Full Text View

Postoperative complications occur in up to 30% of patients within the first 30 days after major abdominal cancer surgery, ^{1, 2} despite an intense focus on improved surgical and anaesthesiology techniques to enhance recovery. ³ These findings emphasise the need for improvements in early detection of complications to allow early interventions and potentially facilitate recovery. Preoperative and intraoperative risk factors have been used to establish risk models for postoperative complications, ^{4, 5} but there is sparce information from the post-anaesthesia care unit (PACU) for identifying at-risk patients, despite the fact that surgery itself may be the ultimate test of how the specific patient reacts to trauma. Thus, PACU admission, with the inherent close monitoring, may serve as an ideal phase to assess patients’ risk and to plan for increased monitoring need after discharge to the general ward. Recent studies have shown that patient deterioration during admission to the PACU in the immediate postoperative period may be an important predictor of subsequent morbidity and mortality. ^{6, 7} However, interpretation is confounded due to heterogeneous procedures and lack of a standardised protocol for optimised postoperative care. In addition, the occurrence of physiological abnormalities in the general ward was only recorded by manual intermittent monitoring and not by continuous wireless monitoring, resulting in a high risk of not detecting the true occurrence of vital signs abnormalities. ⁸

We aimed to assess the correlation between vital signs abnormalities during the PACU stay and vital signs abnormalities within the first 96 hours in the general ward after PACU discharge following pancreaticoduodenectomy or oesophageal resection. Secondly, we assessed the association between vital signs abnormalities in the PACU and serious adverse events (SAEs). We hypothesised that vital signs abnormalities during the PACU stay were positively correlated to the duration of severe vital signs abnormalities and occurrences of SAEs in the general ward.

This was an observational, prospective single-centre study conducted from February 2018 to September 2019 at Rigshospitalet University Hospital, Copenhagen, Denmark. The study was approved by the regional ethics committee (H-17033535) and registered on ClinicalTrials.gov (ID: NCT04188093) as part of the Wireless Assessment of Respiratory and Circulatory Distress (WARD) project (ClinicalTrials.gov ID: NCT03491137).

Patients aged ≥ 60 years undergoing pancreaticoduodenectomy or oesophageal resection were included. The procedures were selected due to a planned long postoperative PACU stay and to an extensive surgery with well known risk for postoperative complications. Patients were excluded if they were assessed unable to cooperate, had reduced cognitive function (Mini-Mental State Examination score < 24), had a pacemaker or implantable cardioverter defibrillator unit, or had an allergy to the study material.

Procedures were performed under general anaesthesia with remifentanil, propofol and cisatracurium, and thoracic epidural. Postoperative analgesia consisted of thoracic epidural, paracetamol and opioid. Baseline demographic characteristics (age, sex, height, weight, American Society of Anesthesiology [ASA] status, comorbid conditions, and Charlson Comorbidity Index [CCI] ⁹ ) and intraoperative data were collected from electronic health care records.

After surgery, patients were admitted to the PACU until discharge the following day from 09:00 am and onwards, upon fulfilling standardised discharge criteria (no single score > 1 and no total score > 4, or assessed ready by the attending anaesthetist). PACU nurses assessed the patient’s recovery every 30 minutes using the modified Aldrete criteria, ¹⁰ as recommended by the Danish Society of Aanesthesia and Intensive Medicine (Dansk Selskab for Anæstesiologi og Intensiv Medicin [DASAIM]) (Online Appendix). The post-anaesthesia recovery score (PACU score) is a modified Aldrete criteria scale, calculated from ten parameters (sedation, oxygen saturation measured by pulse oximetry [SpO₂], respiratory rate, systolic blood pressure, heart rate, pain at rest, motor function, nausea, diuresis and temperature), each given a score from 0 to 3 based on predefined criteria for symptoms. ¹⁰ In case of interventions to improve either blood pressure (vasopressor, blood transfusions, non-standard fluid infusion > 250 mL within 15 minutes) or oxygen saturation (continuous positive airway pressure, oxygen supplementation > 2 L/min, or high flow oxygenation therapy) a score of 3 was recorded for the respective indicator. In the general ward, patients adhered to well implemented procedure-specific enhanced recovery care programs, including multimodal analgesia and early mobilisation. According to usual standard of care in the hospital’s general ward, patients were observed by clinical staff and had vital signs recorded following an Early Warning Score (EWS) protocol. EWS monitoring is by default 8 hours, and vital signs abnormalities start an escalation protocol including increased frequency of measurements. For example, interventions are started and a ward doctor is summoned. If the patient is still unstable, an anaesthetist is called and further rescue is planned (ie, advanced interventions or intensive care unit [ICU] admission).

Vital signs were continuously monitored in the general ward using three wearable wireless devices as part of the WARD project:

Data were transmitted through Bluetooth to a bedside gateway and from the gateway to a hospital server via a secure hospital internet connection. Research staff attended patients daily to ensure proper function of the equipment as well as patient comfort. Clinical staff and patients were blinded to the collected vital signs data.

SAEs during the first 96 hours in the general ward were assessed according to a predefined manual with internationally recognised standards for diagnostic classification, and were rated as SAEs if considered life-threatening, resulting in death, required prolongation of existing hospitalisation, or resulted in significant disability or incapacity. Complications (SAEs) can be diagnosed by on-call medical doctor or assessed by the investigator by review of the patient’s record according to the criteria defined below:

The primary exposure was the maximum PACU score at any time point during the PACU stay.

Secondary exposures included the maximum PACU score at prespecified 6-hour intervals starting on admission and until discharge. Maximum PACU subscores during the PACU stay and at discharge were calculated for circulatory vital parameters (combined scores from pulse and systolic blood pressure) or respiratory vital parameters (combined scores from respiratory rate and oxygen saturation).

The primary outcome was the cumulative duration of any vital signs abnormalities (heart rate, blood pressure, respiratory rate, and arterial oxygen saturation) during the first 96 hours in the general ward or until discharge, whichever came first. Thresholds for circulatory (heart rate > 130 min^-1 or < 30 min^-1 and systolic blood pressure < 90 mmHg or > 220 mmHg) and respiratory (SpO₂ < 85% and respiratory rate > 24 min^-1) vital signs abnormalities were set.

Secondary outcomes were the cumulative duration of circulatory vital signs abnormalities (heart rate, systolic blood pressure), cumulative duration of respiration abnormalities (SpO₂ and respiratory rate), and the frequency of any SAE during the first 96 hours in the general ward or until discharge, whichever came first.

Specific pre-analytical algorithms removed noise and artefacts before analysing continuous data. Heart rate and respiratory rate were recorded with a one-minute sampling frequency derived from automatic detection of the QRS complex and R peaks in the single-lead ECG, digitised at 1000 samples per second. Every minute, 10-second segments of the ECG underwent a computerised filtration process, where signal quality was ensured based on QRS complex analysis. The heart rate, respiratory rate, and SpO₂ included both raw values and a calculated average per minute. An SpO₂ change > 4% per second was considered an artefact and thus removed from the final analysis. The duration of abnormal blood pressure measurements was calculated as the time between the abnormal value and the previous normal value, plus the time until the first subsequent normal value divided by two, thus assuming the abnormal value started and ended midway between the normal and abnormal value.

Due to the lack of similar studies on duration of vital signs abnormalities, we did not perform a formal power calculation in this exploratory study. Based on the existing literature, we expected a 20–30% complication rate.

Data are presented using descriptive statistics either as medians with interquartile ranges (IQRs) or means with 95% confidence intervals (95% CIs), dependent on distribution. The duration of vital signs abnormalities was measured as minutes during the entire monitoring time (96 hours). SAEs were presented with percentages for categorical data. The primary analysis was the association between the maximum PACU score and the cumulative duration of vital signs abnormalities (any). Secondary analyses were the associations between maximum and discharge circulatory and respiratory PACU subscores and duration of circulatory and respiratory vital signs abnormalities respectively. Associations were tested by Spearman rank correlation, with a significance level of P = 0.05. The predictive value of the PACU scores for SAEs in the general surgical ward were calculated as the area under the receiver operating characteristic (AUROC) curve with 95% CIs. All data analyses were conducted using the R Foundation Statistical Program (R v.3.6.2) and R Studio (v.1.2.5001).

One-hundred patients were prospectively included. Eight patients were excluded due to being transferred directly from the PACU to the ICU (n = 4), change of surgical procedure (n = 3), or withdrawn consent (n = 1), leaving 92 patients for analysis. The mean age was 69.6 years (95% CI, 68.5–70.3); 30 patients (33%) were female, 36 (39%) underwent an oesophageal resection, and the median ASA score was 2. During the first 96 hours in the general ward, there was one ICU admission, mortality was zero, and the median hospital length of stay was 9.6 days (IQR, 7.6–15.9 days). All patients received standardised general anaesthetic (six patients received volatile anaesthesia) followed by a multimodal postoperative analgesia plan. Demographic characteristics are detailed in Table 1. The median PACU stay was 19 hours (IQR, 19–21 hours), and patients were successively discharged to the general surgical ward, with 92 patients admitted for 12–18 hours, 62 patients admitted for 18–24 hours, and only seven patients remaining in the ICU after 24 hours. The median maximum PACU score was 7 and occurred predominantly during the first 6 hours after PACU arrival (Figure 1). SpO₂ (median score, 2) and systolic blood pressure (median score, 1) were the most frequent vital parameter abnormalities in the PACU.

The median duration of continuous monitoring was 91 hours (IQR, 71–95 hours) per patient. We recorded a median 84 hours (IQR, 64–94 hours) of respiratory and heart rate, 55 hours (IQR, 30–76 hours) of SpO₂, and 83 (IQR, 32–116) blood pressure measurements per patient for analysis. The primary outcome of duration of cumulative vital signs abnormalities (any) was a median duration of 114 minutes (IQR, 23–324 minutes; range, 0–1755 minutes) and occurred in 88 patients (96%) during the time of continuous monitoring (Table 2).


No significant correlation between the maximum PACU score and duration of (any) vital signs abnormalities during the first 96 hours in the general ward was found (R = 0.13; P = 0.22) (Figure 2). The maximum circulatory PACU score and circulatory discharge score were significantly correlated to the duration of circulatory vital signs abnormalities in the general ward (R = 0.32 [P = 0.0021] and R = 0.26 [P = 0.014] respectively). The maximum respiratory PACU score and the respiratory discharge score were not significantly correlated to the duration of respiratory abnormalities in the ward (R = -0.02 [P = 0.85] and R = -0.056 [P = 0.6] respectively).

Seventeen patients (18%) had at least one SAE within the first 96 hours in the general ward (Table 3). Fifteen patients had one SAE, one patient had two SAEs, and one patient had six SAEs. The most frequent SAEs were delirium, ⁴ pneumonia, ³ surgical site infection ² and new onset of atrial fibrillation. ² Figure 3 illustrates the AUROC curves showing poor association between PACU scores and outcomes at all time points, with all AUROC curves close to 0.50, signifying no predictive value of the PACU scores at the different time points.

The degree of maximum abnormalities during PACU stay was not significantly correlated to overall vital signs abnormalities in the general ward (R = 0.13; P = 0.22). Severe vital signs abnormalities in the ward occurred in almost all patients (96%). Severe circulatory abnormalities in the overall PACU stay and at discharge were significantly correlated to duration of circulatory vital signs abnormalities on the ward (R = 0.32 [P = 0.00021] and R = 0.26 [P = 0.014] respectively). Seventeen patients (18%) experienced SAEs, without significant association to the PACU stay (AUROC, 0.59; 95% CI, 0.46–0.73).

The current study did not detect any correlation between PACU stay abnormalities and subsequent deterioration during the first 96 hours in the general ward, either assessed as duration of vital signs abnormalities by continuous vital signs monitoring or clinical SAEs. However, we found significant correlation between severe circulatory PACU complications and systolic blood pressure and heart rate abnormalities in the general ward, similar to a recent study where hypotension in the PACU was correlated with rapid response team (RRT) activations.7 Severe vital signs abnormalities or SAEs occurred in 96% and 18% of patients respectively despite a good clinical status at the time of PACU discharge and an implemented enhanced recovery protocol, ¹¹ calling for measures to improve and identify these patients.

Our findings are in contrast to a previous study, ⁶ which reported significant associations between prolongation of the PACU stay for clinical reasons and increased risk of RRT activations in the general ward. Our study setup does not allow for a similar analysis due to mandatory PACU stay until the morning after the surgery.

Preoperative scoring systems for calculating risk of postoperative morbidity and mortality are widely used internationally, to help clinicians form an evidence-based decision of whether to operate or not. Such systems are the CCI and the National Surgical Quality Improvement Program (NSQIP), with well proven efficacy for predicting 30-day complications, but not for identifying patients at risk for complications in the first postoperative days. ^{4, 12} Our data indicate that a modified Aldrete criteria PACU score is not useful for identifying patients at-risk of postoperative severe vital signs abnormalities and complications during the early stay in the general wards, calling for other measures to improve patients’ safety. Currently, patient monitoring in general surgical wards often relies on intermittent manual assessments performed by clinical staff at intervals of up to 12 hours with EWS or similar systems. ¹³ However, significant deterioration may occur between these intervals, as previously shown ^{8, 14} and demonstrated in the current study, when a substantial proportion of patients experienced vital signs abnormalities and SAEs during admission to the ward. Given that severe vital signs abnormalities are not detected by routine monitoring, rescue interventions such as RRT or medical emergency team are not alerted and therefore not activated at the start of patient deterioration. One possible solution is the application of wireless monitoring devices recording vital signs continuously, as used in this study, which may decrease the number of ICU admissions, rescue interventions, cardiac arrest, and mortality. ¹⁵

Focusing on the post-PACU phase may be relevant. Recent studies have supported the importance of vital signs abnormalities, where episodes of early postoperative hypoxaemia increased the one-year mortality (odds ratio [OR], 1.2; 95% CI, 1.1–1.3 per 10% increase in episodes of desaturation < 85%, ¹⁶ and systolic episodes < 90 mmHg, especially during the first 4 postoperative days, significantly increased the risk for myocardial infarction and/or death (OR, 2.8; 98.3% CI, 1.3–6.4). ¹⁷ These findings emphasise the need for continued close monitoring of high risk postoperative patients even after discharge from the PACU. However, a recent study on the full 500 patient cohort (various procedures and no systematic PACU scores) from which the patients in this study were included has recently been published.18 This study showed that vital signs abnormalities are frequent in patients without clinical abnormalities, calling for more advanced multimodal analyses of vital signs abnormalities to better predict oncoming complications while reducing false alerts. ¹⁸

As with all studies, ours has limitations. The main weakness of the current study is the relatively small sample size. However, this is ameliorated using continuous vital signs monitoring after discharge to the general wards, allowing for highly sensitive detection of vital signs abnormalities and the strict recording of predefined SAEs following internationally recognised standards. In addition, we chose to include patients with well known high risk of postoperative complications to adjust for the small number of patients. Furthermore, the PACU score contains parameters that are not part of the continuous vital signs monitoring (ie, pain and sedation). Their association to later vital sign abnormalities in the ward cannot be dismissed, but the severity of pain and sedation were low in the PACU and as such would not be predictive for vital signs abnormalities in the ward. The investigations took place at a single centre, and the request for informed consent may have resulted in the cohort being in the more fit end of patient spectrum, although this is speculative as patients with dementia are not deemed fit for surgery in most cases. Our PACU setup resembles that of an ICU at other facilities with the same procedures, but the discharge criteria are generic and as such the results should be transferable when it comes to major abdominal surgery. In contrast, our findings may not reflect other patient categories, which should be explored. The impact of RRT activation and interventions may have resulted in less severe SAEs, but this only applies to the cases where the routine monitoring with EWS detected abnormalities. In that sense, our monitoring was blinded to the staff and should reflect the actual patient condition, well known to be missed by standard clinical rounds. ^{8, 19} All patients fulfilling the inclusion criteria were invited to join the study when study personnel were available to perform informed inclusion. Apart from the inclusion and exclusion criteria, we have no reason to believe the cohort differs from other patients undergoing the same surgical procedures in the hospital.

The overall PACU stay did not show a tendency towards predicting duration of severe vital signs abnormalities or occurrence of SAEs during the first 4 postoperative days in the general ward after major abdominal surgery. Circulatory deterioration during the PACU stay was significantly correlated to circulatory ward complications. Thus, deterioration in the general ward is not detectable from PACU stay, despite a very high frequency of prolonged vital signs abnormalities and clinical SAEs. Future work should focus on detecting patient deterioration in the ward, as patients with an initial uncomplicated PACU stay may develop novel complications and vital signs abnormalities.