Effectiveness of non-pharmacological intervention protocol for prevention of postoperative delirium in the surgical intensive care unit

Abstract

Postoperative delirium (POD) is a common adverse event in patients admitted to the intensive care unit (ICU). We aimed to determine the effectiveness of a multicomponent non-pharmacological intervention protocol to reduce the incidence of POD in elderly patients admitted to the surgical ICU (SICU). This before-and-after cohort study included 300 patients aged ≥ 65 years who were admitted to the SICU within 7 days postoperatively with an anticipated SICU stay > 24 h. During the pre-intervention period, patients received medical care based on the attending physicians. While during the intervention, patients received the same medical care plus a multicomponent non-pharmacological intervention protocol. POD was monitored twice daily using the Confusion Assessment Method for the ICU. Demographic and clinical data during SICU stay were collected and compared between the pre-intervention and intervention periods. The primary outcome was POD incidence. The secondary outcomes were POD duration, delirium-free days, and other clinical outcomes. The incidences of POD during the pre-intervention and intervention periods were not different (40.0% vs. 38.0%, P = 0.723; OR 0.92, 95% CI 0.58–1.46). Multivariate regression analyses with two different models demonstrated that the multicomponent non-pharmacological intervention protocol was not associated with POD prevention (OR 0.70, 95% CI 0.39–1.25 for Model 1 and OR 0.63, 95% CI 0.37–1.08 for Model 2). The protocol was associated with lower incidence of SICU events, particularly self-removal of endotracheal tube and nosocomial infection. Implementation of the multicomponent non-pharmacological intervention protocol was not associated with POD prevention in elderly patients admitted to the SICU.

Trial registration Thai Clinical Trials Registry. Trial No. TCTR20181201001. Retrospective registered 01 December 2018.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-86768-4.

Delirium is a clinical syndrome consisting of acute onset of alteration and fluctuation of the level of consciousness, attention, and cognition¹. Although there is no consensus on the definition, postoperative delirium (POD) is usually referred to as delirium developing between 24 and 72 h postoperatively^1–3. The pathophysiology of POD is a multifactorial process, and the postulated mechanism is acute neurological insult caused by proinflammatory mediators and alteration in the balance of neurotransmitters in the brain in vulnerable patients^2–4. The development of POD is associated with various precipitating factors, such as aging, baseline cognitive function, pre-existing comorbidities, acuity of illness, admission to the intensive care unit (ICU), some types of surgery such as cardiac or major vascular surgery or emergency surgery, pain, some medication, unfamiliar environment, physical restraint, and sleep deprivation^4–6. The incidence of POD ranges between 13% and 29% in patients with a high risk of POD, such as elderly surgical patients^7–10. Moreover, patients who develop delirium are more likely to have adverse outcomes, such as reoperation and readmission⁹, prolonged length of stay (LOS)^7,8, increased hospital mortality^7,8, and long-term cognitive decline¹¹.

For decades, various interventions, including pharmacological and non-pharmacological measures, have been widely investigated to prevent or decrease the risk of development of delirium^4–6. The outcomes of implementing non-pharmacological interventions have been promising^12–14, although some data are conflicting. In 1999, Inouye et al. demonstrated that the multicomponent intervention strategy significantly reduced the incidence of delirium from 15.0 to 9.9% in elderly patients admitted to the general medicine service¹⁵. On the other hand, Bryczkowsk et al. demonstrated no difference in the incidence of delirium in patients admitted to the surgical ICU (SICU) after implementing a delirium prevention program including a single non-pharmacological intervention¹⁶. Therefore, this study aimed to determine the effectiveness of a multicomponent non-pharmacological intervention protocol to reduce the incidence of POD in elderly patients admitted to the SICU.

Methods

Research ethics

This study was approved by the Institutional Review Board of Siriraj Hospital, Faculty of Medicine Siriraj Hospital, Mahidol University, Bangkok, Thailand (COA No. Si 211/2018) and was retrospectively registered at the Thai Clinical Trials Registry (TCTR20181201001). Our research team screened all patients admitted to the SICU for inclusion and exclusion criteria. Eligible patients, or their proxies in cases where patients were sedated, were contacted to provide detailed information about the study and to obtain written informed consent for participation before inclusion. The study was conducted following the guidelines and regulations of our faculty and university. The manuscript was prepared and reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines (https://www.equator-network.org/reporting-guidelines/strobe/).

Setting and patient population

This before-and-after cohort study was conducted at the SICU of a tertiary university-based hospital in Bangkok, Thailand. This SICU is a closed ICU model with 14-bed capacity, staffed 24/7 by anesthesia-based intensivists, critical care fellows, and anesthesiology residents with 1:1 of nurse-to-patient ratio. Patients undergoing vascular, abdominal, urological, head and neck, orthopedic, plastic, otorhinolaryngologic, gynecologic, obstetric, and ophthalmologic surgeries who require intensive perioperative care are admitted to this SICU. Family members are allowed to visit patients daily from 11 a.m. to 8 p.m.

All patients admitted to the SICU between June 2018 and November 2021 were eligible for enrollment. Patients included were aged ≥ 65 years, admitted to the SICU during 7 days following their operation, and had an anticipated SICU stay of > 24 h. We excluded patients who underwent surgery > 7 days prior to SICU admission, those who did not undergo surgery, and those who could not communicate in Thai or had severe visual and auditory impairments that preclude communication.

Study procedure

After an extensive literature review, a multidisciplinary team consisting of intensivists, surgeons, geriatricians, psychiatrists, physiatrists, and critical care nurses was set up and developed a multicomponent non-pharmacological intervention protocol to prevent POD^12–14. The multicomponent non-pharmacological interventions comprised seven tasks, including orientation, cognition, ambulation based on mobilization levels^17–19, pain assessment and management using the Thai version of the Critical Care Pain Observation Tool²⁰, clearing ears and eyes, sleep promotion, and medication review (Supplementary file). The pre-intervention and intervention cohorts included patients before and after the protocol implementation. There was a 3-month interval between the two periods for implementing, educating, and training the protocols to attending staffs, fellows, residents, critical care nurses, and physical therapists who attended patients in the SICU. The education and training program included formal lectures, case-based learning, and bedside hands-on coaching. In the pre-intervention period, all the care processes, such as medication, fluid administration, and other life support, were provided at the discretion of the responsible attending staffs. The pain and sedation protocol implemented in the SICU followed guideline recommendations²¹, and a written protocol for daily assessment and mechanical ventilation weaning²² was also implemented. In the intervention period, included patients received the medical care in the same manner as during the pre-intervention period plus the multicomponent non-pharmacological intervention protocol for POD prevention (Supplementary file) which was implemented within 24 h after SICU admission and continued for 28 consecutive days or until they were discharged from the SICU or death, whichever came first. Trained healthcare providers who were not members of the research team and were assigned to care of patients during each shift performed each intervention based on the multicomponent non-pharmacological intervention protocol. In both pre-intervention and intervention periods, well-trained research assistants screened POD using the Thai version of the Confusion Assessment Method for the ICU (CAM-ICU)²³twice daily between 08:00 and 10:00 in the morning and between 16:00 and 20:00 in the evening. Screening was performed within 24 h after SICU admission and continued for 28 consecutive days or until patients were discharged from the SICU or death, whichever came first. In the case of positive CAM-ICU, POD was further classified as either hyperactive, hypoactive, or mixed type.

Data collection

The research team recorded demographic data including age, sex, comorbidities, smoking and alcohol consumption status, acuity of illness including diagnosis and operation type (either elective or emergency) and site (abdomen, vascular, urologic, orthopedics, gynecologic, head and neck) of surgery, Acute Physiology and Chronic Health Evaluation (APACHE) II score, Sequential Organ Failure Assessment (SOFA) score, requirement of inotropes/vasopressors and ventilator support, sepsis, and laboratory values at SICU admission. In both pre-intervention and intervention cohorts, POD was defined as positive CAM-ICU arising at any time during the screening period and was counted as a binary outcome as either present or absent, as well as the total number of days with POD during the screening period. Delirium-free days were defined as the total number of days without POD during 28 consecutive days after inclusion. If the patient was discharged from the SICU before 28 days, the patient was considered to not develop POD subsequently. If the patient died before 28 days, the delirium-free day was zero. In the intervention cohort, the non-pharmacological intervention protocol performed by trained health care providers would be self-checked in the checklist chart, and the adherence to the protocol was determined as the ratio of the number of times that the protocol was performed to the total number of times that it should be given.

Outcomes and sample size calculation

The primary outcome was the incidence of POD in elderly patients admitted to the SICU before and after the implementation of the multicomponent non-pharmacological intervention protocol. The secondary outcomes were POD duration, delirium-free day, mechanical ventilation duration, presence of adverse events in the SICU (self-removal of tubes, lines and drains, agitation-related self-injury, and SICU-acquired infection), SICU and hospital LOS, and SICU and hospital discharge status. Based on our previous study²⁴, we estimated a 46.5% incidence of POD in this patient population. We expected a 34% reduction in the incidence of POD after implementing the protocol based on the study results of Inouye et al.¹⁵, which was equal to 30.7%. With 80% power and 95% confidence interval (CI), a sample size of 148 subjects per group was required. We planned to include 150 subjects in each cohort.

Statistical analysis

For descriptive statistics, continuous data were presented as mean with standard deviation or median with interquartile range (IQR) depending on their distribution, and categorical data were presented as number with percentage. To compare the pre-intervention and intervention cohorts, unpaired t- or Mann–Whitney U test and Chi-square or Fisher’s exact test were used for continuous and categorical data, respectively. Univariate analyses were used to identify the potential factors associated with POD. To assess the relationship between the implementation of the multicomponent non-pharmacological intervention protocol and the incidence of POD, multivariate logistic regression analyses were conducted. In Model 1, the protocol was included as the dependent variable, along with variables identified with a P value < 0.05 in univariate analyses. Additionally, Model 2 incorporated potential risk factors for POD based on findings from our previous study²⁴, including age, underlying diseases such as dementia and diabetes mellitus, disease severity (SOFA score), perioperative use of benzodiazepines, and mechanical ventilation. To address potential multicollinearity, variance inflation factors (VIFs) were calculated for all independent variables included in the model and variables with a VIF > 10 were excluded. Subsequently, the adjusted odds ratios (OR) for POD with 95% confidence intervals (CI) were calculated and reported. The post-hoc sensitivity analysis was conducted to balance baseline characteristics between patients in the pre-intervention and intervention cohorts using propensity score matching. The propensity score was derived from the multivariate logistic regression model where the protocol was treated as the dependent variable, alongside other variables showing a P value < 0.05 in univariate analyses. Patients in the pre-intervention period were then matched with those in the intervention period based on the closest propensity score in a 1:1 ratio. Subsequently, the association between the protocol and POD was reanalyzed using two previously described multivariate logistic regression models. For all analyses, a two-tailed test was performed, and P value < 0.05 was considered statistically significant. Data were prepared and analyzed using IBM SPSS Statistics, version 28 (IBM Corp, Armonk, NY, USA).

Results

Patent demographic data and SICU data

Between July 2018 and November 2021, 150 and 150 patients were included during the pre-intervention and intervention periods, respectively (Fig. 1). Patients in the intervention cohort had significantly more comorbidities, including hypertension, cardiac diseases, dementia, previous stroke, psychiatric diseases, diabetes mellitus, and chronic kidney disease, and more frequently used statins and benzodiazepine preoperatively (Table 1). Patients in the intervention cohort underwent emergency surgery more frequently than those in the pre-intervention cohort. Most patients in both groups underwent abdominal and vascular surgeries. The operative time, intraoperative fluid balance, and intraoperative adverse events were not significantly different between two cohorts (Table 1).

Fig. 1.

Study flow.

Table 1.

Demographic data and intraoperative data compared between patients included during pre-intervention and intervention periods.

	Entire cohort			Propensity score matching cohort
	Pre-intervention (n = 150)	Intervention (n = 150)	p value	Pre- intervention (n = 78)	Intervention (n = 78)	p value
Age, year	75.1 ± 7.5	76.7 ± 8.1	0.083	75.6 ± 7.7	75.9 ± 7.9	0.773
Male gender	84 (56.0%)	67 (44.7%)	0.050	40 (51.3%)	36 (46.2%)	0.522
Comorbidities
Hypertension	101 (67.3%)	122 (81.3%)	0.006	63 (80.8%)	57 (73.1%)	0.254
Cardiac diseases	37 (24.7%)	57 (38.0%)	0.013	22 (28.2%)	22 (28.2%)	> 0.999
Dementia	20 (13.3%)	54 (36.0%)	< 0.001	16 (20.5%)	15 (19.2%)	0.841
Previous stroke	13 (8.7%)	25 (16.7%)	0.037	8 (10.3%)	7 (9.0%)	0.786
Psychiatric diseases	1 (0.7%)	9 (6.0%)	0.019	1 (1.3%)	1 (1.3%)	> 0.999
Diabetes mellitus	41 (27.3%)	61 (40.7%)	0.015	28 (35.9%)	29 (37.2%)	0.868
Chronic kidney disease	30 (20.0%)	51 (34.0%)	0.006	22 (28.2%)	22 (28.2%)	> 0.999
Cirrhosis	6 (4.0%)	12 (8.0%)	0.145	1 (1.3%)	8 (10.3%)	0.034
Current smoking	29 (19.3%)	30 (20.0%)	0.885	12 (15.4%)	21 (26.9%)	0.078
Current alcohol drinking	12 (8.0%)	6 (4.0%)	0.145	4 (5.1%)	5 (6.4%)	0.731
Preoperative statins used	58 (38.7%)	89 (59.3%)	< 0.001	40 (51.3%)	42 (53.8%)	0.873
Preoperative benzodiazepine used	7 (4.7%)	17 (11.3%)	0.033	5 (6.4%)	8 (10.3%)	0.564
Type of surgery
Elective	86 (57.3%)	68 (45.3%)	0.038	42 (53.8%)	42 (53.8%)	> 0.999
Emergency	64 (42.7%)	82 (54.7%)		36 (46.2%)	36 (46.2%)
Site of surgery
Abdomen	79 (52.7%)	68 (45.3%)	0.019	39 (50.0%)	33 (42.3%)	0.379
Vascular	43 (28.7%)	48 (32.0%)		25 (32.1%)	27 (34.6%)
Urologic	3 (2.0%)	13 (8.7%)		1 (1.3%)	6 (7.7%)
Orthopedics	8 (5.3%)	14 (9.3%)		6 (7.7%)	7 (9.0%)
Gynecologic	1 (0.7%)	1 (0.7%)		1 (1.3%)	0 (0.0%)
Head and Neck	16 (10.7%)	6 (4.0%)		6 (7.7%)	5 (6.4%)
Operative time, min	187.5 (100–325)	175 (105–270)	0.431	187.5 (95–280)	212.5 (118.5–367.5)	0.317
Intraoperative fluid balance
Intake, ml	2613.5 (958–4490)	1999 (900–4457)	0.270	2000 (776.5–3922.5)	2471.5 (1085–5474)	0.129
Estimated blood loss, ml	435 (85 − 1225)	260 (40 − 800)	0.145	300 (45–1100)	400 (50–2000)	0.327
Transfusion	94 (62.7%)	93 (62.0%)	0.905	46 (59.0%)	49 (62.8%)	0.743
Intraoperative events
Hypotension	127 (84.7%)	120 (80.0%)	0.289	63 (80.8%)	62 (79.5%)	> 0.999
Desaturation	9 (6.0%)	8 (5.3%)	0.803	4 (5.1%)	5 (6.4%)	0.731

Data are presented as mean ± standard deviation, median (interquartile range) or number (%) as appropriate.

For the SICU data, patients in the intervention cohort had significantly lower APACHE II score than those in the pre-intervention cohort [median (IQR), 12 (10–15) vs. 14 (11–19), P = 0.001] but more frequently presented with shock (54.0% vs. 26.7%, P < 0.001) and active infection (48.7% vs. 26.0%, P < 0.001). The SOFA score at SICU admission and baseline laboratory data was not significantly different between both cohorts, except for a significantly lower serum sodium level and higher blood urea nitrogen in the intervention cohort (Table 2). The number of patients receiving mechanical ventilation, restraints, and urinary catheterization, and those in a comatose state (defined as Richmond Agitation Sedation Scale − 4 or − 5) was not significantly different. Pain score and opioid administration did not differ between both cohorts. Although benzodiazepine use was significantly higher in the intervention cohort than in the pre-intervention cohort (25.3% vs. 16.0%, P = 0.046), the cumulative benzodiazepine dose was not significantly different between both cohorts (Table 2). The overall adherence to the non-pharmacological intervention protocol was 74.6%.

Table 2.

SICU admission data and SICU daily assessment data compared between patients included during pre-intervention and intervention periods.

	Entire cohort			Propensity score matching cohort
	Pre-intervention (n = 150)	Intervention (n = 150)	p value	Pre- intervention (n = 78)	Intervention (n = 78)	p value
APACHE II score	14 (11–19)	12 (10–15)	0.001	12 (10–14)	12 (10–16)	0.746
SOFA score	4 (2–6)	4 (2–7)	0.842	3.5 (2–6)	4 (2–7)	0.583
Presence of shock	40 (26.7%)	81 (54.0%)	< 0.001	27 (34.6%)	33 (42.3%)	0.411
Active infection	39 (26.0%)	73 (48.7%)	< 0.001	26 (33.3%)	27 (34.6%)	> 0.999
Laboratory values
Hematocrit, %	31.9 (28.0–36.9)	31.0 (27.2–35.0)	0.123	32.0 (28.0–38.0)	30.9 (27.4–34.3)	0.142
Sodium, mmol/L	138 (135–141)	137 (135–140)	0.011	138.5 (136–141)	138 (135–141)	0.929
Bicarbonate, mmol/L	19.0 (16.0–22.0)	19.0 (16.3–21.0)	0.680	19 (15–22)	19 (17–22)	0.862
Blood urea nitrogen, mg/dl	17.8 (12.4–29.0)	21.2 (14.3–34.7)	0.042	19.0 (13.9–33.9)	21.2 (14.0–32.4)	0.645
Creatinine, mg/dl	1.07 (0.72–1.67)	1.12 (0.76–1.66)	0.794	1.24 (0.80–1.96)	1.16 (0.71–1.93)	0.667
Albumin, g/dL	2.7 (2.3–3.1)	2.7 (2.3–3.2)	0.772	2.8 (2.5–3.2)	2.8 (2.4–3.2)	0.882
Serum glucose	143 (120–179)	144 (119–188)	0.667	139 (119–189)	141.5 (116–193)	0.860
Receiving mechanical ventilation	126 (84.0%)	131 (87.3%)	0.410	62 (79.5%)	68 (87.2%)	0.283
Receiving restraint	74 (49.3%)	79 (52.7%)	0.564	35 (44.9%)	41 (52.6%)	0.423
Presence of coma	18 (12.0%)	18 (12.0%)	> 0.999	7 (9.0%)	9 (11.5%)	0.793
Duration of coma	3 (1–5)	2 (1–3)	0.146	2.5 (1–6)	2 (1–3)	0.529
Pain score	1 (0–3)	2 (0.5–3)	0.255	1 (0–3)	2 (1–3)	0.097
Opioid used	139 (92.7%)	140 (93.3%)	0.821	72 (92.3%)	73 (93.6%)	0.754
Fentanyl equivalent cumulative dose, mcg	580 (150–1500)	470 (90–1800)	0.576	400 (101.5–1272.5)	645 (117.5–2455)	0.183
Benzodiazepine used in SCIU	24 (16.0%)	38 (25.3%)	0.046	10 (12.8%)	19 (24.4%)	0.099
Benzodiazepine cumulative dose, mg	3 (2–34)	3 (2–5)	0.315	2.5 (1–5)	4 (2–5)	0.484

Data are presented as median (interquartile range) or number (%) as appropriate.

Coma is defined as RASS − 4 to -5.

APACHE II, Acute Physiology and Chronic Health Evaluation II score; RASS, Richmond Agitation Sedation Scale; SOFA, Sequential Organ Failure Assessment score.

Primary outcome of the incidence of POD

The POD incidence did not significantly differ between the pre-intervention and intervention cohorts (40.0% in the pre-intervention cohort vs. 38.0% in the intervention cohort, P = 0.723; OR 0.92, 95% CI 0.58–1.46) (Table 3; Fig. 2). The multivariate logistic regression analyses showed no association between the implementation of the protocol and the incidence of POD (OR 0.70, 95% CI 0.39–1.25 for Model 1 and OR 0.63, 95% CI 0.37–1.08 for Model 2) (Fig. 2).

Table 3.

Clinical outcomes compared between patients included during pre-intervention and intervention periods.

	Entire cohort			Propensity score matching cohort
	Pre-intervention (n = 150)	Intervention (n = 150)	p value	Pre-intervention (n = 78)	Intervention (n = 78)	p value
Postoperative delirium	60 (40.0%)	57 (38.0%)	0.723	31 (39.7%)	25 (32.1%)	0.404
POD subtype*
Hypoactive	16 (26.7%)	4 (7.0%)	0.007	10 (32.3%)	2 (8.0%)	0.042
Hyperactive	17 (28.3%)	28 (49.1%)		8 (25.8%)	13 (52.0%)
Mixed	27 (45.0%)	25 (43.9%)		13 (41.9%)	10 (40.0%)
POD onset*	2 (1–4)	2 (1–4)	0.732	2 (1–3)	2 (1–4)	0.685
POD duration, day*	3.5 (1–7.8)	2.5 (1–4.5)	0.125	2 (1–7)	2 (1–3.5)	0.306
Delirium-free day*	24.5 (20–27)	25 (23.5–27)	0.201	26 (21–27)	26 (24–27)	0.434
Duration of mechanical ventilation, day	3 (1–7)	3 (1–5)	0.283	3 (1–5)	2 (1–5)	0.730
SICU events
Any event	94 (62.7%)	51 (34.0%)	< 0.001	49 (62.8%)	24 (30.8%)	< 0.001
Self-removal of tube	21 (14.0%)	9 (6.0%)	0.021	13 (16.7%)	4 (5.1%)	0.037
Self-removal of line and drain	29 (19.3%)	22 (14.7%)	0.282	19 (24.4%)	12 (15.4%)	0.228
Nosocomial infection	41 (27.3%)	25 (16.7%)	0.026	15 (19.2%)	11 (14.1%)	0.520
SICU LOS, day	5 (3–10)	5 (3–8)	0.162	5 (3–9)	5 (2.5–8)	0.540
SICU mortality	5 (3.3%)	4 (2.7%)	0.735	3 (3.8%)	3 (3.8%)	> 0.999
Hospital LOS, day	20 (13–34)	16.5 (10–31)	0.034	21 (12–34)	16 (11–26)	0.086
Hospital mortality	25 (16.7%)	15 (10.1%)	0.089	13 (16.7%)	8 (10.3%)	0.348

Data are presented as median (interquartile range) or number (%).

LOS, length of stay; POD, postoperative delirium; SICU, surgical intensive care unit.

*Only patients who developed delirium, n = 117 in entire cohort and n = 56 in propensity score matching cohort.

Fig. 2.

Multicomponent non-pharmacological intervention protocol and postoperative delirium. Univariate and multivariate logistic regression analysis using 2 different model were performed to determine the association between the multicomponent non-pharmacological intervention protocol and the incidence of POD. PSM, propensity score matching.

Secondary outcomes of POD and other clinal outcomes

Hyperactive delirium subtype was significantly higher in the intervention cohort (Table 3). POD onset and duration and delirium-free days were not significantly different between both two cohorts (Table 3). SICU events significantly less occurred in the intervention cohort (34.0% vs. 62.7%, P < 0.001), particularly self-removal of tube (6.0% vs. 14.0%, P = 0.021) and nosocomial infection (16.7% vs. 27.3%, P = 0.026) (Table 3). Hospital LOS was significantly shorter in the intervention cohort [median (IQR), 16.5 (10–31) days vs. 20 (13–34) days, P = 0.034)]. However, SICU LOS and SICU and hospital mortality rates were not significantly different between the cohorts (Table 3). Multivariate logistic analyses demonstrated that the protocol was associated with lower rates of SICU events (adjusted OR 0.17, 95% CI 0.09–0.33), self-removal of tube (adjusted OR 0.35, 95% CI 0.14–0.84) and nosocomial infection (adjusted OR 0.49, 95% CI 0.25–0.97) but not hospital LOS (adjusted coefficient − 3.44, 95% CI -9.37 to 2.48).

Sensitivity analysis

Propensity score matching yielded 78 patients in the pre-intervention cohort matched to 78 patients in the intervention cohort. There was no significant difference in demographic data and SICU data between the two groups, except higher proportion of patients with underlying disease of cirrhosis in the intervention cohort (Tables 1 and 2). The incidence of POD did not significantly differ between the two groups (39.7% in the pre-intervention group vs. 32.1% in the intervention group, P = 0.404; OR 0.72, 95% CI 0.37–1.38) (Table 3; Fig. 2) and the multivariate analyses did not show the association between the protocol and the POD incidence (OR 0.559, 95% CI 0.26–1.20 for Model 1 and OR 0.501, 95% CI 0.23–1.08 for Model 2) (Fig. 2). SICU events, particularly self-removal of tube, significantly less occurred in the intervention group (30.8% vs. 62.8%, P < 0.001 and 5.1% vs. 16.7%, P = 0.037, respectively) (Table 3). Multivariate analyses showed the association between the protocol and lower rate of SICU events (adjusted OR 0.14, 95% CI 0.06–0.32) and self-removal of tube (adjusted OR 0.27, 95% CI 0.08–0.88).

Discussion

The main finding of this quasi-experimental, before-and-after study was that POD incidence in elderly patients admitted to the SICU postoperatively was not reduced after implementing the multicomponent non-pharmacological intervention protocol. However, the rates of SICU events, particularly endotracheal tube self-removal and nosocomial infection, were significantly lower after protocol implementation.

Awareness of adverse outcomes in patients who develop delirium drives the initiation of both pharmacological and non-pharmacological intervention strategies to prevent and manage delirium. A recent meta-analysis demonstrated that pharmacological intervention may not be a suitable strategy to manage delirium²⁵. By contrast, non-pharmacological intervention strategies have shown benefits in patients with delirium^12–14. Currently, many of these interventions are incorporated into our routine patient care, such as encouraging orientation and cognition, early ambulation, pain assessment, and adequate pain control. We hypothesized that multiple predisposing and precipitating risk factors contribute to the development of delirium in critically ill patients in ICU. Therefore, a single intervention targeting only one risk factor may not be sufficient to effectively prevent delirium. Recent systematic reviews demonstrated that, compared with single non-pharmacological intervention strategies, multicomponent non-pharmacological intervention strategies are more effective in preventing and managing delirium^12,13. Therefore, our multidisciplinary team developed a multicomponent non-pharmacological intervention protocol consisting of practical interventions in our setting, and we tested its effectiveness in preventing POD in elderly patients admitted to our SICU postoperatively. Unfortunately, our study did not show a significant reduction in POD incidence using this protocol.

Our negative result could be attributed to the fact that some non-pharmacological interventions in our protocol, such as encouraging orientation and cognition, pain assessment, and adequate pain control, were already performed as routine patient care before protocol implementation. Therefore, the protocol might provide little effect in modifying the risk of POD. In a recent meta-analysis, Bannon et al. demonstrated that multicomponent non-pharmacological intervention did not reduce either delirium incidence or duration in critically ill patients²⁶. We believed that these interventions per se could help reduce the development of delirium, and they have already been performed as standard practice in our SICU. Hence, additional interventions targeting other POD risk factors in the protocol should be required to attenuate POD incidence.

Our finding may also be explained by patients in the intervention cohort were more likely at risk of POD as the presence of more comorbidities, especially dementia, more preoperative use of benzodiazepines, more active infection at SICU admission, and more frequent admissions to the SICU following emergency surgery. Due to the COVID-19 pandemic in early 2020, all elective surgeries were postponed, and only emergency surgeries were performed at our hospital. Additionally, a dedicated cohort ICU was established to care for COVID-19 patients, including surgical cases. However, the SICU where this study was conducted admitted only critically ill surgical patients who were not infected with COVID-19. As a result, the non-pharmacological intervention protocol applied during the pandemic was identical to the protocol used before the pandemic. The main difference was the patient characteristics, as only emergency surgical cases were included during the pandemic period. We excluded COVID-19 patients from the study because they often required deeper sedation to facilitate mechanical ventilation, which could have hindered the implementation of the full non-pharmacological intervention protocol, such as early mobilization. Although patients in the intervention cohort were more likely to be at risk of POD, its incidence did not significantly differ from that of the pre-intervention cohort, which could be due to the effect of our multicomponent non-pharmacological intervention protocol. Additionally, adverse events, including self-removal of endotracheal tube and nosocomial infection, were significantly lower in the intervention cohort, although they had a higher incidence of hyperactive delirium. Multivariate logistic regression analyses confirmed the association of the protocol and lower incidence of SICU events, particularly self-removal of tube and nosocomial infection.

Our study had 74.6% overall adherence to the multicomponent non-pharmacologic intervention protocol. The component that had the lowest adherence rate was sleep promotion, which nursing and medical cares could not be avoided during sleep time. An increasing number of studies indicate that sleep deprivation commonly occurs in patients admitted to the ICU and can potentially be a risk factor for delirium²⁷. Adherence to the intervention protocol may be suboptimal and, consequently, affect the outcomes. Improvement in adherence to this multicomponent non-pharmacological intervention protocol may effectively enhance the clinical outcomes of patients with POD. In this ontext, exploring the reasons for low adherence could provide valuable insights and help in interpreting the results of our study. Unfortunately, we do not have detailed information on adherence to each individual component of the protocol, which is a limitation of this study.

Our study has limitations. First, a significant imbalance was found in the baseline characteristics of patients in both cohorts. To overcome this limitation, a randomized control study may be ideal. Alternatively, we performed a sensitivity analysis using propensity score matching, and no difference was observed in the results. However, the number of patients included in the propensity score-matched cohort was relatively small, which may introduce a potential risk of bias. Second, there may be bias from the Hawthorne effect. As mentioned earlier, some interventions in the protocol were already performed as routine patient care before the protocol implementation. Healthcare providers probably performed some interventions intensively as the awareness of the research study. Consequently, POD incidence before and after protocol implementation did not show demonstrable difference. Third, our study included only elderly patients admitted to the SICU postoperatively. Results from our study might be not able to apply to other patient populations, such as medical patients with different nature of diseases. In addition, the settings from different ICUs are different, and our protocol might be fit for only some ICUs but not all. Fourth, delirium was assessed twice daily during the SICU stay. Due to the fluctuation in symptoms, the diagnosis of POD may be missed. Consequently, this would underestimate the true incidence of POD.

Conclusion

Our quasi-experimental, before-and-after study demonstrated that implementing a multicomponent non-pharmacological intervention protocol was not associated with reducing POD incidence in elderly patients admitted to the SICU postoperatively. Nevertheless, the protocol seemed to help decrease adverse events in the SICU, particularly endotracheal tube self-removal and nosocomial infection. Further research on other interventions aimed at preventing POD would be warranted.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Abbreviations

APACHE

Acute Physiology and Chronic Health Evaluation

CAM-ICU

Confusion Assessment Method for the ICU

CI

Confidence interval

ICU

Intensive care unit

IQR

Interquartile range

LOS

Length of stay

POD

Postoperative delirium

SOFA

Sequential Organ Failure Assessment

SICU

Surgical ICU

Author contributions

AP and OC were responsible for the conceptualization, methodology, and project administration of the study. TY, AN, SL, CT, CP, and NT were involved in data curation. AP and CT conducted the formal analysis. AP made substantial contributions to the analysis and interpretation of the data. TY, AN, SL, and AP contributed to the original draft and to the review and editing of the manuscript. All authors read and approved the final version of the manuscript.

Funding

The study was supported by Siriraj Research Development Fund managed by Routine to Research, Siriraj Hospital on December 20, 2018. Grant number was R016135045 and Prasert Prasarttong-Osoth Scholarship, Medical Association of Thailand, Bangkok, Thailand. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Data availability

The datasets used in this study are available from the corresponding author on reasonable request.

Declarations

Competing interests

The authors declare no competing interests.

Ethics approval and consent to participate

This study was conducted according to the ethical standards established by the 1964 Declaration of Helsinki. The study was approved by the Siriraj Institutional Review Board of the Faculty of Medicine Siriraj Hospital, Mahidol University, Bangkok, Thailand (Si 211/2018, Chairperson Prof. Chairat Shayakul, M.D.) on 11 April 2018. Informed consent was obtained from all individual participants included in the study.