1 Introduction
Several surgery and anesthesia-related factors increase the risk of an adverse respiratory event (ARE) in the perioperative period. Most importantly, AREs are highly associated with the use of opioid therapy for pain, resulting in opioid-induced respiratory depression (OIRD) [
1‐
4]. OIRD is a potentially lethal complication of activation of opioid receptors in brainstem respiratory neuronal network, associated with bradypnea, apnea, hypercapnia and hypoxia. The number of OIRD events in the postoperative period is not known. A recent multicenter observational study showed the occurrence of an OIRD event in 46% of patients in the first 48 h after surgery under general anesthesia [
5,
6]. In that study, continuous capnography was used to detect patterns of OIRD. Still, deterioration of the patient from a capnography-related pattern abnormality to a sentinel event requiring an intervention (e.g., administration of opioid antagonist naloxone, reintubation, mechanical ventilation, cardiopulmonary resuscitation, or transfer to the ICU) is much less common and may occur not more than once in every 200–1000 postoperative patients [
2,
5,
7]. When respiratory deterioration does occur, however, results can be catastrophic and costs for both the patient and the healthcare system are high [
2,
5,
8,
9]. Given the availability of effective interventions, these respiratory catastrophes following routine, elective surgery have been termed ‘never events’ in that they should never be allowed to occur [
9].
The challenge is to predict or identify respiratory events and intervene before any further respiratory deterioration. Available scoring systems, such as the STOP-BANG questionnaire, which is based on patient-related risk factors, predict postoperative AREs poorly [
10,
11]. Moreover, current standard monitoring practices in postoperative patients do not detect many instances of respiratory compromise [
4,
5,
12‐
14]. Sun et al. showed that prolonged episodes of hypoxemia are common in the first 48 h following non-cardiac surgery and that 90% of these events were missed by routine 4-hourly spot checks in the postoperative wards [
15]. Similarly, Lee et al. showed that the time between the discovery of respiratory depression and the last nursing assessment was 2 h in 42% of the cases and a concerning 15 min in 13% of the cases [
13].
Given all of the above, continuous respiratory monitoring in patients receiving parenteral opioids in the first 24 postoperative hours has been advocated by multiple stakeholders (including the Anesthesia Patient Safety Foundation, the Joint Commission, the American Society for Pain Management Nursing) [
16]. A systematic review of studies evaluating continuous monitoring
via pulse oximetry or capnography reported improved detection of oxygen desaturation or OIRD-events compared to routine nursing checks [
17]. However, impact on clinical outcomes has so far not been demonstrated [
12,
17]. Furthermore, continuous monitoring, regardless of which parameter it is based on, has its own limitations, with a potential for false positive alarms disrupting nurse workflow and leading to alarm fatigue [
1,
12,
16‐
18]. Recent developments aim to use multiple parameters to detect AREs. Application of smart algorithms that combine individual physiological variables into one index may increase the ability to detect a true adverse respiratory event while avoiding false alarms and limiting alarm fatigue [
1]. An example of such a multiparameter index is the Integrated Pulmonary Index or IPI™, which integrates oxygen saturation (SpO
2), respiratory rate (RR), end-tidal PCO
2 (P
ETCO
2) and heart rate (HR) into a single integer value of 1–10 that represents adequacy of respiratory condition of the patient using a fuzzy logic inference mathematical model [
18,
19]. So far, the IPI has been validated with retrospective data obtained in a variety of clinical settings, but it has not been studied prospectively as a monitor of postoperative AREs [
19].
In this randomized controlled trial, the use of the IPI was compared to standard continuous respiratory monitoring in postoperative patients. Our aim was to determine whether the IPI enables early detection of postoperative respiratory events and alters clinical interventions.
2 Materials and methods
Initially we performed an observational trial to evaluate the clinical utility of the IPI algorithm in postoperative patients and to determine the incidence of AREs. The results of this study are published elsewhere [
20]. The data generated by this study were used to design and power the current study.
2.1 Ethics and patients
The protocol was approved by the Investigational Review Board (IRB) (Commissie Medische Ethiek, Leiden University Medical Center, the Netherlands) in August 2015. All study procedures were performed in compliance with the 2013 version of the Declaration of Helsinki and Good Clinical Practice guidelines. The study was registered at the trial register of the Dutch Cochrane Center under identifier 5231. Patients were recruited between November 2017 and January 2019. Subjects were enrolled for the study and they gave verbal and written informed consent prior to study procedures.
2.2 Patients
Patients were adult (at least 18 years old), American Society of Anesthesiologists (ASA) class 1–3, scheduled for elective surgery under general anesthesia, expected to receive opioids for treatment of postoperative pain, and requiring an overnight post-anesthesia care unit (PACU) stay following surgery. Exclusion criteria included use of epidural anesthesia, nerve blocks, surgery that would hamper the postoperative application of the IPI sensors, emergency surgery or the inability to give informed consent.
2.3 Study design
The study had a two-arm, parallel, randomized controlled design. Patients were randomized on the day of surgery to an observational arm or an interventional arm using a computer-generated randomization list. Neither patient nor the anesthetic team responsible for clinical care during surgery were informed of the allocation. Due to the nature of the study, patients and PACU nurses were not blinded to the study allocation once data collection commenced. Outcome assessors were blinded to allocation.
2.3.1 Clinical care in both treatment groups
Anesthesia technique (total intravenous anesthesia or volatile anesthesia, opioid use, use of neuromuscular reversal agents) was left to the discretion of the attending anesthesiologist. Once surgery had ended and the patient was extubated and transported to the PACU, the patient was connected to standard monitoring equipment (3-lead ECG, non-invasive blood pressure monitoring using an arm cuff, pulse oximetry via a finger probe). Additionally, the standalone Capnostream 20p monitor (Medtronic, Fridley, Minnesota) was connected to the patient. This monitor collects SpO2 and HR measurements via pulse oximetry using a finger probe. Additionally, it monitors PETCO2 and RR via a nasal cannula that allows oral and nasal sampling of inspired and exhaled air as well as the delivery of supplemental oxygen with flow rates up to 5 L/min (FilterLine®, Medtronic). All patients were admitted to the PACU until 8AM the following morning, when monitoring by Capnostream was discontinued. Any complication or the need for a prolonged PACU stay was noted in the patient electronic health record. Nurses were asked to note instances in which the Capnostream monitor (finger probe or nasal cannula) was disconnected as may have occurred during meals or patient care.
2.3.2 Clinical care in the observational arm
Patients randomized to the observational arm of the study were attached to the Capnostream monitor, but the monitor screen was shielded and Capnostream alarms were silenced. Nurses were instructed to treat their patients according to standard clinical care using standard monitors and clinical experience. The PACU nurses were requested to note every respiratory event such as apnea, hypoxia, respiratory depression or obstructed breathing. For every respiratory event, they were also asked to note the associated intervention to improve respiratory condition such as verbal or tactile patient stimulation, chin lift, administration of supplemental oxygen via the nasal cannula, escalation to involve the attending PACU physician, naloxone administration, or reintubation. The decision to intervene and manner of intervention was based on local protocol, which relies on monitoring of SpO2, respiratory rate and sedation level.
2.3.3 Clinical care in the interventional arm
Patients randomized to the interventional arm of the study were attached to the Capnostream monitor with the screen visible to the nursing staff. The screen displays the capnography trace (including actual PETCO2 values), HR, RR, SpO2 and IPI value. Prior to the start of the study, nurses were trained to use the Capnostream monitor and interpret the IPI values. A prestudy run in 29 patients was performed (results not included) to finetune the set-up and feasibility of the protocol. Based on the prestudy run, the alarm threshold was set to an IPI value of 1 prior to start of the study.
Since it is mandatory in our hospital to collect vital signs in the patient’s electronic health record, the patient was also attached to standard clinical monitors. However, nurses were requested to guide their assessment of the patient’s respiratory condition based on the IPI value. In case of a clinically relevant discrepancy between the monitors, nurses were asked to evaluate the patient and intervene according to their experience and report the discrepancy to the investigators. In case of an alarm at an IPI value of 1, the nurses were instructed to approach the patient, assess the patient’s condition (apnea, respiratory depression, hypoxia, but also whether the low IPI event could be considered an artefact) and intervene as required (see above). The occurrence of a low IPI event was noted as well as the assessment of the patient’s condition.
2.4 Data collection
Data were collected from the Capnostream monitor, the electronic health record database (Healthcare Information X-change (HiX), Chipsoft, the Netherlands) and the case record forms (CRF) containing the notes regarding respiratory events, and interventions. The Capnostream monitor stored data at 0.5 Hz intervals; the HiX database provided information regarding patient history and characteristics, and drugs administered during surgery and during PACU stay. Major complications were recorded in the electronic health record and the CRF, adverse events were collected in the CRF.
2.4.1 Primary and secondary study endpoints
The primary study endpoints were the number of low IPI events and the number and nature of the nurse responses to low IPI values. Secondary endpoints were the duration of IPI events and the main causes of IPI events, as determined by one or more of the 4 individual variables used in the calculation of the IPI value.
2.4.2 Data selection
We manually checked the data for the presence of artefacts at the end of the study. Low IPI events were considered true and clinically relevant adverse respiratory events if (1) the nurse had not noted the event as an artefact or sensor mispositioning and (2) the recording of vital signs from the electronical medical database corroborated the Capnostream data and (3) in case of hypoxia or obstructed breathing (as noted by the nurses), the event was followed by a sympathetic response such as tachycardia. Once a low IPI event was confirmed to be a true ARE, the nature of the event was examined. An event was considered to be associated with hypoxia when the SpO2 was ≤ 90%; an event was considered a respiratory depression event when: (1) RR < 6 breaths/min or (2) at least 1 episode of apnea (RR = 0 breath/min for at least 15 s) and (1) end-tidal PCO2 > 60 mmHg) or (2) end-tidal PCO2 < 15 mmHg.
2.5 Statistical analysis
We assumed that an intervention would be required in all subjects of the interventional study, and assuming an ‘intervention:low IPI event’ ratio of 99.9% in the interventional arm and 70% in the observational arm, 35 patients were required per group to assess if IPI based intervention differs from local protocol (alpha = 0.05, 1-beta = 0.9). Because of anticipated drop-outs, we aimed to include 80 patients in the study (40 per group).
The normal distribution of numerical data was visually assessed and groups were subsequently compared using either independent samples t-tests or Mann-Whitney U tests. Categorical data were compared using Pearson’s chi-squared test. Results were considered significant with a p-value of < 0.05. Statistical analyses were performed using IBM SPSS statistics for Windows v25.0 (IBM Corp., Armonk, NY).
4 Discussion
To our knowledge, this is the first randomized controlled trial that studied the use of a multiparameter monitoring system to track the respiratory status of postoperative patients. We found that compared to continuous monitoring using respiratory rate and pulse oximetry alone, the use of the IPI monitor led to an increase in the number of interventions performed by nurses to improve the respiratory condition of the patient. This did not lead to a reduction in the number of patients that experienced an ARE, but did cause a significant reduction of the number of events per patient combined with a shorter duration of respiratory events.
Our study shows that 47% of patients experienced one or more adverse respiratory events in the first 24 h following surgery, with an equal distribution among randomization arms. This is consistent with the findings in the PRODIGY trial [
5]. In that prospective, observational, multicenter trial, patients receiving opioids for postoperative pain relief were monitored with the Capnostream monitor and separate signals (end-tidal CO
2, RR, SpO
2 and HR) were collected and analyzed. Respiratory depression was defined by an end-tidal PCO
2 ≤ 15 or ≥ 60 mmHg for at least 3 min, RR < 6 breaths/min for at least 3 min, apnea lasting more than 30 s, or any opioid-related adverse respiratory event. A respiratory depression event occurred in 46% of 1,335 patients.
The clinical relevance of many of these events remains unknown, as deterioration towards a serious adverse event requiring major interventions was fortunately infrequent in our study. However, both studies suggest that nearly half of our patients spend some of the early postoperative period in an state of potential respiratory compromise.
Aside from male sex, we did not detect any risk factors for the occurrence of a postoperative AREs. Other studies, including the aforementioned PRODIGY trial, detected several predictors of a respiratory event, including age, sleep disorders, opioid naivety or high blood pressure [
5,
6,
14]. Our relatively small sample size precluded detection of additional risk factors.
Although monitoring of the IPI did not lead to a reduction in the number of patients experiencing a respiratory event, the number and duration of events per patient was decreased. This is not surprising. Respiratory monitoring
per se does not affect the propensity of a patient to experience adverse respiratory events, however, when a respiratory alarm leads to a nurse intervention, a cascade of events is interrupted that might otherwise have led to more frequent or prolonged events. Since we allowed use of supplemental oxygen, low SpO
2 values contributed to the IPI alarm in only 17% of events. In all other cases the events were triggered by low respiratory rate (or apnea). It is important to realize that supplemental oxygen may mask or may even exacerbate opioid induced respiratory depression when only SpO
2 is monitored [
4,
18,
21]. Patients on oxygen will have some oxygen reserve causing a delay in the detection of an obstructive or central apneic event. Additionally, the peripheral respiratory drive is blunted by supplemental oxygen which will cause a further depression of ventilation [
21]. Our study indicates that it is a challenge for nurses to identify patients experiencing opioid induced respiratory depression in the absence of hypoxia, even with continuous monitoring of respiratory rate in a high care setting with experienced nurses and a low (1:2) nurse to patient ratio. This suggests that continuous monitoring of just SpO
2 is not effective for the identification of opioid-induced respiratory depression.
After a prestudy run we decided to put the IPI alarm threshold at 1, because many of the alarms caused by an IPI of 1–4 represented subtle ventilatory disturbances that were clinically irrelevant. As a consequence, the majority of the alarms in our study were caused by low RR or apnea. It is therefore unclear whether IPI-based monitoring confers a benefit over monitoring of respiratory rate alone. The difference lies in the fact that with simple continuous RR monitoring, the alarm threshold is fixed (e.g. ≤5 breaths/min) whereas with a fuzzy logic algorithm, a RR of 5 breaths/min can trigger an alarm when SpO2 is 92% but not when SpO2 is 100%. It would be interesting to compare IPI to continuous respiratory rate monitoring alone in a future study.
We observed a large number of IPI alarm artefacts. These artefacts were related to the mispositioning of the nasal sensor causing sensing artefacts or the result of minor clinical events. One such frequent alarm that we considered an artefact was related to mildly obstructed breathing during sleep which caused low end-tidal PCO2 values with low but otherwise normal breathing. Still, while our study suggests that lowering the triggering alarm to an IPI value of 1 can be done safely (no serious respiratory events were missed by the monitor), more than half of the alarms did not represent actual respiratory compromise. For successful implementation of future monitoring systems without disrupting nurse workflow, the issue of false alarms and alarm fatigue needs to be addressed. Interestingly, in case of a true alarm, some of our patients did not require actual interventions because their bedside monitor alarm had aroused them before the nurse had intervened. It is imaginable that future monitoring systems could trigger bedside alarms first and only be transmitted to a nursing pager when the respiratory compromise persists or worsens.
This study has some methodological issues that warrant comment. First, we tested the IPI in the PACU setting. Consequently, extrapolation of our results to the general surgical ward should be done with caution. We chose the current approach in order to have two comparable arms, one arm with intervention based on the IPI monitor, and one arm with intervention based on standard PACU monitors (RR and SpO
2) allowing reliable comparison of event occurrences and interventions. We observed that patients receiving respiratory intervention based on the IPI monitor had less and shorter adverse respiratory events, compared to control patients. We expect the benefit of the IPI monitor to be higher on the general surgical ward, where opioids are given, respiratory events can occur but clinical monitoring is infrequent[
22].
Second, we studied the IPI monitor in a population of patients expected to require opioids in the first 24 h after surgery. Several studies have shown that events are most likely to occur within this timeframe and that the use of opioids is a major risk factor for adverse respiratory events [
1‐
3,
13]. However, half of our patients did not experience any respiratory event and not all patients received an opioid postoperatively. It is likely that more benefit and less harm (e.g., disruptions because of alarm artefacts) could be achieved if the monitor was tested in a population at intrinsically higher risk for postoperative respiratory events (e.g. patients receiving parenteral opioids).
Given all of the above, we suggest that future studies of continuous respiratory monitoring focus on devices using algorithms that rely on multiple parameters and that most benefit is to be expected in general surgical wards in patients that are expected to be at a high risk of AREs.
In conclusion, the use of the IPI monitor in postoperative patients did not result in a reduction of the number of patients experiencing adverse respiratory events, compared to standard clinical care. However, use of the IPI monitor did lead to an increase in the number of nurse interventions and a decrease in the number and duration of respiratory events in patients that experienced postoperative AREs.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.