Early outcome prediction (NEWS, EWSO2 and HRV) for spontaneously breathing ICU patients - A data-warehousing analysis

doi:10.21203/rs.3.rs-3400410/v1

Download PDF

Research Article

Early outcome prediction (NEWS, EWSO2 and HRV) for spontaneously breathing ICU patients - A data-warehousing analysis

https://doi.org/10.21203/rs.3.rs-3400410/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Purpose

Predicting models using physiological parameters have been developed for use in the emergency department. The aim of our study was to evaluate NEWS (New Early Warning Score), EWSO2 (Early Warning Score Oxygen), and Heart rate variability (HRV) as outcome predictors for spontaneously breathing patients attending the ICU.

Methods

A post-hoc analysis of a prospective datawarehousing project (ReaSTOC clinicalTrials identifier NCT 02893462). Physiological data and raw PPG curves were prospectively collected during the first 24-hrs of ICU admission. NEWS, EWSO2 and HRV were calculated and the best thresholds to predict the need for invasive ventilation and mortality were determined.

Results

From 112 patients, 8 and 18 patients died within the ICU and during their hospitalization respectively. Patients who died in ICU had a highest NEWS score as compared to ICU survivors (9,0 [5,0–11,6] vs. 6,0 [5,0–7,0]; p = 0.03), highest EWSO2 index (18,4 [15, 7–19, 8] vs. 9,0 [5, 8–20, 9], p = 0.008). Considering a composite prognostic outcome including ICU mortality and/or the need of invasive ventilation and/or a length of stay superior to the median length of stay of patients without invasive ventilation, there was also a significantly difference for NEWS, HRV and EWSO2.

Conclusion

For spontaneously breathing patients attending the ICU, NEWS, EWSO2 and HRV were significantly different in between ICU and hospital survivors and non-survivors. These parameters also enabled an early detection of subsequent additional respiratory assistance (e.g high flow oxygen, noninvasive ventilation) but not the requirement for endotracheal intubation.

Monitoring

Intensive care

arly warning score

Outcome

Vital signs

Physiology

Triage scales have initially been developed to identify patients at risk of poor outcome on emergency department (ED) admission, to prioritize treatment or monitoring. However, these triage scales did not yet permit to detect patients at risk of deterioration [1, 2]. Several predictive risk models using physiological parameters, so-called Early Warning Scores (EWS) have therefore been developed, in EDs [3–5] but also in pediatric populations ] or obstetrics wards [7].

In the ED, the NEWS seems to be an accurate score to predict 30-days mortality, the occurrence of a cardiac arrest or an unanticipated intensive care unit (ICU) admission [8]. Use of the NEWS is therefore recommended in clinical routine by the UK National Health Service to adapt clinical response for each patient [9]. The NEWS uses six physiological parameters, but implementation of additional parameters such as the inspired fraction of oxygen (FiO2)[10] or biologics parameters (serum lactate or D-dimer levels) have also been proposed to predict short- and long-term mortality in various pathological conditions [11, 12]. The SOFA score has been developed to assess organ failure in critically ill patients [13], and a high SOFA score is associated with increased ICU mortality. It sums 6 organ dysfunction variables and requires several biologics parameters to be measured, such as oxygenation parameters. To the best of our knowledge there is only one study that used the NEWS in ICU patients to help decision making [14].

Respiratory parameters are mainly analyzed in case of acute respiratory failure. In a recent post hoc analysis of the FLORALI study, predictive factors for mechanical ventilation were very different depending of the oxygenation technique [15]. Respiratory rate (RR) is correlated to the need of respiratory assistance and its increase might be a good indicator of a risk of adverse event occurrence [15–18]. PaO₂/FIO₂ (P/F) is used to calculate the SOFA [13], and to evaluate gravity in case of acute respiratory distress syndrome (ARDS). Previous studies have demonstrated that SpO₂/FIO₂ (S/F) was well correlated with P/F for ARDS patients’ follow-up [19]. A recent retrospective study also demonstrated that S/F was a good predictor of ICU transfer when used within respiratory ward [20]. Others respiratory scores have been developed during the past years such as the oxygenation index in case of ARDS [21]. The ROX index was recently used to identify patients who need invasive ventilation in case of hypoxemic acute respiratory failure treated with high flow nasal canula (HFNC) [22]. The EWSO₂ is a quite similar predictor that considers either oxygenation, respiratory rate and oxygen flow, but in spontaneously breathing patients, with or without low flow oxygen rates. It has been recently validated to predict outcome in patients attending the ED [23].

Heart rate variability (HRV) reflects the autonomic nervous system activity (ANS) [24, 25]. Measuring HRV enables to explore the influence of the ANS on the cardiovascular system and reflects the ability of the heart to modulate its rhythm, according to unpredictable internal and external solicitation. Prior studies have demonstrated that HRV could be used as an outcome predictor in various pathological, cardiovascular and non-cardiovascular conditions [26].

Based on a prospective data-warehousing project using data-mining and artificial intelligence promoted by Brest University Hospital and LATIM INSERM UMR 1101, the present study was conducted to determine the respective accuracy of either NEWS, EWSO2, or HRV parameters in terms of outcome prediction and respiratory assistance requirement for spontaneously breathing patients attending the ICU.

Study design and population

Data analysis was retrospectively performed on the ReaSTOC database, which is an ongoing prospective physiological and biomedical signals datawarehousing project, including all consecutive patients admitted to our adult medical ICU (ClinicalTrials.gov identifier NCT02893462). A previous publication have described the design and conduct of the ReaSTOC study [15]. The protocol was approved by our local ethics committee (Comité d’Ethique du CHRU de Brest 2018CE.27) and written informed consent was waived according to French legislation. All physiological data were collected by the same research assistant, using a standardized protocol. Clinical metadata and outcomes were prospectively recorded using standard monitoring devices and the computerized medical file (Intellivue MP70® and ICCA® Philips Healthcare, Amsterdam, Netherlands). No specific intervention or interventional measures was required during data recording.

All spontaneously breathing patients aged 18 years or older admitted within our ICU were considered eligible. if they were not requiring invasive mechanical ventilation on admission. The only exclusion criteria were the use of invasive mechanical ventilation on admission and the patient and/or relatives’ refusal to participate.

Measurements

General data recording

Each raw physiological PPG plot was recorded at native resolution (75 Hz) during 2 consecutive hours, within the first 24h following ICU admission, using a private data network and a dedicated software (Synapse, LTSI INSERM UMR 1099). All measurements were performed with the patients lying in a supine position, with a 30° angulation of the upper part of the body.

Clinical metadata were collected prospectively and included general and physiological parameters that may interfere with HRV measurement (such as sedation, chronic cardiac state, and medications) to assess internal validity. Outcome parameters were also collected prospectively and included length-of-stay (LOS), either in the ICU, in the hospital and at Day-28.

Hemodynamic instability was defined as the need for a fluid challenge (≥ 0.5L crystalloids) and/or introduction or modification of vasopressive agents within the past 2h of care.

HRV features extraction

According to previous publication using the same database [15], HRV measurements were performed on PPG. All tracings were analyzed using Kubios HRV Premium 3.0 (Kuopio, Finland), using standard methods. For each marker, we considered the average value of 3 data samples of at least 3 consecutive minutes, over the two hours of recording.

Time based Analysis

Time-based analysis was performed using statistical and geometrical methods. Parameters to be analysed were the R-R intervals, Standard Deviation of R-R intervals (SDNN), Root Mean Square of Successive Differences (RMSSD), the proportion of R-R intervals differing from 50ms divided by total number of R-R (pNN50) and the triangular interpolation of R-R intervals histogram (Triangular Index).

Frequency-based Analysis

Spectral analysis using the discrete Fourier transform identified powers of high frequency (HF 0.15Hz to 0.40Hz), low frequency (LF 0.04Hz to 0.15Hz), and very-low frequency (VLF 0.0033 to 0.4 Hz) bands [17]. LF/HF ratio was also calculated.

Non-linear Analysis

Non-linear chaotic evaluation method was performed using the Poincaré plotting, thus providing the standard deviation of instant beat-to-beat time variation (SD1) and the standard deviation of long-term beat-to-beat time variation (SD2) [16].

Several entropy parameters such as Approximate Entropy [18], Sample Entropy, and Shannon Entropy were also measured [19].

Early Warning Scoring systems (NEWS and EWSO2)

All data used to calculate early warning scores were extracted from the ReaSTOC database. To calculate NEWS, instead mental status determined using the alert, verbal, painful, unresponsive (APVU) scale, we used the GCS as it is routinely performed in most ICUs. An APVU score of 3 was attributed when GCS was ≤ 12. EWSO₂ was defined as the ratio of respiratory rate over SpO₂/FIO₂ (23). It has been proposed to be equivalent or superior to common early warning scores to predict poor outcome at first medical contact, while focusing on simple clinical and respiratory parameters. While included patients were spontaneous breathing, FiO₂ was estimated from the oxygen flow rate and determined according to the following equation FIO₂ = 4xQO₂ + 21% (27). For methodological reasons related to the unavailability of an exact value within the database, we chose to use a simplified EWSO₂ version, not considering the influence of the heart rate.

Outcomes

All outcome parameters were collected prospectively, and mortality was assessed either within the ICU, the entire hospital stay, and at Day-28. The need for invasive MV, or any form of respiratory assistance (ie. Invasive MV, NIV or HFNC) were collected over the entire ICU stay. We used a composite prognosis index for a bad outcome, which was considered as the occurrence of at least one of the following criteria: death, and/or the need for any form of respiratory assistance during the ICU stay, and/or an ICU length-of-stay higher than 3-days.

Statistical analysis

Continuous variables were expressed as mean ± SD in case of normal distribution and median (interquartile range (IQR)) otherwise; categorical variables were expressed as numbers and percentages. Univariate analysis was performed using a non-parametric Kruskall–Wallis test for quantitative and Fisher’s exact test for qualitative datasets.

Independent outcome predictors were identified by logistic regression using a stepwise model; performance of these predictors for a specific cut-off value was determined while calculating the area under the receiving operating characteristics curve (AUC for ROC). In all cases, we considered a minimal 60% sensitivity and specificity to select the threshold value. All variables with a P value < 0.25 within the univariate analysis were included within the multivariate analysis.

All analysis were performed using R++ (v1.5.03; Zebrys, Toulouse, France). A p-value equal or less than 0.05 was considered statistically significant.

Between May 2014 to April 2019, 500 patients were included within the ReaSTOC database. 388 patients were considered ineligible based on the exclusion criteria (317 patients under MV, 71 patients without registration during the first 24 hours of the ICU stay). Consequently, 112 patients were included in the study (Fig. 1).

Patients’ characteristics

Patients’ physiological and pathological characteristics on admission are provided within Table 1. Amongst the 112 patients, 94 patients (83.9%) were breathing room air oxygen, 11 patients (9.8%) were treated with high flow nasal cannula (HNFC), and 7 patients (6.3%) were receiving noninvasive ventilation (NIV). HRV and physiological scores values are provided within Tables 2 and 3.

Table 1

Patients’ Baseline Characteristics
Patients (n = 112)
Physiological characteristics
Age (years)	63,5 ± 14,0–66,0 [60,9–66,1]
Sex	36 Female (32,1%) / 76 Male
BMI (kg/m²)	25,4 ± 8,3–25,8 [23, 7–26, 6]
SAPS II	35,8 ± 15,2–35,0 [30, 2–37, 8]
Glasgow Coma Scale	14,3 ± 1,7–15,0 [15,0–15,0]
Temperature (°C)	36,9 ± 1,0–36,7 [36,3–37,0]
Heart rate (c/mn)	89,9 ± 21,3–88,0 [83,7–92,0]
Mean arterial pressure (mm Hg)	82,3 ± 16,4–82,0 [78,0–86,0]
Respiratory rate (c/mn)	22,1 ± 6,4–21,0 [19,0–23,0]
SpO₂ (%)	94,6 ± 4,0–95,0 [94,0–95,0]
Pathological characteristics
Actual ICU admission cause
Respiratory failure – n (%)	46 (41,1)
Cardiovasculary failure – n (%)	32 (28,6)
Neurological failure – n (%)	14 (12,5)
Others – n (%)	20 (17,9)
Associated Pathological Status
Unstable hemodynamics – n (%)	10 (8,9)
Atrial Fibrillation or Sustained Arrythmia – n (%)	16 (14,3)
Dialysis– n (%)	34 (30,4)
Immunosupression – n (%)	10 (8,9)
Cardiac – n (%)	61 (54,5)
Respiratory – n (%)	46 (41,1)
Respiratory Support on ICU Admission
Spontaneous ventilation – n (%)	94 (83,9)
High Nasal Flow Oxygen – n (%)	11 (9,8)
Non invasive ventilation – n (%)	7 (6,3)
O2 flow (L/mn), excluding High Nasal Flow Oxygen	1,2 ± 2,0–0 [0–0,7]
Outcome
Respiratory Support during ICU Stay
Spontaneous ventilation – n (%)	71 (63,4)
High Nasal Flow Oxygen – n (%)	12 (10,7)
Non invasive ventilation – n (%)	23 (20,5)
Invasive ventilation – n (%)	6 (5,4)
Length of Stay
ICU LOS(jours)	4,7 ± 5,7–3,0 [3,0–4,0]
Hospital LOS (jours)	21,6 ± 23,2–15,0 [11,4–20,0]
Respiratory assistance (jours)	11,0 ± 12,9–6,5 [3, 2–31, 6]
Death
ICU – n (%)	8 (7,1)
Day 28 – n (%)	14 (12,5)
Hospital - n (%)	18 (16,1)
Results are provided as mean ± SD – median [IQR] or n (%)
BMI: Body Mass Index; SAPS II: Simplified Acute Physiological Score version 2; SpO2:pulse oximetry; ICU: Intensive Care Unit; Day 28: outcome parameter measured 28 days after admission.

Table 2

NEWS, EWSO2 and Heart Rate Variability comparison according to the need for any form of respiratory assistance
	All (n = 112)	Respiratory assistance (n = 37)	No respiratory assistance (n = 75)	P value
SAPS II	35,0 [30, 2–37, 8]	35.0 [29.7–40.0]	33.0 [29.0-39.8]	0.9
EWSO2	5,6 [5, 1–6, 5]	7.6 [5.8–10.3]	5.1 [4.4–5.8]	0.002
NEWS	6,0 [5,0–6,5]	8.0 [6.0-8.9]	5.0 [4.0–6.0]	0.01
Heart Rate Variability
Time Domain
SDNN (ms)	39,7 [90,1–61,8]	51.8 [27.9–95.7]	37.9 [25.9–54.2]	0.9
RMSSD (ms)	43,9 [32,0–57,9]	52.1 [35.1–96.6]	39.5 [26.5–59.4]	0.9
pNN50 (%)	12,3 [6, 5–19, 9]	20.3 [8.3–29.7]	7.5 [3.7–17.2]	0.5
Triangular Index	6,6 [5, 6–7, 8]	8.0 [5.0-10.6]	6.2 [5.0-7.1]	0.9
Frequency Domain
VLF (0.0-0.04 Hz) - (ms²)	129 [91–257]	202 [55–511]	115 [68–249]	0.4
LF (0.04–0.15 Hz) - (ms²)	1168 [526–2626]	1508 [474–5652]	1135 [447–2130]	0.5
HF (0.15–0.4 Hz) - (ms²)	743 [408–2087]	938 [456–7036]	636 [244–1889]	0.3
LF/HF	1,1 [1,0–1,4]	1.1 [0.8–1.2]	1.2 [1.0-1.6]	0.2
Non Linear Domain
Entropy
Approximate entropy	1,1 [1, 1–1, 2]	1.1 [1.1–1.3]	1.2 [1.1–1.2]	0.4
Sample entropy	1,4 [1, 3–1, 5]	1.5 [1.1–1.7]	1.3 [1.2–1.5]	0.4
Shanon entropy	3,3 [3, 2–3, 4]	3.2 [3.1–3.4]	3.4 [3.2–3.5]	0.3
Poincaré Analysis
SD1 (ms)	32,9 [23, 4–43, 5]	36.9 [24.8–68.4]	29.1 [19.2–50.1]	1.0
SD2 (ms)	46,3 [39,0–70,4]	58.9 [32.4-114.3]	44.9 [32.4–70.4]	0.5
SD2/SD1	1,4 [1, 4–1, 6]	1.3 [1.2–1.4]	1.5 [1.4–1.6]	0.02
Results are provided as median [IQR]. A P value equal or below 0.05 was considered statistically significant (in bold character).
SAPS II: Simplified Acute Physiological Score version 2; EWSO2: Early Warning Score Oxygen; NEWS: National Early Warning Score; SDNN: Standard deviation of NN intervals; RMSSD: Root mean square of successive RR interval differences; pNN50: Percentage of successive RR intervals that differ by more than 50 ms; Triangular Index: Integral of the density of the RR interval histogram divided by its height; VLF: Absolute power of the very low frequency; LF: Absolute power of the low frequency; HF: Absolute power of the high frequency; LF/HF: low over high frequency ratio; SD1: Poincaré plot standard deviation perpendicular the line of identity; SD2: Poincaré plot standard deviation along the line of identity; SD2/SD1: SD2 over SD1 ratio.

Table 3

NEWS, EWSO2 and Heart Rate Variability comparison according to ICU survival
	All (n = 112)	Death (n = 8)	Survival (n = 104)	P value
IGS II	35,0 [30, 2–37, 8]	48,0 [41,1–70,4]	33,0 [29,0–36,0]	0,001
Predicted Mortality	16.7%
NEWS	6,0 [5,0–6,5]	6,0 [5,0–7,0]	9,0 [5,0–11,6]	0,03
EWSO2	5,6 [5, 1–6, 5]	18,4 [15, 7–19, 8]	9,0 [5, 8–20, 9]	0,008
Heart Rate Variability
Time Domain
SDNN (ms)	39,7 [90,1–61,8]	60,1 [8,5–190,4]	39,7 [31,0–57,9]	0,65
RMSSD (ms)	43,9 [32,0–57,9]	60,3 [11,0–218,6]	43,9 [32,4–57,5]	0,77
pNN50 (%)	12,3 [6, 5–19, 9]	11,3 [0,1–74,4]	12,3 [6, 8–19, 6]	0,67
Triangular Index	6,6 [5, 6–7, 8]	5,1 [2, 8–28, 4]	6,6 [5,7–8,0]	0,73
Frequency Domain
VLF (0.0-0.04 Hz) - (ms²)	129 [91–257]	201 [2–5265]	129 [97–255]	0,56
LF (0.04–0.15 Hz) - (ms²)	1168 [526–2626]	2924 [15–29632]	1168 [560–2342]	0,50
HF (0.15–0.4 Hz) - (ms²)	743 [408–2087]	3485 [14–26867]	743 [419–2028]	0,58
LF/HF	1,1 [1,0–1,4]	0,9 [0,7–1,3]	1,1 [1,0–1,4]	0,25
Non Linear Domain
Entropy
Approximate entropy	1,1 [1, 1–1, 2]	1,1 [0,9–1,3]	1,1 [1, 1–1, 2]	0,74
Sample entropy	1,4 [1, 3–1, 5]	1,8 [1,2–2,0]	1,4 [1, 2–1, 5]	0,04
Shanon entropy	3,3 [3, 2–3, 4]	2,8 [2, 5–3, 6]	3,3 [3, 2–3, 4]	0,03
Poincaré Analysis
SD1 (ms)	32,9 [23, 4–43, 5]	42,7 [7,8–154,7]	32,9 [23, 8–42, 2]	0,75
SD2 (ms)	46,3 [39,0–70,4]	71,0 [8,1–220,1]	46,3 [39,6–70,4]	0,53
SD2/SD1	1,4 [1, 4–1, 6]	1,3 [1,0–1,6]	1,4 [1, 4–1, 5]	0,12
Results are provided as median [IQR]. A P value equal or below 0.05 was considered statistically significant (in bold character).
SAPS II: Simplified Acute Physiological Score version 2; EWSO2: Early Warning Score Oxygen; NEWS: National Early Warning Score; SDNN: Standard deviation of NN intervals; RMSSD: Root mean square of successive RR interval differences; pNN50: Percentage of successive RR intervals that differ by more than 50 ms; Triangular Index: Integral of the density of the RR interval histogram divided by its height; VLF: Absolute power of the very low frequency; LF: Absolute power of the low frequency; HF: Absolute power of the high frequency; LF/HF: low over high frequency ratio; SD1: Poincaré plot standard deviation perpendicular the line of identity; SD2: Poincaré plot standard deviation along the line of identity; SD2/SD1: SD2 over SD1 ratio.

Outcomes

Endotracheal intubation and invasive mechanical ventilation were required for 6/112 patients (5.4%). EWSO2 and NEWS were significantly higher for patients requiring any form of respiratory assistance, and the SD2/SD1 ratio was significantly lower (Table 2 and Table S-1).

Eight over 112 patients (7.1%) died within the ICU, and 28-day and hospital mortality were respectively 12.5% and 16.1%. NEWS and EWSO2 scores were higher for patients that died within the ICU, as compared with survivors (Table 3). When considering hospital mortality, EWSO2 index was higher in the group of patients who died during their hospital stay. SD2/SD1 was lower on admission for patients deemed to die during their hospital stay (Table S-2).

When considering the composite outcome parameter, higher NEWS, EWSO2, sample entropy and SD2/SD1 values were associated with a better outcome (Table S-3).

Multivariate Analysis and Predictive Models

Independent predictors for ICU and hospital survival, need for further respiratory assistance are provided within Table 4.

Table 4

Independent predictors for ICU survival and need for further respiratory assistance
	Cut-off	Sensibility	Specificity	AUC	P value
ICU Survival
NEWS	8	77.7	62.5	0.73	0.005
Shanon	2.9	72.9	75	0.74	0.04
Need for respiratory assistance
EWSO2	6.4	64.7	70.6	0.72	< 0.001
NEWS	5.3	65.8	58.2	0.65	0.004
SD2/SD1	1.4	63.9	61.1	0.65	0.007
Results are provided exact value for Cut-off and AUC and as percentage for Sensibility and Specificity. A P value equal or below 0.05 was considered statistically significant (in bold character).
AUC: Area under the curve; ICU:Intensive care unit; NEWS: National early warning score; Shanon: Shanon entropy; EWSO2: Early warning score oxygen; SD2/SD1: SD2 Poincaré plot over SD1 Poincaré plot ratio.

NEWS and Shanon entropy were considered as independent predictors for ICU survival. EWSO2, NEWS and SD2/SD1 were considered as independent predictors for the need of respiratory assistance.

EWSO2 and SD2/SD1 ratio were considered as independent predictors for hospital survival (p = 0.006).

A first model combining EWSO2 and SD2/SD1 enabled to predict the need for any form of respiratory assistance (AUC = 0.63, p = 0.006) and ICU survival (AUC = 0.61, p = 0.02). A second model combining EWSO2, SD2/SD1 and Shanon entropy enabled to predict the need for any form of respiratory assistance (AUC = 0.62, p = 0.01) and ICU survival (AUC = 0.72, p = 0.02).

In this warehousing study including data that were prospectively acquired from 112 spontaneously breathing patients attending the ICU, EWSO2 and HRV on admission were considered as independent predictors for the need of respiratory assistance. NEWS score and EWSO2 index were respectively independent predictors for ICU survival or hospital survival, while HRV (Shanon entropy and SD2/SD1) was an independent predictor for both. When considering the need for respiratory assistance prediction on admission, the best model determined by AI included EWSO2 and SD2/SD1. To predict ICU survival, the best model was the one combining EWSO2, SD2/SD1 and Shannon entropy.

Many Early Warning Scores have been developed to detect patients at risk of poor outcome within emergency departments [28–30]. The choice of the NEWS as a comparator within our study was performed while it seemed to be best discriminating scoring system for patients at risk of clinical deterioration and acute mortality [8]. Furthermore, the NEWS has been proposed for use in various acute medical situations like sepsis, acute trauma, surgical patients, and also within the ICU for a lot of various critical illnesses. NEWS AUC for outcome prediction tended to be lower within our study database, as compared to previous studies [8, 31], with a best cut-off prediction value for ICU mortality prediction equal or higher than 8. The Royal College of Physicians recommended an assessment by a critical care team and usually a transfer in ICU, when the NEWS is superior or equal to 7 [9]. In a recent study, NEWS was also used in an empirical plan to discriminate patients according to discharge location, and its performance seemed to be adequate when used within various ICU wards [14].

In the study by Roca et al. using the ROX, all included patients had a pneumonia and presented clinical signs of acute respiratory failure, requiring respiratory assistance [22]. Within our study population, Beside the fact that all were spontaneously breathing, the physiological characteristics of our patients were quite heterogeneous, thus representing a more realistic ICU admission panel. Most of them were spontaneously breathing with and without oxygen, 9.8% were treated with HNFC, 6.3% were treated using NIV. When considering the use of a respiratory score as an outcome predictor, we were interested in its use among a heterogenous patient population with various pathological conditions. While several authors depicted that the ROX index may help detect patients deemed to succeed HFNC [32], or those deemed to be successfully separated from the assistance, such parameter has not yet been proposed as an outcome predictor in the early clinical phase following ICU admission in spontaneously breathing patients. EWSO2 was initially developed by our team as an outcome predictor for emergency department patients, without any form of respiratory assistance [23]. If EWSO2 and ROX concepts and objectives remains quite similar, nevertheless they differ on several other methodological aspects: while EWSO2 uses a FIO2 estimation under standard oxygen administration, ROX uses the FIO2 measurements or estimates derived from the HFNC devices.

Similarly to previous studies [33], HRV on admission significantly differs in between patients who will die, or those that will survive within the ICU. LH/LF ratio was inversely associated with hospital mortality. This result was consistent with others studies in which the authors shown a lower LH/LF ratio who died or required invasive critical care support [34, 35]. In the last decades, many studies had shown a link between autonomous nervous system and HRV [24, 25]. The variation of HRV parameters was exemplify of the balance between sympathetic and vagal system. The usefulness of HRV in anesthetic and resuscitative care is wide since it has been evaluated both in physiological and pathological situations. Several studies have shown an interest of this measure in the pain or anesthesia management [36–38]. In various pathological situations HRV was used as a predictor of poor outcomes or mortality especially in case of sepsis, septic shock, multiple organ dysfunction syndrome [35, 39–41] or stroke [26]. With the goal of establishing a successful model, a recent study has proposed the use of HRV with biological variables with better results than HRV alone to predict deterioration in patients with sepsis [42]. HRV measures were performed through time-based, spectral and nonlinear Poincaré analyses and even with signal entropy evaluations, indeed, there are many HRV evaluation parameters with different measurements and there are no reference ranges of different HRV parameters clearly described. Moreover, each parameter analyses differently and more or less accurately the sympathetic or the parasympathetic activity. This is the main reason why we did consider all HRV parameters, while promoting frequency and nonlinear domains.

To the best of our knowledge, this data-warehousing study using artificial intelligence is the first to evaluate the combination of respiratory scores and HRV as ICU predictors. While NEWS was initially described to determine clinical responses and any escalation of care [9] but when considering the required number of clinical parameters to calculate NEWS, as compared to the relative equivalence of a more simplistic tool like EWSO2, we chose not to include it within our predictive model derivation process. If confirmed clinically, such simple and dynamic models may help clinicians better discriminate patients that will benefit from aggressive respiratory care at the early phase of ICU admission.

Some limitations of our study may be discussed. First, the relatively small number of patients and the monocentric characteristics of the study, and thus it’s applicability to other centers, may be considered as its main limitation. Second, data collection was performed for 2 hours within the first 24h following admission, and they were not reassessed afterward. However, the aim of the study was to find predictor of deterioration at patient admissions. Third, some physiological or clinical conditions such as congestive heart failure, diabetes, medications that may affect HRV measures were not evaluated [43]. For this reason, we chose to promote frequency and nonlinear measurements domains over time domains HRV evaluations [44].

In conclusion, our results suggest that either NEWS, EWSO2 and HRV enables outcome prediction in spontaneously breathing ICU patients. Predictive models combining EWSO2 and HRV parameters developed using AI may predict the need for any form of respiratory assistance and ICU survival. These results might require validation of the models in a prospective and multicenter study.

Funding

he authors declare that no funds, grants, or other support were received during the preparation of this manuscript.

Competing Interests

The authors have no relevant financial or non-financial interests to disclose.

Author Contributions

NF, LB and KdL collected data, have read and approved the manuscript. NF and ELH conceived the study, wrote the manuscript. ELH analyzed data. FL have read, corrected and approved the final version of the manuscript.

Ethics approval and consent to participate

The protocol was approved by our local ethics committee (Comité d’Ethique du CHRU de Brest 2018CE.27) and written informed consent was waived according to French legislation.

Information was provided to all patients and their relatives, enabling them to retrieve their consent to include data in the analysis.

Taboulet P, Moreira V, Haas L, Porcher R, Braganca A, Fontaine J-P, Poncet MC (2009) Triage with the French Emergency Nurses Classification in Hospital scale: reliability and validity. Eur J Emerg Med 16:61‑67. https://doi: 10.1097/MEJ.0b013e328304ae57.
Kuriyama A, Urushidani S, Nakayama T (2017) Five-level emergency triage systems: variation in assessment of validity. Emerg Med J 34:703‑710. https:// doi: 10.1136/emermed-2016-206295.
Churpek MM, Snyder A, Han X, Sokol S, Pettit N, Howell MD, Edelson DP (2017) Quick Sepsis-related Organ Failure Assessment, Systemic Inflammatory Response Syndrome, and Early Warning Scores for Detecting Clinical Deterioration in Infected Patients outside the Intensive Care Unit. Am J Respir Crit Care Med 195:906‑911. https:// doi: 10.1164/rccm.201604-0854OC.
Alam N, Vegting IL, Houben E, van Berkel B, Vaughan L, Kramer MHH, Nanayakkara PWB (2015) Exploring the performance of the National Early Warning Score (NEWS) in a European emergency department. Resuscitation 90:111‑115. https:// doi:10.1016/j.resuscitation.2015.02.011.
Goulden R, Hoyle M-C, Monis J, Railton D, Riley V, Martin P, Martina R, Nsutebu E (2018) qSOFA, SIRS and NEWS for predicting inhospital mortality and ICU admission in emergency admissions treated as sepsis. Emerg Med J 35:345‑349. https:// doi: 10.1136/emermed-2017-207120.
Akre M, Finkelstein M, Erickson M, Liu M, Vanderbilt L, Billman G (2010) Sensitivity of the pediatric early warning score to identify patient deterioration. Pediatrics 125:e763-769. https:// doi: 10.1542/peds.2009-0338.
Singh S, McGlennan A, England A, Simons R (2012) A validation study of the CEMACH recommended modified early obstetric warning system (MEOWS). Anaesthesia 67:12‑18. https:// doi: 10.1111/j.1365-2044.2011.06896.x.
Smith GB, Prytherch DR, Meredith P, Schmidt PE, Featherstone PI (2013) The ability of the National Early Warning Score (NEWS) to discriminate patients at risk of early cardiac arrest, unanticipated intensive care unit admission, and death. Resuscitation 84:465‑470. https:// doi: 10.1016/j.resuscitation.2012.12.016.
National Early Warning Score (NEWS) 2 [Internet]. RCP London. 2017. available on https://www.rcplondon.ac.uk/projects/outputs/national-early-warning-score-news-2
Malycha J, Farajidavar N, Pimentel MAF, Redfern O, Clifton DA, Tarassenko L, Meredith P, Prytherch D, Ludbrook G, Duncan young J, Watkinson PJ (2019) The effect of fractional inspired oxygen concentration on early warning score performance: A database analysis. Resuscitation 2019;139:192‑199. https:// doi: 10.1016/j.resuscitation.2019.04.002.
Nickel CH, Kellett J, Cooksley T, Bingisser R, Henriksen DP, Brabrand M (2016) Combined use of the National Early Warning Score and D-dimer levels to predict 30-day and 365-day mortality in medical patients. Resuscitation 106:49‑52. https:// doi: 10.1016/j.resuscitation.2016.06.012.
Jo S, Jeong T, Lee JB, Jin Y, Yoon J, Park B (2016) Validation of modified early warning score using serum lactate level in community-acquired pneumonia patients. The National Early Warning Score-Lactate score. Am J Emerg Med 34:536‑541. https:// doi: 10.1016/j.ajem.2015.12.067.
Vincent JL, Moreno R, Takala J, Willatts S, De Mendonça A, Bruining H, Reinhart CK, Suter PM, Thijs LG (1996) The SOFA (Sepsis-related Organ Failure Assessment) score to describe organ dysfunction/failure. On behalf of the Working Group on Sepsis-Related Problems of the European Society of Intensive Care Medicine. Intensive Care Med 22:707‑710. https:// doi: 10.1007/BF01709751.
Zaidi H, Bader-El-Den M, McNicholas J. Using the National Early Warning Score (NEWS/NEWS 2) in different Intensive Care Units (ICUs) to predict the discharge location of patients. BMC Public Health 19:1231. https:// doi: 10.1186/s12889-019-7541-3.
Frat J-P, Ragot S, Coudroy R, Constantin J-M, Girault C, Prat G, et al (2018) Predictors of Intubation in Patients With Acute Hypoxemic Respiratory Failure Treated With a Noninvasive Oxygenation Strategy. Crit Care Med 46:208‑215. https:// doi: 10.1097/CCM.0000000000002818.
Fieselmann JF, Hendryx MS, Helms CM, Wakefield DS (1993) Respiratory rate predicts cardiopulmonary arrest for internal medicine inpatients. J Gen Intern Med 8:354‑360. https:// doi: 10.1007/BF02600071.
Goldhill DR, McNarry AF, Mandersloot G, McGinley A (2005) A physiologically-based early warning score for ward patients: the association between score and outcome. Anaesthesia 60:547‑553. https:// doi: 10.1111/j.1365-2044.2005.04186.x.
Subbe CP, Davies RG, Williams E, Rutherford P, Gemmell L (2003) Effect of introducing the Modified Early Warning score on clinical outcomes, cardio-pulmonary arrests and intensive care utilisation in acute medical admissions. Anaesthesia 58:797‑802. https:// doi: 10.1046/j.1365-2044.2003.03258.x.
Rice TW, Wheeler AP, Bernard GR, Hayden DL, Schoenfeld DA, Ware LB, National Institutes of Health, National Heart, Lung, and Blood Institute ARDS Network (2007) Comparison of the SpO2/FIO2 ratio and the PaO2/FIO2 ratio in patients with acute lung injury or ARDS. Chest 132:410‑417. https:// doi: 10.1378/chest.07-0617.
Kwack WG, Lee DS, Min H, Choi YY, Yun M, Kim Y, et al (2018) Evaluation of the SpO2/FiO2 ratio as a predictor of intensive care unit transfers in respiratory ward patients for whom the rapid response system has been activated. PloS One. 2018;13(7):e0201632. https:// doi: 10.1371/journal.pone.0201632.
DesPrez K, McNeil JB, Wang C, Bastarache JA, Shaver CM, Ware LB (2017) Oxygenation Saturation Index Predicts Clinical Outcomes in ARDS. Chest 152:1151‑1158. https:// doi: 10.1016/j.chest.2017.08.002.
Roca O, Caralt B, Messika J, Samper M, Sztrymf B, Hernández G, Garcia-de-Acilu M, Frat JP, Masclans JR, Ricard JD (2019) An Index Combining Respiratory Rate and Oxygenation to Predict Outcome of Nasal High-Flow Therapy. Am J Respir Crit Care Med 199:1368‑1376. https:// doi: 10.1164/rccm.201803-0589OC.
Viglino D, L'her E, Maltais F,Maignan M, Lellouche F (2020) Evaluation of a new respiratory monitoring tool "Early Warning ScoreO2" for patients admitted at the emergency department with dyspnea. Resuscitation 148:59–65. https://doi: 10.1016/j.resuscitation.2020.01.004
Cysarz D, Lange S, Matthiessen PF, van Leeuwen P (2007) Regular heartbeat dynamics are associated with cardiac health. Am J Physiol Regul Integr Comp Physiol 292:R368-372. https:// doi: 10.1152/ajpregu.00161.2006.
Lombardi F, Stein PK (2011) Origin of heart rate variability and turbulence: an appraisal of autonomic modulation of cardiovascular function. Front Physiol 2:95. https:// doi: 10.3389/fphys.2011.00095.
Tang S-C, Jen H-I, Lin Y-H, Hung C-S, Jou W-J, Huang P-W, et al (2015) Complexity of heart rate variability predicts outcome in intensive care unit admitted patients with acute stroke. J Neurol Neurosurg Psychiatry 86:95‑100. https:// doi: 10.1136/jnnp-2014-308389.
Wettstein RB, Shelledy DC, Peters JI (2005) Delivered oxygen concentrations using low-flow and high-flow nasal cannulas. Respir Care 50:604-609. https:// doi: 10.1136/jnnp-2014-308389.
Forster S, Housley G, McKeever TM, Shaw DE (2018) Investigating the discriminative value of Early Warning Scores in patients with respiratory disease using a retrospective cohort analysis of admissions to Nottingham University Hospitals Trust over a 2-year period. BMJ Open 8:e020269. https:// doi: 10.1136/bmjopen-2017-020269.
Dundar ZD, Ergin M, Karamercan MA, Ayranci K, Colak T, Tuncar A, Cander B, Gul M (2016) Modified Early Warning Score and VitalPac Early Warning Score in geriatric patients admitted to emergency department. Eur J Emerg Med 23:406-412. doi: 10.1097/MEJ.0000000000000274.
Nannan Panday RS, Minderhoud TC, Alam N, Nanayakkara PWB (2017) Prognostic value of early warning scores in the emergency department (ED) and acute medical unit (AMU): A narrative review. Eur J Intern Med 45:20-31.https://doi: 10.1016/j.ejim.2017.09.027.
Smith GB, Prytherch DR, Jarvis S, Kovacs C, Meredith P, Schmidt PE, Briggs J (2016) Comparison of the Ability of the Physiologic Components of Medical Emergency Team Criteria and the U.K. National Early Warning Score to Discriminate Patients at Risk of a Range of Adverse Clinical Outcomes. Crit Care Med 44:2171-2181. https://doi: 10.1097/CCM.0000000000002000.
Rodriguez M, Thille AW, Boissier F, Veinstein A, Chatellier D, Robert R, Le Pape S, Frat JP, Coudroy R (2019) Predictors of successful separation from high-flow nasal oxygen therapy in patients with acute respiratory failure: a retrospective monocenter study. Ann Intensive Care 9:101. https://doi: 10.1186/s13613-019-0578-8.S
Bodenes L, N'Guyen QT, Le Mao R, Ferrière N, Pateau V, Lellouche F, L'Her E (2022) Early heart rate variability evaluation enables to predict ICU patients' outcome. Sci Rep 12:2498. https://doi: 10.1038/s41598-022-06301-9.
Barnaby D, Ferrick K, Kaplan DT, Shah S, Bijur P, Gallagher EJ (2002) Heart rate variability in emergency department patients with sepsis. Acad Emerg Med 9:661‑670. https://doi: 10.1111/j.1553-2712.2002.tb02143.x.
de Castilho FM, Ribeiro ALP, da Silva JLP, Nobre V, de Sousa MR (2017) Heart rate variability as predictor of mortality in sepsis: A prospective cohort study. PloS One 12:e0180060. https:// doi: 10.1371/journal.pone.0180060.
Papaioannou V, Chouvarda I, Gaertner E, Benyamina M, Ferry A, Maurel V, Soussi S, Blet A, Chaouat M, Plaud B, Mebazaa A, Legrand M; PRONOBURN Group (2016) Heart rate variability and cardiac baroreflex inhibition-derived index predicts pain perception in burn patients. Burns 42:1445‑1454. https://doi: 10.1016/j.burns.2016.04.017.
Broucqsault-Dédrie C, De Jonckheere J, Jeanne M, Nseir S (2016) Measurement of Heart Rate Variability to Assess Pain in Sedated Critically Ill Patients: A Prospective Observational Study. PloS One 11:e0147720. https://doi: 10.1371/journal.pone.0147720.
Nagaraj SB, Biswal S, Boyle EJ, Zhou DW, McClain LM, Bajwa EK, Quraishi SA, Akeju O, Barbieri R, Purdon PL, Westover MB (2017) Patient-Specific Classification of ICU Sedation Levels From Heart Rate Variability. Crit Care Med 45:e683‑90. https://doi: 10.1097/CCM.0000000000002364.
Annane D, Trabold F, Sharshar T, Jarrin I, Blanc AS, Raphael JC, Gajdos P (1999) Inappropriate sympathetic activation at onset of septic shock: a spectral analysis approach. Am J Respir Crit Care Med 160:458‑465. https:// doi: 10.1164/ajrccm.160.2.9810073.
Chen W-L, Kuo C-D (2007) Characteristics of heart rate variability can predict impending septic shock in emergency department patients with sepsis. Acad Emerg Med 14:392‑397. https://doi: 10.1197/j.aem.2006.12.015.
Schmidt H, Hoyer D, Hennen R, Heinroth K, Rauchhaus M, Prondzinsky R, Hottenrott K, Buerke M, Müller-Werdan U, Werdan K (2008) Autonomic dysfunction predicts both 1- and 2-month mortality in middle-aged patients with multiple organ dysfunction syndrome. Crit Care Med 36:967‑970. https://doi: 10.1097/CCM.0B013E3181653263.
Barnaby DP, Fernando SM, Herry CL, Scales NB, Gallagher EJ, Seely AJE (2019) Heart Rate Variability, Clinical and Laboratory Measures to Predict Future Deterioration in Patients Presenting With Sepsis. Shock 51:416‑422. https://doi: 10.1097/SHK.0000000000001192.
Task Force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology (1996) Heart rate variability: standards of measurement, physiological interpretation and clinical use. Circulation 93:1043‑1065.
Singh N, Moneghetti KJ, Christle JW, Hadley D, Plews D, Froelicher V (2018) Heart Rate Variability: An Old Metric with New Meaning in the Era of using mHealth Technologies for Health and Exercise Training Guidance. Part One: Physiology and Methods. Arrhythmia Electrophysiol Rev 7:193‑198. https:// doi: 10.15420/aer.2018.27.2.

No competing interests reported.

SupplementaryAppendixEWSO2FinalELH26092023.docx

Download PDF

Version 1

posted

You are reading this latest preprint version

Early outcome prediction (NEWS, EWSO2 and HRV) for spontaneously breathing ICU patients - A data-warehousing analysis

Status:

Version 1

Abstract

Purpose

Methods

Results

Conclusion

Figures

INTRODUCTION

METHODS

Study design and population

Measurements

General data recording

HRV features extraction

Time based Analysis

Frequency-based Analysis

Non-linear Analysis

Early Warning Scoring systems (NEWS and EWSO2)

Outcomes

Statistical analysis

RESULTS

Patients’ characteristics

Outcomes

Multivariate Analysis and Predictive Models

DISCUSSION

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1