Continuous Vital Sign Monitoring at the Surgical Ward for Improved Outcomes After Major Noncardiac Surgery: A Randomized Clinical Trial

Abstract

BACKGROUND:

Complications occur in a third of patients after major noncardiac surgery and are often preceded by vital sign deviations undetected by current vital sign monitoring practice, despite major advances in surgical and perioperative care. Continuous wireless vital sign monitoring with real-time alerts may allow for a reduction of vital sign abnormalities and complications.

METHODS:

Adult patients undergoing major noncardiac surgery were included and randomized to either standard of care (manual intermittent vital sign monitoring) vs standard of care plus continuous wireless vital sign monitoring with real-time vital sign alerts to staff smartphones at the general postoperative ward. The primary outcome was cumulative duration of severe vital sign deviations, including desaturation, tachy- and bradycardia, tachy- and bradypnea, hypo- and hypertension. Secondary outcomes included adverse events within 30 days. Patients and outcome assessors were blinded to the randomization.

RESULTS:

Four hundred patients were randomized, with 200 in the intervention and 200 in the control group, respectively. Median [interquartile range (IQR)] duration of severely deviating vital signs was 60 [25–136] vs 76 [28–192] min/d in the intervention versus control group, respectively (P = .19). Duration of Spo₂ <88% had a mean reduction of 47 minutes per day (95% confidence interval [CI], 18–80, P = .02). Adverse events occurred in 42.5% vs 31.5% of patients within 30 days (P = .02), and serious adverse events in 34.5% vs 29.5% (P = .39).

CONCLUSIONS:

Continuous vital sign monitoring with real-time staff alerts did not significantly reduce cumulative severe vital sign deviations in this setup. Significant reductions in desaturations and adverse events were found, giving evidence to future studies in the use of continuous vital sign monitoring to improve patient outcomes.

KEY POINTS.

• Question: Does continuous vital sign monitoring of patients after major surgery in the postoperative ward improve patient physiology using vital signs as a proxy?

• Findings: While there was no significant reduction in the total duration of combined severe vital sign deviations, the intervention group experienced fewer and shorter critical desaturation events. Additionally, there was a reduction in overall adverse events within 30 days.

• Meaning: Continuous monitoring in the postoperative ward improves specific physiological parameters and reduces adverse events.

Despite the recent decades’ major efforts to improve surgical outcomes, including perioperative strategies such as Enhanced Recovery After Surgery (ERAS) protocols,¹ complications are a considerable concern after major surgery, with serious adverse events (SAE) occurring in up to approximately 33% of patients.^2–4 In addition to increased morbidity and mortality, complications result in prolonged hospital admissions⁵ and increase health care expenses.⁶

As complications often are preceded by vital sign deviations, manual intermittent vital sign monitoring, such as the National Early Warning Score (NEWS)⁷ has been widely implemented to allow for early detection and interventions, which in turn should decrease the severity and consequences. However, NEWS has never been proven effective in improving patient outcomes,^8,9 with limitations including absent or incorrect measurements,^10,11 delays in escalation protocol interventions,¹² and deteriorations occurring between the manual monitoring rounds occurring up to 12 hours apart.^13,14

In contrast to the intermittent nature of NEWS monitoring rounds, continuous vital sign monitoring with real-time alerts (CVSMA) may provide a solution for timely recognition and intervention in severe physiological deviations, with the ultimate aim of reducing the occurrence of SAEs. This is supported by results from a meta-analysis,¹⁵ retrospective studies^16,17 and most recently a large propensity matched study¹⁸ showing reductions of various complications, length of stay (LOS)¹⁹ and intensive care unit (ICU) admissions¹⁵ when using CVSMA.

Despite these promising results, there is a paucity of large randomized clinical trials testing CVSMA versus manual intermittent monitoring, and increasing the evidence level within this field to support clinical implementation and the research agenda.

An important factor for the clinical uptake of CVSMA is a focus on reducing irrelevant alerts, meaning alerts that may reflect vital sign deviations, but are without relation to clinical outcomes, in contrast to false alerts that are the result of wrong device function. CVSMA may induce alert fatigue caused by numerous false and irrelevant or nonactionable alerts.^20–22 Thus, the tested WARD-Clinical Support System included an AI-augmented combination of severity and duration of vital sign deviations with real-time alerts to staff, with proven alert reduction.²³ The current trial aimed to assess the effect of CVSMA on the duration of severely deviating vital signs in patients after major noncardiac surgery and secondarily on the frequency of adverse events (AE) and SAE, comparing NEWS monitoring versus NEWS+WARD-CSS. We hypothesized that introducing CVSMA would reduce the cumulative duration of severely deviating vital signs compared with NEWS monitoring alone.

METHODS

Trial Design

The trial protocol was approved by the Danish National Committee on Health Research Ethic (Study ID: H-20034555), and the Danish Medicines Agency (Study ID: 2020080707) and the study was registered on ClinicalTrials.gov (NCT04640415, principal investigator: J. Mølgaard, date of registration: November 19, 2020) before patient enrollment. The trial was conducted in surgical departments of 2 major university hospitals (Copenhagen University Hospital – Rigshospitalet and Copenhagen University Hospital – Bispebjerg and Frederiksberg) with a combined 1.700 beds capacity. Staff at all participating departments were given information and instruction in how to use the system before patient randomization.

Written informed consent was obtained from all subjects, and the study was conducted in accordance with the principles of the Helsinki Declaration. Further, the study was performed in alignment with Good Clinical Practice (GCP) guidelines, which was monitored by the GCP monitoring unit of the Capital Region of Denmark. Safety aspects of the trial were submitted as yearly reports to the relevant oversight authorities. A predefined interim analysis after randomization of 200 (50%) patients was performed to assess safety and study feasibility.

Participants

Patients ≥50 years were eligible for inclusion if they were admitted for elective major abdominal, vascular arterial, urologic, or orthopedic surgery, with an expected surgical duration of ≥2 hours in general anesthesia and ≥2 expected overnight stays. Patients were excluded if they were unable/unwilling to give informed consent, had a pacemaker or implantable cardioverter defibrillator, or were allergic to sensor materials. See Supplemental Digital Content, Table 1, https://links.lww.com/AA/F351 for a complete list of inclusion and exclusion criteria.

All patients were treated according to departmental standard-of-care, including well-established enhanced recovery after surgery (ERAS) procedure-specific protocols.²⁴ In the department, there was a nurse-patient ratio of approximately 1:3 (daytime) and 1:8 (night).

Procedures

Patients provided oral and written informed consent during their preoperative visit. A 6-block randomization key was generated via www.sealedenvelope.com and included stratification by study site. Due to the nature of the study, hospital staff were not blinded to the randomization. Patients were setup with monitoring equipment from arrival at the surgical ward, and we did not use the system during the patients’ stay in the postanesthesia care unit (see Supplemental Digital Content, Figure 1, https://links.lww.com/AA/F351). The patients not randomized within 24 hours of transfer from postanesthesia care unit, or those declining monitoring after surgery for any reason, were pulled out and never randomized.

Patients were randomized in a 1:1 ratio to either:

• Control group: At least 2 scheduled manual vital sign measurements per day, with a defined NEWS escalation protocol.^8,11 The control group had vital signs recorded continuously and wirelessly, but staff were blinded to this information.

• Intervention group: Standard of care with the same NEWS paradigm as the control group but in addition CVSMA, including real-time data and alerts to the clinical staff as described in the Supplemental Digital Content, Table 2, https://links.lww.com/AA/F351. Data were transmitted in real-time via Bluetooth and hospital Wi-Fi to a server for processing.

Intervention

The CVSMA system consisted of 3 overall components: wireless vital sign sensors, an algorithmic bundle analyzing vital sign (an artificial intelligence), and a graphic user interface installed on the staff mobile devices receiving vital sign information and alerts in real-time.

Vital Sign Sensors

Three wireless, patient-worn devices were used, all CE/FDA approved, and validated in a previous study²⁵: a single-lead ECG patch (Lifetouch, Isansys) measuring heart rate (HR) and respiration frequency (RF) every minute, a pulse-oximeter (Nonin Wristox 3150, Nonin Medical Inc) measuring peripheral oxygen saturation (Spo₂) every minute, and a cuff-based blood pressure monitor (TM-2441, A&D Medical) measuring blood pressure (BP) every 30 minutes from 7 am- 10 pm and every hour from 10 pm to 7 a. m (See Supplemental Digital Content, Figure 2, https://links.lww.com/AA/F351 for an illustration of how devices were attached to a patient).

Analysis

Data were transmitted via Bluetooth and hospital Wi-Fi to a server for algorithmic processing (proprietary). The algorithm bundle continuously monitors vital signs and issues an alert if a parameter remains outside predefined thresholds for a specified duration, while filtering out transient or spurious deviations. The WARD-CSS alert criteria are based on NEWS criteria in combination with a duration dimension, to allow for spontaneous resolution of deviations, which in turn reduces the number of alerts. The alert criteria are detailed in Supplemental Digital Content, Table 2, https://links.lww.com/AA/F351.

Mobile Device

After data processing, both real-time vital sign values and alerts for patients in the intervention group were sent to a purpose-built, smartphone-based app that gives clinical staff continuous access to patients’ vital sign status in real-time and issues alerts if sustained deviations criteria are fulfilled (see Supplemental Digital Content, Table 2, https://links.lww.com/AA/F351). Clinical staff were used to working with smartphone-based devices for documenting medicine administrations and registering results of manual vital sign monitoring, and the WARD-CSS was primarily running on the same devices.

The WARD-CSS is a clinical support system and not a decision tool, and hence, did not specifically suggest corrective actions is case of alerts, but nurses were advised to use standard clinical escalation-of-care protocols (eg. in case of desaturation: to improve patient positioning, give oxygen supplement, suction if needed, or further diagnostic work-up including requesting assistance). All patients were allowed to remove devices for comfort, and to give indications of data quality and patient satisfaction with a comprehensive system.

Data Sources

Vital sign data were collected from the CVSMA system server hosted by the hospital. Patient electronic health records were reviewed manually for adverse events by physicians blinded to the randomization allocation using a prespecified outcome-manual based on international diagnostic criteria (Supplemental Digital Content, Table 3, https://links.lww.com/AA/F351), where SAE was defined according to International Conference on Harmonization-GCP criteria.²⁶

Outcomes

The primary outcome was the total duration of severely deviating vital signs in minutes per 24 hours outside the defined composite outcome: either Desaturation (Spo₂ < 85%), Bradypnea (RF ≤ 5 min^–1), tachypnea (RF ≥ 24 min^–1), bradycardia (HR < 30), tachycardia (HR > 130 min^–1), hypotension (SBP <91), hypertension (SBP ≥220 mm Hg), OR a combination of hypotension (SBP <100 mm Hg) with simultaneous tachycardia (HR > 110 min^–1) OR bradycardia (HR < 50 min^–1) OR desaturation (Spo₂ < 92%) for the entire monitoring period.

One secondary outcome was the frequency of sustained vital sign deviations, and their duration, as prespecified in the WARD-CSS alert criteria (Supplemental Digital Content, Table 3, https://links.lww.com/AA/F351). The other secondary outcome was the incidence of adverse events (both adverse events and SAE), comprehensive complication index (CCI),²⁷ ICU admission, and mortality within 30 postoperative days. Tertiary outcomes were hospital length of stay and AE and SAE within 7 days.

Data Processing

All data were analyzed as -minute time-series data. Vital sign values were preprocessed before analysis to reduce the number of artifacts. For instance, noisy data evaluated using raw pulse-plethysmograph data were discarded, impossible values (SBP < DBP), and values with large disagreements between sensors (comparing heart rate information from the ECG and PPG signals). BP measurements were only done intermittently, and followingly each BP measurement was extrapolated forward for a maximum of 30 minutes.

To account for varying lengths of monitoring, vital sign deviations were calculated per 24 hours of monitoring time. Similarly, the number of alerts was calculated as the number of alarms per 24 hours.

Statistical Analysis

Sample size calculation was based on data from previous observational studies,²⁸ from which we constructed a synthetic dataset with a 50% reduction in duration of severely deviating vital signs—from median [interquartile range {IQR}] 63 minutes [8–3251] to 32 minutes [4–1626]. These data were used as input in a nonparametric model for sample size calculation by Happ et al.²⁹ with a power of 80% and an alpha of 0.05, resulting in a sample size of 384 patients in a 1:1 intervention/control ratio.

Analyses were performed on an intention-to-treat (ITT) basis. In addition, a per-protocol analysis evaluated patients with at least 12 hours of monitoring. To be included in the per-protocol intervention-population, patients with vital value alerts had to have at least 1 alert responded to by the staff, which was used as a proxy for nurse engagement with the WARD-CSS. This criterion was chosen as an objective measure that WARD-CSS was integrated into the care process, as the use of the monitoring app represented a new element in the nurses’ workflow.

Continuous data were tested with the Mann-Whitney U test, categorical alert data with Fisher exact test, and time-to-event with Cox proportional-hazards ratio.

Results are reported as either median with interquartile ranges or as mean-difference with 95% confidence intervals.

Data analyses were performed in R v. 4.2.2, and Python v. 3.9.7 (packages pandas v. 1.5.2 and lifeline v. 0.27.4, with matplotlib v. 3.5.3) and were validated by an independent, blinded researcher.

RESULTS

Patients

Between January 2021 and October 2022, 1391 patients fulfilled the inclusion criteria and had no exclusion criteria, of which 400 patients were randomized and included in the final analysis (figure 1). The median age was 70 years, and 65% were males. The most frequent procedure was abdominal surgery (72%), followed by arterial vascular (15%), urologic (12%) and orthopedic surgery (1%). Other baseline characteristics are described in Table 1.

Figure 1.

Patient flow chart. Consort flow diagram. A total of 341 patients could not be asked for informed consent before surgery due to personnel constraints, including brief preoperative visits and insufficient research personnel. Additionally, 164 patients were not given informed consent due planned low staffing (eg, holidays and weekends). The target for the study was 400 randomized patients, rather than the total number of included patients.

Table 1.

Demographics

	Intervention (n = 200)	Control (n = 200)
Age, median [IQR], y	68 [60–74]	70 [64–75]
Male sex, no. (%)	125 (63)	134 (67)
Type of surgery, no. (%)
Colonic	48 (24)	56 (28)
Pancreatic	29 (15)	30 (15)
Esophageal	12 (6.0)	19 (9.5)
Rectal	16 (8.0)	12 (6.0)
Other abdominal	37 (19)	28 (14)
Arterial	33 (17)	28 (14)
Urologic	24 (12)	24 (12)
Orthopedic	1 (0.5)	3 (1.5)
Smoking history, no. (%)
Never smoked	52 (26)	52 (26)
Former smoker	108 (54)	109 (55)
Current smoker	40 (20)	37 (19)
Alcohol, no. (%)
None	29 (15)	26 (13)
Not more than MAAa	143 (72)	138 (69)
More than MAAa	28 (14)	34 (17)
Functional status, no. (%)
Independent	182 (91)	178 (89)
Partially dependent	14 (7.0)	22 (11)
ASA physical status, no. (%)
I	11 (5.5)	2 (1)
II	86 (43)	92 (46)
III	97 (49)	103 (52)
IV	3 (1.5)	3 (1.5)
CFS, no. (%)
1–3	147 (74)	133 (67)
4–5	52 (26)	63 (32)
6+	0	3 (1.5)
Surgery duration, median [IQR]	3 h 14 min [2 h 16 min–4 h 24 min]	3 h 19 min [2 h 33 min–4 h 29 min]
Duration from surgery to randomization, median [IQR]	20 h 17 m [16 h 35 m–21 h 57 m]	20 h 35 m [15 h 15 m–22 h 25 m]

Median [IQR] or absolute numbers (%).

Abbreviations: ASA, American Society of Anesthesiologists physical status; CFS, Canadian Study of Health and Aging 9-point Clinical Frailty Scale; IQR, interquartile range; MAA, maximum advisable amount.

^aMaximum advisable amount of alcohol according to the Danish health authority is 168 g per week for men and 84 grams per week for women.

Data Availability

The patient monitoring intervention was started a median of 13 hours [IQR 1 hour–17 hours] after arrival at the general ward, and the median monitoring duration was 2 days 20 hours [IQR, 1 day 20 hours–4 days 1 hours] and 2 days 3 hours [1 day 1 hour–4 days 2 hours] for the intervention group and control group respectively. The median amount of data acquisition (the duration of registered data compared with the time the device was active) was 95% of the total monitored duration [75%–99%] from the single-lead ECG patch, 50% [34%–70%] from the pulse-oximeter, and 20% [0%–46%] from the blood pressure monitor.

Primary Outcome

The median duration of severely deviating vital signs was 60 minutes per 24 hours [25–136] for the intervention group and 76 minutes per 24 hours [28–192] for the control group (mean difference 20 minutes per 24 hours (95% confidence interval [CI], −20 to 60); P = .19).

Secondary Outcomes

For the individual components in the primary outcome, there were no significant differences between the groups (Table 2). A significantly shorter duration of Spo₂ < 92% (mean difference 90 minutes per 24 hours (95% CI, 28–152); P = .008) and Spo₂ < 88% (mean difference 47 minutes per 24 hours (95% CI, 18–80); P = .02) were recorded in the intervention group. (Table 2). There was no difference in the number of patients with sustained vital sign deviations (see Table 3).

Table 2.

Cumulative Duration of Deviating Vital Sign Values Per 24 h

	Intention-to-treat population
	Intervention (n = 200)	Control (n = 200)	Mean difference (95% CI)	P value
Primary outcome, combined cumulative	60 [25–136]	76 [28–192]	20 (−20 to 60)	.19
Respiratory
Spo2 < 92%b	216 [86–472]	353 [142–627]	90 (28–152)	.008
Spo2 < 88%	37 [13–82]	51 [16–148]	47 (18–80)	.02
Spo2 < 85%a	12 [3–31]	14 [4–50]	23 (2–44)	.23
Spo2 < 80%	2 [0–5]	2 [0–6]	5 (−1 to 12)	.60
RF ≤ 5 min–1a	0 [0–1]	0 [0–1]	1 (0–3)	.29
RF < 11 min–1	34 [7–102]	36 [8–106]	4 (−20 to 27)	.84
RF ≥ 24 min–1a	14 [2–47]	15 [2–69]	2 (−25 to 29)	.60
HR < 30 min–1	0 [0–0]	0 [0–0]	1 (0–2)	.45
Circulatory
HR < 30 min–1	0 [0–0]	0 [0–0]	1 (0–2)	.45
HR < 40 min–1a	0 [0–0]	0 [0–0]	3 (−1 to 7)	.36
HR > 110 min–1a	7 [0–40]	6 [0–38]	–5 (−30 to 18)	.71
HR > 130 min–1	0 [0–2]	0 [0–2]	–3 (−11 to 4)	.34
SBP <70 mm Hg	0 [0–0]	0 [0–0]	4 (−1 to 9)	.52
SBP <90 mm Hga	22 [0–83]	18 [0–106]	18 (−15 to 50)	.74
SBP ≥180 mm Hg	32 [0–75]	33 [0–85]	17 (−6 to 39)	1.0
SBP ≥220 mm Hga	0 [0–0]	0 [0–0]	0 (−6 to 5)	.41
Combination—SBP <100 mm Hg AND
HR > 110 min–1a	0 [0–0]	0 [0–0]	2 (−1 to 4)	.30
HR < 50 min–1a	0 [0–0]	0 [0–0]	3 (−1 to 8)	.15
Spo2 < 92%a	0 [0–4]	0 [0–7]	5 (−5 to 14)	.29

The table presents the primary outcome in the first row, which is the cumulative duration of deviations outside the combined thresholds: Spo₂ < 85% or 5 > RF > 24 min^–1, or 40 > HR > 110 min^–1, 91 > SBP >219 mm Hg, or a combination of SBP <100 mm Hg AND (HR > 110 or HR < 50 or Spo₂ < 92%). All values are presented as median [interquartile range] or mean difference with (95% CI). All values are calculated as the number of deviating minutes per 24 h of monitoring.

Abbreviations: CI, confidence interval; COPD, chronic obstructive pulmonary disease; HR, heart rate; RF, respiration frequency; SBP, systolic blood pressure; Spo₂, peripheral oxygen saturation.

^aAll rows are categories that are part of the primary outcome, which is considered secondary outcomes.

^bPatients with BMI ≥40 or registered COPD are excluded from this analysis.

Table 3.

Number of Episodes of Sustained Deviations Per 24 h

	Intention-to-treat population
	Intervention (n = 200)	Control (n = 200)
	n (%)	n (%)
Desaturation
Spo2 < 92% for ≥60 min	105 (53)	109 (55)
Spo2 < 88% for ≥10 min	117 (59)	124 (62)
Spo2 < 85% for ≥5 min	90 (45)	99 (50)
Spo2 < 80% for ≥1 min	52 (26)	51 (26)
Bradypnea
RF ≤ 5 min–1 for >1 min	4 (2.0)	7 (3.5)
RF < 11 min–1 for ≥5 min	16 (8.0)	17 (8.5)
Tachypnea
RF ≥ 24 min–1 for ≥5 min	54 (27)	70 (35)
Bradycardia
HR < 30 min–1 for ≥1 min	1 (0.5)	6 (3.0)
HR [30–39] min–1 for ≥5 min	2 (1.0)	7 (3.5)
Tachycardia
HR ≥ 111 min–1 for ≥60 min	12 (6.0)	15 (7.5)
HR > 130 min–1 for ≥30 min	5 (2.5)	6 (3.0)
Hypotension
SBP <70 mm Hg for >30 min	10 (5.0)	11 (5.5)
SBP <91 mm Hg for >60 min	62 (31)	64 (32)
Hypertension
SBP >180 mm Hg for ≥60 min	69 (35)	68 (34)
SBP ≥220 mm Hg for >30 min	9 (4.5)	12 (6.0)
Combined deviations—  SBP <100 mm Hg AND
Spo2 < 92% for >10 min	47 (24)	61 (31)
HR > 110 min–1 for >30 min	4 (2.0)	9 (3.5)
HR < 50 min–1 for >30 min	1 (0.5)	5 (2.5)
Any episode	185 (93)	189 (95)

Episodes of Spo₂ < 92% were not calculated for patients with COPD or BMI ≥40 kg/m². To trigger an episode of bradypnea ≤5 min^–1 a heart rate >20 min^–1 was also needed to ensure that critical bradypnea was not a result of sensor disconnection. To trigger an episode of bradypnea <11 min^–1 an Spo₂ < 88% was also needed to ensure that bradypnea was a symptom of respiratory distress. “Any episode” is the number of patients with at least 1 episode of sustained deviations from any of the above definitions.

Abbreviations: HR, heart rate; RF, respiration frequency; SBP, systolic blood pressure; Spo₂, peripheral arterial oxygen saturation.

At postoperative day 30, 1 or more AE had occurred in 31.5 % of participants in the intervention group vs 42.5 % of participants in the control group (P = .03), and similarly at day 30, SAEs had occurred in 29.5 % of patients in the intervention group vs 34.5 % of patients in the control group (P = .39).

The median length of stay was 5 days [3 days–7 days] for the intervention group vs 6 days [3 days–8 days] for the control group. Median CCI for the intervention group was 0 [IQR 0–22.6] vs 8.7 [IQR 0–24.7] for the control group (P = .18).

SAEs and tertiary outcomes are detailed in figure 2a and Supplemental Digital Content, Tables 5–7, https://links.lww.com/AA/F351.

Figure 2.

Time to first serious adverse event or adverse event. Kaplan-Meier cumulative density plot of time-to-first-event for the intention-to-treat population. A, Time to first SAE. Hazard ratio: 0.82 (95% CI, 0.58–1.16); P = .26. B, Time to first AE. Hazard ratio: 0.69 (95% CI, 0.50–0.95); P = .02. x-axis: Time from arrival at the ward and start of monitoring to the first event. y-axis: Proportion of patients with an event. AE indicates adverse event; CI, confidence interval; SAE, serious adverse event.

Perprotocol Analysis

Sixty-one (30.5%) intervention and 14 (7%) control patients were excluded from the protocol analysis patients due to premature termination of monitoring, and/or no documentation of nurses engaging with the app (n = 56) (see figure 1 for details), leaving 139 intervention and 186 control patients for analysis.

For the primary outcome, there was a mean difference of 9 minutes per 24 hours (95% CI, −33 to 52). For details, see Supplemental Digital Content, Tables 8a and 8b, https://links.lww.com/AA/F351.

On postoperative day 30, 46 (33.1%) vs 78 (41.9%) had developed an AE. For SAEs, they occurred in 39 (28.1%) vs 64 (34.4%)—intervention and control group, respectively. See Supplemental Digital Content, Figure 3 and 4, https://links.lww.com/AA/F351.

DISCUSSION

This randomized clinical trial found no significant difference between patients monitored with CVSMA plus standard care versus standard care alone regarding the cumulated duration of severely deviating vital signs after major surgery.

However, significantly shorter durations of desaturations of median 90 minutes for Spo₂ < 92% and 60 minutes for Spo₂ < 88% were seen in the intervention group. Significantly fewer AEs also occurred in the intervention group 31.5 % vs 42.5%, but not of SAEs (29.5% vs 34.0% intervention versus control, respectively)

As opposed to other studies of CVSMA focusing on clinical outcomes such as length of stay and ICU admission,¹⁵ our primary outcome was a composite measure of duration of severe vital sign derangement, to reflect the intervention effect on a heterogenous group of cardio-respiratory vital signs. Since the WARD -CSS system relays information of deviations in vital sign values to patient-care staff, we expected that vital signs would be impacted the most. One of the clinically important secondary outcomes was a composite measure of “any SAE,” a patient endured, which was chosen to capture patient-related outcome at an earlier stage than ICU admission. Although the relationship between vital sign deviations and patient-related outcomes has not been fully elucidated, this trial attempts to take a step in this direction.

The logic behind CVSMA stems from the assumption that postoperative SAEs are preceded by increasing physiologic instability before clinical detection of SAEs,^30–32 or that deviating physiology in itself inflicts injury such as hypotension leading to myocardial injury.^33,34 In either case, directing staff attention to undesirable physiology could have a dual impact by either helping to detect early signs of SAEs or to avoid deteriorating patient physiology. The presented findings supplement the growing body of evidence on the use of CVSMA to potentially improve postoperative outcomes.^15,35,36 A recent propensity-matched retrospective study of more than 34,000 surgical patients found that wireless monitoring of vital sign values resulted in reduced ICU admission or mortality (12.8% vs 37.9%), and shorter hospital length of stay (3.31 vs 4.19 days).¹⁸

These findings, together with this study, support a recent meta-analysis, where CVSMA showed a trend toward benefit in the form of reduced risk of rapid response or cardiac arrest team activation(risk ratio: 0.84 (95% CI, 0.69–1.01)), although no definitive conclusion were reached due to the lack of large and rigorous RCTs.¹⁵

CVMSA systems are complex interventions with many potential points-of-failure such as data-connectivity,^37,38 artifact removal,³⁸ staff compliance^38,39 and adverse impact on perceived staff workload.^37,38 Thus, even though the technology may collect all vital sign data, the algorithms correctly detect important conditions, and data is sent correctly to the nurses; the engagement of the staff may be low, and the chosen interventions may not be effective. A study found that while patients were positive about monitoring, staff involvement and engagement remained challenging,³⁹ and limited evidence exists concerning how hospital-staff prefer to use CVSMA systems. Based on human-factors testing, nursing staff considered the WARD-CSS app intuitive. In the current study, nurses did not mark in the app that they had acknowledged an alert in 28% of patient cases, and although we cannot exclude that the alert had resulted in an action in these cases, it highlights an area for improvement. The importance of engagement is supported by the increased reduction of SAE’s in the “per protocol analysis,” compared to the “intention-to-treat analysis” when “nonresponders” were removed. These observations underscore the need to refine user interfaces and integrate such systems more seamlessly into clinical workflows, ensuring that they meet the dynamic demands of the work environment.” Another point to consider is the choice of outcome measures in future trials. In hindsight, the primary outcome of combined vital sign deviations may have been suboptimal because some vital signs—such as heart rate in cases of infection or hypovolemia—require more time to normalize even with adequate corrective interventions, while others, like desaturations, can be corrected more quickly. On the other hand, end points such as ICU admission or death, may lack the sensitivity needed to detect clinically meaningful improvements or demand very large sample sizes. This study reports the number of SAE and AE, which may provide a more sensitive measure of the intervention’s impact.

A key strength of this study is the randomized design and testing in a normal clinical setting with well-implemented enhanced recovery after surgery (ERAS) protocols. To further improve the external validity for a CVSMA that is thought to work alongside normal clinical practice, this trial included a wide range of patient types, and the trial did not include specific suggestions for interventions, leaving nurses free to use the app, and intervene however they saw fit. We conducted both pretraining of nurses and a run-in training phase on the use of WARD-CSS before randomizing patients, where the WARD-CSS was reported as being easy to use.

The high SAE rate despite being in an ERAS setting, reflects the fact that most procedures were major abdominal, with similar SAE rates as in previous publications.^4,28

We used thresholds for vital value deterioration that are based on the NEWS thresholds, with a minimum duration criteria adapted for continuous monitoring systems to improve the interpretability of the system,⁴⁰ as staff would be acquainted with the majority of thresholds.

Limitations included a relatively high number of patients not consenting to the study, opting-out before randomization, or after a short monitoring duration, where the main reason for opting-out was equipment discomfort. Data acquisition was not optimal for the oximeter and cuff-based blood pressure monitor due to patient discomfort. We tried to address each patient’s concern, including allowing them to keep only 1 device on if the alternative was to opt-out completely, we also missed early postoperative data, by not starting monitoring in the moment patients arrive in the ward. Future trials should take great care in deciding which devices agree with patients, since choosing the wrong device can make patients withdraw from the study altogether. In the current study, the monitoring duration was restricted to a maximum of 5 days due to device battery constraints, and additional effects from longer monitoring periods can be hypothesized. Sample size calculation was based on a 50% reduction in duration of severely deviating vital sign values which was not based on prior pilot studies and larger future trials are needed to establish this finding. Using the information gained in this trial to do a post hoc power calculation on the secondary outcome of SAE at day 7 and 30 results in a power of 39.9% and 18.6%, respectively, reflecting that the trial was underpowered to detect a clinically relevant difference.

In conclusion, continuous vital sign monitoring with real-time staff-alerts did not significantly reduce cumulative severe vital sign deviations, however, significant reductions in desaturations and AEs were found. These findings provide valuable insights into the effect of continuous monitoring in postoperative care. They highlight the need for larger trials with extended follow-up to further evaluate the impact of continuous monitoring on longer-term postoperative patient outcomes that may not be apparent in the immediate postoperative period.

DISCLOSURES

Conflicts of Interest: C. S. Meyhoff and E. K. Aasvang have founded WARD24/7 ApS, Copenhagen, Denmark, with the aim of pursuing the regulatory and commercial activities of the WARD-project (developing a clinical support system for continuous wireless monitoring of vital signs). WARD24/7 ApS has signed a license agreement for any WARD-project software and patents. One patent has been filed: “Wireless Assessment of Respiratory and circulatory Distress (WARD), EP 21184712.4 and EP 21205557.8.” K. K. Grønbæk, T.-H. Hoang , and S. S. Rasmussen have since the writing of this manuscript become employees of WARD 24/7 ApS. No other authors declared Conflicts of Interest. Funding: This project received funding from: The innovation fund Denmark (8056-00055B). The funder had no influence on the study design. This manuscript was handled by: Jennifer M. Weller, MBBS, MD, MClinEd, FANZCA, FRCA.

Supplementary Material

ane-141-807-s001.docx ^{(2.4MB, docx)}