Dataset Information

Using machine learning models to predict oxygen saturation following ventilator support adjustment in critically ill children: A single center pilot study.

ABSTRACT:

Background

In an intensive care units, experts in mechanical ventilation are not continuously at patient's bedside to adjust ventilation settings and to analyze the impact of these adjustments on gas exchange. The development of clinical decision support systems analyzing patients' data in real time offers an opportunity to fill this gap.

Objective

The objective of this study was to determine whether a machine learning predictive model could be trained on a set of clinical data and used to predict transcutaneous hemoglobin oxygen saturation 5 min (5min SpO2) after a ventilator setting change.

Data sources

Data of mechanically ventilated children admitted between May 2015 and April 2017 were included and extracted from a high-resolution research database. More than 776,727 data rows were obtained from 610 patients, discretized into 3 class labels (< 84%, 85% to 91% and c92% to 100%).

Performance metrics of predictive models

Due to data imbalance, four different data balancing processes were applied. Then, two machine learning models (artificial neural network and Bootstrap aggregation of complex decision trees) were trained and tested on these four different balanced datasets. The best model predicted SpO2 with area under the curves < 0.75.

Conclusion

This single center pilot study using machine learning predictive model resulted in an algorithm with poor accuracy. The comparison of machine learning models showed that bagged complex trees was a promising approach. However, there is a need to improve these models before incorporating them into a clinical decision support systems. One potentially solution for improving predictive model, would be to increase the amount of data available to limit over-fitting that is potentially one of the cause for poor classification performances for 2 of the three class labels.

SUBMITTER: Ghazal S

PROVIDER: S-EPMC6382156 | biostudies-literature | 2019

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Using machine learning models to predict oxygen saturation following ventilator support adjustment in critically ill children: A single center pilot study.

Ghazal Sam S Sauthier Michael M Brossier David D Bouachir Wassim W Jouvet Philippe A PA Noumeir Rita R

PloS one 20190220 2

<h4>Background</h4>In an intensive care units, experts in mechanical ventilation are not continuously at patient's bedside to adjust ventilation settings and to analyze the impact of these adjustments on gas exchange. The development of clinical decision support systems analyzing patients' data in real time offers an opportunity to fill this gap.<h4>Objective</h4>The objective of this study was to determine whether a machine learning predictive model could be trained on a set of clinical data an ...[more]

PMID: 30785881

Similar Datasets

Project description:BackgroundCentral venous oxygen saturation (ScvO2) is often used to help to guide resuscitation of critically ill patients. The standard gold technique for ScvO2 measurement is the co-oximetry (Co-oximetry_ScvO2), which is usually incorporated in most recent blood gas analyzers. However, in some hospitals, those machines are not available and only calculated ScvO2 (Calc_ScvO2) is provided. Therefore, we aimed to investigate the agreement between Co-oximetry_ScvO2 and Calc_ScvO2 in a general population of critically ill patients and septic shock patients.MethodsA total of 100 patients with a central venous catheter were included in the study. One hundred central venous blood samples were collected and analyzed using the same point-of-care blood gas analyzer, which provides both the calculated and measured ScvO2 values. Bland and Altman plot, intra-class correlation coefficient (ICC), and Cohen's Kappa coefficient were used to assess the agreement between Co-oximetry_ScvO2 and Calc_ScvO2. Multiple linear regression analysis was performed to investigate the independent explanatory variables of the difference between Co-oximetry_ScvO2 and Calc_ScvO2.ResultsIn all population, Bland and Altman's analysis showed poor agreement (+4.5 [-7.1, +16.1]%) between the two techniques. The ICC was 0.754 [(95% CI: 0.393-0.880), P< 0.001], and the Cohen's Kappa coefficient, after categorizing the two variables into two groups using a cutoff value of 70%, was 0.470 (P <0.001). In septic shock patients (49%), Bland and Altman's analysis also showed poor agreement (+5.6 [-6.7 to 17.8]%). The ICC was 0.720 [95% CI: 0.222-0.881], and the Cohen's Kappa coefficient was 0.501 (P <0.001). Four independent variables (PcvO2, Co-oximetry_ScvO2, venous pH, and Hb) were found to be associated with the difference between the measured and calculated ScvO2 (adjusted R2 = 0.8, P<0.001), with PcvO2 being the main independent explanatory variable because of its highest absolute standardized coefficient. The area under the receiver operator characteristic curves (AUC) of PcvO2 to predict Co-oximetry_ScvO2 ? 70% was 0.911 [95% CI: 0.837-0.959], in all patients, and 0.903 [95% CI: 0.784-0.969], in septic shock patients. The best cutoff value was ? 36 mmHg (sensitivity, 88%; specificity, 83%), in all patients, and ? 35 mmHg (sensitivity, 94%; specificity, 71%) in septic shock patients.ConclusionsThe discrepancy between the measured and calculated ScvO2 is clinically not acceptable. We do not recommend the use of calculated ScvO2 to guide resuscitation in critically ill patients. In situations where the Co-oximetry technique is not available, relying on PcvO2 to predict the measured ScvO2 value above or below 70% could be an option.

Project description:BackgroundSerum lactate and central venous oxygen saturation (ScvO2) are commonly used and commonly recommended as markers of tissue oxygenation in shock states. Medical literature has both explicitly stated and implied that the two biomarkers are interchangeable in the management of patients with shock. However, there have been relatively few direct comparisons of these tests in clinical circumstances, and the relationship between them is uncertain. The objective of our study was to evaluate whether simultaneous or near-simultaneous measurements of lactate and ScvO2 reveal a consistent relationship between these two biomarkers.MethodsA retrospective cohort study was conducted in an urban, academic US hospital. All adults in ICUs between March 2007 and March 2017 who had a lactate measurement and ScvO2 or mixed venous oxygen saturation (SvO2) measurement made +/- 1 h from the lactate were included. Linear and non-linear correlations of ScvO2 and lactate were assessed in a variety of shock states.ResultsTwo thousand sixty-two patients were included. Lactate and ScvO2 correlated poorly (r 2 = 0.0041, p = 0.0019). This was true for patients with ScvO2 ≤ 65% (r 2 = 0.0431, p < 0.001), patients with normal kidney and liver function (r 2 = 0.0517, p < 0.001), and septic shock patients (r 2 = 0.0037, p = 0.17). For patients with an O2 extraction ratio ≥ 50%, lactate and ScvO2 were strongly correlated (r 2 = 0.93, p = 0.0019), but these patients represented only 2.8% of patients in whom the ratio could be calculated.ConclusionsLactate can predict ScvO2 when patients are at or below the critical oxygen delivery threshold, but relatively few shock patients meet this criterion. In the overall population of critically ill patients, serum lactate predicts ScvO2 poorly, even after controlling for factors that may affect lactate production. Lactate and ScvO2 should not be assumed to be interchangeable markers of tissue oxygenation/perfusion.

Project description:ObjectiveThe aim of the current study was to investigate how cerebral and splanchnic oxygen saturation (rSO2-C and rSO2-A) in critically ill children transported in air ambulance was affected by flight with cabin pressurization corresponding to ≥ 5000 feet. A second aim was to investigate any differences between cyanotic and non-cyanotic children in relation to cerebral and splanchnic oxygen saturation during flight ≥ 5000 feet. The variability of the cerebral and splanchnic Near Infrared Spectroscopy (NIRS) sensors was evaluated.DesignNIRS was used to measure rSO2-C and rSO2-A during transport of critically ill children in air ambulance. rSO2 data was collected and stored by the NIRS monitor and extracted and analyzed off-line after the transport. Prior to evaluation of the NIRS signals all zero and floor-effect values were removed.SettingThe Pediatric Intensive Care Unit (PICU) at Astrid Lindgren Children's Hospital, Karolinska University Hospital in Stockholm, Sweden.PatientsIn total, 44 critically ill children scheduled for inter-hospital transport by a specialized pediatric transport team were included in the study between January 2014 and January 2019 (convenience sampling).InterventionNo interventions were conducted.MeasurementsAll study patients were monitored with a cerebral NIRS-sensor placed over the forehead and an abdominal NIRS-sensor placed in the infra-umbilical area for cerebral and splanchnic regional oxygen saturation monitoring, rSO2-C and rSO2-A, respectively.Main resultsComplete rSO2-C and rSO2-A data was obtained in 39 patients. Median age was 12 days. Cyanotic congenital heart malformations were present in 9 patients (23%). In 22 patients (56%) rSO2-C decreased at altitude ≥ 5000 feet and in 24 patients (61%) rSO2-A decreased at altitude ≥ 5000 feet compared to baseline (p<0.0001). In 25 patients (64%) the rSO2-C/rSO2-A ratio was greater at altitude ≥ 5000 feet than at baseline. A ratio ≥ 1 was seen in 77% of patients at altitude ≥ 5000 feet compared to in 67% of patients at baseline.ConclusionBoth cerebral and splanchnic oxygen saturation decreased at altitude ≥ 5000 feet compared to baseline. In most patients, both cyanotic and non-cyanotic, cerebral oxygen saturation was preserved more than splanchnic oxygen saturation.

Project description:ImportanceAlthough most critically ill patients receive invasive mechanical ventilation (IMV), few studies have characterized how IMV is discontinued in practice.ObjectiveTo describe practice variation in IMV discontinuation internationally, associations between initial discontinuation events and outcomes, and factors associated with the use of select discontinuation strategies and failed initial spontaneous breathing trials (SBTs).Design, setting, and participantsProspective, multinational, observational study of critically ill adults who received IMV for at least 24 hours from 142 intensive care units (ICUs) in 19 countries within 6 regions (27 in Canada, 23 in India, 22 in the UK, 26 in Europe, 21 in Australia/New Zealand, and 23 in the US).ExposuresReceiving IMV.Main outcomes and measuresPrimary analyses characterized types of initial IMV discontinuation events (extubation, SBT, or tracheostomy) and associations with clinical outcomes (including duration of ventilation, ICU and hospital mortality, and ICU and hospital length of stay). Secondary analyses examined the associations between SBT outcome and SBT timing and clinical outcomes.ResultsAmong 1868 patients (median [interquartile range] age, 61.8 [48.9-73.1] years; 1173 [62.8%] men) 424 (22.7%) underwent direct extubation, 930 (49.8%) had an initial SBT (761 [81.8%] successful), 150 (8.0%) underwent direct tracheostomy, and 364 (19.5%) died before a weaning attempt. Across regions, there was variation in the use of written directives to guide care, daily screening, SBT techniques, ventilator modes, and the roles played by clinicians involved in weaning. Compared with initial direct extubation, patients who had an initial SBT had higher ICU mortality (20 [4.7%] vs 96 [10.3%]; absolute difference, 5.6% [95% CI, 2.6%-8.6%]), longer duration of ventilation (median of 2.9 vs 4.1 days; absolute difference, 1.2 days [95% CI, 0.7-1.6]), and longer ICU stay (median of 6.7 vs 8.1 days; absolute difference, 1.4 days [95% CI, 0.8-2.4]). Patients whose initial SBT failed (vs passed) had higher ICU mortality (29 [17.2%] vs 67 [8.8%]; absolute difference, 8.4% [95% CI, 2.0%-14.7%]), longer duration of ventilation (median of 6.1 vs 3.5 days; absolute difference, 2.6 days [95% CI, 1.6-3.6]), and longer ICU stay (median of 10.6 vs 7.7 days; absolute difference, 2.8 days [95% CI, 1.1-5.2]). Compared with patients who underwent early initial SBTs, patients who underwent late initial SBTs (>2.3 days after intubation) had longer duration of ventilation (median of 2.1 vs 6.1 days; absolute difference, 4.0 days [95% CI, 3.7-4.5]), longer ICU stay (median of 5.9 vs 10.8 days; absolute difference, 4.9 days [95% CI, 4.0-6.3]), and longer hospital stay (median of 14.3 vs 22.8 days; absolute difference, 8.5 days [95% CI, 6.0-11.0]).Conclusions and relevanceIn this observational study of invasive mechanical ventilation discontinuation in 142 ICUs in Canada, India, the UK, Europe, Australia/New Zealand, and the US from 2013 to 2016, weaning practices varied internationally.Trial registrationClinicalTrials.gov Identifier: NCT03955874.

Dataset Information

Using machine learning models to predict oxygen saturation following ventilator support adjustment in critically ill children: A single center pilot study.

Background

Objective

Data sources

Performance metrics of predictive models

Conclusion

Publications

Using machine learning models to predict oxygen saturation following ventilator support adjustment in critically ill children: A single center pilot study.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets