Appendix

Appendix

Contents

Data tables

Table 1: Summary of the Ben‑Ari et al. (2010) study

Table 2: Summary of the Sorkine et al. (2008) abstract

Table 3: Summary of the Frendl et al. (2013) abstract

Table 4: Summary of the Zimlichman et al. (2012) study

Table 5: Summary of the Brown et al. (2014) study

Table 6: Summary of the Zimlichman et al. (2009) company report

Table 1 Summary of the Ben‑Ari et al. (2010) study

Study component

Description

Objectives hypotheses

To study the accuracy of the EarlySense system for heart and respiratory rate measurements.

Study design

Non‑controlled case series.

Intervention

EarlySense.

Setting

Sleep laboratory and ICU settings.

Inclusion/exclusion criteria

Sleep laboratory inclusion:

Adults (age above 18 years) referred to the sleep laboratory for any indication and who were willing to sign an informed consent. Children (aged 4–18 years) referred to the sleep laboratory for any indication and consented by legal guardian. Exclusions not stated.

ICU inclusion:

Hospitalisation in critical care unit and consent obtained from patients or their next of kin (for intubated and ventilated patients).

ICU exclusion:

Missing or compromised data (due to noise from external sources).

Primary outcomes

Respiratory rate and heart rate.

Methods

Sleep laboratory:

Reference standards: RR ‑ Embla N7000 with Somnologica Studio Software System (Embla Systems Inc. Iceland). HR ‑ Embla Sleep Lab System. RR and HR were measured simultaneously using the reference device and the EarlySense for the entire night (21:00–06:00 hours).

ICU:

HR measured with standard ECG monitoring (Datex/Ohmeda GE Medical). RR for ventilated patients was measured by end‑tidal CO2 (ET CO2) module. The RR of non‑ventilated patients was measured manually by trained research assistants. RR was also measured using a standard impedance technique (Datex‑Ohmeda GE Medical).

Data and statistical analysis:

EarlySense and standard measurements were compared to calculate the accuracy and the detection rate of EarlySense. The absolute relative error rate (aRE) was computed as: aRE=(reference – EarlySense)/reference. For the linear regression and for the BMI‑accuracy correlation, Pearson correlation coefficient was used.

Participants

Sleep laboratory: 16 adults (8 men, 8 women); mean age ± SD=30.6±5.3 years; BMI (± SD)=24.2±4.7 kg/m2

41 children (32 boys, 9 girls); mean age ± SD=7.6±3.6 years; BMI ± SD=18.4± 5.7 kg/m2

ICU: 42 adults (25 men, 17 women); mean age ± SD=54.8±17.4 years; BMI ± SD=26.9±6.6 kg/m2

Results

Comparison of EarlySense readings with the standard device for HR and RR in the sleep laboratory patients:

RR:

Adults (n=16): 1249/1341 (93.1%) accurate data points; aRE±SD=0.04±0.01

Children (n=37): 3346/3646 (91.8%) accurate data points; aRE±SD=0.04±0.02. Correlation=0.93

HR:

Adults (n=16) 5475/5792 (94.4%) accurate data points; aRE±SD=0.03±0.03

Children (n=37):10,348/11309 (91.5%) accurate data points; aRE±SD=0.05±0.03. Correlation=0.97

Comparison of EarlySense and standard impedance readings with the standard device for HR and RR in the ICU:

EarlySense compared with reference standard

HR:

42 patients; 42,752/45,470 (94%) accurate data points; aRE±SD=0.03±0.003; correlation=0.91

RR (ET CO2):

13 patients; 6288/7625 (82%) accurate points; aRE±SD=0.07±0.06; correlation=0.82

RR (manual):

35 patients; 547/734 (75%) accurate points; aRE±SD=0.08±0.08; correlation=0.93

Impedance compared with reference standard

RR (ET CO2):

13 patients; 3310/6388 (52%) accurate points, aRE±SD=0.22±0.11;correlation=0.37

RR (manual):

35* patients; 352/635 (55%) accurate points; aRE±SD=0.16±0.14; correlation=0.81

Adverse events

No adverse events occurring in the sleep laboratory or ICU were related to EarlySense.

Conclusions

The monitoring system was considered to be sufficiently accurate in accordance with standard regulatory and industry criteria. Specifically, RR measured using EarlySense was more accurate than impedance‑based technologies widely used currently. The authors noted that further research is needed, but that this device could allow earlier recognition and response to changes in a hospitalised patient's condition.

Abbreviations: aRE, absolute relative error rate; BMI, body mass index; BPM, beats per minute; Br, breaths; ECG, electrocardiogram; ET CO2, end tidal CO2; HR, heart rate; ICU, intensive care unit; RR, respiratory rate; SD, standard deviation.

Table 2 Summary of the Sorkine et al. (2008) abstract

Study component

Description

Objectives/hypotheses

To evaluate the accuracy of EarlySense for measuring HR and RR and its ability to alert staff to significant changes, in comparison to standard ICU methods.

Study design

Case series.

Intervention

EarlySense.

Setting

Hospital ICU.

Inclusion exclusion criteria

No inclusion/exclusion criteria were stated.

Primary outcomes

HR and RR.

Methods

EarlySense's sensor was placed under the mattress with the data displayed on its control unit. Patients were simultaneously monitored using the ICU's standard of care of ECG for HR and electrical plethysmography/end tidal CO2 for RR (makes and models of these comparator devices not given). Trained technicians also manually measured RR.

Participants

38 critically ill patients (23 male and 15 female, age 16–87 years).

Results

Range was 35–180 BPM for HR and 6–45 breaths/min for RR. Paired data points (n=40,719) were evaluated for HR accuracy. EarlySense had a 92.1% accuracy with an aRE=3.6%. RR compared with end‑tidal and manual RR counts showed an accuracy of 79.4% and 80.1% with an aRE=9.6% and 12.6%, respectively. The number of alerts relating to extreme changes in HR (for example atrial fibrillation and arrhythmia) and RR were consistent between the 2 methods.

Adverse events

None reported by the authors.

Conclusions

EarlySense is accurate and easy to use for measuring HR, RR and trends. It enables staff to continuously assess patients that are currently not monitored, without interfering with their comfort.

Abbreviations: aRE, absolute relative error rate; BPM, beats per minute; Br, breaths; ECG, electrocardiogram; HR, heart rate; ICU, intensive care unit; RR, respiratory rate.

Table 3 Summary of the Frendl et al. (2013) abstract

Study component

Description

Objectives/hypotheses

To identify respiratory patterns characteristic or predictive of respiratory failure.

Study design

Prospective, observational study.

Intervention

EarlySense.

Setting

Not stated however ICU in hospital seems likely.

Inclusion/exclusion criteria

Inclusion:

Adult, critically ill surgical patients.

Exclusion:

No exclusion criteria were noted.

Primary outcomes

Respiratory patterns.

Methods

HR and RR were recorded continuously with EarlySense while the patients were intubated/mechanically ventilated with PSV and followed for up to 24 hours after extubation.

Two patient cohorts were studied: those who successfully extubated and those who were not (end‑points of failure: reintubation, tracheostomy or death). The periods (6.5–24 hours) prior to extubation and 9–24 hours following extubation were studied. The abstract does not clarify whether this was the range studied for all patients individually or whether it is the range of periods studied across the cohort. The recordings were analysed by an algorithm to detect abnormal patterns that correlate with respiratory outcomes and were confirmed by visual examination of the raw signals.

Participants

37 adult critically ill surgical patients (consent taken by surrogates).

Results

A healthy respiratory pattern (normal rate, minimally variable depth) was observed in those patients who were later successfully extubated. Three forms of non‑reassuring respiratory patterns were seen with those who failed extubation:

1. Generalised disorganised respirations;

2. Frequent occurrences (every 30–120 seconds) of a single deep gasping breath;

3. Multiple periods of apnea >30 seconds.

Patients were considered to have abnormal respiration when their non‑reassuring respiration patterns were at least twice as long as their periods with normal respiratory patterns. After extubation, abnormal respiratory patterns identified those patients requiring additional ventilatory support (90% specificity, 50% sensitivity, PPV: 60%, NPV: 86%). Abnormal respiratory patterns prior to extubation showed 91% specificity, 44% sensitivity, PPV: 78%, NPV: 71%.

Adverse events

None reported by the authors.

Conclusions

A pattern of unstable respirations (apparent in inconsistent chest wall motion amplitude) was found to precede and correlate with respiratory failure. For ventilated patients this abnormal pattern (seen in PSV mode) may indicate that a patient is not ready for extubation. In the case of recently extubated patients, the pattern may indicate a need for interventions to prevent re‑intubation.

Abbreviations: ICU, intensive care unit; NPV, negative predictive value; PPV, positive predictive value; PSV, pressure support ventilation.

Table 4 Summary of the Zimlichman et al. (2012) study

Study component

Description

Objectives/hypotheses

To establish the accuracy of the EarlySense continuous monitoring system in predicting clinical deterioration.

Study design

Non‑interventional prospective study with retrospective data analysis.

Intervention

EarlySense.

Setting

2 hospitals: Sheba and Sapir Medical Centres in Israel.

Inclusion/exclusion criteria

Inclusion:

Patients hospitalised with an acute respiratory condition including pneumonia, chronic obstructive pulmonary disease or asthma exacerbation, congestive heart failure with pulmonary oedema or congestion, and patients who needed supplemental oxygen on admission. Patients were enrolled only if they were initially assessed within 24 hours of hospitalisation.

Exclusion:

Dementia and inability to sign informed consent.

Primary outcomes

HR and RR

Methods

Patient enrolment took place during the period of January to December of 2008 in Sheba Medical Center and July to November of 2008 in Sapir Medical Center. Since the study was non‑interventional, and the EarlySense monitor alarms were not responded to by healthcare professionals during the study, patients were also monitored by other monitoring devices, such as telemetry and pulse oximeter, as indicated clinically.

Patients were monitored for the full extent of their stay in the 2 medical departments. Only 2 monitors were available at each site, so in cases where patients stayed for more than 2 weeks, a decision was made by the principal investigator whether to continue monitoring or disconnect the sensor and enrol another patient. This was based on an assessment of clinical stability and on how much longer patients were expected to stay in the unit, in an attempt to avoid a situation where a patient was utilising 1 monitor for a very long period of time. This was done to boost patient recruitment numbers.

Participants

One hundred and forty nine patients were enrolled, mean age 69±6 years, 45.1% were female. Mean BMI 27 (17–52.7), mean Charlson comorbidity score 1.58, mean LOS 7.3±5.9 days. No significant differences were seen in the groups in any of these parameters. One hundred and thirteen had the required ≥30 hours monitoring required for this study.

Results

9/113 (8.0%) patients had a major clinical event, including 2 patients who were transferred to an ICU, 1 patient who was intubated and ventilated in the study unit and later had a cardiac arrest and died, and 6 more patients who had cardiac arrests and died in the study units (overall, 10 major clinical events).

The only significant difference between patients with and without major clinical events was length of stay: 6.9±6.0 days without a major clinician event, compared to 11.9±7.8 days with a major clinical event (p=0.02).

Threshold alerts:

HR: sensitivity (sens) 82%, specificity (spec) 67 %, positive predictive value (PPV) 21%, negative predictive value (NPV) 97%

RR: sens 64%, spec 81%, PPV 26%, NPV 95%

HR and RR: sens 55%, spec 94%, PPV 50%, NPV 95%

Trend alerts:

Change in HR ≥20 beats/minute: sensitivity 78%, specificity 90%, PPV 41%, NPV 97%

Change in RR ≥5 breaths/minute: sensitivity 100%, specificity 64%, PPV 20%, NPV 100%

Change in both HR ≥20 beats/minute and RR ≥5 breaths/minute: sensitivity 78%, specificity 94%, PPV 54%, NPV 98%

Sensitivity and specificity were reasonably high for all threshold and trend alerts, and negative predictive value was very high for EarlySense. However, positive predictive value was consistently low for all types of alerts studied. Retrospective analysis showed that the optimal cut‑offs for the threshold alerts were HR below 40 or above 115 beats/min, and RR below 8 or above 40 breaths/min.

Adverse events

None reported by the authors.

Conclusions

This study found that the EarlySense monitor is able to continuously measure RR and HR, providing low alert frequency. The current study demonstrates the potential of this system to provide timely prediction of patient deterioration. Utilising a trend algorithm has been shown to improve the device's accuracy and reduce associated alert burden and false‑positive alerts.

Abbreviations: CHF, congestive heart failure; COPD, chronic obstructive pulmonary disease; HR, heart rate; ICU, intensive care unit; LOS, length of stay; NPV, negative predictive value; PPV, positive predictive value; RR, respiratory rate; SD, standard deviation.

Table 5 Summary of the Brown et al. (2014) study

Study component

Description

Objectives/hypotheses

To assess the effects of continuous heart rate and respiration rate monitoring in a medical‑surgical unit on unplanned transfers, length of stay in the intensive care unit and length of stay in the medical‑surgical unit.

Study design

Controlled 9 month prospective intervention period and a 9 month retrospective baseline period.

Intervention

EarlySense system.

Setting

Medical‑surgical unit.

Inclusion/exclusion criteria

None stated by the authors.

Primary outcomes

Unplanned ICU transfers, average ICU length of stay (LOS) for transferred patients, and medical‑surgical unit LOS.

Methods

A 33‑bed medical‑surgical unit (intervention unit, using EarlySense monitoring) was compared to a similar 'sister' control unit, and patients were monitored over a 9 month period. Results were also compared with 9 months of retrospective patient data for each ward, referred to as 'the pre‑implementation phase'. Patients were admitted to 1 of the 2 units by the hospital's admissions office in an alternating manner (the authors state that the similar patient populations, supervision and service levels make this 'almost random'). Following the intervention, all beds in the intervention unit were equipped with monitors that allowed for continuous assessment of heart and respiration rate.

Participants

Participants were compared using their mean age, sex, baseline acuity levels and Charlson scores, with no statistically significantly differences.

Results

The study population included 7643 patients, of which 2314 patients in the intervention unit were placed under continuous monitoring using EarlySense. The remaining 5329 patients were in the control arms. Results are presented in a 3‑armed fashion: with control unit and intervention unit arms both compared to pre‑implementation baseline and each other. Statistically significant differences in favour of EarlySense are seen in mean length of stay on the medical/surgical ward, the mean length of stay in ICU and the improvement in the numbers of code blue events.

Control unit (CU)

Intervention unit (IU)

p values

Pre

Post

p

Pre

Post

p

CU‑IU Pre

CU‑IU Post

3 arms

LOS in medical /surgical unit, mean

(25%–75% IQR)

3.80 (1.26–4.25)

3.61 (1.19–4.12)

0.07

4.00 (1.26–4.66)

3.63 (1.19–4.22)

0.02

0.19

0.37

<0.01

ICU transfers/ 1000 patients

18.89

19.06

1.00

26.52

25.93

0.92

0.17

0.12

0.19

Days in ICU/1000 patients

32.69

85.36

0.01

120.11

63.44

0.10

0.06

0.02

0.04

ICU LOS, mean (25%–75% IQR)

1.73 (1.06–2.28)

4.48 (0.94–4.09)

4.53 (0.91–4.39)

2.45 (0.86–2.82)

APACHE II score, mean (25–75% IQR)

13.08 (7.75–18.00)

14.0 (6.00–19.00)

0.59

15.19 (10.00–18.00)

13.38 (7.00–18.25)

0.25

0.29

0.61

0.53

LOS in unit before transfer to ICU, mean (25%–75% IQR)

13.07 (7.26–19.46)

17.03 (6.97–19.9)

0.29

16.82 (5.97–19.95)

11.94 (5.77–14.40)

0.07

0.19

0.13

0.14

Code blue events/1000 patients

3.9

2.1

0.36

6.3

0.9

<0.01

0.44

0.45

0.01

Adverse events

None reported by the authors.

Conclusions

Continuous monitoring on a medical‑surgical unit was associated with a significant decrease in total length of stay in the hospital and in intensive care unit days for transferred patients, as well as lower rates of code blue events.

Abbreviations: APACHE, acute physiology and chronic health evaluation; ICU, intensive care unit; IQR, interquartile range; LOS, length of stay; p, p value; SD, standard deviation.

Table 6 Summary of the Zimlichman et al. (2009) company report

Study component

Description

Objectives/hypotheses

The study objectives were to evaluate the capability of the EarlySense to monitor and alert staff to deteriorations in patient condition and to evaluate the usability of the EarlySense device during routine activities in sub‑acute care units such as medical/surgical wards. Specifically, it sought to identify if and how clinical deterioration was detected by the EarlySense device in medical/surgical departments and to evaluate the system's level of false alarms.

Study design

Multi‑centre, prospective study.

Intervention

EarlySense.

Setting

Hospitals in the USA (Metro West, Massachusetts) and Israel (Sheba, Tel Aviv District and Sapir, Centre District).

Inclusion/exclusion criteria

Inclusion:

None.

Exclusion:

Patients who were unwilling to sign the consent form, patients with Parkinson's disease, intubated patients or patients who regularly used a CPAP or BIPAP device during the night, were excluded from the study.

Primary outcomes

HR and RR.

Methods

Clinicians set 'high' and 'low' heart and respiration rate thresholds for each patient individually, so EarlySense would alert staff if the patient's heart or respiratory rate crossed the pre‑defined threshold. For each alert in the US site, the nurses indicated whether it was a true alert or a false alarm, allowing analysis of the false alarm rate. In addition, the system displayed the trends of the measured parameters enabling clinicians to identify unexpected increases or decreases in heart or respiratory rate. Patients' clinical conditions were recorded on CRFs as well as logged in the patients' clinical charts.

EarlySense reports were reviewed and compared with patients' CRFs and clinical charts. Three levels of deterioration severity were defined: 'major/severe', 'moderate' and 'minor'. The severity level of each deterioration event was assessed mainly according to the medical intervention required, for example ICU transferral or intubation was classified as a 'major' intervention. Utilisation of BIPAP was considered as a moderate intervention. EarlySense reports of the patients whose condition deteriorated were reviewed. Heart and respiratory rate trends as well as system's alerts were reviewed and compared to the patients' condition.

Participants

Participant demographics: 204 patients (99 males, 105 females), mean age 66.1±18.5 years, mean weight 78.6±20.2 kg, mean BMI 28.7±7.8 kg/m2. Patients were not compared for statistical significance at baseline.

Results

A total of 204 patients (99 males and 105 females) from all sites were continuously monitored for over 14,000 cumulative hours. Worsening of clinical conditions (where an intervention of any level was necessary) occurred in 29 patients (14%), with a total of 35 deterioration events. By monitoring heart or respiratory rates using EarlySense, signs of worsening (alerts and trends) were detected in 31 (88.6%) of these events. Of the 35 total events, 11 were defined as major. EarlySense detected signs of worsening in 100% of these major events.

Adverse events

None reported by the authors.

Conclusions

These are preliminary results. The authors suggest that continuous monitoring of RR and HR for patients hospitalised in general medical/surgical units can provide an early warning for deterioration and to allow better clinical data‑driven decisions.

Abbreviations: BIPAP, bilevel positive airway pressure; BMI, body mass index; CPAP, continuous positive airway pressure; CRF(s), case report forms; HR, heart rate; ICU, intensive care unit; RR, respiratory rate; SD, standard deviation.