Association Between Preoperative Malnutrition Assessed by the Controlling Nutritional Status (CONUT) Score and Postoperative Delirium in Elderly Patients with Hip Fractures: A Retrospective Cohort Study

Abstract

Background

Postoperative delirium (POD) is associated with impaired cognitive function, increased morbidity, and elevated mortality. Although preoperative nutritional impairment has been associated with POD, the specific association between the objective, laboratory-based Controlling Nutritional Status (CONUT) score and POD, including its dose-response pattern, remains insufficiently characterized in elderly patients with hip fractures. This study aimed to evaluate the association between preoperative CONUT score and POD in this population.

Methods

This retrospective cohort study included patients aged ≥ 65 years who underwent surgery for hip fracture at a university-affiliated hospital between January 2020 and December 2024. Preoperative nutritional status was assessed using the CONUT score, and POD was identified retrospectively from medical records with reference to the Confusion Assessment Method. Restricted cubic spline analysis was used to characterize the dose-response association between continuous CONUT score and POD, and multivariable logistic regression was performed as the primary adjusted analysis. Exploratory cutoff-based analyses included propensity score matching, conditional logistic regression, subgroup analyses, and sensitivity analyses.

Results

A total of 2030 patients were included, of whom 337 (16.6%) developed POD. Restricted cubic spline analysis showed a positive dose-response association between increasing CONUT score and POD. In the primary multivariable analysis, each 1-point increase in CONUT score was associated with higher odds of POD (adjusted OR, 1.08; 95% confidence interval [CI], 1.02–1.14; P = 0.009). In the exploratory categorical analysis, patients with a CONUT score ≥ 5 had higher odds of POD than those with lower scores after propensity score matching (OR, 1.63; 95% CI, 1.20–2.21; P = 0.002). No significant effect modification was observed across predefined subgroups, and sensitivity analyses yielded consistent findings.

Conclusion

Higher preoperative CONUT scores were associated with greater odds of POD in elderly patients with hip fractures. The CONUT score may serve as a practical adjunctive marker for identifying patients with greater perioperative vulnerability to POD.

Introduction

Globally, over 10 million hip fractures occur each year, a number projected to nearly double by 2050, and 1-year mortality can approach 30%.¹ The incidence of postoperative delirium (POD) in elderly patients with hip fractures ranges from 10.1% to 51.3%,² typically occurring between postoperative days 2 and 5.³ It not only significantly increases the complication rates and mortality, delays rehabilitation, impairs long-term cognitive function, but also prolongs hospital stays and elevates medical costs.³ Given its severity, the American Society for Enhanced Recovery and the Perioperative Quality Initiative have identified POD prevention as a key target for optimizing perioperative care.⁴ In fact, roughly one-third of POD cases are preventable, thereby making the early identification of modifiable risk factors crucial.⁵

Malnutrition is biologically plausible, clinically relevant, and potentially modifiable in relation to POD, but prior studies, despite establishing a stepwise evidence base, have left important gaps. In a prospective observational study of 415 older patients with hip fractures assessed using the Mini Nutritional Assessment-Short Form (MNA-SF), preoperative nutritional risk and overt malnutrition were independently associated with POD, establishing a clinical link between poor nutrition and delirium.⁶ However, the exposure relied on a questionnaire with rater dependence and bedside time cost, which limits scalability and standardization compared with a laboratory-based screen. A retrospective analysis showed that preoperative hypoalbuminemia was associated with a higher likelihood of POD, with a graded pattern and a data-driven threshold. Albumin, though objective and widely available, is a single marker that is sensitive to inflammation, hydration shifts, and acute illness;⁷ it does not capture multi-dimensional immune–nutritional reserve. Notably, the Controlling Nutritional Status (CONUT) score, which incorporates albumin, total lymphocyte count, and total cholesterol, possesses a sensitivity of 92.3% and a specificity of 85.0% in comprehensive nutritional assessment for inpatients.⁸ It allows assessors to evaluate nutritional status independently, retrospectively, and reproducibly. A retrospective study of 211 patients with hip fractures showed that the CONUT score was associated with postoperative complications, supporting its potential utility in perioperative risk assessment.⁹ However, POD was not included as an outcome in that study, and the specific association between preoperative CONUT score and POD in elderly patients with hip fractures remains insufficiently characterized. Importantly, although previous studies have linked general nutritional impairment or individual nutritional biomarkers to POD, evidence specifically evaluating the multidimensional CONUT score in relation to POD, including its dose-response association, remains limited. Because nutritional impairment is potentially modifiable, its early identification may provide an opportunity for timely nutritional assessment and supportive management; however, whether nutritional intervention can reduce POD requires confirmation in prospective interventional studies.

Therefore, the present study aimed to evaluate the association between preoperative nutritional status assessed by the CONUT score and POD in elderly patients undergoing hip fracture surgery and to characterize the dose-response relationship between CONUT score and POD.

Materials and Methods

General Information

This study retrospectively enrolled inpatients aged 65 years and older who underwent surgical intervention for hip fracture at Central Hospital Affiliated to Shenyang Medical College between January 2020 and December 2024. Patients were excluded if they met any of the following criteria: (1) multiple trauma; (2) pathological fractures; (3) preoperative central nervous system or neuropsychiatric disorders, including preoperative delirium, dementia (Alzheimer’s disease or dementia with Lewy bodies), Parkinson’s disease, stroke within the previous 6 months, or other central nervous system diseases; (4) preoperative or immediate postoperative admission to the intensive care unit (ICU); or (5) incomplete clinical data. The study was performed in accordance with the Strengthening the Reporting of Cohort Studies in Surgery (STROCSS) guidelines and the ethical principles of the Declaration of Helsinki, with prior approval from the Ethics Committee of Central Hospital Affiliated to Shenyang Medical College (approval No. K-2026-031-02). The requirement for informed consent was waived due to the retrospective design and anonymized data analysis. It should be noted that no formal a priori sample size calculation was performed in this study, and all eligible patients during the predefined study period were included. With 337 POD events and approximately 13.5 events per candidate predictor in the primary multivariable model, the study exceeded the conventional rule of thumb of 10 events per variable for logistic regression. Therefore, the sample size was considered adequate for the planned multivariable association analysis.

Data Collection

Clinical data were collected in four fields: demographic characteristics, comorbidities, perioperative variables, and laboratory biomarkers. Demographic characteristics included sex, age, calculated body mass index (BMI), living place (urban/rural), marital status, occupation, primary payer, current smoking, and alcohol consumption. Comorbidities encompassed hypertension, diabetes mellitus, heart disease, cerebrovascular disease, chronic obstructive pulmonary disease (COPD), tumors, liver disease, kidney disease, autoimmune diseases, and peripheral vascular disease. Past history variables included surgical history and allergic history. Perioperative variables included preoperative sleep difficulty, surgical delay (days), anesthesia method, injury energy, fracture type, operation type, surgical duration, intraoperative blood loss, and blood transfusion. Preoperative physical status and comorbidity burden were assessed using the American Society of Anesthesiologists (ASA) score and the Charlson Comorbidity Index (CCI), respectively. Laboratory biomarkers were obtained from routine preoperative testing and included red blood cell count (RBC), hemoglobin concentration (HGB), white blood cell count (WBC), platelet count (PLT), total protein (TP), alanine aminotransferase (ALT), aspartate aminotransferase (AST), hypersensitive C-reactive protein (HCRP), serum sodium (Na⁺), potassium (K⁺), calcium (Ca²⁺), fasting blood glucose (FBG), urea, creatinine (CREA), anion gap (GAP), total cholesterol, lymphocytes, and serum albumin. If multiple laboratory measurements were available prior to surgery, the value closest to the operation was collected for analysis.

Nutritional Status Assessment

In this study, we used the Controlling Nutritional Status (CONUT) score to evaluate the preoperative nutritional status of included patients. The CONUT score was computed from three key nutritional markers: serum albumin, total cholesterol, and lymphocyte count (Table 1). All laboratory tests were conducted in accordance with the manufacturer’s recommended protocols, with biochemical assays performed on a Beckman Coulter AU5800 chemistry analyzer (Beckman Coulter, Brea, CA, USA) and hematological detection completed using a UniCel DXI 800 system (Beckman Coulter).

Table 1.

The CONUT Score for Nutritional Status Assessment

Variables	Malnutrition Status
	Normal	Mild	Moderate	Severe
Serum albumin concentration (g/L)	≥ 35.0	34.9–30.0	29.9–25.0	< 25.0
Albumin points	0	2	4	6
Peripheral blood lymphocyte count (/L)	> 1.6	1.2–1.6	0.8–1.2	< 0.8
Lymphocyte points	0	1	2	3
Total cholesterol concentration (mg/L)	≥ 18.0	14.0–17.9	10.0–13.9	< 10.0
Cholesterol points	0	1	2	3
Screening total score	0-1	2-4	5-8	9-12

Notes: Bold text indicates column headings and CONUT scoring rows, including component point values and total score categories.

Abbreviation: CONUT, Controlling Nutritional Status.

Diagnosis of POD

In this study, POD was retrospectively ascertained from electronic medical records (EMRs) with reference to the core features of the Confusion Assessment Method (CAM).^10,11 Because standardized prospective CAM assessments were not routinely performed at predefined postoperative time points, the four core CAM features (I. acute onset and fluctuating course, II. inattention, III. altered level of consciousness, and IV. disorganized thinking) were operationalized into predefined chart-extractable indicators. Specifically, acute onset and fluctuating course were identified from documentation of sudden postoperative changes in mental status, such as agitation or incoherent speech; inattention was identified from records indicating inability to follow simple instructions or impaired orientation; altered level of consciousness was identified from documentation of abnormal arousal, including drowsiness, difficulty in awakening, hyperalertness, or marked agitation; and disorganized thinking was identified from records of illogical speech, irrelevant responses, or incoherent thought processes. POD was classified when the documented features fulfilled the combination of Criterion I + Criterion II + (Criterion III or Criterion IV).

The medical records reviewed included postoperative nursing records, physician progress notes, and specialist consultation records during postoperative days 1–7. To improve the consistency of POD ascertainment, two trained orthopedic surgeons (SY and LFJ) independently reviewed the EMRs and extracted POD-related information. Discrepancies were resolved by cross-verification and joint review of the original records, and any remaining ambiguous cases were adjudicated by a chief neurologist blinded to the patients’ nutritional status. This retrospective chart-based ascertainment was not equivalent to standardized prospective and repeated CAM assessments performed at uniform postoperative time points and may therefore have resulted in outcome misclassification, particularly underrecognition of hypoactive delirium.

Statistical Analysis

For continuous variables, the Kolmogorov–Smirnov test was applied to verify the normality of data distribution. Normally distributed variables were presented as mean ± standard deviation (SD), and intergroup comparisons were performed using the Student’s t-test. Non-normally distributed continuous variables were described as median (interquartile range, IQR, Q1–Q3), with between-group differences tested by the Mann–Whitney U-test. Categorical variables were summarized as counts (percentages), and comparisons were conducted using the Pearson’s chi-square test or Fisher’s exact test, as appropriate.

The CONUT score was treated as a continuous exposure in the primary analysis. To characterize the dose-response relationship between preoperative CONUT score and POD, a restricted cubic spline (RCS) model with four knots placed at the 5th, 35th, 65th, and 95th percentiles of the CONUT score distribution was constructed within a univariable logistic regression framework. This analysis was performed using the rms package in R software (version 4.3.1, R Foundation for Statistical Computing, Vienna, Austria). The P-overall value was used to assess the overall association between CONUT score and POD, whereas the P-nonlinearity value was used to evaluate evidence of a nonlinear association.

To quantify the adjusted association between continuous CONUT score and POD, multivariable logistic regression analysis was performed, with odds ratios (ORs) and corresponding 95% confidence intervals (CIs) reported for each 1-point increase in the CONUT score. Candidate covariates were selected based on both their associations with POD in univariate analyses and their clinical relevance. Variables with P < 0.05 in univariate analyses were considered candidate covariates, while sex and fracture type were additionally included because of their clinical relevance to POD and perioperative outcomes despite not reaching the univariate significance threshold. The initial multivariable model therefore included age, sex, BMI, current smoking, preoperative sleep difficulty, cerebrovascular disease, COPD, kidney disease, surgical delay, anesthesia method, injury energy, fracture type, operation type, surgical duration, ASA score, CCI, hemoglobin, total protein, ALT, HCRP, serum calcium, urea, creatinine, and anion gap, together with the continuous CONUT score. Backward stepwise logistic regression was subsequently applied to obtain the final multivariable model. Serum albumin, peripheral blood lymphocyte count, and total cholesterol were not included as covariates because they are direct components of the CONUT score. In addition, RBC was not simultaneously entered with hemoglobin because of their close biological and statistical overlap.

For exploratory analyses, the optimal cutoff value of the CONUT score for discriminating POD was determined by maximizing the Youden index derived from receiver operating characteristic (ROC) curve analysis.¹² Based on the identified threshold, patients were categorized into the low CONUT score (LCS) group and the high CONUT score (HCS) group. All subsequent cutoff-based analyses, including the LCS versus HCS comparison, propensity score matching, subgroup analyses, and sensitivity analyses, were considered exploratory analyses.

To minimize confounding in the exploratory comparison between the LCS and HCS groups, propensity scores were estimated using logistic regression. Covariates were selected based on their clinical relevance and their potential to confound the association between preoperative nutritional status and POD rather than solely on statistical significance in baseline comparisons. The propensity score model included sex, age, BMI, living place, marital status, occupation, primary payer, current smoking, alcohol consumption, preoperative sleep difficulty, hypertension, diabetes mellitus, heart disease, cerebrovascular disease, COPD, tumors, liver disease, kidney disease, autoimmune diseases, peripheral vascular disease, surgical history, allergic history, surgical delay, anesthesia method, injury energy, fracture type, operation type, ASA score, CCI, RBC, hemoglobin, WBC, platelet count, total protein, ALT, AST, HCRP, serum sodium, potassium, calcium, fasting blood glucose, urea, creatinine, and anion gap. These variables were selected to provide broad adjustment for demographic characteristics, comorbidity burden, physiological status, fracture and treatment characteristics, systemic inflammation, metabolic status, hematologic status, and renal function that could potentially confound the association between CONUT-defined nutritional status and POD. Serum albumin, peripheral blood lymphocyte count, and total cholesterol were not included because they are direct components of the CONUT score. Surgical duration, intraoperative blood loss, and blood transfusion were also not included in the propensity score model because they occurred after assessment of the preoperative CONUT exposure.

Propensity score matching (PSM) was performed using the MatchIt package in R software with a 1:1 nearest-neighbor matching algorithm. A caliper width of 0.2 times the standard deviation of the logit of the propensity score was used. Covariate balance before and after matching was assessed using standardized mean differences (SMDs), with an absolute SMD < 0.10 considered indicative of satisfactory balance.¹³

After PSM, conditional logistic regression analysis was used to evaluate the association between HCS and POD, with results presented as ORs and corresponding 95% CIs. Subgroup analyses were subsequently performed to assess potential heterogeneity in the association between CONUT category and POD across predefined strata, including age, sex, BMI, anesthesia type, fracture type, operation type, ASA score, and CCI. Interaction effects were evaluated by including the cross-product term between CONUT category and the corresponding subgroup variable in the regression model. A P-value for interaction < 0.05 was considered indicative of statistically significant effect modification.

To assess the robustness of the exploratory cutoff-based findings, a series of sensitivity analyses was conducted. First, the cutoff-based analysis was repeated after excluding patients whose peripheral blood lymphocyte count, serum albumin level, or total cholesterol level was below the 2.5th percentile or above the 97.5th percentile of the overall study population to minimize the potential influence of extreme values. This analysis represented sample restriction rather than covariate adjustment; these three variables were not entered as adjustment covariates in any model. Second, asymmetric trimming of the propensity score was performed by excluding patients with a propensity score below the 2.5th percentile of the HCS group and above the 97.5th percentile of the LCS group to further improve comparability between groups. Third, patients with tumors, liver disease, or kidney disease were excluded because these conditions may affect nutritional metabolism and serum biomarker levels, potentially influencing CONUT-based nutritional assessment. Fourth, patients with autoimmune diseases were excluded because autoimmune disorders may alter immune status and lymphocyte counts, potentially affecting CONUT score calculation and confounding its association with POD.

All statistical tests were two-sided, and a P-value < 0.05 was considered to indicate a statistically significant difference. All statistical analyses were performed using R software (version 4.3.1, R Foundation for Statistical Computing, Vienna, Austria).

Results

As shown in Figure 1, a total of 2030 eligible elderly patients with hip fractures were included in the final analysis, of whom 337 (16.6%) developed POD. The mean age was 75.08 ± 7.51 years, and 62.76% of the cohort were women. Compared with patients without POD, those who developed POD were generally older and leaner and more frequently had current smoking, preoperative sleep difficulty, cerebrovascular disease, COPD, kidney disease, prolonged surgical delay, general anesthesia, high-energy injury, arthroplasty, longer operative duration, higher ASA score, higher CCI, and a higher CONUT score. Several laboratory variables, including RBC count, HGB, TP, HCRP, Ca²⁺, UREA, CREA, and GAP, also differed significantly between the two groups (P < 0.05, Table 2).

Figure 1.

Flow diagram of patient selection. Arrows indicate the sequence of patient screening, exclusion, CONUT-based grouping, postoperative outcome ascertainment, and propensity score matching.

Abbreviations: CONUT, Controlling Nutritional Status; LCS, low CONUT score; HCS, high CONUT score; POD, postoperative delirium; PSM, propensity score matching.

Table 2.

Univariate Analysis of Variables of Interest Between Patients with POD and without POD

Variables	Overall (n=2030)	Patients Without POD (n=1693)	Patients with POD (n=337)	P-value
Sex (female)	1274 (62.76%)	1064 (62.85%)	210 (62.31%)	0.854
Age	75.08 ± 7.51	74.37 ± 7.36	78.65 ± 7.24	<0.001
> 75	645 (31.77%)	445 (26.28%)	200 (59.35%)	<0.001
BMI (kg/m2)	23.57 ± 3.70	23.74 ± 3.59	22.70 ± 4.10	<0.001
18.5–23.9	991 (48.82%)	827 (48.85%)	164 (48.66%)	<0.001
<18.5	161 (7.93%)	108 (6.38%)	53 (15.73%)
24.0–27.9	682 (33.60%)	585 (34.55%)	97 (28.78%)
≥28.0	196 (9.66%)	173 (10.22%)	23 (6.82%)
Living place (Urban)	939 (46.26%)	774 (45.72%)	165 (48.96%)	0.275
Marital status				0.501
Spouse	1787 (88.03%)	1494 (88.25%)	293 (86.94%)
No spouse	243 (11.97%)	199 (11.75%)	44 (13.06%)
Occupation				0.659
Manual	1039 (51.18%)	865 (51.09%)	174 (51.63%)
Nonmanual	500 (24.63%)	423 (24.99%)	77 (22.85%)
Unavailable	491 (24.19%)	405 (23.92%)	86 (25.52%)
Primary payer				0.359
Insurance	1720 (84.73%)	1440 (85.06%)	280 (83.09%)
Self-pay	310 (15.27%)	253 (14.94%)	57 (16.91%)
Current smoking	283 (13.94%)	224 (13.23%)	59 (17.51%)	0.038
Alcohol consumption	275 (13.55%)	223 (13.17%)	52 (15.43%)	0.269
Difficulty in sleeping	527 (25.96%)	384 (22.68%)	143 (42.43%)	<0.001
Hypertension	986 (48.57%)	809 (47.78%)	177 (52.52%)	0.112
Diabetes mellitus	451 (22.22%)	364 (21.50%)	87 (25.82%)	0.082
Heart disease	555 (27.34%)	456 (26.93%)	99 (29.38%)	0.358
Cerebrovascular disease	647 (31.87%)	508 (30.01%)	139 (41.25%)	<0.001
COPD	82 (4.04%)	60 (3.54%)	22 (6.53%)	0.011
Tumors	43 (2.12%)	32 (1.89%)	11 (3.26%)	0.110
Liver disease	68 (3.35%)	51 (3.01%)	17 (5.04%)	0.058
Kidney disease	105 (5.17%)	80 (4.73%)	25 (7.42%)	0.041
Autoimmune diseases	51 (2.51%)	45 (2.66%)	6 (1.78%)	0.347
Peripheral vascular disease	51 (2.51%)	41 (2.42%)	10 (2.97%)	0.559
Surgical history	609 (30.00%)	496 (29.30%)	113 (33.53%)	0.121
Allergic history	405 (19.95%)	339 (20.02%)	66 (19.58%)	0.854
Surgical delay (days)				0.014
< 2	215 (10.59%)	185 (10.93%)	30 (8.90%)
2-5	812 (40.00%)	696 (41.11%)	116 (34.42%)
≥ 6	1003 (49.41%)	812 (47.96%)	191 (56.68%)
Anesthesia method (General)	957 (47.14%)	777 (45.89%)	180 (53.41%)	0.012
Injury energy (high energy)	192 (9.46%)	150 (8.86%)	42 (12.46%)	0.039
Fracture type				0.077
Femoral-neck fracture	1011 (49.80%)	858 (50.68%)	153 (45.40%)
Intertrochanteric fracture	1019 (50.20%)	835 (49.32%)	184 (54.60%)
Operation type				0.014
Osteosynthesis	1235 (60.84%)	1050 (62.02%)	185 (54.90%)
Arthroplasty	795 (39.16%)	643 (37.98%)	152 (45.10%)
Surgical duration (> 180 min)	150 (7.39%)	108 (6.38%)	42 (12.46%)	<0.001
Intraoperative blood loss (mL)	200.0 (100.0, 400.0)	200.0 (100.0, 400.0)	200.0 (100.0, 300.0)	0.297
ASA score				<0.001
I-II	1129 (55.62%)	997 (58.89%)	132 (39.17%)
III-IV	901 (44.38%)	696 (41.11%)	205 (60.83%)
CCI				<0.001
0	781 (38.47%)	768 (45.36%)	13 (3.86%)
1-2	932 (45.91%)	704 (41.58%)	228 (67.66%)
≥3	317 (15.62%)	221 (13.05%)	96 (28.49%)
Blood transfusion	652 (32.12%)	533 (31.48%)	119 (35.31%)	0.169
CONUT score	4.00 (2.00, 6.00)	4.00 (2.00, 6.00)	4.00 (2.00, 6.00)	<0.001
RBC	3.64 ± 0.62	3.67 ± 0.63	3.48 ± 0.60	<0.001
< lower limitation	1060 (52.22%)	842 (49.73%)	218 (64.69%)	<0.001
HGB	112.63 ± 18.15	113.65 ± 18.34	107.52 ± 16.27	<0.001
< lower limitation	1070 (52.71%)	844 (49.85%)	226 (67.06%)	<0.001
WBC	8.87 ± 2.85	8.85 ± 2.75	8.96 ± 3.30	0.593
> 9.5 × 109/L	730 (35.96%)	604 (35.68%)	126 (37.39%)	0.550
PLT (109/L)	211.54 ± 79.87	211.25 ± 78.33	213.01 ± 87.31	0.732
125-350 × 109/L	1749 (86.16%)	1471 (86.89%)	278 (82.49%)	0.083
< 125 × 109/L	176 (8.67%)	141 (8.33%)	35 (10.39%)
> 350 × 109/L	105 (5.17%)	81 (4.78%)	24 (7.12%)
TP	60.22 ± 7.49	60.60 ± 7.57	58.35 ± 6.79	<0.001
< 65 g/L	1476 (72.71%)	1221 (72.12%)	255 (75.67%)	0.182
ALT	23.36 ± 24.44	23.81 ± 21.78	21.12 ± 34.79	0.065
> upper limitation	165 (8.13%)	151 (8.92%)	14 (4.15%)	0.003
AST	24.66 ± 17.43	24.72 ± 16.36	24.35 ± 22.07	0.721
> upper limitation	245 (12.07%)	194 (11.46%)	51 (15.13%)	0.059
HCRP	50.50 ± 42.94	48.65 ± 41.89	59.81 ± 46.87	<0.001
> 52.1 mg/L	958 (47.19%)	758 (44.77%)	200 (59.35%)	<0.001
Na+	137.46 ± 3.85	137.50 ± 3.70	137.26 ± 4.51	0.371
< 137 mmol/L	840 (41.38%)	695 (41.05%)	145 (43.03%)	0.501
K+	4.01 ± 0.49	4.01 ± 0.48	4.03 ± 0.55	0.456
< 3.5 mmol/L	266 (13.10%)	216 (12.76%)	50 (14.84%)	0.302
Ca2+	2.13 ± 0.17	2.14 ± 0.18	2.08 ± 0.14	<0.001
< 2.11 mmol/L	896 (44.14%)	709 (41.88%)	187 (55.49%)	<0.001
FBG	6.79 ± 2.40	6.73 ± 2.01	7.12 ± 3.79	0.061
> 6.1 mmol/L	1077 (53.05%)	884 (52.22%)	193 (57.27%)	0.089
UREA	6.33 ± 2.86	6.15 ± 2.66	7.25 ± 3.56	<0.001
> 8 mmol/L	358 (17.64%)	260 (15.36%)	98 (29.08%)	<0.001
CREA	65.82 ± 46.77	64.42 ± 43.80	72.87 ± 59.09	0.013
> upper limitation	423 (20.84%)	324 (19.14%)	99 (29.38%)	<0.001
GAP	9.51 ± 2.69	9.38 ± 2.68	10.18 ± 2.64	<0.001
< 8 mmol/L	1504 (74.09%)	1233 (72.83%)	271 (80.42%)	0.007
8-16 mmol/L	497 (24.48%)	437 (25.81%)	60 (17.80%)
> 16 mmol/L	29 (1.43%)	23 (1.36%)	6 (1.78%)

Note: Bold text indicates column headings.

Abbreviations: POD, postoperative delirium; BMI, body mass index; COPD, chronic obstructive pulmonary disease; ASA, American Society of Anesthesiologists; CCI, Charlson Comorbidity Index; CONUT, Controlling Nutritional Status; RBC, red blood cell, reference range: Female, 3.5–5.0×10¹²/L; Males, 4.0–5.5×10¹²/L; HGB, hemoglobin, reference range: Females, 110–150g/L; Males, 120–160g/L; WBC, white blood cell; PLT, platelet; TP, total protein; ALT, alanine aminotransferase, reference range: Female, 7–40 U/L; Males, 9–50 U/L; AST, aspartate aminotransferase, reference range: Female, 13–35 U/L; Males, 15–40 U/L; HCRP, high-sensitivity C-reactive protein, A new cutoff value of 52.1 mg/L was identified by maximizing the Youden index; FBG, fasting blood glucose; UREA, blood urea; CREA, serum creatinine, reference range: Female, 41–73 μmol/L; Males, 57–97 μmol/L; GAP, anion gap.

As shown in Figure 2, RCS analysis demonstrated a positive dose-response association between increasing preoperative CONUT score and POD, without evidence of a nonlinear association (P-overall < 0.001; P-nonlinearity = 0.287). In univariable logistic regression analysis, each 1-point increase in the CONUT score was associated with higher odds of POD (OR, 1.19; 95% CI, 1.13–1.24; P < 0.001). In the primary multivariable logistic regression analysis, this association remained statistically significant after adjustment for covariates selected through backward stepwise regression, with each 1-point increase in CONUT score associated with higher odds of POD (adjusted OR, 1.08; 95% CI, 1.02–1.14; P = 0.009; Table 3).

Figure 2.

Restricted cubic spline (RCS) analysis of the association between preoperative CONUT score and postoperative delirium in elderly patients with Hip fractures. The RCS model was fitted within a univariable logistic regression framework using four knots placed at the 5th, 35th, 65th, and 95th percentiles of the CONUT score distribution. The solid red line represents the estimated odds ratio, and the shaded area represents the corresponding 95% confidence interval. The horizontal dashed line indicates an odds ratio of 1.0. The histogram shows the density distribution of CONUT scores and corresponds to the right y-axis. P-overall indicates the significance of the overall association, and P-nonlinearity indicates the test for nonlinearity.

Abbreviations: CONUT, Controlling Nutritional Status; POD, postoperative delirium; RCS, restricted cubic spline; OR, odds ratio; CI, confidence interval.

Table 3.

Logistic Regression Analysis of the Association Between Continuous CONUT Score and POD

Variables	Unadjusted OR (95% CI)	P-value	Adjusted OR (95% CI)	P-value
CONUT score	1.19 (1.13–1.24)	<0.001	1.08 (1.02–1.14)	0.009

Notes: Bold text indicates column headings. ORs for the CONUT score are presented per 1-point increase. Candidate covariates entered into the initial multivariable model included age, sex, BMI, current smoking, preoperative sleep difficulty, cerebrovascular disease, COPD, kidney disease, surgical delay, anesthesia method, injury energy, fracture type, operation type, surgical duration, ASA score, CCI, hemoglobin, total protein, ALT, HCRP, serum calcium, urea, creatinine, and anion gap. Backward stepwise selection was subsequently applied to obtain the final multivariable model. Serum albumin, lymphocyte count, and total cholesterol, which are components of the CONUT score, were not included as covariates.

Abbreviations: CONUT, Controlling Nutritional Status; POD, postoperative delirium; OR, odds ratio; CI, confidence interval.

For the exploratory categorical analyses, the optimal cutoff value of the CONUT score determined by the maximum Youden index was ≥ 5 (Figure 3). Patients were therefore categorized into the LCS and HCS groups for subsequent cutoff-based analyses. Before PSM, patients in the HCS group more frequently had advanced age, underweight status, sleep difficulty, hypertension, cerebrovascular disease, heart disease, COPD, tumors, peripheral vascular disease, general anesthesia, blood transfusion, delayed surgery, and higher CCI. Several laboratory variables, including RBC, HGB, PLT, TP, ALT, AST, HCRP, Na⁺, K⁺, Ca²⁺, FBG, and GAP, also differed significantly between the LCS and HCS groups (Table 4).

Figure 3.

Youden index, sensitivity, and specificity across different CONUT score cutoff values for discriminating postoperative delirium. The purple line with open circles represents the Youden index and corresponds to the left y-axis. The green and Orange lines represent sensitivity and specificity, respectively, and correspond to the right y-axis. The vertical red dashed line indicates the selected cutoff value of CONUT ≥ 5, corresponding to the maximum Youden index. The bold annotation identifies the selected cutoff value.

Abbreviation: CONUT, Controlling Nutritional Status.

Table 4.

Comparison of Baseline Characteristics Between the LCS and HCS Groups Before and After PSM

Variable	Before PSM					After PSM
	LCS (n = 1075)	HCS (n = 955)		P	SMD	LCS (n = 562)	HCS (n = 562)	P	SMD
Sex (female)	677 (62.98%)	597 (62.51%)		0.829	−0.010	354 (62.99%)	357 (63.52%)	0.853	0.011
Age (> 75years)	238 (22.14%)	407 (42.62%)		<0.001	0.414	189 (33.63%)	205 (36.48%)	0.317	0.059
BMI				<0.001				0.947
18.5–23.9	506 (47.07%)	485 (50.79%)			0.074	277 (49.29%)	274 (48.75%)		−0.011
<18.5	61 (5.67%)	100 (10.47%)			0.157	42 (7.47%)	46 (8.19%)		0.026
24.0–27.9	392 (36.47%)	290 (30.37%)			−0.133	186 (33.10%)	189 (33.63%)		0.011
≥28.0	116 (10.79%)	80 (8.38%)			−0.087	57 (10.14%)	53 (9.43%)		−0.024
Live place				0.639				0.676
Rural	583 (54.23%)	508 (53.19%)			−0.021	300 (53.38%)	293 (52.14%)		−0.025
Urban	492 (45.77%)	447 (46.81%)			0.021	262 (46.62%)	269 (47.86%)		0.025
Marital status				0.521				1.000
Spouse	951 (88.47%)	836 (87.54%)			−0.028	492 (87.54%)	492 (87.54%)		0.000
No spouse	124 (11.53%)	119 (12.46%)			0.028	70 (12.46%)	70 (12.46%)		0.000
Occupation				0.143				0.587
Manual	572 (53.21%)	467 (48.90%)			−0.086	294 (52.31%)	277 (49.29%)		−0.061
Nonmanual	251 (23.35%)	249 (26.07%)			0.062	131 (23.31%)	137 (24.38%)		0.025
Unavailable	252 (23.44%)	239 (25.03%)			0.037	137 (24.38%)	148 (26.33%)		0.044
Primary payer				0.313				1.000
Insurance	919 (85.49%)	801 (83.87%)			−0.044	479 (85.23%)	479 (85.23%)		0.000
Self-pay	156 (14.51%)	154 (16.13%)			0.044	83 (14.77%)	83 (14.77%)		0.000
Current smoking	152 (14.14%)	131 (13.72%)		0.784	−0.012	76 (13.52%)	84 (14.95%)	0.495	0.040
Alcohol consumption	143 (13.30%)	132 (13.82%)		0.733	0.015	71 (12.63%)	74 (13.17%)	0.790	0.016
Difficulty in sleeping	242 (22.51%)	285 (29.84%)		<0.001	0.160	164 (29.18%)	155 (27.58%)	0.552	−0.036
Hypertension	497 (46.23%)	489 (51.20%)		0.025	0.099	297 (52.85%)	288 (51.25%)	0.591	−0.032
Diabetes	228 (21.21%)	223 (23.35%)		0.247	0.051	141 (25.09%)	140 (24.91%)	0.945	−0.004
Heart disease	250 (23.26%)	305 (31.94%)		<0.001	0.186	168 (29.89%)	163 (29.00%)	0.744	−0.020
Cerebrovascular disease	317 (29.49%)	330 (34.55%)		0.014	0.107	198 (35.23%)	182 (32.38%)	0.313	−0.061
COPD	31 (2.88%)	51 (5.34%)		0.005	0.109	20 (3.56%)	26 (4.63%)	0.366	0.051
Tumors	16 (1.49%)	27 (2.83%)		0.037	0.081	13 (2.31%)	15 (2.67%)	0.702	0.022
Liver disease	29 (2.70%)	39 (4.08%)		0.083	0.070	19 (3.38%)	19 (3.38%)	1.000	0.000
Kidney disease	49 (4.56%)	56 (5.86%)		0.185	0.056	34 (6.05%)	32 (5.69%)	0.800	−0.015
Autoimmune diseases	31 (2.88%)	20 (2.09%)		0.257	−0.055	17 (3.02%)	14 (2.49%)	0.585	−0.034
Peripheral vascular disease	14 (1.30%)	37 (3.87%)		<0.001	0.133	13 (2.31%)	15 (2.67%)	0.702	0.022
Surgical history	311 (28.93%)	298 (31.20%)		0.264	0.049	178 (31.67%)	166 (29.54%)	0.437	−0.047
Allergic history	228 (21.21%)	177 (18.53%)		0.132	−0.069	118 (21.00%)	108 (19.22%)	0.457	−0.045
Surgical delay (days)				<0.001				0.888
< 2	176 (16.37%)	39 (4.08%)			−0.621	38 (6.76%)	34 (6.05%)		−0.030
2-5	439 (40.84%)	373 (39.06%)			−0.036	221 (39.32%)	223 (39.68%)		0.007
≥ 6	460 (42.79%)	543 (56.86%)			0.284	303 (53.91%)	305 (54.27%)		0.007
Anesthesia method (General)	469 (43.63%)	488 (51.10%)		<0.001	0.149	253 (45.02%)	278 (49.47%)	0.135	0.089
Injury energy (high energy)	110 (10.23%)	82 (8.59%)		0.206	−0.059	52 (9.25%)	48 (8.54%)	0.675	−0.025
Fracture type				0.262				0.551
Femoral-neck fracture	548 (50.98%)	463 (48.48%)			−0.050	284 (50.53%)	274 (48.75%)		−0.036
Intertrochanteric fracture	527 (49.02%)	492 (51.52%)			0.050	278 (49.47%)	288 (51.25%)		0.036
Operation type				0.649				0.462
Osteosynthesis	659 (61.30%)	576 (60.31%)			−0.020	339 (60.32%)	351 (62.46%)		0.044
Arthroplasty	416 (38.70%)	379 (39.69%)			0.020	223 (39.68%)	211 (37.54%)		−0.044
Surgical duration (> 180 min)	68 (6.33%)	82 (8.59%)		0.052	0.081	43 (7.65%)	39 (6.94%)	0.646	−0.028
Intraoperative blood loss (mL)	200.00 (100.00, 300.00)	260.00 (150.00, 400.00)		<0.001	0.150	200.00 (150.00, 400.00)	200.00 (150.00, 400.00)	0.891	−0.010
ASA score				0.072				0.549
I-II	618 (57.49%)	511 (53.51%)			−0.080	300 (53.38%)	310 (55.16%)		0.036
III-IV	457 (42.51%)	444 (46.49%)			0.080	262 (46.62%)	252 (44.84%)		−0.036
CCI				<0.001				0.903
0	521 (48.47%)	260 (27.23%)			−0.477	179 (31.85%)	172 (30.60%)		−0.027
1-2	420 (39.07%)	512 (53.61%)			0.292	284 (50.53%)	289 (51.42%)		0.018
≥3	134 (12.47%)	183 (19.16%)			0.170	99 (17.62%)	101 (17.97%)		0.009
Blood transfusion	315 (29.30%)	337 (35.29%)		0.004	0.125	200 (35.59%)	196 (34.88%)	0.803	−0.015
RBC < lower limitation	384 (35.72%)	676 (70.79%)		<0.001	0.771	319 (56.76%)	332 (59.07%)	0.432	0.047
HGB < lower limitation	372 (34.60%)	698 (73.09%)		<0.001	0.868	322 (57.30%)	341 (60.68%)	0.249	0.069
WBC > 9.5 × 109/L	381 (35.44%)	349 (36.54%)		0.605	0.023	210 (37.37%)	209 (37.19%)	0.951	−0.004
PLT				<0.001				0.985
125-350 × 109/L	947 (88.09%)	802 (83.98%)			−0.112	482 (85.77%)	480 (85.41%)		−0.010
< 125 × 109/L	68 (6.33%)	108 (11.31%)			0.157	48 (8.54%)	49 (8.72%)		0.006
> 350 × 109/L	60 (5.58%)	45 (4.71%)			−0.041	32 (5.69%)	33 (5.87%)		0.008
TP < 65 g/L	694 (64.56%)	782 (81.88%)		<0.001	0.450	448 (79.72%)	441 (78.47%)	0.608	−0.030
ALT > upper limitation	75 (6.98%)	90 (9.42%)		0.044	0.084	43 (7.65%)	45 (8.01%)	0.824	0.013
AST > upper limitation	105 (9.77%)	140 (14.66%)		<0.001	0.138	73 (12.99%)	77 (13.70%)	0.726	0.021
HCRP > 52.1 mg/L	393 (36.56%)	565 (59.16%)		<0.001	0.460	282 (50.18%)	307 (54.63%)	0.135	0.089
Na+ < 137 mmol/L	355 (33.02%)	485 (50.79%)		<0.001	0.355	252 (44.84%)	251 (44.66%)	0.952	−0.004
K+ < 3.5 mmol/L	112 (10.42%)	154 (16.13%)		<0.001	0.155	65 (11.57%)	69 (12.28%)	0.713	0.022
Ca2+< 2.11 mmol/L	416 (38.70%)	480 (50.26%)		<0.001	0.231	270 (48.04%)	255 (45.37%)	0.370	−0.054
FBG > 6.1 mmol/L	519 (48.28%)	558 (58.43%)		<0.001	0.206	323 (57.47%)	318 (56.58%)	0.763	−0.018
UREA > 8 mmol/L	174 (16.19%)	184 (19.27%)		0.069	0.078	116 (20.64%)	119 (21.17%)	0.069	0.013
CREA > 97 mmol/L	217 (20.19%)	206 (21.57%)		0.443	0.034	126 (22.42%)	130 (23.13%)	0.776	0.017
GAP				<0.001				0.477
< 8 mmol/L	848 (78.88%)	656 (68.69%)			−0.220	436 (77.58%)	418 (74.38%)		−0.073
8-16 mmol/L	203 (18.88%)	294 (30.79%)			0.258	122 (21.71%)	140 (24.91%)		0.074
> 16 mmol/L	24 (2.23%)	5 (0.52%)			−0.237	4 (0.71%)	4 (0.71%)		0.000

Note: Bold text indicates column headings.

Abbreviations: LCS, low CONUT score; HCS, high CONUT score; PSM, propensity score matching; BMI, body mass index; COPD, chronic obstructive pulmonary disease; ASA, American Society of Anesthesiologists; CCI, Charlson Comorbidity Index; RBC, red blood cell, reference range: Female, 3.5–5.0×10¹²/L; Males, 4.0–5.5×10¹²/L; HGB, hemoglobin, reference range: Females, 110–150g/L; Males, 120–160g/L; WBC, white blood cell; PLT, platelet; TP, total protein; ALT, alanine aminotransferase, reference range: Female, 7–40 U/L; Males, 9–50 U/L; AST, aspartate aminotransferase, reference range: Female, 13–35 U/L; Males, 15–40 U/L; HCRP, high-sensitivity C-reactive protein; FBG, fasting blood glucose; UREA, blood urea; CREA, serum creatinine, reference range: Female, 41–73 μmol/L; Males, 57–97 μmol/L; GAP, anion gap; CONUT, Controlling Nutritional Status.

After PSM, 562 matched pairs were retained, and post-matching covariate balance was satisfactory, with all absolute SMDs < 0.10 (Figure 4 and Table 4). In the matched cohort, the incidence of POD was higher in the HCS group than in the LCS group (22.2% [125/562] vs 14.9% [84/562], P = 0.002). Conditional logistic regression analysis showed that patients with a CONUT score ≥ 5 had higher odds of POD than those with a lower CONUT score (OR, 1.63; 95% CI, 1.20–2.21; P = 0.002).

Figure 4.

Standardized mean differences of baseline covariates before and after propensity score matching between the LCS and HCS groups. Red circles labeled “Unadjusted” represent the absolute standardized mean differences before PSM, whereas blue triangles labeled “Adjusted” represent the absolute standardized mean differences after PSM. The vertical dashed line at 0.10 indicates the threshold for acceptable covariate balance, with absolute SMD values <0.10 considered indicative of satisfactory balance.

Abbreviations: LCS, low CONUT score; HCS, high CONUT score; SMD, standardized mean difference; PSM, propensity score matching; HGB, hemoglobin; RBC, red blood cell; HCRP, high-sensitivity C-reactive protein; CCI, Charlson Comorbidity Index; TP, total protein; GAP, anion gap; FBG, fasting blood glucose; BMI, body mass index; PLT, platelet; AST, aspartate aminotransferase; ASA, American Society of Anesthesiologists; UREA, blood urea; COPD, chronic obstructive pulmonary disease; ALT, alanine aminotransferase; CREA, serum creatinine; WBC, white blood cell; CONUT, Controlling Nutritional Status.

Subgroup analyses showed no significant effect modification across strata of age, sex, BMI, anesthesia type, fracture type, operation type, ASA score, and CCI (all P for interaction > 0.05), indicating that the association between CONUT category and POD was generally consistent across the predefined subgroups (Figure 5). The exploratory cutoff-based findings also remained consistent across sensitivity analyses. Compared with the matched analysis (OR, 1.63; 95% CI, 1.20–2.21), only minor changes in effect estimates were observed after excluding patients with extreme values of CONUT components (OR, 1.62; 95% CI, 1.17–2.22), after asymmetric propensity score trimming (OR, 1.66; 95% CI, 1.22–2.27), after excluding patients with tumors, liver disease, or kidney disease (OR, 1.76; 95% CI, 1.26–2.46), and after excluding patients with autoimmune diseases (OR, 1.65; 95% CI, 1.21–2.25). Across these sensitivity analyses, a CONUT score ≥ 5 remained associated with higher odds of POD (Table 5).

Figure 5.

Forest plot of subgroup analyses showing the association between a high CONUT score and POD. Black squares represent the estimated odds ratios, and horizontal lines represent the corresponding 95% confidence intervals. Arrowheads indicate confidence intervals extending beyond the displayed plotting range. The vertical dashed line indicates an odds ratio of 1.0. Bold text indicates the overall cohort and subgroup variable headings. P values for interaction were used to assess potential effect modification across subgroup categories.

Abbreviations: LCS, low CONUT score; HCS, high CONUT score; OR, odds ratio; CI, confidence interval; BMI, body mass index; ASA, American Society of Anesthesiologists; CCI, Charlson Comorbidity Index; CONUT, Controlling Nutritional Status; POD, postoperative delirium.

Table 5.

Sensitivity Analysis Was Performed to Test the Association Between HCS and the Risk of POD After PSM

Group	LCS	HCS	OR (95% CI)	P-value
	POD/Total	POD/Total
All patients	14.95% (84/562)	22.24% (125/562)	1.63 (1.20–2.21)	0.002
Main exposure factors trimming*	15.01% (77/513)	22.20% (115/518)	1.62 (1.17–2.22)	0.003
PS asymmetric trimming#	15.27% (82/537)	23.06% (122/529)	1.66 (1.22–2.27)	0.001
Patients without tumors, liver or kidney diseases	13.40% (67/500)	21.44% (107/499)	1.76 (1.26–2.46)	<0.001
Patients without autoimmune diseases	15.05% (82/545)	22.63% (124/548)	1.65 (1.21–2.25)	0.001

Notes: Bold text indicates column headings. *Excluding patients with peripheral blood lymphocyte count, serum albumin concentration, and total cholesterol concentration below the 2.5th percentile or above the 97.5th percentile compared to the overall. #Asymmetric trimming was used to exclude patients whose PS was below the 2.5th percentile of the PS of the HCS group and above the 97.5th percentile of the PS of the LCS group.

Abbreviations: HCS, high CONUT score; POD, postoperative delirium; PSM, propensity score-matched; LCS, low CONUT score; HCS, high CONUT score; OR, odds ratio; CI, confidence interval; PS, propensity score.

Discussion

This retrospective cohort study demonstrated a positive dose-response association between preoperative CONUT score and POD in elderly patients with hip fractures. In the primary multivariable analysis, each 1-point increase in the CONUT score was associated with higher odds of POD after adjustment for relevant covariates (adjusted OR, 1.08; 95% CI, 1.02–1.14). In the exploratory categorical analysis, patients with a CONUT score ≥5 had higher odds of POD than those with lower scores after PSM (OR, 1.63; 95% CI, 1.20–2.21). Subgroup analyses showed no significant effect modification, and the exploratory cutoff-based findings remained generally consistent across sensitivity analyses. These findings suggest that the CONUT score may serve as an adjunctive marker for identifying elderly patients with hip fractures who have greater perioperative vulnerability to POD.

Many instruments have been developed for nutritional screening and assessment in clinical practice, including the Mini Nutritional Assessment (MNA), the Malnutrition Universal Screening Tool (MUST), and the Nutritional Risk Screening 2002 (NRS-2002).¹⁴ However, in older adults with hip fractures, the practical utility of these approaches may be constrained by their reliance on patient-reported information and relatively complex scoring procedures. This limitation is particularly relevant in geriatric populations, in whom recent weight change, dietary intake, and other historical details are often difficult to ascertain accurately. In this context, the CONUT score has several pragmatic advantages. Because it is derived entirely from routinely available biochemical and hematologic parameters, it provides a more standardized and reproducible assessment while reducing the influence of subjective judgment and recall inaccuracy. In addition, its component variables can be retrieved directly from routine preoperative laboratory tests, making the tool readily applicable even in frail patients with limited communication ability or poor cooperation. Another notable strength is that the CONUT score can be repeatedly recalculated during the perioperative period, allowing clinicians to monitor temporal changes in nutritional status and adjust supportive strategies accordingly.¹⁵ Taken together, these features make CONUT a practical and scalable tool for longitudinal nutritional surveillance and perioperative risk stratification in elderly patients with hip fractures.

The CONUT score, integrating serum albumin, peripheral lymphocyte count, and total cholesterol, may be biologically linked to several pathways relevant to POD in elderly hip fracture patients: First, hypoalbuminemia indicates reduced protein reserve and impaired antioxidant/anti-inflammatory capacity, and is independently associated with POD.^7,16 Decreased albumin may be accompanied by microcirculatory disorders and increased vascular permeability, enabling peripheral inflammatory factors (IL-6, TNF-α, IL-1β) to cross/act on the blood-brain barrier (BBB) more easily, triggering microglial activation and neuronal damage, thereby enhancing delirium susceptibility.^17,18 Second, lymphocyte depletion reflects immunosenescence and reduced capacity to resolve postoperative inflammation; as a result, persistent peripheral inflammation propagates central glial activation and network instability, key mechanisms in delirium pathophysiology.¹⁹ Third, low cholesterol compromises neuronal membrane/lipid-raft organization; consequently, receptor clustering and synaptic transmission (including cholinergic/dopaminergic signaling) become vulnerable during perioperative stress.^20,21 Taken together, a higher CONUT score may reflect a combined vulnerability involving impaired nutritional reserve, immune dysregulation, and altered neuronal homeostasis, providing a biologically plausible context for the graded association between worsening preoperative nutritional status and POD. However, the CONUT score should not necessarily be interpreted as an independent causal determinant of delirium. Its components may also be influenced by frailty, systemic inflammation, acute illness severity, and comorbidity burden, particularly in patients with acute hip fractures who may already exhibit substantial inflammatory and metabolic changes before surgery.

Notably, subgroup analyses demonstrated no significant interactions (all P for interaction > 0.05), suggesting that the association between CONUT category and POD was generally consistent across the predefined subgroups. Together with the sensitivity analyses, these exploratory findings support the consistency of the observed cutoff-based association. Given that the CONUT score can be calculated from routinely available preoperative laboratory tests, it may serve as a simple adjunctive marker for identifying patients with greater perioperative vulnerability to POD. However, the present observational study did not evaluate whether nutritional screening or nutritional intervention improves clinical outcomes. Therefore, high CONUT scores should primarily be interpreted as signals prompting further clinical assessment rather than as evidence that CONUT-guided nutritional intervention will reduce POD.

A nationwide retrospective cohort study based on China’s urban basic medical insurance databases reported that the annual number of hip fractures among adults aged 55 years and older rose from 16,587 in 2012 to 66,575 in 2016, while total hospitalization costs increased from US$60 million to US$380 million over the same period.²² More recent national evidence on inpatient nutrition (China Nutrition Fundamental Data 2020 Project; 24,139 older inpatients) showed that 18.9% of older inpatients had malnutrition on admission, and 32.26% of inpatients with malnutrition did not receive nutritional support during hospitalization.²³ These findings suggest that nutritional care remains insufficiently integrated into routine inpatient management in older adults and may be even more critical in geriatric hip-fracture patients, who often have pre-existing comorbidities and are exposed to acute trauma-related metabolic stress. Accordingly, incorporating nutritional assessment into preoperative evaluation may provide additional information for perioperative risk stratification in elderly patients with hip fractures. Future prospective multicenter studies involving more diverse patient populations and standardized, repeated delirium assessments are needed to validate the observed association and improve the generalizability of these findings. In addition, prospective interventional studies are warranted to determine whether targeted nutritional management can reduce the occurrence of POD in elderly patients with hip fractures.

Despite the relatively large sample size and extensive adjustment for potential confounders using propensity score matching and multivariable models, several limitations should be acknowledged. First, regarding exposure assessment, although the CONUT score is an objective and practical screening tool, it may not fully capture the multidimensional nature of malnutrition in older patients with hip fractures. Specifically, it does not incorporate important domains such as micronutrient or vitamin deficiencies, sarcopenia, frailty status, or functional reserve. Moreover, because serum albumin, peripheral blood lymphocyte count, and total cholesterol may be influenced by systemic inflammation, acute illness severity, and comorbidity burden, the CONUT score may partly reflect overall physiological vulnerability rather than nutritional status alone. Second, the CONUT cutoff of ≥ 5 was derived from the present dataset using the maximum Youden index and has not been externally validated in an independent cohort. Therefore, the cutoff-based analyses were considered exploratory, and results based on this threshold should be interpreted cautiously. Third, with respect to outcome ascertainment, POD was identified retrospectively from inpatient medical records with reference to the CAM, and we could not ensure that all patients underwent standardized, repeated delirium screening at uniform time points during hospitalization. In particular, hypoactive delirium is often clinically subtle and may have been under-recognized, potentially resulting in underestimation of POD incidence and bias in effect-size estimates. Fourth, although we adjusted for a broad range of measured confounders, residual confounding and unmeasured confounding cannot be completely excluded in an observational study. Therefore, our findings should be interpreted primarily as evidence of association rather than causation. Finally, this was a single-center retrospective study conducted in a university-affiliated tertiary referral hospital, where case complexity, perioperative management, and nursing resources may differ from those in other settings. As a result, the generalizability of our findings to primary hospitals or populations with different healthcare resources may be limited.

Conclusion

Higher preoperative CONUT scores were associated with greater odds of postoperative delirium in elderly patients with hip fractures. In the exploratory categorical analysis, patients with a CONUT score ≥ 5 had higher odds of POD than those with lower scores (OR = 1.63; 95% CI, 1.20–2.21). These findings suggest that the CONUT score may serve as an adjunctive marker for identifying patients with greater perioperative vulnerability to POD. However, given the observational nature of this study, the findings should not be interpreted as evidence of a causal relationship, and whether nutritional intervention can reduce the occurrence of POD requires further investigation in prospective multicenter interventional studies.

Funding Statement

The authors received no external funding to support this project.

Data Sharing Statement

All data in this study can be obtained from the authors based on reasonable demand.

Ethical Approval

This study was approved by the Ethics Committee of Central Hospital Affiliated to Shenyang Medical College (approval No. K-2026-031-02).

Informed Consent

The requirement for informed consent was waived due to the retrospective design and anonymized data analysis.

Author Contributions

All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work.

Disclosure

All authors declared that they have no conflicts of interest in this work.