Sleep staging in the ICU with heart rate variability and breathing signals. An exploratory cross-sectional study using deep neural networks.

Introduction: To measure sleep in the intensive care unit (ICU), full polysomnography is impractical, while activity monitoring and subjective assessments are severely confounded. However, sleep is an intensely networked state, and reflected in numerous signals. Here, we explore the feasibility of estimating conventional sleep indices in the ICU with heart rate variability (HRV) and respiration signals using artificial intelligence methods Methods: We used deep learning models to stage sleep with HRV (through electrocardiogram) and respiratory effort (through a wearable belt) signals in critically ill adult patients admitted to surgical and medical ICUs, and in age and sex-matched sleep laboratory patients Results: We studied 102 adult patients in the ICU across multiple days and nights, and 220 patients in a clinical sleep laboratory. We found that sleep stages predicted by HRV- and breathing-based models showed agreement in 60% of the ICU data and in 81% of the sleep laboratory data. In the ICU, deep NREM (N2 + N3) proportion of total sleep duration was reduced (ICU 39%, sleep laboratory 57%, p < 0.01), REM proportion showed heavy-tailed distribution, and the number of wake transitions per hour of sleep (median 3.6) was comparable to sleep laboratory patients with sleep-disordered breathing (median 3.9). Sleep in the ICU was also fragmented, with 38% of sleep occurring during daytime hours. Finally, patients in the ICU showed faster and less variable breathing patterns compared to sleep laboratory patients Conclusion: The cardiovascular and respiratory networks encode sleep state information, which, together with artificial intelligence methods, can be utilized to measure sleep state in the ICU.

High prevalence of sleep-disordered breathing in the intensive care unit - a cross-sectional study.

PURPOSE: Sleep-disordered breathing may be induced by, exacerbate, or complicate recovery from critical illness. Disordered breathing during sleep, which itself is often fragmented, can go unrecognized in the intensive care unit (ICU). The objective of this study was to investigate the prevalence, severity, and risk factors of sleep-disordered breathing in ICU patients using a single respiratory belt and oxygen saturation signals. METHODS: Patients in three ICUs at Massachusetts General Hospital wore a thoracic respiratory effort belt as part of a clinical trial for up to 7 days and nights. Using a previously developed machine learning algorithm, we processed respiratory and oximetry signals to measure the 3% apnea-hypopnea index (AHI) and estimate AH-specific hypoxic burden and periodic breathing. We trained models to predict AHI categories for 12-h segments from risk factors, including admission variables and bio-signals data, available at the start of these segments. RESULTS: Of 129 patients, 68% had an AHI ≥ 5; 40% an AHI > 15, and 19% had an AHI > 30 while critically ill. Median [interquartile range] hypoxic burden was 2.8 [0.5, 9.8] at night and 4.2 [1.0, 13.7] %min/h during the day. Of patients with AHI ≥ 5, 26% had periodic breathing. Performance of predicting AHI-categories from risk factors was poor. CONCLUSIONS: Sleep-disordered breathing and sleep apnea events while in the ICU are common and are associated with substantial burden of hypoxia and periodic breathing. Detection is feasible using limited bio-signals, such as respiratory effort and SpO2 signals, while risk factors were insufficient to predict AHI severity.

Automated Annotation of Epileptiform Burden and Its Association with Outcomes.

OBJECTIVE: This study was undertaken to determine the dose-response relation between epileptiform activity burden and outcomes in acutely ill patients. METHODS: A single center retrospective analysis was made of 1,967 neurologic, medical, and surgical patients who underwent >16 hours of continuous electroencephalography (EEG) between 2011 and 2017. We developed an artificial intelligence algorithm to annotate 11.02 terabytes of EEG and quantify epileptiform activity burden within 72 hours of recording. We evaluated burden (1) in the first 24 hours of recording, (2) in the 12-hours epoch with highest burden (peak burden), and (3) cumulatively through the first 72 hours of monitoring. Machine learning was applied to estimate the effect of epileptiform burden on outcome. Outcome measure was discharge modified Rankin Scale, dichotomized as good (0-4) versus poor (5-6). RESULTS: Peak epileptiform burden was independently associated with poor outcomes (p < 0.0001). Other independent associations included age, Acute Physiology and Chronic Health Evaluation II score, seizure on presentation, and diagnosis of hypoxic-ischemic encephalopathy. Model calibration error was calculated across 3 strata based on the time interval between last EEG measurement (up to 72 hours of monitoring) and discharge: (1) <5 days between last measurement and discharge, 0.0941 (95% confidence interval [CI] = 0.0706-0.1191); 5 to 10 days between last measurement and discharge, 0.0946 (95% CI = 0.0631-0.1290); >10 days between last measurement and discharge, 0.0998 (95% CI = 0.0698-0.1335). After adjusting for covariates, increase in peak epileptiform activity burden from 0 to 100% increased the probability of poor outcome by 35%. INTERPRETATION: Automated measurement of peak epileptiform activity burden affords a convenient, consistent, and quantifiable target for future multicenter randomized trials investigating whether suppressing epileptiform activity improves outcomes. ANN NEUROL 2021;90:300-311.

Prolonged Intubation in Patients With Prior Cerebrovascular Disease and COVID-19.

Objectives: Patients with comorbidities are at increased risk for poor outcomes in COVID-19, yet data on patients with prior neurological disease remains limited. Our objective was to determine the odds of critical illness and duration of mechanical ventilation in patients with prior cerebrovascular disease and COVID-19. Methods: A observational study of 1,128 consecutive adult patients admitted to an academic center in Boston, Massachusetts, and diagnosed with laboratory-confirmed COVID-19. We tested the association between prior cerebrovascular disease and critical illness, defined as mechanical ventilation (MV) or death by day 28, using logistic regression with inverse probability weighting of the propensity score. Among intubated patients, we estimated the cumulative incidence of successful extubation without death over 45 days using competing risk analysis. Results: Of the 1,128 adults with COVID-19, 350 (36%) were critically ill by day 28. The median age of patients was 59 years (SD: 18 years) and 640 (57%) were men. As of June 2nd, 2020, 127 (11%) patients had died. A total of 177 patients (16%) had a prior cerebrovascular disease. Prior cerebrovascular disease was significantly associated with critical illness (OR = 1.54, 95% CI = 1.14-2.07), lower rate of successful extubation (cause-specific HR = 0.57, 95% CI = 0.33-0.98), and increased duration of intubation (restricted mean time difference = 4.02 days, 95% CI = 0.34-10.92) compared to patients without cerebrovascular disease. Interpretation: Prior cerebrovascular disease adversely affects COVID-19 outcomes in hospitalized patients. Further study is required to determine if this subpopulation requires closer monitoring for disease progression during COVID-19.

Development and Prospective Validation of a Deep Learning Algorithm for Predicting Need for Mechanical Ventilation.

BACKGROUND: Objective and early identification of hospitalized patients, and particularly those with novel coronavirus disease 2019 (COVID-19), who may require mechanical ventilation (MV) may aid in delivering timely treatment. RESEARCH QUESTION: Can a transparent deep learning (DL) model predict the need for MV in hospitalized patients and those with COVID-19 up to 24 h in advance? STUDY DESIGN AND METHODS: We trained and externally validated a transparent DL algorithm to predict the future need for MV in hospitalized patients, including those with COVID-19, using commonly available data in electronic health records. Additionally, commonly used clinical criteria (heart rate, oxygen saturation, respiratory rate, Fio2, and pH) were used to assess future need for MV. Performance of the algorithm was evaluated using the area under receiver operating characteristic curve (AUC), sensitivity, specificity, and positive predictive value. RESULTS: We obtained data from more than 30,000 ICU patients (including more than 700 patients with COVID-19) from two academic medical centers. The performance of the model with a 24-h prediction horizon at the development and validation sites was comparable (AUC, 0.895 vs 0.882, respectively), providing significant improvement over traditional clinical criteria (P < .001). Prospective validation of the algorithm among patients with COVID-19 yielded AUCs in the range of 0.918 to 0.943. INTERPRETATION: A transparent deep learning algorithm improves on traditional clinical criteria to predict the need for MV in hospitalized patients, including in those with COVID-19. Such an algorithm may help clinicians to optimize timing of tracheal intubation, to allocate resources and staff better, and to improve patient care.

CoVA: An Acuity Score for Outpatient Screening that Predicts Coronavirus Disease 2019 Prognosis.

BACKGROUND: We sought to develop an automatable score to predict hospitalization, critical illness, or death for patients at risk for coronavirus disease 2019 (COVID-19) presenting for urgent care. METHODS: We developed the COVID-19 Acuity Score (CoVA) based on a single-center study of adult outpatients seen in respiratory illness clinics or the emergency department. Data were extracted from the Partners Enterprise Data Warehouse, and split into development (n = 9381, 7 March-2 May) and prospective (n = 2205, 3-14 May) cohorts. Outcomes were hospitalization, critical illness (intensive care unit or ventilation), or death within 7 days. Calibration was assessed using the expected-to-observed event ratio (E/O). Discrimination was assessed by area under the receiver operating curve (AUC). RESULTS: In the prospective cohort, 26.1%, 6.3%, and 0.5% of patients experienced hospitalization, critical illness, or death, respectively. CoVA showed excellent performance in prospective validation for hospitalization (expected-to-observed ratio [E/O]: 1.01; AUC: 0.76), for critical illness (E/O: 1.03; AUC: 0.79), and for death (E/O: 1.63; AUC: 0.93). Among 30 predictors, the top 5 were age, diastolic blood pressure, blood oxygen saturation, COVID-19 testing status, and respiratory rate. CONCLUSIONS: CoVA is a prospectively validated automatable score for the outpatient setting to predict adverse events related to COVID-19 infection.

COVID-19 Outpatient Screening: a Prediction Score for Adverse Events.

BACKGROUND: We sought to develop an automatable score to predict hospitalization, critical illness, or death in patients at risk for COVID-19 presenting for urgent care during the Massachusetts outbreak. METHODS: Single-center study of adult outpatients seen in respiratory illness clinics (RICs) or the emergency department (ED), including development (n = 9381, March 7-May 2) and prospective (n = 2205, May 3-14) cohorts. Data was queried from Partners Enterprise Data Warehouse. Outcomes were hospitalization, critical illness or death within 7 days. We developed the COVID-19 Acuity Score (CoVA) using automatically extracted data from the electronic medical record and learning-to-rank ordinal logistic regression modeling. Calibration was assessed using predicted-to-observed event ratio (E/O). Discrimination was assessed by C-statistics (AUC). RESULTS: In the development cohort, 27.3%, 7.2%, and 1.1% of patients experienced hospitalization, critical illness, or death, respectively; and in the prospective cohort, 26.1%, 6.3%, and 0.5%. CoVA showed excellent performance in the development cohort (concurrent validation) for hospitalization (E/O: 1.00, AUC: 0.80); for critical illness (E/O: 1.00, AUC: 0.82); and for death (E/O: 1.00, AUC: 0.87). Performance in the prospective cohort (prospective validation) was similar for hospitalization (E/O: 1.01, AUC: 0.76); for critical illness (E/O 1.03, AUC: 0.79); and for death (E/O: 1.63, AUC=0.93). Among 30 predictors, the top five were age, diastolic blood pressure, blood oxygen saturation, COVID-19 testing status, and respiratory rate. CONCLUSIONS: CoVA is a prospectively validated automatable score to assessing risk for adverse outcomes related to COVID-19 infection in the outpatient setting.

Development and Prospective Validation of a Transparent Deep Learning Algorithm for Predicting Need for Mechanical Ventilation.

IMPORTANCE: Objective and early identification of hospitalized patients, and particularly those with novel coronavirus disease 2019 (COVID-19), who may require mechanical ventilation is of great importance and may aid in delivering timely treatment. OBJECTIVE: To develop, externally validate and prospectively test a transparent deep learning algorithm for predicting 24 hours in advance the need for mechanical ventilation in hospitalized patients and those with COVID-19. DESIGN: Observational cohort study SETTING: Two academic medical centers from January 01, 2016 to December 31, 2019 (Retrospective cohorts) and February 10, 2020 to May 4, 2020 (Prospective cohorts). PARTICIPANTS: Over 31,000 admissions to the intensive care units (ICUs) at two hospitals. Additionally, 777 patients with COVID-19 patients were used for prospective validation. Patients who were placed on mechanical ventilation within four hours of their admission were excluded. MAIN OUTCOME(S) and MEASURE(S): Electronic health record (EHR) data were extracted on an hourly basis, and a set of 40 features were calculated and passed to an interpretable deep-learning algorithm to predict the future need for mechanical ventilation 24 hours in advance. Additionally, commonly used clinical criteria (based on heart rate, oxygen saturation, respiratory rate, FiO2 and pH) was used to assess future need for mechanical ventilation. Performance of the algorithms were evaluated using the area under receiver-operating characteristic curve (AUC), sensitivity, specificity and positive predictive value. RESULTS: After applying exclusion criteria, the external validation cohort included 3,888 general ICU and 402 COVID-19 patients. The performance of the model (AUC) with a 24-hour prediction horizon at the validation site was 0.882 for the general ICU population and 0.918 for patients with COVID-19. In comparison, commonly used clinical criteria and the ROX score achieved AUCs in the range of 0.773 - 0.782 and 0.768 - 0.810 for the general ICU population and patients with COVID-19, respectively. CONCLUSIONS AND RELEVANCE: A generalizable and transparent deep-learning algorithm improves on traditional clinical criteria to predict the need for mechanical ventilation in hospitalized patients, including those with COVID-19. Such an algorithm may help clinicians with optimizing timing of tracheal intubation, better allocation of mechanical ventilation resources and staff, and improve patient care.

Burden of Epileptiform Activity Predicts Discharge Neurologic Outcomes in Severe Acute Ischemic Stroke.

BACKGROUND/OBJECTIVES: Clinical seizures following acute ischemic stroke (AIS) appear to contribute to worse neurologic outcomes. However, the effect of electrographic epileptiform abnormalities (EAs) more broadly is less clear. Here, we evaluate the impact of EAs, including electrographic seizures and periodic and rhythmic patterns, on outcomes in patients with AIS. METHODS: This is a retrospective study of all patients with AIS aged ≥ 18 years who underwent at least 18 h of continuous electroencephalogram (EEG) monitoring at a single center between 2012 and 2017. EAs were classified according to American Clinical Neurophysiology Society (ACNS) nomenclature and included seizures and periodic and rhythmic patterns. EA burden for each 24-h epoch was defined using the following cutoffs: EA presence, maximum daily burden < 10% versus > 10%, maximum daily burden < 50% versus > 50%, and maximum daily burden using categories from ACNS nomenclature ("rare" < 1%; "occasional" 1-9%; "frequent" 10-49%; "abundant" 50-89%; "continuous" > 90%). Maximum EA frequency for each epoch was dichotomized into ≥ 1.5 Hz versus < 1.5 Hz. Poor neurologic outcome was defined as a modified Rankin Scale score of 4-6 (vs. 0-3 as good outcome) at hospital discharge. RESULTS: One hundred and forty-three patients met study inclusion criteria. Sixty-seven patients (46.9%) had EAs. One hundred and twenty-four patients (86.7%) had poor outcome. On univariate analysis, the presence of EAs (OR 3.87 [1.27-11.71], p = 0.024) and maximum daily burden > 10% (OR 12.34 [2.34-210], p = 0.001) and > 50% (OR 8.26 [1.34-122], p = 0.035) were associated with worse outcomes. On multivariate analysis, after adjusting for clinical covariates (age, gender, NIHSS, APACHE II, stroke location, stroke treatment, hemorrhagic transformation, Charlson comorbidity index, history of epilepsy), EA presence (OR 5.78 [1.36-24.56], p = 0.017), maximum daily burden > 10% (OR 23.69 [2.43-230.7], p = 0.006), and maximum daily burden > 50% (OR 9.34 [1.01-86.72], p = 0.049) were associated with worse outcomes. After adjusting for covariates, we also found a dose-dependent association between increasing EA burden and increasing probability of poor outcomes (OR 1.89 [1.18-3.03] p = 0.009). We did not find an independent association between EA frequency and outcomes (OR: 4.43 [.98-20.03] p = 0.053). However, the combined effect of increasing EA burden and frequency ≥ 1.5 Hz (EA burden * frequency) was significantly associated with worse outcomes (OR 1.64 [1.03-2.63] p = 0.039). CONCLUSIONS: Electrographic seizures and periodic and rhythmic patterns in patients with AIS are associated with worse outcomes in a dose-dependent manner. Future studies are needed to assess whether treatment of this EEG activity can improve outcomes.

Burst Suppression: Causes and Effects on Mortality in Critical Illness.

BACKGROUND: Burst suppression in mechanically ventilated intensive care unit (ICU) patients is associated with increased mortality. However, the relative contributions of propofol use and critical illness itself to burst suppression; of burst suppression, propofol, and critical illness to mortality; and whether preventing burst suppression might reduce mortality, have not been quantified. METHODS: The dataset contains 471 adults from seven ICUs, after excluding anoxic encephalopathy due to cardiac arrest or intentional burst suppression for therapeutic reasons. We used multiple prediction and causal inference methods to estimate the effects connecting burst suppression, propofol, critical illness, and in-hospital mortality in an observational retrospective study. We also estimated the effects mediated by burst suppression. Sensitivity analysis was used to assess for unmeasured confounding. RESULTS: The expected outcomes in a "counterfactual" randomized controlled trial (cRCT) that assigned patients to mild versus severe illness are expected to show a difference in burst suppression burden of 39%, 95% CI [8-66]%, and in mortality of 35% [29-41]%. Assigning patients to maximal (100%) burst suppression burden is expected to increase mortality by 12% [7-17]% compared to 0% burden. Burst suppression mediates 10% [2-21]% of the effect of critical illness on mortality. A high cumulative propofol dose (1316 mg/kg) is expected to increase burst suppression burden by 6% [0.8-12]% compared to a low dose (284 mg/kg). Propofol exposure has no significant direct effect on mortality; its effect is entirely mediated through burst suppression. CONCLUSIONS: Our analysis clarifies how important factors contribute to mortality in ICU patients. Burst suppression appears to contribute to mortality but is primarily an effect of critical illness rather than iatrogenic use of propofol.

Real-Time, Automated Detection of Ventilator-Associated Events: Avoiding Missed Detections, Misclassifications, and False Detections Due to Human Error.

OBJECTIVETo validate a system to detect ventilator associated events (VAEs) autonomously and in real time.DESIGNRetrospective review of ventilated patients using a secure informatics platform to identify VAEs (ie, automated surveillance) compared to surveillance by infection control (IC) staff (ie, manual surveillance), including development and validation cohorts.SETTINGThe Massachusetts General Hospital, a tertiary-care academic health center, during January-March 2015 (development cohort) and January-March 2016 (validation cohort).PATIENTSVentilated patients in 4 intensive care units.METHODSThe automated process included (1) analysis of physiologic data to detect increases in positive end-expiratory pressure (PEEP) and fraction of inspired oxygen (FiO2); (2) querying the electronic health record (EHR) for leukopenia or leukocytosis and antibiotic initiation data; and (3) retrieval and interpretation of microbiology reports. The cohorts were evaluated as follows: (1) manual surveillance by IC staff with independent chart review; (2) automated surveillance detection of ventilator-associated condition (VAC), infection-related ventilator-associated complication (IVAC), and possible VAP (PVAP); (3) senior IC staff adjudicated manual surveillance-automated surveillance discordance. Outcomes included sensitivity, specificity, positive predictive value (PPV), and manual surveillance detection errors. Errors detected during the development cohort resulted in algorithm updates applied to the validation cohort.RESULTSIn the development cohort, there were 1,325 admissions, 479 ventilated patients, 2,539 ventilator days, and 47 VAEs. In the validation cohort, there were 1,234 admissions, 431 ventilated patients, 2,604 ventilator days, and 56 VAEs. With manual surveillance, in the development cohort, sensitivity was 40%, specificity was 98%, and PPV was 70%. In the validation cohort, sensitivity was 71%, specificity was 98%, and PPV was 87%. With automated surveillance, in the development cohort, sensitivity was 100%, specificity was 100%, and PPV was 100%. In the validation cohort, sensitivity was 85%, specificity was 99%, and PPV was 100%. Manual surveillance detection errors included missed detections, misclassifications, and false detections.CONCLUSIONSManual surveillance is vulnerable to human error. Automated surveillance is more accurate and more efficient for VAE surveillance.Infect Control Hosp Epidemiol 2018;826-833.