Incidence change of postoperative delirium after implementation of processed electroencephalography monitoring during surgery: a retrospective evaluation study

Data were retrieved from the postoperative intravenous patient-controlled analgesia (IVPCA) dataset of Chi Mei medical center from January 1, 2013, to December 31, 2017. This dataset was established for quality management to monitor side effects, including delirium. Chi Mei medical center is a tertiary health service unit and teaching hospital in southern Taiwan. Patients who were willing to use IVPCA for pain management after surgery were evaluated by the pain management group to ensure that their cognitive function was adequate to understand and operate the PCA pump correctly. Patients unable to operate the pump independently or who required prolonged sedation were not enrolled. Once the pump was set, the pain management team would visit the patient twice daily and follow up for 3 days. Records of 7425 eligible individuals were extracted from the dataset during this period. Patients who dropped out within 3 days or those without a complete hospitalization record were excluded (N = 110). Patients with missing detailed information were excluded (N = 24). Therefore, 7318 participants who used IVPCA after a noncardiac surgery under general anesthesia were enrolled in this study.

We performed a retrospective plausibility evaluation to assess the change in the incidence of POD after intraoperative pEEG during general anesthesia administered in July 2015. A plausibility evaluation controls confounders by comparing post-intervention and baseline groups to examine the improvements in target indicators [20]. The incidence of POD is represented per 1000 persons and divided into four quarters of the year. We analyzed the quarterly incidence of POD for distinct periods before and after intraoperative pEEG administration to examine if there was a significant change. Before the introduction of pEEG, minimal alveolar concentration measured from end-tidal anesthetic had been used as a general guide for immobility and decreasing awareness [21]. Processed electroencephalogram was introduced in July 2015 to monitor anesthetic depth and applied to all patients undergoing surgery under general anesthesia. The monitor started before induction throughout the surgery until the patient fully awakened, and all the staff was educated to adjust anesthetics according to the value and clinical presentation.

In this study, the patients using an IVPCA set were followed up to 3 days to ensure adequate pain control and evaluated side effects, including POD occurrence during the postoperative period. POD was initially screened by nurses in the pain management team using the Nursing Delirium Screening Scale. Based on this scale, patients were assigned a score of 0 to 2 for each of the five symptoms: disorientation, inappropriate behavior, inappropriate communication, illusion/hallucination, and psychomotor retardation (0 = not present, 1 = mild, and 2 = severe) [22]. We used ≥ 1 as the threshold, which provided higher sensitivity (67%) and adequate specificity (93%) [23]. After the initial assessment, an anesthesiologist further evaluated patients with delirium scores ≥ 1 and suspected delirium episodes. The diagnosis of POD was confirmed according to the Confusion Assessment Method diagnostic algorithm [24].

Given the sample sizes collected in the dataset and the incidences of POD reported in previous meta-analysis, the power to detect the difference in the primary outcome was calculated to be more than 90% on the basis of 2-sided ɑ<0.05 [19]. The occurrence of POD is presented as the frequency with percentage and compared using Pearson’s chi-square test between the periods before and after the implementation of intraoperative pEEG monitoring. The delirium score of patients within 72 h is presented as median with quartiles, and non-parametric testing was performed to examine the difference. Quarterly data of POD in PCA populations are shown as incidence per 1000 persons and the number of patients. Piecewise linear regression was used to estimate the mean and slope of the quarterly POD trend across the two periods. Subgroup analysis was performed to elucidate the incidence of POD stratified by four age groups: <45, 45–59, 60–74, and ≥ 75 years.

To analyze potential confounders for POD, patients with and without delirium were categorized into case and control groups in a case-control design. Participant information, including intraoperative pEEG monitoring, sex, the formula of PCA, ASA, and emergency, is presented as the frequency with percentage and age as mean with standard deviation. The differences in covariates between the case and control groups were examined using Pearson’s chi-square test and Fisher’s exact test for categorical variables, and Student’s t-test for age. Logistic regression analysis was used to estimate the risk factors for POD, including pEEG-guided anesthesia, age per 10 years, sex, PCA formula, ASA, emergency, and surgery type. Statistical analyses were conducted using SAS statistical software (version 9.4; SAS Institute, Inc., Cary, NC, USA) and StataCorp. 2017 (StataCorp LLC, Stata: Release 15. Statistical Software. College Station, TX, USA). Statistical significance was set at a P-value of < 0.05.

In this retrospective evaluation study, we examined the incidence change of POD after the implementation of intraoperative pEEG monitoring using the IVPCA dataset. From 2013 to 2017, we found that the incidence of POD within 3 days after surgery decreased from 1.18 to 0.41% after the implementation of intraoperative pEEG for all enrollees and from 5.1 to 1.56% for those aged ≥ 75 years.

The major strength of our study lies in the data retrieved from an IVPCA databank. Since all participants involved were IVPCA users, they had been adjudged to have an adequate cognitive function capable of operating the self-controlled analgesia set and making decisions independently. This ensured the baseline cognitive status of the study population. Another strength is that IVPCA users are more likely to achieve adequate pain control [26]. Targeting a pain score ≤ 3, the pain management group visited patients twice daily and aggressively modified the pain management plan. Consequently, we minimized the risk of biased results from poor cognitive function and uncontrolled pain, as either could be a strong factor contributing to delirium [7, 27, 28].

This study has some limitations. First, secular effects could not be excluded. This is inevitable when comparing postoperative outcomes before and after introducing a new monitor or concept. General anesthesia in the medical center was performed by approved anesthesiologists in line with the guideline. Standardized mean differences in demographic characteristics between the two periods show a mild difference (< 0.2). The significant association between pEEG-guided anesthesia and a lower POD rate is reconfirmed by multivariable analysis under a case-control design. Second, some factors—such as dynamic pEEG values fluctuating during surgery—are difficult to simplify into a single variable, which makes it difficult to control or adjust in the multivariable regression model. Therefore, this study can only infer the association between POD and pEEG administration but not the value maintained during surgery. Third, POD might be underdiagnosed in patients with hypoactive presentation [29, 30]. Although this was a retrospective observational study, it was built for quality management and monitoring of side effects, including delirium events, during the use of PCA. As a result, the pain management team was vigilant for any symptoms of POD. Fourth, intraoperative hemodynamic was not controlled as a covariate. However, our faculty was expected to maintain stability under general anesthesia following the same protocol, regardless of whether intraoperative pEEG monitoring was applied or not. Since the target population were intravenous patient-controlled analgesia users, patients who were likely to have poor outcomes requiring postoperative intensive care and/or prolonged sedation were not included in this study. In the second part of the study, we employed a case-control setting and adjusted for variables related to patient conditions, such as ASA classification, emergency cases, and surgery type, using a multivariable regression model. While these variables provided valuable insights, we acknowledge that they may not serve as a perfect proxy for intraoperative hemodynamic status. Finally, since the population in this study was entirely PCA users, the results may not be generalizable to patients with cognitive dysfunction or other populations. For example, minimally invasive surgeries, such as a ureteroscopy, may rarely require a PCA service.

Electroencephalogram is an objective and non-invasive method for assessing neurophysiological function. A processed electroencephalogram simplifies complex brain electrical activity into a numerical index, which allows anesthesiologists to monitor anesthetic depth during surgery and minimize burst suppression [31]. Burst suppression consisting of episodes of isoelectric waves stems from deep brain inactivation, such as that caused by general anesthesia [32]. This EEG pattern was deleterious and associated with several adverse clinical outcomes, including POD [8, 33, 34]. The use of pEEG monitoring during general anesthesia has been reported to reduce anesthetic consumption and improve postanesthetic recovery [13, 35]. Furthermore, pEEG-guided anesthesia was associated with a lower incidence of POD in randomized controlled trials [36, 37]. A meta-analysis in 2018 concluded that optimized anesthesia achieved with the assistance of pEEG could reduce the incidence of POD in patients aged ≥ 60 years after noncardiac procedure [12]. Nevertheless, the ENGAGES randomized controlled trial in 2019 did not reproduce the same result. Despite reducing anesthetic exposure and duration of EEG suppression, this study showed no preventive effect on the occurrence of POD with pEEG-guided anesthetic administration [18].

Our study found that the overall incidence of POD decreased from 1.18 to 0.41% after implementing pEEG-guided anesthesia (Table 1). No event of intraoperative awareness was noted. The incidence of POD was pretty low but still within the wide range reported in previous literature [38]. One reason for the low incidence of POD is the inclusion of relatively young patients. Age is a strong factor in POD occurrence [39]. The incidences of POD in the age group ≧ 75 years are 5.1% and 1.56% before and after intraoperative pEEG, respectively. The low incidence could partly stem from the fact that patients with preexisting cognitive dysfunction were not involved as IVPCA users and adequate pain management. The trends of incidence change is presented in Fig. 1. Although no instant decrease in POD incidence in the first quarter after intraoperative pEEG-guided anesthesia administration, it reduced gradually in the following quarters. Despite well educated, clinicians reserve the flexibility of making decision according to clinical scenarios; that is, it might require some time for practitioners to recognize pEEG as a reliable tool and involve the monitor in the clinical decision-making process. Essentially, more than the pEEG results in interpreting the results and subsequent prompt management provide benefits to the patients [40].

We further analyzed the risk factors for POD in a case-control design (Table 2). After adjusting for other covariates, the odds ratio for POD would increase by 2.53 (95% CI 1.96–3.27) for every 10-year increase in age. This result is compatible with the odds ratio of 2.0 per 10 years reported by Raats et al. in 2015 [41]. While ASA classification was also mentioned as a risk factor, its association with POD was not significant in our study after adjustment for other covariates. Furthermore, patients who underwent emergency surgery had higher odds of POD (aOR 3.59, 95% CI 1.73–7.47) than those who underwent regular procedures, which is concordant with 1.5 to three times increased risk revealed in previous studies [42, 43]. Even under the setting of a case-control design, intraoperative pEEG monitoring was associated with a significantly lower risk of POD (aOR 0.33, 95% CI 0.18–0.60) after adjustment in a multivariable regression model.

Age is widely recognized as an important predisposing factor for POD [44]. Thus, we stratified patients by age groups to see if the association between intraoperative pEEG monitoring and lower incidence of POD differs in each subgroup. We found that patients of advanced age (≥ 75 years) were the only subgroup associated with a significantly lower incidence of POD (Fig. 2). One reason could be that the effect size in this subgroup was large enough to be detected. While the incidences of POD in other subgroups were less than 1%, it occurred in 50.97 per 1000 persons in the group ≥ 75 years before pEEG administration and reduced to 15.56 per 1000 persons after pEEG administration. Further, the tendency of the higher benefit of intraoperative pEEG to a more susceptible population could have been another factor. Lower brain anesthetic resistance detected by a pEEG is associated with POD [45]. As a result, anesthesia guided by pEEG might have benefitted this vulnerable population by reducing unnecessary anesthetics exposure.

This retrospective observational study was approved by the Institutional Review Board (IRB-10009-012) of Chi Mei Medical Center, Tainan, Taiwan (Chairperson Prof Chung-Hsi Hsing) on 10 August 2018. The study participants provided verbal informed consent for using their medical information, which was approved by the IRB of Chi Mei Medical Center, Tainan, Taiwan (IRB-10009-012, Chairperson Prof Chung-Hsi Hsing) due to the noninvasive nature of the procedure. Medical records in the IVPCA dataset were de-identified and rebuilt for quality management and research purposes. All methods were carried out in accordance with relevant guidelines and regulations.

Not applicable.

The authors declare no competing interests.

The datasets used and analysed during the current study are available from the corresponding author on reasonable request.