surgery. In this observational study, we performed blinded continuous non-invasive arterial pressure monitoring with finger-cuffs (ClearSight system; Edwards Lifesciences, Irvine, CA, USA) in 100 patients recovering from non-cardiac surgery in the PACU. We defined PACU hypotension as a mean arterial pressure (MAP) < 65 mmHg. Patients had continuous finger-cuff monitoring for a median (25th percentile, 75th percentile) of 64 (44 to 91) minutes. Only three patients (3%) had PACU hypotension for at least one consecutive minute. These three patients had 4, 4, and 2 cumulative minutes of PACU hypotension; areas under a MAP of 65 mmHg of 17, 9, and 9 mmHg x minute; and time-weighted averages MAP less than 65 mmHg of 0.5, 0.3, and 0.2 mmHg. The median volume of crystalloid fluid patients were given during PACU treatment was 200 (100 to 400) ml. None was given colloids or a vasopressor during PACU treatment. In low-risk patients recovering from non-cardiac surgery, the incidence of PACU hypotension was very low and the few episodes of PACU hypotension were short and of modest severity.

Intraoperative haemodynamic monitoring and management of adults having non-cardiac surgery: Guidelines of the German Society of Anaesthesiology and Intensive Care Medicine in collaboration with the German Association of the Scientific Medical Societies.

Haemodynamic monitoring and management are cornerstones of perioperative care. The goal of haemodynamic management is to maintain organ function by ensuring adequate perfusion pressure, blood flow, and oxygen delivery. We here present guidelines on "Intraoperative haemodynamic monitoring and management of adults having non-cardiac surgery" that were prepared by 18 experts on behalf of the German Society of Anaesthesiology and Intensive Care Medicine (Deutsche Gesellschaft für Anästhesiologie und lntensivmedizin; DGAI).

Continuous Finger-cuff versus Intermittent Oscillometric Arterial Pressure Monitoring and Hypotension during Induction of Anesthesia and Noncardiac Surgery: The DETECT Randomized Trial.

BACKGROUND: Finger-cuff methods allow noninvasive continuous arterial pressure monitoring. This study aimed to determine whether continuous finger-cuff arterial pressure monitoring helps clinicians reduce hypotension within 15 min after starting induction of anesthesia and during noncardiac surgery. Specifically, this study tested the hypotheses that continuous finger-cuff-compared to intermittent oscillometric-arterial pressure monitoring helps clinicians reduce the area under a mean arterial pressure of 65 mmHg within 15 min after starting induction of anesthesia and the time-weighted average mean arterial pressure less than 65 mmHg during noncardiac surgery. METHODS: In this single-center trial, 242 noncardiac surgery patients were randomized to unblinded continuous finger-cuff arterial pressure monitoring or to intermittent oscillometric arterial pressure monitoring (with blinded continuous finger-cuff arterial pressure monitoring). The first of two hierarchical primary endpoints was the area under a mean arterial pressure of 65 mmHg within 15 min after starting induction of anesthesia; the second primary endpoint was the time-weighted average mean arterial pressure less than 65 mmHg during surgery. RESULTS: Within 15 min after starting induction of anesthesia, the median (interquartile range) area under a mean arterial pressure of 65 mmHg was 7 (0, 24) mmHg × min in 109 patients assigned to continuous finger-cuff monitoring versus 19 (0.3, 60) mmHg × min in 113 patients assigned to intermittent oscillometric monitoring (P = 0.004; estimated location shift: -6 [95% CI: -15 to -0.3] mmHg × min). During surgery, the median (interquartile range) time-weighted average mean arterial pressure less than 65 mmHg was 0.04 (0, 0.27) mmHg in 112 patients assigned to continuous finger-cuff monitoring and 0.40 (0.03, 1.74) mmHg in 115 patients assigned to intermittent oscillometric monitoring (P < 0.001; estimated location shift: -0.17 [95% CI: -0.41 to -0.05] mmHg). CONCLUSIONS: Continuous finger-cuff arterial pressure monitoring helps clinicians reduce hypotension within 15 min after starting induction of anesthesia and during noncardiac surgery compared to intermittent oscillometric arterial pressure monitoring.

Pulse wave analysis: basic concepts and clinical application in intensive care medicine.

PURPOSE OF REVIEW: The measurement of cardiac output ( CO ) is important in patients with circulatory shock. Pulse wave analysis (PWA) estimates CO continuously and in real-time using the mathematical analysis of the arterial pressure waveform. We describe different PWA methods and provide a framework for CO monitoring using PWA in critically ill patients. RECENT FINDINGS: PWA monitoring systems can be classified according to their invasiveness (into invasive, minimally invasive, and noninvasive systems) and their calibration method (into externally calibrated, internally calibrated, and uncalibrated systems). PWA requires optimal arterial pressure waveform signals. Marked alterations and rapid changes in systemic vascular resistance and vasomotor tone can impair the measurement performance of PWA. SUMMARY: Noninvasive PWA methods are generally not recommended in critically ill patients (who have arterial catheters anyway). PWA systems can be used to continuously track stroke volume and CO in real-time during tests of fluid responsiveness or during therapeutic interventions. During fluid challenges, continuous CO monitoring is important because - if CO decreases - a fluid challenge can be stopped early to avoid further unnecessary fluid administration. PWA externally calibrated to indicator dilution methods can be used - in addition to echocardiography - to diagnose the type of shock.

A classification system for pulmonary artery catheters.

Flow-directed, balloon-tipped pulmonary artery catheters allow measuring cardiac output and other haemodynamic variables including intracardiac pressures. We propose classifying pulmonary artery catheters by generations and specifying additional measurement modalities. Based on the method used to measure cardiac output, pulmonary artery catheters can be classified into three generations: first-generation using intermittent pulmonary artery thermodilution; second-generation using a thermal filament for automated pulmonary artery thermodilution; and third-generation combining thermal filament-based automated pulmonary artery thermodilution and pulmonary artery pulse wave analysis. Each of these pulmonary artery catheter generations can include additional measurements, such as continuous mixed venous oxygen saturation, right ventricular ejection fraction and end-diastolic volume, and right ventricular pressure. This classification should help define indications for pulmonary artery catheters in clinical practice and research.

Goal-directed haemodynamic therapy: an imprecise umbrella term to avoid.

'Goal-directed haemodynamic therapy' describes various haemodynamic treatment strategies that have in common that interventions are titrated to achieve predefined haemodynamic targets. However, the treatment strategies differ substantially regarding the underlying haemodynamic target variables and target values, and thus presumably have different effects on outcome. It is an over-simplifying approach to lump complex and substantially differing haemodynamic treatment strategies together under the term 'goal-directed haemodynamic therapy', an imprecise umbrella term that we should thus stop using.

Perioperative troponin surveillance in major noncardiac surgery: a narrative review.

Myocardial injury is now an acknowledged complication in patients undergoing noncardiac surgery. Heterogeneity in the definitions of myocardial injury contributes to difficulty in evaluating the value of cardiac troponins (cTns) measurement in perioperative care. Pre-, post-, and peri-operatively increased cTns are encompassed by the umbrella term 'myocardial injury' and are likely to reflect different pathophysiological mechanisms. Increased cTns are independently associated with cardiovascular and non-cardiovascular complications, poor short-term and long-term cardiovascular outcomes, and increased mortality. Preoperative measurement of cTns aids preoperative risk stratification beyond the Revised Cardiac Risk Index. Systematic measurement detects acute perioperative increases and allows early identification of acute myocardial injury. Common definitions and standards for reporting are a prerequisite for designing impactful future trials and perioperative management strategies.

Intraoperative arterial pressure management: knowns and unknowns.

Preventing postoperative organ dysfunction is integral to the practice of anaesthesia. Although intraoperative hypotension is associated with postoperative end organ dysfunction, there remains ambiguity with regards to its definition, targets, thresholds for initiating treatment, and ideal treatment modalities.

Endotypes of intraoperative hypotension during major abdominal surgery: a retrospective machine learning analysis of an observational cohort study.

BACKGROUND: Intraoperative hypotension is associated with myocardial injury, acute kidney injury, and death. In routine practice, specific causes of intraoperative hypotension are often unclear. A more detailed understanding of underlying haemodynamic alterations of intraoperative hypotension may identify specific treatments. We thus aimed to use machine learning - specifically, hierarchical clustering - to identify underlying haemodynamic alterations causing intraoperative hypotension in major abdominal surgery patients. Specifically, we tested the hypothesis that there are distinct endotypes of intraoperative hypotension, which may help refine therapeutic interventions. METHODS: We conducted a secondary analysis of intraoperative haemodynamic measurements from a prospective observational study in 100 patients who had major abdominal surgery under general anaesthesia. We used stroke volume index, heart rate, cardiac index, systemic vascular resistance index, and pulse pressure variation measurements. Intraoperative hypotension was defined as any mean arterial pressure ≤65 mm Hg or a mean arterial pressure between 66 and 75 mm Hg requiring a norepinephrine infusion rate exceeding 0.1 µg kg-1 min-1. To identify endotypes of intraoperative hypotension, we used hierarchical clustering (Ward's method). RESULTS: A total of 615 episodes of intraoperative hypotension occurred in 82 patients (46 [56%] female; median age: 64 [57, 73] yr) who had surgery of a median duration of 270 (195, 335) min. Hierarchical clustering revealed six distinct intraoperative hypotension endotypes. Based on their clinical characteristics, we labelled these endotypes as (1) myocardial depression, (2) bradycardia, (3) vasodilation with cardiac index increase, (4) vasodilation without cardiac index increase, (5) hypovolaemia, and (6) mixed type. CONCLUSION: Hierarchical clustering identified six endotypes of intraoperative hypotension. If validated, considering these intraoperative hypotension endotypes may enable causal treatment of intraoperative hypotension.

Energy Expenditure Under General Anesthesia: An Observational Study Using Indirect Calorimetry in Patients Having Noncardiac Surgery.

BACKGROUND: Perioperative hemodynamic management aims to optimize organ perfusion pressure and blood flow-assuming this ensures that oxygen delivery meets cellular metabolic needs. Cellular metabolic needs are reflected by energy expenditure. A better understanding of energy expenditure under general anesthesia could help tailor perioperative hemodynamic management to actual demands. We thus sought to assess energy expenditure under general anesthesia. Our primary hypothesis was that energy expenditure under general anesthesia is lower than preoperative awake resting energy expenditure. METHODS: We conducted an observational study on patients having elective noncardiac surgery at the University Medical Center Hamburg-Eppendorf (Germany) between September 2019 and March 2020. We assessed preoperative awake resting energy expenditure, energy expenditure under general anesthesia, and energy expenditure after surgery using indirect calorimetry. We compared energy expenditure under general anesthesia at incision to preoperative awake resting energy expenditure using a Wilcoxon signed-rank test for paired measurements. RESULTS: We analyzed 60 patients. Median (95% confidence interval [CI]) preoperative awake resting energy expenditure was 953 (95% CI, 906-962) kcal d -1 m -2 . Median energy expenditure under general anesthesia was 680 (95% CI, 642-711) kcal d -1 m -2 -and thus 263 (95% CI, 223-307) kcal d -1 m -2 or 27% (95% CI, 23%-30%) lower than preoperative awake resting energy expenditure ( P < .001). CONCLUSIONS: Median energy expenditure under general anesthesia is about one-quarter lower than preoperative awake resting energy expenditure in patients having noncardiac surgery.

Agreement between cardiac output measurements by pulse wave analysis using the Pressure Recording Analytical Method and transthoracic echocardiography in patients with veno-venous extracorporeal membrane oxygenation therapy: An observational method comparison.

BACKGROUND: Measuring cardiac output (CO) is important in patients treated with veno-venous extracorporeal membrane oxygenation (vvECMO) because vvECMO flow and CO need to be balanced. Uncalibrated pulse wave analysis with the Pressure Recording Analytical Method (PRAM) may be suitable to measure CO in patients with vvECMO therapy. OBJECTIVE: To assess the agreement between CO measured by PRAM (PRAM-CO; test method) and CO measured by transthoracic echocardiography (TTE-CO; reference method). DESIGN: A prospective observational method comparison study. SETTING: The ICU of a German university hospital between March and December 2021. PATIENTS: Thirty one adult patients with respiratory failure requiring vvECMO therapy: 29 of the 31 patients (94%) were treated for COVID-19 related respiratory failure. MAIN OUTCOME MEASURES: PRAM-CO and TTE-CO were measured simultaneously at two time points in each patient with at least 20 min between measurements. A radial or femoral arterial catheter-derived blood pressure waveform was used for PRAM-CO measurements. TTE-CO measurements were conducted using the pulsed wave Doppler-derived velocity time integral of the left ventricular outflow tract (LVOT) and the corresponding LVOT diameter. PRAM-CO and TTE-CO were compared using Bland-Altman analysis and the percentage error (PE). We defined a PE of <30% as clinically acceptable. RESULTS: Mean ±â€ŠSD PRAM-CO was 6.86 ±â€Š1.49 l min -1 and mean TTE-CO was 6.94 ±â€Š1.58 l min -1 . The mean of the differences between PRAM-CO and TTE-CO was 0.09 ±â€Š0.73 l min -1 with a lower 95% limit of agreement of -1.34 l min -1 and an upper 95% limit of agreement of 1.51 l min -1 . The PE was 21%. CONCLUSIONS: The agreement between PRAM-CO and TTE-CO is clinically acceptable in adult patients with vvECMO therapy.

Continuous Blood Pressure Monitoring in Patients Having Surgery: A Narrative Review.

Hypotension can occur before, during, and after surgery and is associated with postoperative complications. Anesthesiologists should thus avoid profound and prolonged hypotension. A crucial part of avoiding hypotension is accurate and tight blood pressure monitoring. In this narrative review, we briefly describe methods for continuous blood pressure monitoring, discuss current evidence for continuous blood pressure monitoring in patients having surgery to reduce perioperative hypotension, and expand on future directions and innovations in this field. In summary, continuous blood pressure monitoring with arterial catheters or noninvasive sensors enables clinicians to detect and treat hypotension immediately. Furthermore, advanced hemodynamic monitoring technologies and artificial intelligence-in combination with continuous blood pressure monitoring-may help clinicians identify underlying causes of hypotension or even predict hypotension before it occurs.

Intraoperative Hypotension and Acute Kidney Injury, Stroke, and Mortality during and outside Cardiopulmonary Bypass: A Retrospective Observational Cohort Study.

BACKGROUND: In cardiac surgery, the association between hypotension during specific intraoperative phases or vasopressor-inotropes with adverse outcomes remains unclear. This study's hypothesis was that intraoperative hypotension duration throughout the surgery or when separated into hypotension during and outside cardiopulmonary bypass may be associated with postoperative major adverse events. METHODS: This retrospective observational cohort study included data for adults who had cardiac surgery between 2008 and 2016 in a tertiary hospital. Intraoperative hypotension was defined as mean arterial pressure of less than 65 mmHg. The total duration of hypotension was divided into three categories based on the fraction of overall hypotension duration that occurred during cardiopulmonary bypass (more than 80%, 80 to 60%, and less than 60%). The primary outcome was a composite of stroke, acute kidney injury, or mortality during the index hospitalization. The association with the composite outcome was evaluated for duration of hypotension during the entire surgery, outside cardiopulmonary bypass, and during cardiopulmonary bypass and the fraction of hypotension during cardiopulmonary bypass adjusting for vasopressor-inotrope dose, milrinone dose, patient, and surgical factors. RESULTS: The composite outcome occurred in 256 (5.1%) of 4,984 included patient records; 66 (1.3%) patients suffered stroke, 125 (2.5%) had acute kidney injury, and 109 (2.2%) died. The primary outcome was associated with total duration of hypotension (adjusted odds ratio, 1.05; 95% CI, 1.02 to 1.08; P = 0.032), hypotension outside cardiopulmonary bypass (adjusted odds ratio, 1.06; 95% CI, 1.03 to 1.10; P = 0.001) per 10-min exposure to mean arterial pressure of less than 65 mmHg, and fraction of hypotension duration during cardiopulmonary bypass of less than 60% (reference greater than 80%; adjusted odds ratio, 1.67; 95% CI, 1.10 to 2.60; P = 0.019) but not with each 10-min period hypotension during cardiopulmonary bypass (adjusted odds ratio, 1.04; 95% CI, 0.99 to 1.09; P = 0.118), fraction of hypotension during cardiopulmonary bypass of 60 to 80% (adjusted odds ratio, 1.45; 95% CI, 0.97 to 2.23; P = 0.082), or total vasopressor-inotrope dose (adjusted odds ratio, 1.00; 95% CI, 1.00 to 1.00; P = 0.247). CONCLUSIONS: This study confirms previous single-center findings that intraoperative hypotension throughout cardiac surgery is associated with an increased risk of acute kidney injury, mortality, or stroke.

Microcirculatory alterations in critically ill COVID-19 patients analyzed using artificial intelligence.

BACKGROUND: The sublingual microcirculation presumably exhibits disease-specific changes in function and morphology. Algorithm-based quantification of functional microcirculatory hemodynamic variables in handheld vital microscopy (HVM) has recently allowed identification of hemodynamic alterations in the microcirculation associated with COVID-19. In the present study we hypothesized that supervised deep machine learning could be used to identify previously unknown microcirculatory alterations, and combination with algorithmically quantified functional variables increases the model's performance to differentiate critically ill COVID-19 patients from healthy volunteers. METHODS: Four international, multi-central cohorts of critically ill COVID-19 patients and healthy volunteers (n = 59/n = 40) were used for neuronal network training and internal validation, alongside quantification of functional microcirculatory hemodynamic variables. Independent verification of the models was performed in a second cohort (n = 25/n = 33). RESULTS: Six thousand ninety-two image sequences in 157 individuals were included. Bootstrapped internal validation yielded AUROC(CI) for detection of COVID-19 status of 0.75 (0.69-0.79), 0.74 (0.69-0.79) and 0.84 (0.80-0.89) for the algorithm-based, deep learning-based and combined models. Individual model performance in external validation was 0.73 (0.71-0.76) and 0.61 (0.58-0.63). Combined neuronal network and algorithm-based identification yielded the highest externally validated AUROC of 0.75 (0.73-0.78) (P < 0.0001 versus internal validation and individual models). CONCLUSIONS: We successfully trained a deep learning-based model to differentiate critically ill COVID-19 patients from heathy volunteers in sublingual HVM image sequences. Internally validated, deep learning was superior to the algorithmic approach. However, combining the deep learning method with an algorithm-based approach to quantify the functional state of the microcirculation markedly increased the sensitivity and specificity as compared to either approach alone, and enabled successful external validation of the identification of the presence of microcirculatory alterations associated with COVID-19 status.