Postoperative delirium is an independent factor influencing the length of stay of elderly patients in the intensive care unit and in hospital

This is a subgroup analysis of an observational prospective single-centre trial on patients from different surgical disciplines of the University Hospital Bonn. The entire study, conducted from September 2018 to October 2019 under the title "PRe-Operative Prediction of postoperative DElirium by appropriate SCreening (PROPDESC)" included 1097 patients [25]. It was registered in the German Registry for Clinical Studies under the number DRKS00015715 and was approved by the local institutional Ethics Committee at the Medical Faculty of the Rheinische Friedrich-Wilhelms-University of Bonn. Written informed consent was obtained from each patient. Patients with age 60 and older and with a planned surgery duration of at least 60 min were eligible for the PROPDESC study. Exclusion criteria were emergency procedures, language barriers or missing compliance with the study protocol.

The subgroup analyzed here included all enrolled patients aged 70 and older. The patient data pertain to the inpatient period and the discharge date.

In this subgroup analysis, 15 variables were included. Preoperative data collected include the following: age, sex, body-mass-index (BMI), cognitive impairment tested with the Montreal Cognitive Assessment (MoCA), hearing impairment (yes or no), POD in the medical history (yes or no), the number of long-term medication, American Society of Anesthesiologists (ASA) Physical Status Classification System, Revised Cardiac Risk Index (rCRI), New York Heart Association Classification (NYHA), Metabolic Equivalent of Tasks (MET), surgical risk and surgical discipline. Surgical risk was transformed from a 5-level Johns-Hopkins classification to the 3-level modified Johns-Hopkins surgical criteria [26, 27]. Intraoperative data collected include red blood cell transfusion and ventilation time. Postoperative data collected include surgery duration, length of stay (LOS) in the intensive care unit (ICU) and LOS in hospital.

The primary endpoint of POD was assessed on the first five consecutive days after surgery, alternatively after the end of sedation. Sedated patients with RASS [28] score < − 3 were considered as not assessable and therefore their testing for POD was initiated after exceeding this level of sedation according to Confusion Assessment Method for ICU (CAM-ICU) [29].

Trained doctoral students performed the testing. In order not to miss a positive POD diagnosis, different tests were applied in the PROPDESC study. CAM-ICU was used for intensive care patients and Confusion Assessment Method (CAM) and the 4 ‘A’s (4AT) were conducted in patients on the normal ward [29–31]. To avoid missing delirium diagnosis in the context of spot examinations, the Delirium Observation Scale (DOS) was additionally applied by interviewing the nursing staff to assess the previous 24 h [32]. The positive endpoint POD was considered if one of the applied delirium assessments was positive on at least one visit day. The definition of completed POD assessment required a valid conduct of at least three of the five scheduled postoperative visits. Discharge home before the third visit was accepted as an exception to this rule, on the assumption that patients would not subsequently become delirious in their familiar environment.

Statistical analysis was performed using the statistical programming environment R. Continuous and ordinal variables are presented with mean and standard deviation (SD±). Nominal variables are displayed as numbers and percentages. Patients were divided into two groups (non-POD vs. POD group) based on the POD endpoint. The difference between these groups regarding the characteristics was analyzed using the non-parametric Wilcoxon rank-sum test for continuous variables. For categorical variables, Fisher´s exact test was computed to check for independence.

To evaluate the independent effect of POD on LOS in ICU and in hospital, a multi-variable linear regression analysis was performed to adjust for various perioperative potential confounders. The LOS outcomes were log-transformed to ensure approximate normality of residuals. POD was entered as a binary variable while adjusting for perioperative risk factors for POD. These covariates were preoperative age, ASA, NYHA, MET, rCRI classification levels, MoCA sum score, hearing impairment, history of delirium, number of medication and intra-/postoperative surgical risk, surgical discipline, duration of surgery, red cell blood transfusion and ventilation time. In conjunction with the multivariable linear regression analysis related to the effect of POD on LOS in ICU, only patients in the cohort who actually had an ICU stay were included and ventilation time was removed in the risk adjustment because it contains part of the outcome parameter LOS in ICU. To ensure the interpretability, the coefficients of POD from modelling the log LOS were re-transformed and are presented in the percent increase (compared to non-POD) with a corresponding 95% confidence interval. For sensitivity reasons, the regression analyses were repeated also without covariates, which contained more than 5% missing values to check for potential biases induced due to the missing observations.

The mean age of the subgroup analyzed here was 77.2 (± 4.7) years and the gender distribution was 248 (42.2%) women and 340 (57.8%) men. POD incidence was 25.9% (152). Table 1 shows the variables collected preoperatively and postoperatively divided into the non-POD and POD groups. Anesthetic classifications (ASA, p < 0.001; NYHA, p < 0.001; MET, p = 0.009 and rCRI, p < 0.001) were significantly different between the delirious and non-delirious groups. Furthermore, POD patients (21.4 ± 4.1) showed a significantly lower MoCA sum score than non-POD patients (22.8 ± 3.7; p < 0.001). The surgical risk of the cohort was significantly higher in patients who developed POD postoperatively. As expected the highest POD incidence occurred in patients after cardiac surgery with 52.6% (80). Postoperative variables also differed significantly between the delirious and non-delirious groups. POD patients had an average of 78 min longer surgery duration (non-POD 188.9 ± 115.1; POD 266.8 ± 121.9; p < 0.001). Furthermore, the duration of ventilation differed by about 25 h. In the non-delirious patients, the mean value of ventilation time was 7 h (7.44 ± 13.4) and in the delirious patients about 32 h (32.2 ± 24.8; p < 0.001). The length of intensive care stay differed by an average of 138 h between delirious (21.4 ± 62.4) and non-delirious patients (159.3 ± 525.4; p < 0.001). This makes a difference in the LOS in ICU of about 6 days. Also shown was a significant difference in total LOS in the hospital of about 8 days. Patients who developed POD during the visit period stayed about 26 days (25.6 ± 17.2) and patients without POD stayed an average of 17 days (17.2 ± 25.7; p < 0.001).

To obtain a more accurate overview of the cohort that was postoperatively in intensive care, the subgroup was divided into two groups (non-ICU and ICU). Table 2 compares the intraoperative variables and the postoperative POD-Outcome for the ICU (267; 45.4%) and non-ICU (313; 53.2%) group. The average stay of intensive care patients was 123.4 h (± 399.4). Fifty-five point four percent (148) of the patients on ICU underwent cardiac surgery and 71.5% (191) had a high-risk surgery. Patients with a postoperative intensive care stay showed an average of 149 min longer surgery duration (non-ICU 140.9 ± 77.6 min; ICU 290.1 ± 114.9 min; p < 0.001). Furthermore, the two groups differed significantly in red blood cell transfusion (non-ICU 51.1 ± 277.4 ml; ICU 856.0 ± 2136.6 ml; p < 0.001) and ventilation time (non-ICU 3.6 ± 1.5 h; ICU 26.0 ± 95.7 h; p < 0.001). The non-ICU and ICU groups also differed significantly in postoperative outcomes related to POD and overall LOS in hospital. Forty-two point three percent (113) of intensive care patients developed POD and only 11.5% (36) of non-intensive care patients were tested positive. Patients with ICU stay showed an average of 9 days longer total hospital stay (non-ICU 15.4 ± 20.6 days; ICU 24.0 ± 30.3 days; p < 0.001).

Linear regression results confirm POD as an independent predictor of LOS in the ICU after risk adjustment with perioperative variables (Table 3). Following our model, patients with POD have a 36% (95% CI 4–78%; p < 0.001) increase in LOS in ICU independently from their perioperative risk factors. A sensitivity analysis fitting the same regression model without the variable red blood cell transfusion (which contains n = 66 missing values) led to very similar results (43% increase; 95% CI 10–86%; p < 0.001). Furthermore, the linear regression model confirms that patients with a POD have a 22% (95% CI 4–43%; p < 0.001) increase in LOS in hospital after adjusting with perioperative variables. Again, the sensitivity analysis without adjusting for red blood cell transfusion supports this (25% increase; 95% CI 8–45%; p < 0.001).

The POD incidence in this subgroup analyses with patients aged 70 and older was 25.9%. Several preoperative variables showed a significant difference between the POD and non-POD groups. As confirmed in the literature, there was a significant difference in preoperative cognitive testing with the MoCA and positive POD assessment [3–5, 33, 34]. In addition, further studies have shown that cognitive impairment may also have an impact on prolonged hospital stay [35]. However, cognitive impairment as a major preoperative risk marker for POD has been strongly described in systematic reviews as well as in the ESAIC Guideline. Based on the preoperative anesthetic classifications (ASA, NYHA, MET, rCRI), the patients who developed POD were also classified with more pre-existing clinically relevant conditions. Furthermore, the preoperative assessed surgical risk was on average higher in the POD group than in the non-POD group. The ESAIC guidelines recommend, based on their systematic analysis of the study evidence, that ASA classification should be considered a pre-operative risk marker for POD. Furthermore, it is recommended that the factor of surgical risks should also be considered in the risk analysis for POD. These findings are congruent with the POD risk factors described in literature and guidelines [5, 36]. Contrary to what has been reported in the literature, the POD patients in this subgroup did not show significant differences in hearing impairment and a prior POD experience, relative to the non-POD group [37].

POD is a multifactorial complication in which both preoperative predisposing factors as well as intraoperative and postoperative precipitating factors contribute to its development. Significant contributors to the development of POD are the duration time of surgery and the period of ventilation. Patients who developed POD showed a significantly longer operation time of 78 min on average, a longer ventilation time of 25 h and a longer stay in the intensive care unit of 138 h in this subgroup analysis. It should be noted here that outliers, especially in the POD group, influence the time values of ventilation duration and intensive care stay.

The literature describes, in particular, the large influence of the ICU stays on POD. Due to this significant influence of the ICU stay, this cohort of the subgroup analyzed here was considered in more detail. According to the existing results in the literature, the patients with a subsequent ICU stay had a significantly longer operation time of 149 min more on average, a significantly larger amount of blood transfusion of 805 ml, and a longer ventilation time of 22 h. As mentioned above again a few outliers, especially in the ICU group, characterize the values for ventilation duration and transfusion volume. Furthermore, the results showed that the cohort of patients with an ICU stay also developed POD significantly more often than the opposite group. These results also confirm the findings of other studies that patients with an ICU stay are much more likely to develop POD [38–41]. Another observational study also looked at the occurrence of POD in the ICU and found that POD monitoring alone improved patient outcome [42].

There is various evidence in the literature that patients with a POD or ICU stay have a longer LOS in the hospital [3, 18, 19]. In this regard, we wanted to use this subgroup analysis to show more precisely whether the total length of hospital stay was influenced more by the fact of a necessary ICU stay or primarily by the secondary diagnosis of POD. Patients who developed POD had an average longer hospital stay of about 8 days. However, patients with an ICU stay had a longer average hospital stay of 9 days compared to patients without an ICU stay.

To test whether the occurrence of POD influences LOS in the ICU and in the hospital, a linear regression model was performed, risk-adjusted for perioperative risk factors in both cases. The results of our model showed that the development of POD resulted in a 36% increase in LOS in the ICU independent of perioperative risk factors. Furthermore, the results confirm that patients with POD had a 22% increase in-hospital LOS after risk adjustment. Confirming our findings, another study also found that POD is a robust predictor of LOS in ICU and also has a significant impact on the morbidity and mortality of patients undergoing surgery [43]. From these results, it can be concluded that POD has an independent impact on LOS in ICU and in hospital. An intervention study addressed the problem of POD and prolonged ICU stay and found that a more extended ICU visit model can reduce both POD incidence and LOS [44]. Through the results of our analyses and the supporting findings of the literature, the importance of POD issues for elderly patients overall and specifically for ICU patients is demonstrated. Based on the known risk factors for POD and prolonged ICU stay, risk screening and interventions for prevention need to be further explored and applied in routine clinical practice.

This study has several limitations. A limitation is that the positive delirium diagnosis is based on the results of the delirium tests and not on a diagnosis by a psychologist. Another limitation to be mentioned is that although the regression analysis has included certain risk factors for postoperative delirium, there may be other unobserved confounders. In addition, it has to be considered that the analysis carried out here is a subgroup that exclusively observes patients over 70 years of age.