Performance of a machine-learning algorithm to predict hypotension in mechanically ventilated patients with COVID-19 admitted to the intensive care unit: a cohort study.

The Hypotension Prediction Index (HPI) is a commercially available machine-learning algorithm that provides warnings for impending hypotension, based on real-time arterial waveform analysis. The HPI was developed with arterial waveform data of surgical and intensive care unit (ICU) patients, but has never been externally validated in the latter group. In this study, we evaluated diagnostic ability of the HPI with invasively collected arterial blood pressure data in 41 patients with COVID-19 admitted to the ICU for mechanical ventilation. Predictive ability was evaluated at HPI thresholds from 0 to 100, at incremental intervals of 5. After exceeding the studied threshold, the next 20 min were screened for positive (mean arterial pressure (MAP) < 65 mmHg for at least 1 min) or negative (absence of MAP < 65 mmHg for at least 1 min) events. Subsequently, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and time to event were determined for every threshold. Almost all patients (93%) experienced at least one hypotensive event. Median number of events was 21 [7-54] and time spent in hypotension was 114 min [20-303]. The optimal threshold was 90, with a sensitivity of 0.91 (95% confidence interval 0.81-0.98), specificity of 0.87 (0.81-0.92), PPV of 0.69 (0.61-0.77), NPV of 0.99 (0.97-1.00), and median time to event of 3.93 min (3.72-4.15). Discrimination ability of the HPI was excellent, with an area under the curve of 0.95 (0.93-0.97). This validation study shows that the HPI correctly predicts hypotension in mechanically ventilated COVID-19 patients in the ICU, and provides a basis for future studies to assess whether hypotension can be reduced in ICU patients using this algorithm.

Hypotension during intensive care stay and mortality and morbidity: a systematic review and meta-analysis.

PURPOSE: The aim of this study is to provide a summary of the existing literature on the association between hypotension during intensive care unit (ICU) stay and mortality and morbidity, and to assess whether there is an exposure-severity relationship between hypotension exposure and patient outcomes. METHODS: CENTRAL, Embase, and PubMed were searched up to October 2022 for articles that reported an association between hypotension during ICU stay and at least one of the 11 predefined outcomes. Two independent reviewers extracted the data and assessed the risk of bias. Results were gathered in a summary table and studies designed to investigate the hypotension-outcome relationship were included in the meta-analyses. RESULTS: A total of 122 studies (176,329 patients) were included, with the number of studies varying per outcome between 0 and 82. The majority of articles reported associations in favor of 'no hypotension' for the outcomes mortality and acute kidney injury (AKI), and the strength of the association was related to the severity of hypotension in the majority of studies. Using meta-analysis, a significant association was found between hypotension and mortality (odds ratio: 1.45; 95% confidence interval (CI) 1.12-1.88; based on 13 studies and 34,829 patients), but not for AKI. CONCLUSION: Exposure to hypotension during ICU stay was associated with increased mortality and AKI in the majority of included studies, and associations for both outcomes increased with increasing hypotension severity. The meta-analysis reinforced the descriptive findings regarding mortality but did not yield similar support for AKI.

Effect of the machine learning-derived Hypotension Prediction Index (HPI) combined with diagnostic guidance versus standard care on depth and duration of intraoperative and postoperative hypotension in elective cardiac surgery patients: HYPE-2 - study protocol of a randomised clinical trial.

INTRODUCTION: Hypotension is common during cardiac surgery and often persists postoperatively in the intensive care unit (ICU). Still, treatment is mainly reactive, causing a delay in its management. The Hypotension Prediction Index (HPI) can predict hypotension with high accuracy. Using the HPI combined with a guidance protocol resulted in a significant reduction in the severity of hypotension in four non-cardiac surgery trials. This randomised trial aims to evaluate the effectiveness of the HPI in combination with a diagnostic guidance protocol on reducing the occurrence and severity of hypotension during coronary artery bypass grafting (CABG) surgery and subsequent ICU admission. METHODS AND ANALYSIS: This is a single-centre, randomised clinical trial in adult patients undergoing elective on-pump CABG surgery with a target mean arterial pressure of 65 mm Hg. One hundred and thirty patients will be randomly allocated in a 1:1 ratio to either the intervention or control group. In both groups, a HemoSphere patient monitor with embedded HPI software will be connected to the arterial line. In the intervention group, HPI values of 75 or above will initiate the diagnostic guidance protocol, both intraoperatively and postoperatively in the ICU during mechanical ventilation. In the control group, the HemoSphere patient monitor will be covered and silenced. The primary outcome is the time-weighted average of hypotension during the combined study phases. ETHICS AND DISSEMINATION: The medical research ethics committee and the institutional review board of the Amsterdam UMC, location AMC, the Netherlands, approved the trial protocol (NL76236.018.21). No publication restrictions apply, and the study results will be disseminated through a peer-reviewed journal. TRIAL REGISTRATION NUMBER: The Netherlands Trial Register (NL9449), ClinicalTrials.gov (NCT05821647).

Clinical agreement of a novel algorithm to estimate radial artery blood pressure from the non-invasive finger blood pressure.

STUDY OBJECTIVE: A new algorithm was developed that transforms the non-invasive finger blood pressure (BP) into a radial artery BP (B̂PRad), whereas the original algorithm estimated brachial BP (B̂PBra). In this study we determined whether this new algorithm shows better agreement with invasive radial BP than the original one and whether in the operating room this algorithm can be used safely. DESIGN, SETTING AND PATIENTS: This observational study was conducted on thirty-three non-cardiac surgery patients. INTERVENTION AND MEASUREMENTS: Invasive radial and non-invasive finger BP were measured, of the latter B̂PRad and B̂PBra were transformed. Agreement of systolic, mean, and diastolic arterial BP (SAP, MAP, and DAP, respectively) was assessed traditionally with Bland-Altman and trend analysis and clinically safety was quantified with error grid analyses. A bias (precision) of 5 (8) mmHg or less was considered adequate. MAIN RESULTS: Thirty-three patients were included with an average of 676 (314) 20 s segments. For both comparisons, bias (precision) of MAP was within specified criteria, whereas for SAP, precision was higher than 8 mmHg. B̂PRad showed a better agreement than B̂PBra with BPRad for DAP values (bias (precision): 0.7 (6.0) and - 6.4 (4.3) mmHg, respectively). B̂PRad and B̂PBra both showed good concordance in following changes in BPRad (for all parameters overall degree was <7°). There were slightly more measurement pairs of MAP within the no-risk zone for B̂PRad than for B̂PBra (96 vs 77%, respectively). CONCLUSIONS: In this cohort of non-cardiac surgery patients, we found good agreement between BPRad and B̂PRad. Compared to B̂PBra, B̂PRad shows better agreement although clinical implications are small. This trial was registered with ClinicalTrials.gov (https://clinicaltrials.gov/ct2/show/NCT03795831).

Machine learning methods for perioperative anesthetic management in cardiac surgery patients: a scoping review.

BACKGROUND: Machine learning (ML) is developing fast with promising prospects within medicine and already has several applications in perioperative care. We conducted a scoping review to examine the extent and potential limitations of ML implementation in perioperative anesthetic care, specifically in cardiac surgery patients. METHODS: We mapped the current literature by searching three databases: MEDLINE (Ovid), EMBASE (Ovid), and Cochrane Library. Articles were eligible if they reported on perioperative ML use in the field of cardiac surgery with relevance to anesthetic practices. Data on the applicability of ML and comparability to conventional statistical methods were extracted. RESULTS: Forty-six articles on ML relevant to the work of the anesthesiologist in cardiac surgery were identified. Three main categories emerged: (I) event and risk prediction, (II) hemodynamic monitoring, and (III) automation of echocardiography. Prediction models based on ML tend to behave similarly to conventional statistical methods. Using dynamic hemodynamic or ultrasound data in ML models, however, shifts the potential to promising results. CONCLUSIONS: ML in cardiac surgery is increasingly used in perioperative anesthetic management. The majority is used for prediction purposes similar to conventional clinical scores. Remarkable ML model performances are achieved when using real-time dynamic parameters. However, beneficial clinical outcomes of ML integration have yet to be determined. Nonetheless, the first steps introducing ML in perioperative anesthetic care for cardiac surgery have been taken.

One of the first validations of an artificial intelligence algorithm for clinical use: The impact on intraoperative hypotension prediction and clinical decision-making.

This review describes the steps and conclusions from the development and validation of an artificial intelligence algorithm (the Hypotension Prediction Index), one of the first machine learning predictive algorithms used in the operating room environment. The algorithm has been demonstrated to reduce intraoperative hypotension in two randomized controlled trials via real-time prediction of upcoming hypotensive events prompting anesthesiologists to act earlier, more often, and differently in managing impending hypotension. However, the algorithm entails no dynamic learning process that evolves from use in clinical patient care, meaning the algorithm is fixed, and furthermore provides no insight into the decisional process that leads to an early warning for intraoperative hypotension, which makes the algorithm a "black box." Many other artificial intelligence machine learning algorithms have these same disadvantages. Clinical validation of such algorithms is relatively new and requires more standardization, as guidelines are lacking or only now start to be drafted. Before adaptation in clinical practice, impact of artificial intelligence algorithms on clinical behavior, outcomes and economic advantages should be studied too.

Practice of adjunctive treatments in critically ill COVID-19 patients-rational for the multicenter observational PRoAcT-COVID study in The Netherlands.

BACKGROUND: Patients with coronavirus disease 2019 (COVID-19) may need hospitalization for supplemental oxygen, and some need intensive care unit (ICU) admission for escalation of care. Practice of adjunctive and supportive treatments remain uncertain and may vary widely between countries, within countries between hospitals, and possibly even within ICUs. We aim to investigate practice of adjunctive and supportive treatments, and their associations with outcome, in critically ill COVID-19 patients. METHODS: The 'PRactice of Adjunctive Treatments in Intensive Care Unit Patients with Coronavirus Disease 2019' (PRoAcT-COVID) study is a national, observational study to be undertaken in a large set of ICUs in The Netherlands. The PRoAcT-COVID includes consecutive ICU patients, admitted because of COVID-19 to one of the participating ICUs during a 3-month period. Daily follow-up lasts 28 days. The primary endpoint is a combination of adjunctive treatments, including types of oxygen support, ventilation, rescue therapies for hypoxemia refractory to supplementary oxygen or during invasive ventilation, other adjunctive and supportive treatments, and experimental therapies. We will also collect tracheostomy rate, duration of invasive ventilation and ventilator-free days and alive at day 28 (VFD-28), ICU and hospital length of stay, and the mortality rates in the ICU, hospital and at day 90. DISCUSSION: The PRoAcT-COVID study is an observational study combining high density treatment data with relevant clinical outcomes. Information on treatment practices, and their associations with outcomes in COVID-19 patients in highly and urgently needed. The results of the PRoAcT-COVID study will be rapidly available, and circulated through online presentations, such as webinars and electronic conferences, and publications in peer-reviewed journals-findings will also be presented at a dedicated website. At request, and after agreement of the PRoAcT-COVID steering committee, source data will be made available through local, regional and national anonymized datasets. TRIAL REGISTRATION: The PRoAcT-COVID study is registered at clinicaltrials.gov (study identifier NCT04719182).