Impact of delirium on clinical outcomes in critically ill patients with acute pancreatitis: A propensity score-matched study

Abstract

Objective

Delirium is an established predictor of adverse outcomes in general ICU populations, but its specific prognostic impact in critically ill patients with acute pancreatitis (AP) remains unclear. This study aimed to evaluate the independent association between ICU-acquired delirium and clinical outcomes in this high-risk population.

Methods

This retrospective cohort study utilized data from the MIMIC-IV database (2008–2022). Critically ill adults with AP were included and stratified by the presence of ICU-acquired delirium, assessed using the CAM-ICU. The primary outcome was 90-day all-cause mortality. Secondary outcomes included 90-day unplanned readmission, emergency department revisits, and a composite adverse outcome. Propensity score matching (PSM) was performed to balance baseline characteristics, generating 178 matched pairs. Multivariable Cox regression with four sequential models and sensitivity analyses were conducted to assess robustness.

Results

Among 594 included patients, 44.6% (265/594) developed delirium. After PSM, baseline characteristics were well-balanced. Delirium was independently associated with increased 90-day all-cause mortality (aHR=1.91, 95% CI: 1.04–3.50; P=0.038) and a higher risk of the composite adverse outcome (aHR=1.84, 95% CI: 1.24–2.73; P=0.002). The association with unplanned readmission remained significant after full adjustment (aHR=1.87, 95% CI: 1.20–2.92; P=0.006), while the association with ED revisits did not reach statistical significance. Sensitivity analyses confirmed the robustness of the primary findings.

Conclusions

In critically ill patients with AP, ICU-acquired delirium was an independent predictor of increased 90-day mortality, unplanned readmission and composite adverse outcomes. These findings highlight delirium as a significant prognostic factor, underscoring the importance of routine screening and targeted management in this vulnerable population.

Keywords

delirium critically ill patients acute pancreatitis propensity score-matched intensive care unit MIMIC-IV database

1. Introduction

Acute pancreatitis (AP) is a common gastrointestinal emergency characterized by rapid onset, rapid progression, and high mortality.¹ The global incidence of AP has increased steadily over recent decades, with approximately 20% of cases classified as severe acute pancreatitis (SAP) that necessitates intensive care unit (ICU) admission.² Critically ill patients with AP face mortality rates ranging from 20% to 30%, primarily due to systemic inflammatory response syndrome and multiple organ dysfunction.³

Delirium, an acute brain dysfunction characterized by inattention and disorganized thinking, is a frequent complication in critically ill patients, affecting 20–50% of the general ICU population.^4–6 It is strongly associated with prolonged mechanical ventilation, extended hospital stays, and increased mortality.^7–9 In AP patients, the risk of delirium is heightened by factors including severe systemic inflammation, metabolic disturbances, and the potential development of pancreatic encephalopathy—a severe neurological complication.^10,11 Shared inflammatory pathways, such as phospholipase A2 activation, may contribute to both pancreatic injury and cerebral dysfunction, although direct human evidence linking this specific mechanism to delirium remains limited.¹²

Despite the established prognostic role of delirium in general ICU cohorts, its disease-specific impact in critically ill AP patients remains poorly characterized. Existing studies have predominantly focused on heterogeneous ICU populations, and the independent association between delirium and clinical outcomes in AP has not been rigorously quantified using advanced methods to control for confounding. To address this gap, we utilized the MIMIC-IV database to evaluate the independent association between ICU-acquired delirium and clinical outcomes in critically ill patients with AP. We hypothesized that delirium would be independently associated with increased 90-day mortality and a composite of adverse events, even after comprehensive adjustment for baseline severity and confounders using propensity score matching. This study represents the first large-scale, propensity score-matched analysis to specifically investigate the prognostic impact of delirium in this high-risk patient population.

2. Materials and methods

2.1. Study design and data source

This retrospective cohort study utilized data from the Medical Information Mart for Intensive Care IV (MIMIC-IV) database (version 3.0), a publicly available repository containing de-identified clinical data of over 50,000 patients admitted to ICUs at the Beth Israel Deaconess Medical Center (2008–2022).¹³ The use of this database was approved by the institutional review boards of both the Massachusetts Institute of Technology and Beth Israel Deaconess Medical Center. As the data are de-identified, the requirement for informed consent was waived. One author (Xia Zhou) completed the required CITI training and was granted access to the database (Record ID: 54611342). This study was conducted in accordance with the Declaration of Helsinki as revised in 2024 and reported in conformity with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline.¹⁴

2.2. Study population and selection criteria

A total of 180,720 admissions were recorded in the MIMIC-IV database, including 72,793 ICU admissions. The ICD-9 code 577.0 (Acute pancreatitis) and ICD-10 codes K85 (Acute pancreatitis) and its sub-codes (K85.0–K85.9) were used for case identification. A total of 5,978 patients with AP were identified, of whom 1,401 were admitted to the ICU. For this study, data from the first ICU admission of patients aged 18 years or older were included. The exclusion criteria were: (1) age under 18 years; (2) non-initial ICU admissions (only the first ICU stay was considered); (3) ICU stay of less than 24 hours; (4) absence of CAM-ICU assessment; (5) missing key clinical data; and (6) documented history of schizophrenia, dementia, or pre-existing delirium before ICU admission. After applying these criteria, a total of 594 patients were included in the final analysis (Figure 1).

Figure 1.

Study flowchart.

2.3. Data collection

Data were extracted using PostgreSQL (version 14) and Navicat Premium (version 16.3.3). All variables were collected from the first 24 hours of ICU admission. The extracted data included: Demographics: age, sex, ethnicity, body mass index (BMI). Vital signs: heart rate, arterial pressure, oxygen saturation, respiratory rate. Laboratory parameters: white blood cell count, hemoglobin, platelet count, blood urea nitrogen, creatinine, albumin, international normalized ratio (INR). Comorbidities: assessed using the Elixhauser comorbidity index components (e.g., myocardial infarction, congestive heart failure, chronic pulmonary disease, diabetes, renal disease, mild/severe liver disease). Severity scores: Glasgow Coma Scale (GCS), Simplified Acute Physiology Score II (SAPS II), Acute Physiology Score III (APS III), Sequential Organ Failure Assessment (SOFA), and APACHE II score.^15,16 The Revised Atlanta Classification was not applied due to the absence of structured CT-based imaging variables and time-stamped organ failure duration records in the MIMIC-IV database; APACHE II was therefore adopted as the primary severity adjustment variable, supplemented by SOFA score and SAPS II.

Interventions and treatments: mechanical ventilation, use of sedatives (propofol, midazolam, dexmedetomidine) and analgesics (fentanyl, morphine, hydromorphone).

Handling of Missing Data: The proportion of missing data was calculated for each variable (Table S1). For continuous variables with <10% missingness, multiple imputations were performed using a random forest algorithm (via the ‘missForest’ package in R) to minimize bias. With the exception of serum albumin (missing rate: 32.5%), which was retained and imputed given its established clinical importance as a marker of nutritional status and disease severity in acute pancreatitis, variables with >10% missing data were excluded from the model. For categorical variables, missing values were treated as a separate “unknown” category to retain all observations in the analysis. Multicollinearity was assessed using variance inflation factors (VIF), and variables with VIF >5 were removed from multivariable models.^17,18

2.4. Exposure and outcome definitions

Delirium Assessment: The exposure was ICU-acquired delirium, assessed using the CAM-ICU. CAM-ICU is a validated bedside screening instrument that evaluates four core features: (1) acute onset with fluctuating course; (2) inattention; (3) disorganized thinking; and (4) altered level of consciousness.¹⁹ A patient is diagnosed as delirium (i.e., CAM-ICU positive) if they exhibit features 1 and 2, along with either feature 3 or 4.⁴ CAM-ICU assessments were extracted from the MIMIC-IV chartevents table, encompassing all four CAM-ICU components and the Richmond Agitation-Sedation Scale (RASS); the specific itemids used for each component are detailed in Table S2. All assessments recorded between ICU admission and discharge were included, covering the entire ICU stay. Assessments were performed as part of routine nursing documentation, typically every 8–12 hours throughout the ICU stay. Patients with a RASS score of ≤ −3 were considered unarousable and were not assessed for delirium, consistent with standard CAM-ICU administration guidelines; the delirium status of these patients cannot be determined and represents an inherent limitation of the database.

Based on CAM-ICU assessment results, patients were stratified into two groups: the delirium group comprised patients with at least one positive CAM-ICU evaluation during their ICU stay — a threshold deliberately chosen to maximize case detection and minimize false-negative classification — while the non-delirium group consisted of patients with consistently negative CAM-ICU assessments throughout their ICU admission.To establish a clear temporal relationship between ICU admission and delirium onset, patients with documented delirium prior to ICU entry were excluded from the analysis.

Outcome Variables: The primary outcome was all-cause mortality within 90 days of hospital discharge. Secondary outcomes included all-cause unplanned hospital readmission and all-cause unscheduled emergency department (ED) visits, both ascertained within 90 days of hospital discharge; in-hospital events were not counted. The composite outcome was defined as the occurrence of any one of these three events (all-cause death, all-cause unplanned readmission, or all-cause unscheduled ED visit) within 90 days of hospital discharge.

2.5. Statistical analysis

Baseline characteristics were summarized using descriptive statistics. Propensity score matching (PSM) was employed to control confounding. The propensity score was estimated using a logistic regression model that included all baseline variables listed in Section 2.3 (demographics, vitals, labs, comorbidities, severity scores, and treatments). A 1:1 nearest-neighbor matching algorithm was used without replacement, with a caliper width set to 0.05 of the standard deviation of the logit of the propensity score, a common and conservative value to ensure good matches while minimizing bias. Balance between matched groups was assessed using standardized mean differences (SMD), with SMD <0.1 indicating good balance. Sedation-related variables with residual imbalance after PSM were additionally incorporated into the Cox regression models to minimize potential residual confounding.

The association between delirium and outcomes was analyzed using Cox proportional hazards models in the matched cohort. We built four sequential models: unadjusted, adjusted for age and sex (Model I), a fully adjusted model (Model II) incorporating age, sex, BMI, SOFA score, APACHE II score, mechanical ventilation, sepsis, and sedation-related variables with residual imbalance after PSM (dexmedetomidine, fentanyl, and hydromorphone), and a further adjusted model (Model III) additionally incorporating AP etiology to evaluate the potential confounding effect of etiology-related differences in inflammatory burden and ICU course.

Sensitivity Analyses: To test the robustness of our findings, we conducted several sensitivity analyses: 1) Crude Cox regression in the unmatched cohort; 2) Cox regression adjusted for the propensity score (CAPS); 3) Inverse probability of treatment weighting (IPTW); and 4) Standardized mortality ratio weighting (SMRW).²⁰ The E-value was calculated to quantify the potential impact of unmeasured confounding. All analyses were performed using R software (version 4.3.3), with a two-tailed P-value <0.05 considered statistically significant.

3. Results

3.1. Baseline characteristics and propensity score matching

After applying the predefined exclusion criteria, 594 critically ill patients with acute pancreatitis (AP) were included in the final cohort. Of these, 265 patients (44.6%) developed ICU-acquired delirium, while 329 (55.4%) did not. The distribution of AP etiology was as follows: biliary (n=152, 25.6%), alcoholic (n=143, 24.1%), unspecified (n=239, 40.2%), hypertriglyceridemia-induced (n=40, 6.7%), and other (n=20, 3.4%). Prior to matching, the delirium group exhibited significantly greater illness severity, as reflected by higher APACHE II scores, elevated serum creatinine levels, and more frequent use of sedative medications (e.g., propofol, midazolam, fentanyl), with all standardized mean differences (SMD) >0.3 (Table 1).

Table 1.

Baseline characteristics before and after propensity score matching in the MIMIC-IV data.

	Before matching				After matching
	Overall	NO	Yes	SMD	Overall	NO	Yes	SMD
Delirium	594	329	265		356	178	178
Age (mean (SD))	57.71 (16.53)	57.89 (16.64)	57.48 (16.43)	0.025	56.81 (16.42)	57.20 (16.50)	56.42 (16.38)	0.047
Gender (%) Female	238 (40.1)	134 (40.7)	104 (39.2)	0.03	138 (38.8)	65 (36.5)	73 (41.0)	0.092
Race (%)				0.295				0.065
WHITE	344 (57.9)	208 (63.2)	136 (51.3)		203 (57.0)	99 (55.6)	104 (58.4)
BLACK/AFRICAN AMERICAN	68 (11.4)	40 (12.2)	28 (10.6)		41 (11.5)	22 (12.4)	19 (10.7)
UNKNOWN	182 (30.6)	81 (24.6)	101 (38.1)		112 (31.5)	57 (32.0)	55 (30.9)
BMI (mean (SD))	29.96 (7.44)	29.86 (7.63)	30.08 (7.22)	0.031	29.86 (7.71)	29.98 (7.22)	29.74 (8.19)	0.031
Myocardial infarct (%)	59 (9.9)	37 (11.2)	22 (8.3)	0.099	32 (9.0)	17 (9.6)	15 (8.4)	0.039
Congestive heart failure (%)	102 (17.2)	60 (18.2)	42 (15.8)	0.064	60 (16.9)	30 (16.9)	30 (16.9)	<0.001
Peripheral vascular disease (%)	33 (5.6)	14 (4.3)	19 (7.2)	0.126	20 (5.6)	11 (6.2)	9 (5.1)	0.049
Cerebrovascular disease (%)	38 (6.4)	13 (4.0)	25 (9.4)	0.221	26 (7.3)	14 (7.9)	12 (6.7)	0.043
Chronic pulmonary disease (%)	135 (22.7)	71 (21.6)	64 (24.2)	0.061	78 (21.9)	40 (22.5)	38 (21.3)	0.027
Rheumatic disease (%)	19 (3.2)	12 (3.6)	7 (2.6)	0.058	9 (2.5)	6 (3.4)	3 (1.7)	0.107
Peptic ulcer disease (%)	32 (5.4)	15 (4.6)	17 (6.4)	0.082	15 (4.2)	8 (4.5)	7 (3.9)	0.028
Mild liver disease (%)	171 (28.8)	83 (25.2)	88 (33.2)	0.176	94 (26.4)	50 (28.1)	44 (24.7)	0.076
Diabetes (%)	192 (32.3)	116 (35.3)	76 (28.7)	0.141	117 (32.9)	57 (32.0)	60 (33.7)	0.036
Paraplegia (%)	9 (1.5)	2 (0.6)	7 (2.6)	0.161	7 (2.0)	3 (1.7)	4 (2.2)	0.04
Renal disease (%)	91 (15.3)	54 (16.4)	37 (14.0)	0.068	53 (14.9)	26 (14.6)	27 (15.2)	0.016
Severe liver disease (%)	68 (11.4)	27 (8.2)	41 (15.5)	0.226	39 (11.0)	20 (11.2)	19 (10.7)	0.018
Cancer (%)	52 (8.8)	33 (10.0)	19 (7.2)	0.102	28 (7.9)	14 (7.9)	14 (7.9)	<0.001
Aids (%)	7 (1.2)	2 (0.6)	5 (1.9)	0.115	3 (0.8)	3 (1.7)	0 (0.0)	0.185
Apsiii (mean (SD))	56.82 (28.19)	47.70 (23.04)	68.14 (29.90)	0.766	58.76 (29.37)	57.76 (28.55)	59.75 (30.21)	0.067
ApacheII(mean (SD))	23.12 (6.66)	15.13 (8.70)	23.11 (9.44)	0.878	19.20 (10.13)	18.79 (10.62)	19.62 (9.63)	0.082
AP etiology (%)				0.30				0.176
Alcoholic	143 (24.1)	70 (21.3)	73 (27.5)		92 (25.8)	50 (28.1)	42 (23.6)
Biliary	152 (25.6)	98 (29.8)	54 (20.4)		88 (24.7)	40 (22.5)	48 (27.0)
Hypertriglyceridemia	40 (6.7)	26 (7.9)	14 (5.3)		22 (6.2)	13 (7.3)	9 (5.1)
Unspecified	239 ( 40.2)	124 ( 37.7)	115 (43.4)		140 ( 39.3)	69 ( 38.8)	71 (39.9)
Other	20 (3.4)	11 (3.3)	9 (3.4)		14 (3.9)	6 (3.4)	8 (4.5)
Heart rate (mean (SD))	96.59 (17.98)	94.73 (17.54)	98.89 (18.29)	0.232	96.31 (17.28)	96.28 (16.42)	96.34 (18.15)	0.004
SBP (mean (SD))	121.74 (18.53)	124.27 (18.84)	118.61 (17.67)	0.31	121.21 (18.49)	121.48 (18.47)	120.94 (18.57)	0.029
DBP (mean (SD))	69.23 (13.39)	70.52 (13.93)	67.63 (12.54)	0.218	69.52 (13.31)	69.54 (13.13)	69.50 (13.53)	0.003
Resp rate (mean (SD))	20.98 (4.30)	20.54 (4.18)	21.53 (4.40)	0.231	20.98 (4.20)	20.90 (4.25)	21.05 (4.15)	0.037
Temperature (mean (SD))	37.03 (0.59)	36.99 (0.51)	37.09 (0.67)	0.164	37.04 (0.61)	37.03 (0.65)	37.05 (0.58)	0.023
Spo2 (mean (SD))	96.23 (2.20)	96.19 (2.09)	96.28 (2.34)	0.038	96.24 (2.27)	96.31 (2.37)	96.17 (2.18)	0.062
Glucose_mean (mean (SD))	8.30 (3.10)	8.16 (3.02)	8.48 (3.20)	0.101	8.30 (3.04)	8.28 (2.86)	8.31 (3.21)	0.01
Hematocrit (mean (SD))	36.00 (7.23)	36.22 (6.91)	35.73 (7.60)	0.067	36.14 (7.01)	36.08 (7.02)	36.21 (7.02)	0.018
Platelets (mean (SD))	231.82 (131.98)	236.44 (138.22)	226.08 (123.81)	0.079	240.58 (142.85)	236.94 (134.79)	244.21 (150.77)	0.051
WBC(mean (SD))	16.12 (12.41)	16.11 (14.54)	16.13 (9.12)	0.002	16.45 (14.17)	16.27 (9.51)	16.63 (17.67)	0.026
Albumin (mean (SD))	3.20 (0.70)	3.28 (0.67)	3.10 (0.73)	0.248	3.17 (0.73)	3.20 (0.72)	3.14 (0.74)	0.086
Aniongap (mean (SD))	18.80 (6.65)	17.84 (5.82)	19.98 (7.40)	0.322	19.14 (7.27)	19.11 (7.48)	19.17 (7.07)	0.008
Bun (mean (SD))	30.01 (27.17)	25.49 (22.74)	35.63 (30.96)	0.374	29.16 (24.81)	29.16 (23.86)	29.16 (25.80)	<0.001
Creatinine (mean (SD))	1.90 (2.22)	1.58 (1.77)	2.30 (2.63)	0.325	1.84 (2.15)	1.90 (2.37)	1.77 (1.91)	0.062
Ventilation (%)	479 (80.6)	239 (72.6)	240 (90.6)	0.475	289 (81.2)	144 (80.9)	145 (81.5)	0.014
Propofol (%)	238 (40.1)	82 (24.9)	156 (58.9)	0.733	159 (44.7)	78 (43.8)	81 (45.5)	0.034
Midazolam (%)	177 (29.8)	59 (17.9)	118 (44.5)	0.599	122 (34.3)	60 (33.7)	62 (34.8)	0.024
Dexmedetomidine (%)	120 (20.2)	25 (7.6)	95 (35.8)	0.729	78 (21.9)	38 (21.3)	40 (22.5)	0.027
Fentanyl (%)	268 (45.1)	97 (29.5)	171 (64.5)	0.75	176 (49.4)	86 (48.3)	90 (50.6)	0.045
Morphine (%)	103 (17.3)	53 (16.1)	50 (18.9)	0.073	70 (19.7)	34 (19.1)	36 (20.2)	0.028
Hydromorphone (%)	266 (44.8)	152 (46.2)	114 (43.0)	0.064	166 (46.6)	74 (41.6)	92 (51.7)	0.203

To mitigate these substantial baseline imbalances, propensity score matching (PSM) was performed, resulting in 178 well-matched patient pairs. After PSM, all measured covariates were well-balanced between groups, with the vast majority of SMDs below 0.1. Hydromorphone showed modest residual imbalance (SMD=0.204); dexmedetomidine (SMD=0.027) and fentanyl (SMD=0.045) were well-balanced. Given their known clinical association with delirium, all three sedative agents were conservatively incorporated into the Cox regression models (Table 1, Figure S1).

3.2. Association between delirium and clinical outcomes

Kaplan-Meier survival analysis demonstrated a significant divergence in outcomes between the matched groups (Figure 2). Patients with delirium had a significantly higher cumulative incidence of the primary and composite endpoints.

Figure 2.

Kaplan-Meier curves showing the cumulative incidence of clinical outcomes stratified by delirium status. (a) All-cause in-hospital mortality, (b) 90-day unplanned readmission, (c) 90-day emergency department revisit, and (d) 90-day composite outcome. Log-rank test P-values are shown for each comparison.

For the primary outcome of 90-day all-cause mortality, delirium was associated with a significantly increased risk in the unadjusted model (HR = 1.77, 95% CI: 1.07–2.92; P = 0.025). This association remained statistically significant after sequential adjustment for age and sex (Model I: aHR=2.03, 95% CI: 1.12–3.68; P=0.020), further adjustment incorporating SOFA score, APACHE II score, mechanical ventilation, sepsis, and sedation-related variables (Model II: aHR=1.91, 95% CI: 1.04–3.50; P=0.038), and additional adjustment for AP etiology (Model III: aHR=1.95, 95% CI: 1.06–3.61; P=0.033) (Table 2).

Table 2.

Association between delirium and outcomes.

	Unadjusted model HR (95% CI)	Model I aHR (95% CI)	Model II aHR (95% CI)	Model III aHR (95% CI)
All-Cause Mortality	1.77 (1.07-2.92)	2.03 (1.12-3.68)	HR=1.91 (1.04-3.50)	HR=1.95 (1.06-3.61)
All-Cause Mortality	P=0.025	P=0.020	p=0.038	p=0.033
90-Day Unplanned Readmission	2.54 (1.78-3.63)	1.86 (1.20-2.89)	HR=1.87 (1.20-2.92)	HR=1.70 (1.09-2.66)
90-Day Unplanned Readmission	P<0.001	P=0.005	p=0.006	p=0.020
90-Day ED Revisit	2.15 (1.27-3.65)	1.48 (0.76-2.85)	HR=1.40 (0.72-2.72)	HR=1.41 (0.72-2.76)
90-Day ED Revisit	P=0.004	P=0.240	p=0.328	p=0.320
90-Day Composite Outcome	2.18 (1.59-3.00)	1.85 (1.25-2.72)	HR=1.84 (1.24-2.73)	HR=1.71 (1.15-2.54)
90-Day Composite Outcome	P<0.001	P=0.002	p=0.002	p=0.008

Model I was adjusted for age and sex. Model II was further adjusted for SOFA score, APACHE II score, mechanical ventilation, sepsis, and sedation-related variables with residual imbalance after PSM (dexmedetomidine, fentanyl, and hydromorphone). Model III was further adjusted for AP etiology (biliary, alcoholic, hypertriglyceridemia-induced, unspecified, and other) in addition to all Model II covariates. HR, Hazard Ratio; aHR, Adjusted Hazard Ratio; CI, Confidence Interval.

Analysis of the individual secondary outcomes yielded nuanced results. While unadjusted models showed a strong association between delirium and both 90-day unplanned readmission (HR = 2.54) and emergency department (ED) revisits (HR = 2.15), the association with ED revisits did not reach statistical significance in any adjusted model (Model II: aHR=1.40, 95% CI: 0.72–2.72; P=0.328; Model III: aHR=1.41, 95% CI: 0.72–2.76; P=0.320), while the association with unplanned readmission remained significant across all adjusted models (Model II: aHR=1.87, 95% CI: 1.20–2.92; P=0.006; Model III: aHR=1.70, 95% CI: 1.09–2.66; P=0.020). In contrast, the 90-day composite outcome (encompassing mortality, readmission, or ED revisit) demonstrated a robust and consistent association with delirium across all models (Model II: aHR=1.84, 95% CI: 1.24–2.73; P=0.002; Model III: aHR=1.71, 95% CI: 1.15–2.54; P=0.008) (Table 2).

3.3. Sensitivity and subgroup analyses

The robustness of the primary findings was confirmed through extensive sensitivity analyses using alternative statistical approaches, including Cox regression adjusted for the propensity score (CAPS), inverse probability of treatment weighting (IPTW), and standardized mortality ratio weighting (SMRW). For final inference, we prioritized the SMRW analysis due to its efficiency in retaining the entire original cohort while reweighting to achieve balance. All methods consistently showed a significant association between delirium and both 90-day mortality and the composite outcome, reinforcing the primary results (Table 3).

Table 3.

Sensitivity analysis of the association between delirium and clinical outcomes using multiple statistical approaches.

	Unadjusted cox regression HR (95% CI)	CAPS HR (95% CI)	IPTW HR (95% CI)	SMRW HR (95% CI)
N	356	356	594	594
All-Cause Mortality	1.77 (1.07-2.92)	1.51 (1.00-2.28)	1.81 (1.01-3.23)	1.93 (1.20-3.11)
All-Cause Mortality	P=0.025	P=0.05	P=0.045	P=0.007
90-Day Unplanned Readmission	2.54 (1.78-3.63)	2.39 (1.67-3.42)	3.06 (2.01- 4.66)	3.30 (2.13-5.11)
90-Day Unplanned Readmission	P<0.001	P<0.001	P<0.001	P<0.001
90-Day ED Revisit	2.15 (1.27-3.65)	2.06 (1.22-3.50)	1.98 (1.10-3.56)	1.92 (1.04-3.55)
90-Day ED Revisit	P=0.004	P=0.007	P=0.022	P=0.038
90-Day Composite Outcome	2.18 (1.59-3.00)	2.08 (1.51-2.85)	2.58 (1.75-3.80)	2.69 (1.79-4.04)
90-Day Composite Outcome	P<0.001	P<0.001	P<0.001	P<0.001

HR, hazard ratio; CI, confidence interval; CAPS, covariate adjustment through propensity score; IPTW, inverse probability of treatment weighting; SMRW, standardized mortality ratio weighting; ED, emergency department.

Furthermore, E-value analysis suggested that unmeasured confounding was unlikely to fully explain the observed associations for mortality (E-value = 2.76) and the composite outcome (E-value = 3.24) (Table S3). To assess the potential confounding effect of alcohol withdrawal syndrome, a comparative analysis between alcoholic (n=143) and non-alcoholic (n=451) patients was performed. Delirium incidence (51.0% vs 42.6%, p=0.093) and clinical outcomes did not differ significantly between groups (Table S4). After excluding all alcoholic pancreatitis patients, delirium remained significantly associated with composite adverse outcomes in the remaining 451 non-alcoholic patients (HR=2.09, 95% CI: 1.27–3.42; p=0.004), confirming that the observed association was not driven by alcohol withdrawal syndrome.

3.4. Subgroup analysis

Subgroup analyses were performed to assess the consistency of the association between delirium and the composite adverse outcome across various patient characteristics (Figure 3, Table S5).

Figure 3.

Subgroup analysis of the association between delirium and clinical outcomes.

The detrimental association between delirium and the composite outcome was notably stronger in several high-risk subgroups. A significantly elevated risk was observed in patients with impaired renal function, particularly those with serum creatinine levels between 2–5 mg/dL (HR = 3.13, 95% CI: 1.72–5.70). Similarly, the use of specific sedative agents was associated with a markedly higher hazard, including midazolam (HR = 3.39, 95% CI: 2.10–5.48), fentanyl (HR = 3.00, 95% CI: 2.02–4.47), and propofol (HR = 2.73, 95% CI: 1.74–4.28). Furthermore, patients with high disease severity, as indicated by top-tertile APS III scores, faced a substantially increased risk (HR = 3.28, 95% CI: 2.12–5.06).

Formal tests for interaction were statistically significant (P-interaction < 0.001) for the majority of these subgroups, including age, renal function, sedative exposure, and disease severity, suggesting that these factors may modify the effect of delirium on patient outcomes. In contrast, the association remained positive and significant across other key demographics and comorbidities, such as gender and the presence of severe liver disease, indicating the robustness of the primary finding.

3.5. Post-hoc analysis: Inflammatory and renal biomarkers

To provide empirical context for the proposed neuroimmune mechanisms, a post-hoc analysis was conducted examining the association between ICU-acquired delirium and key inflammatory and renal biomarkers (Table S6). Patients with delirium exhibited significantly higher creatinine (1.4 [IQR 0.8–2.7] vs 1.0 [IQR 0.7–1.6] mg/dL, p<0.001), higher BUN (25 [IQR 15–47] vs 18 [IQR 12–30] mg/dL, p<0.001), and lower albumin (3.1 [IQR 2.5–3.6] vs 3.3 [IQR 2.8–3.7] g/dL, p=0.003) compared with non-delirious patients, suggesting that renal dysfunction and nutritional depletion accompany delirium development. WBC did not differ significantly between groups (14.0 [IQR 9.7–19.7] vs 13.7 [IQR 10.0–18.8] ×10⁹/L, p=0.506). However, interaction analyses revealed that none of these biomarkers significantly modified the delirium-outcome association (all P-interaction >0.05: WBC p=0.116, creatinine p=0.259, albumin p=0.277, BUN p=0.615), indicating that the prognostic effect of delirium was consistent regardless of inflammatory or renal status. (Table S6).

4. Discussion

This propensity score-matched analysis, drawing from a large critical care database, establishes delirium as an independent determinant of adverse clinical outcomes in critically ill patients with acute pancreatitis. Our findings demonstrate that ICU-acquired delirium is associated with a significantly increased risk of 90-day all-cause mortality and unplanned readmission, as well as a significantly elevated risk of composite adverse events, even after comprehensive adjustment for baseline severity and therapeutic interventions. These results necessitate a conceptual evolution in how we perceive delirium in this population—transitioning from viewing it merely as an epiphenomenon of critical illness to recognizing its potential role as an active contributor to the disease trajectory.^21,22

The pre-match disparities observed in our cohort, where delirious patients exhibited higher illness severity scores and greater exposure to sedative agents, align with established literature on delirium epidemiology in critical care settings.^7,23 The methodological strength of our approach lies in the application of propensity score matching, which effectively mitigated these substantial baseline imbalances and created comparable groups for outcome assessment.

A key concern in this field is whether delirium represents an independent prognostic factor or merely an epiphenomenon of disease severity. Our data provide several lines of evidence against the latter interpretation. First, PSM comprehensively balanced objective severity indicators — including SOFA score, APACHE II score, SAPS II score, mechanical ventilation, and sedative exposure — between groups, and the significant association between delirium and composite adverse outcomes persisted (Model II: aHR=1.84, 95% CI: 1.24–2.73; P=0.002). Second, the fully adjusted Model II directly incorporated SOFA score, APACHE II score, mechanical ventilation, sepsis, and sedation-related variables as covariates, and the association remained robust across all four sequential models and all four sensitivity analyses (crude Cox, CAPS, IPTW, SMRW). Third, E-value analysis indicated that an unmeasured confounder would need a risk ratio of at least 3.24 with both delirium and the composite outcome to fully explain away the observed association — a threshold unlikely to be met given the comprehensive adjustment already performed.

The differential impact of delirium across outcome metrics warrants careful consideration.^24–27 The significant association with unplanned readmission but not ED revisits after full adjustment suggests that these two healthcare utilization outcomes may be driven by distinct pathways, with unplanned readmission more directly attributable to the sequelae of delirium itself. This pattern implies that the apparent link between delirium and healthcare utilization may be largely mediated by underlying disease severity and complication profiles characteristic of severe pancreatitis, such as pancreatic necrosis and fistula formation.^28,29 In contrast, the robust association with the composite outcome across all analytical approaches indicates that delirium exerts broader detrimental effects on patient recovery that extend beyond acute mortality.

The pathophysiological pathways underlying this association may involve complex neuroimmune interactions, though direct mechanistic evidence from the present study is limited.³⁰ The profound systemic inflammatory response characteristic of severe AP may create a vulnerable cerebral environment predisposing to delirium, as evidenced by the connection between systemic inflammation and acute brain dysfunction.¹¹ Shared inflammatory mediators, including phospholipase A2, have been implicated in both pancreatic injury and neurological impairment,¹² suggesting potential mechanistic convergence. Furthermore, delirium itself may perpetuate a harmful feedback cycle through neuroinflammation, with elevated levels of cytokines such as IL-6 and TNF-α potentially exacerbating systemic inflammatory response syndrome and organ dysfunction.^31–33 In support of this framework, our post-hoc analysis revealed significant differences in renal and nutritional biomarkers between groups, including higher creatinine, BUN, and lower albumin in delirious patients; however, none of these biomarkers significantly modified the delirium-outcome association (all P-interaction >0.05), suggesting that the prognostic effect of delirium operates independently of inflammatory or renal status. The cognitive and behavioral manifestations of delirium may also compromise patient engagement with treatment and rehabilitation, indirectly contributing to adverse outcomes.^34,35 It should be noted that the neuroinflammatory hypothesis presented here is largely informed by existing literature and was not directly tested in the present study; prospective studies incorporating dedicated inflammatory cytokines and neurological biomarkers are warranted to validate these proposed mechanisms.

From a clinical perspective, our findings underscore the imperative for systematic delirium screening in critically ill AP patients using validated instruments like the CAM-ICU.^18,36,37 The identification of high-risk subgroups—particularly older patients, those with renal impairment, and individuals receiving specific sedative regimens—should prompt heightened vigilance and targeted preventive strategies.^38–41 Clinical protocols should emphasize judicious sedation practices, including the consideration of delirium-sparing agents where feasible, and aggressive management of modifiable risk factors such as renal dysfunction and metabolic disturbances.^42,43 Future investigation should focus on developing disease-specific delirium prevention bundles and evaluating their impact on long-term recovery in AP survivors.

Several limitations warrant acknowledgment. The observational design, despite robust methodology, cannot definitively establish causality. While comprehensive adjustment and E-value analysis suggest that residual confounding is unlikely to fully explain the primary findings, it cannot be entirely excluded. The use of MIMIC-IV, a single-center database derived from a tertiary academic medical center in the United States, raises important considerations regarding external validity and generalizability. Notably, the AP etiology profile in our cohort — with biliary (25.6%) and alcoholic (24.1%) causes predominating — differs substantially from that typically reported in Asian populations, where biliary pancreatitis is more prevalent and hypertriglyceridemia-induced pancreatitis accounts for a considerably higher proportion of cases. Furthermore, differences in comorbidity profiles, ICU management protocols, sedation practices, and healthcare system structures between Western and Asian settings may influence both delirium incidence and clinical outcomes. As the authors are affiliated with a Chinese institution, we recognize the particular importance of validating these findings in Asian and non-Western cohorts. Future multicenter studies incorporating diverse ethnic and geographic populations are warranted to determine whether the observed association between ICU-acquired delirium and adverse outcomes is generalizable beyond the Western critical care context. Additionally, as delirium diagnosis relied on CAM-ICU assessments documented in routine nursing records (typically every 8-12 hours), hypoactive delirium — characterized by reduced psychomotor activity and potentially masked by sedation — may have been underdetected, potentially leading to misclassification of some delirious patients into the non-delirium group. Such non-differential misclassification would bias the observed association toward the null, suggesting that the true association between delirium and adverse outcomes may be stronger than reported. Furthermore, patients with a RASS score of ≤ −3 were considered unarousable and could not undergo CAM-ICU assessment; the delirium status of this subgroup remains indeterminate and represents an inherent limitation of the MIMIC-IV database. Pain intensity, assessed by the Numeric Rating Scale (NRS), could not be incorporated into the propensity score model due to a missing rate exceeding 30% (33.5% overall; 31.5% post-PSM), representing a potential source of residual confounding that future studies with more complete pain documentation should address.

The proportion of hypertriglyceridemia-induced pancreatitis in our cohort (6.7%) was lower than typically reported in Asian populations, and a dedicated subgroup analysis was not feasible due to insufficient statistical power. Additionally, the Revised Atlanta Classification could not be implemented due to inherent limitations of the MIMIC-IV database, including the absence of structured CT-based imaging variables and time-stamped organ failure duration records necessary to distinguish transient from persistent organ failure. Finally, despite being among the largest studies specifically addressing delirium in AP, statistical power for certain subgroup analyses remained limited.

In conclusion, this study provides evidence suggesting that delirium independently predicts mortality and composite adverse outcomes in critically ill AP patients. These findings position delirium not merely as a marker of disease severity but as a potentially modifiable determinant of patient recovery. Future prospective studies should focus on elucidating the underlying mechanisms and evaluating targeted interventions to mitigate the burden of delirium in this vulnerable population.

5. Conclusion

In critically ill patients with acute pancreatitis, ICU-acquired delirium is independently associated with significantly increased 90-day mortality and composite adverse outcomes. These findings underscore the importance of routine delirium screening and targeted management—such as judicious sedation and early mobilization—in this high-risk population. Further prospective studies are warranted to validate these results and develop effective delirium-prevention strategies.

Supplemental material

Supplemental material - Impact of delirium on clinical outcomes in critically ill patients with acute pancreatitis: A propensity score-matched study

Supplemental material for Impact of delirium on clinical outcomes in critically ill patients with acute pancreatitis: A propensity score-matched study by Xia Zhou, Haoxuan Xu, Liping Chen, Linling Zhu, Lingyan Dong, Shaoming Wei, Zuwei Wang, Huiren Zhuang in Science Progress

Supplemental material

Supplemental material - Impact of delirium on clinical outcomes in critically ill patients with acute pancreatitis: A propensity score-matched study

Footnotes

ORCID iD

Huiren Zhuang

Ethical considerations

This retrospective cohort study utilized data from the Medical Information Mart for Intensive Care IV (MIMIC-IV) database. The establishment of the MIMIC-IV database was approved by the Institutional Review Boards (IRB) of the Massachusetts Institute of Technology (MIT) and Beth Israel Deaconess Medical Center (BIDMC).

Consent to participate

As the database contains de-identified patient data, the requirement for individual informed consent was waived for this study. The author (Xia Zhou) who accessed the database completed the required Collaborative Institutional Training Initiative (CITI) program course “Data or Specimens Only Research” (Record ID: 54611342).

Consent for publication

Not applicable. This study uses only de-identified data from a publicly available database.

Author contributions

Xia Zhou: Conceptualization, Methodology, Investigation, Formal Analysis, Writing – Original Draft.

Haoxuan Xu: Investigation, Data Curation, Validation, Visualization, Writing – Original Draft.

Liping Chen: Resources, Software, Validation, Investigation.

Linling Zhu: Investigation, Formal Analysis.

Lingyan Dong: Data Curation, Writing – Review & Editing.

Shaoming Wei: Supervision, Funding Acquisition, Project Administration, Writing – Review & Editing.

Zuwei Wang: Supervision, Conceptualization, Writing – Review & Editing.

Huiren Zhuang: Supervision, Project Administration, Resources, Writing – Review & Editing.

All authors have read and approved the final manuscript.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data Availability Statement

The datasets generated and/or analyzed during the current study are derived from the MIMIC-IV database, which is publicly available upon completion of the required training and data use agreement. The data that support the findings of this study are available from the corresponding author upon reasonable request.

Supplemental material

Supplemental material for this article is available online.

References

Lankisch

Apte

Banks

. Acute pancreatitis. Lancet 2015; 386(9988): 85–96. https://doi.org/10.1016/s0140-6736(14)60649-8

Porter

Cason

Morgan

. Acute Pancreatitis: How Can MR Imaging Help. Magn Reson Imaging Clin N Am 2018; 26(3): 439–450. https://doi.org/10.1016/j.mric.2018.03.011

Petrov

Yadav

. Global epidemiology and holistic prevention of pancreatitis. Nat Rev Gastroenterol Hepatol 2019; 16(3): 175–184. https://doi.org/10.1038/s41575-018-0087-5

Hipp

Ely

. Pharmacological and nonpharmacological management of delirium in critically ill patients. Neurotherapeutics 2012; 9(1): 158–175. https://doi.org/10.1007/s13311-011-0102-9

Mattison

MLP

. Delirium. Ann Intern Med 2020; 173(7): Itc49–itc64. https://doi.org/10.7326/aitc202010060

Wilcox

Girard

Hough

. Delirium and long term cognition in critically ill patients. BMJ 2021; 373(n1007): n1007. https://doi.org/10.1136/bmj.n1007

Salluh

Wang

Schneider

, et al. Outcome of delirium in critically ill patients: systematic review and meta-analysis. BMJ 2015; 350(h2538): h2538. https://doi.org/10.1136/bmj.h2538

Fiest

Soo

Hee

, et al. Long-Term Outcomes in ICU Patients with Delirium: A Population-based Cohort Study. Am J Respir Crit Care Med 2021; 204(4): 412–420. https://doi.org/10.1164/rccm.202002-0320OC

Cavallazzi

Saad

Marik

. Delirium in the ICU: an overview. Ann Intensive Care 2012; 2(1): 49. https://doi.org/10.1186/2110-5820-2-49

10.

Mao

. Intensive management of severe acute pancreatitis. Annals of translational medicine 2019; 7(22): 687. https://doi.org/10.21037/atm.2019.10.58

11.

Brummel

Hughes

McNeil

, et al. Systemic inflammation and delirium during critical illness. Intensive Care Med 2024; 50(5): 687–696. https://doi.org/10.1007/s00134-024-07388-6

12.

Lindroth

Fiest

Lee

, et al. Delirium Severity Trajectories in Critically Ill Adults Using the Intensive Care Delirium Screening Checklist: A Population-Based Cohort. Crit Care Med 2025; 53: e1748–e1758. https://doi.org/10.1097/ccm.0000000000006768

13.

Johnson

AEW

Bulgarelli

Shen

, et al. Author Correction: MIMIC-IV, a freely accessible electronic health record dataset. Scientific data 2023; 10(1): 219. https://doi.org/10.1038/s41597-023-02136-9

14.

von

Altman

Egger

, et al. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. Ann Intern Med 2007; 147(8): 573–577. https://doi.org/10.7326/0003-4819-147-8-200710160-00010

15.

Knaus

Draper

Wagner

, et al. APACHE II: a severity of disease classification system. Crit Care Med 1985; 13(10): 818–829.

16.

Knaus

Wagner

Draper

, et al. The APACHE III prognostic system. Risk prediction of hospital mortality for critically ill hospitalized adults. Chest 1991; 100(6): 1619–1636. https://doi.org/10.1378/chest.100.6.1619

17.

Inouye

van Dyck

Alessi

, et al. Clarifying confusion: the confusion assessment method. A new method for detection of delirium. Ann Intern Med 1990; 113(12): 941–948. https://doi.org/10.7326/0003-4819-113-12-941

18.

Ely

Margolin

Francis

, et al. Evaluation of delirium in critically ill patients: validation of the Confusion Assessment Method for the Intensive Care Unit (CAM-ICU). Crit Care Med 2001; 29(7): 1370–1379. https://doi.org/10.1097/00003246-200107000-00012

19.

Wang

Liao

Lin

, et al. YL-109 attenuates sepsis-associated multiple organ injury through inhibiting the ERK/AP-1 axis and pyroptosis by upregulating CHIP. Biomed Pharmacother 2024; 175: 116633. https://doi.org/10.1016/j.biopha.2024.116633

20.

Schuler

Chu

Coffman

. Propensity score weighting for a continuous exposure with multilevel data. Health services & outcomes research methodology 2016; 16(4): 271–292. https://doi.org/10.1007/s10742-016-0157-5

21.

Kang

Cho

Lee

, et al. Effects of nonpharmacological interventions on sleep improvement and delirium prevention in critically ill patients: A systematic review and meta-analysis. Aust Crit Care 2023; 36(4): 640–649. https://doi.org/10.1016/j.aucc.2022.04.006

22.

Herling

Greve

Vasilevskis

, et al. Interventions for preventing intensive care unit delirium in adults. Cochrane Database Syst Rev 2018; 11(11): CD009783. https://doi.org/10.1002/14651858.CD009783.pub2

23.

Siegel

Traube

. Pediatric delirium: epidemiology and outcomes. Curr Opin Pediatr 2020; 32(6): 743–749. https://doi.org/10.1097/MOP.0000000000000960

24.

Girard

Thompson

Pandharipande

, et al. Clinical phenotypes of delirium during critical illness and severity of subsequent long-term cognitive impairment: a prospective cohort study. Lancet Respir Med 2018; 6(3): 213–222. https://doi.org/10.1016/S2213-2600(18)30062-6

25.

Moskowitz

Overbey

Jones

, et al. Post-operative delirium is associated with increased 5-year mortality. Am J Surg 2017; 214(6): 1036–1038. https://doi.org/10.1016/j.amjsurg.2017.08.034

26.

Roden

Simmons

. Delirium superimposed on dementia and mild cognitive impairment. Postgrad Med 2014; 126(6): 129–137. https://doi.org/10.3810/pgm.2014.10.2827

27.

Schnorr

Fleiner

Schroeder

, et al. Post-discharge Mortality in Patients With Delirium and Dementia: A 3-Year Follow Up Study. Frontiers in psychiatry 2022; 13: 835696. https://doi.org/10.3389/fpsyt.2022.835696

28.

Bogan

McGuire

Maatman

. Readmission in acute pancreatitis: Etiology, risk factors, and opportunities for improvement. Surgery open science 2022; 10: 232–237. https://doi.org/10.1016/j.sopen.2022.10.010

29.

Garg

Campbell

Anugwom

, et al. Incidence and Predictors of Readmissions in Acute Pancreatitis: A Nationwide Analysis. Pancreas 2018; 47(1): 46–54. https://doi.org/10.1097/mpa.0000000000000952

30.

Heffernan

ÁB

Steinruecke

Dempsey

, et al. Role of glia in delirium: proposed mechanisms and translational implications. Mol Psychiatry 2025; 30(3): 1138–1147. https://doi.org/10.1038/s41380-024-02801-4

31.

Zhang

Luo

, et al. Correlation of Serum BACE1 With Emergence Delirium in Postoperative Patients: A Preliminary Study. Front Aging Neurosci 2020; 12: 555594. https://doi.org/10.3389/fnagi.2020.555594

32.

Kotfis

Ślozowska

Safranow

, et al. The Practical Use of White Cell Inflammatory Biomarkers in Prediction of Postoperative Delirium after Cardiac Surgery. Brain sciences 2019; 9(11): ■■■. https://doi.org/10.3390/brainsci9110308

33.

Cabrera-Rivera

Madera-Sandoval

León-Pedroza

, et al. Increased TNF-α production in response to IL-6 in patients with systemic inflammation without infection. Clin Exp Immunol 2022; 209(2): 225–235. https://doi.org/10.1093/cei/uxac055

34.

Lin

Zhang

Xiao

, et al. Delirium is a Potential Predictor of Unfavorable Long-term Functional Outcomes in Patients with Acute Ischemic Stroke: A Prospective Observational Study. Journal of inflammation research 2025; 18: 4019–4035. https://doi.org/10.2147/jir.S505038

35.

Rollo

Brunetti

Scala

, et al. Impact of delirium on the outcome of stroke: a prospective, observational, cohort study. J Neurol 2022; 269(12): 6467–6475. https://doi.org/10.1007/s00415-022-11309-2

36.

Chen

Chung

Chang

, et al. Diagnostic accuracy of the CAM-ICU and ICDSC in detecting intensive care unit delirium: A bivariate meta-analysis. Int J Nurs Stud 2021; 113: 103782. https://doi.org/10.1016/j.ijnurstu.2020.103782

37.

Miranda

Gonzalez

Plana

, et al. Confusion Assessment Method for the Intensive Care Unit (CAM-ICU) for the diagnosis of delirium in adults in critical care settings. Cochrane Database Syst Rev 2023; 11(11): CD013126. https://doi.org/10.1002/14651858.CD013126.pub2

38.

Bramley

McArthur

Blayney

, et al. Risk factors for postoperative delirium: An umbrella review of systematic reviews. Int J Surg 2021; 93: 106063. https://doi.org/10.1016/j.ijsu.2021.106063

39.

Bugiani

. Why is delirium more frequent in the elderly? Neurol Sci 2021; 42(8): 3491–3503. https://doi.org/10.1007/s10072-021-05339-3

40.

Siew

Fissell

Tripp

, et al. Acute Kidney Injury as a Risk Factor for Delirium and Coma during Critical Illness. Am J Respir Crit Care Med 2017; 195(12): 1597–1607. https://doi.org/10.1164/rccm.201603-0476OC

41.

Pang

Kumar

Ely

, et al. Acute kidney injury-associated delirium: a review of clinical and pathophysiological mechanisms. Crit Care 2022; 26(1): 258. https://doi.org/10.1186/s13054-022-04131-9

42.

Duprey

Dijkstra-Kersten

SMA

Zaal

, et al. Opioid Use Increases the Risk of Delirium in Critically Ill Adults Independently of Pain. Am J Respir Crit Care Med 2021; 204(5): 566–572. https://doi.org/10.1164/rccm.202010-3794OC

43.

Clegg

Young

. Which medications to avoid in people at risk of delirium: a systematic review. Age Ageing 2011; 40(1): 23–29. https://doi.org/10.1093/ageing/afq140

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.22 MB

4.07 MB