Abstract
Background
Hospital-acquired malnutrition (HAM) is prevalent and harmful. Global Leadership Initiative on Malnutrition (GLIM) criteria lack lipid metabolism information, impeding admission risk identification. TCBI integrates lipids and energy reserve, but its predictive utility for HAM remains unclear.
Methods
This multicenter prospective study (June-September 2014, 34 tertiary hospitals in mainland China) consecutively enrolled 3,382 non-critically ill adults without malnutrition at admission and with hospital stays of 7–30 days. TCBI was calculated as triglyceride (mg/dL) × total cholesterol (mg/dL) × body weight (kg) / 1000 and natural log-transformed (TCBI-LN) due to skewness. The primary outcome was HAM diagnosed at discharge by adapted GLIM framework. Multivariable logistic regression, restricted cubic splines, and subgroup analyses with interaction testing were used to assess the independent association and dose‑response relationship between TCBI-LN and HAM.
Results
Among 3,382 patients without admission malnutrition, 21.6% developed HAM at discharge. After full adjustment, each one-unit increase in log-transformed TCBI was independently associated with a 14.7% lower HAM risk (OR = 0.853), and restricted cubic spline analysis revealed a significant nonlinear dose-response relationship with two inflection points (TCBI-LN 7.27 and 8.50), suggesting three exploratory RCS-derived strata (HAM incidence: 26.1%, 17.0%, and 18.3%, respectively). The high-TCBI high-risk group had a small sample size and wide confidence intervals; findings are hypothesis‑generating only and require validation in larger cohorts enriched for high TCBI. Subgroup analyses further showed that the protective effect was more pronounced in surgical patients and those without hypertension, with significant effect modifications by hospital department and hypertension history.
Conclusion
Admission TCBI is independently associated with the risk of HAM, exhibiting a significant nonlinear dose-response relationship. These findings suggest that TCBI may serve as a complementary metabolic-anthropometric marker for risk stratification for identifying patients at elevated risk for HAM at admission, warranting further prospective validation and interventional studies to confirm its clinical utility.
Supplementary Information
The online version contains supplementary material available at https://doi.org/10.1186/s41043-026-01394-8.
Keywords: Triglyceride-cholesterol-body weight index, Hospital-acquired malnutrition, GLIM criteria, Multicenter study
Introduction
Malnutrition represents a pervasive public health challenge worldwide, with prevalence estimates among hospitalized patients ranging from 20% to 50% [1]. A substantial body of evidence has established malnutrition as a critical independent risk factor for a constellation of adverse clinical outcomes, including heightened susceptibility to infection, impaired wound healing, prolonged length of hospital stay, and elevated rates of readmission and mortality [2–5]. These events, in turn, impose a considerable incremental burden on healthcare expenditures [6]. Of note, malnutrition observed in the inpatient setting encompasses not only pre-existing conditions present at the time of admission but also malnutrition that emerges de novo or worsens during the course of hospitalization. The latter scenario is formally designated as hospital-acquired malnutrition (HAM). HAM is associated with poor prognosis and increased healthcare utilization, and may develop during hospitalization through reduced intake, inflammation, catabolic stress, immobility, and treatment-related adverse effects [7]. In recent years, the dissemination of nutritional science concepts and advances in clinical nutritional support have led to increased clinical awareness of pre-admission malnutrition. Nonetheless, the identification and management of HAM remain frequently overlooked. Critically, irrespective of whether malnutrition is present upon admission or acquired during the hospital stay, both conditions markedly elevate the risk of hospital-acquired complications and adversely impact patient prognosis [8]. Therefore, for patients who are not diagnosed with malnutrition upon admission, the early prediction and dynamic assessment of HAM risk hold substantial clinical significance.
The current Global Leadership Initiative on Malnutrition (GLIM) criteria, which integrate both phenotypic and etiologic dimensions within their framework, are widely acknowledged for their authority and clinical utility. Nevertheless, their core components remain predominantly centered on anthropometric measurements, dietary intake, and inflammatory burden, with limited integration of critical metabolic information—most notably, lipid metabolism. Consequently, the GLIM framework may not fully capture the imbalance between an individual’s energy reserves and metabolic stress [9]. Prior investigations have demonstrated that nutritional status is intimately coupled with metabolic phenotype. Notably, aberrations in lipid and glucose metabolism often precede observable alterations in conventional indicators such as body weight (BW) or muscle mass, underscoring the potential utility of metabolic indices in the early identification of nutritional risk [10]. Therefore, even among patients who do not fulfill the diagnostic criteria for malnutrition upon admission, systematic attention to lipid metabolism during routine admission assessment is warranted. This necessitates the exploration of a simple and sensitive tool—one that synthesizes readily accessible lipid profiles and BW data—to aid in the early risk stratification of patients susceptible to HAM.
Recently, the triglyceride-cholesterol-body weight index (TCBI = TG (mg/dL) × TC (mg/dL) × BW (kg) / 1000) has emerged as a novel composite metric integrating lipid metabolism and energy reserve. Unlike single indicators, TCBI captures both nutritional and metabolic dimensions, and its components are simple and commonly obtainable in many hospitalized patients when lipid profiling is performed at admission or during early hospitalization[11, 12]. However, robust large‑scale prospective evidence on TCBI’s ability to predict hospital-acquired malnutrition (HAM) in a broad inpatient population is still lacking. Therefore, the present study aimed to: (1) evaluate whether admission TCBI is independently associated with GLIM-defined HAM in a large multicenter inpatient cohort; (2) characterize the dose-response relationship using restricted cubic splines; and (3) explore potential effect modifications across clinical subgroups.
Methods
Patients
The study was a multicenter investigation carried out from June to September 2014 across 34 tertiary hospitals in mainland China, during which eligible patients were consecutively enrolled. All enrolled patients underwent comprehensive medical record tracking and nutritional assessment at the time of admission. The inclusion criteria were as follows: (1) alert and oriented mental status; (2) no emergency surgery performed; and (3) length of hospital stay ranging from 7 to 30 days. The exclusion criteria were: (1) confirmed diagnosis of malnutrition at admission; (2) refusal to participate and provide informed consent; (3) incomplete clinical data during hospitalization; and (4) admission via the emergency department. A detailed flowchart of the patient screening and enrollment process is presented in Fig. 1. The analysis of TCBI was conducted post hoc, as TCBI emerged as a candidate marker after the primary study was completed. For enrolled patients’ hospital stay duration, the 7-day minimum was imposed because HAM diagnosis requires sufficient time for nutritional deterioration to manifest and be detectable by adapted GLIM framework; shorter stays would not allow meaningful assessment of incident malnutrition. The 30-day upper limit excluded patients with prolonged complex hospitalizations where nutritional trajectories may be driven by multiple sequential clinical events rather than admission metabolic reserve. These restrictions define the target population as non-critically ill patients with intermediate-length hospitalizations. This study was approved by the Ethics Committee of Beijing Hospital (Approval No. 2014BJYYEC-022-02) and was conducted in strict accordance with the Declaration of Helsinki. The study was registered with the Chinese Clinical Trial Registry (Registration No.: ChiCTR-EPC14005253).
Fig. 1.
Flowchart of patient selection for the present study
Data collection
This study followed a unified standardized research protocol and systematically collected the following types of data: (1) Demographic data, including sex, age, marital status, and educational level; (2) Clinical information, covering the primary reason for this admission, past medical history, payment method, weight changes before and during hospitalization, and daily dietary intake. Payment method was collected as a proxy indicator of socioeconomic status and healthcare access; (3) Anthropometric indicators, including height, BW, calf circumference (CC), and handgrip strength. All measurements were strictly performed according to standard operating procedures, and all researchers involved in data collection received uniform training. For CC measurement, the participant maintained a sitting position with the knee flexed at 90° and both feet flat on the ground. The circumference at the thickest part of the calf was measured using a non-elastic flexible tape. Handgrip strength was measured using an electronic dynamometer (EH101, Xiangshan County, Guangdong, China). During measurement, the participant was seated with the shoulder in neutral adduction and the elbow flexed at 90°. Three measurements were taken for each hand, and the maximum value was recorded with a precision of 0.1 kg. (4) Laboratory test indicators: All patients completed routine blood and blood biochemistry examinations within 24 h of admission. Test items included total protein (TP), albumin (ALB), prealbumin, TG, and TC. Additional laboratory variables included white blood cell count, lymphocyte count, hemoglobin, platelet count, fasting glucose, alanine aminotransferase, total bilirubin, blood urea nitrogen, and creatinine. The TyG index was calculated as ln [fasting triglycerides (mg/dL) × fasting glucose (mg/dL) / 2]. The platelet-to-lymphocyte ratio (PLR) was calculated based on blood cell analysis results. The prognostic nutritional index (PNI) was calculated using the formula: PNI = 10 × ALB (g/dL) + 5 × Lymphocyte Count (×10⁹/L). In addition, all patients completed the Nutritional Risk Screening 2002 (NRS 2002) assessment within 24 h of admission [13].
In this study, HAM at discharge was defined as patients who did not meet the diagnostic criteria for malnutrition at admission but met the diagnostic criteria at discharge. Malnutrition was diagnosed using an adapted operationalization of the GLIM framework based on the variables available in the dataset [14]. The diagnostic process was divided into two steps: Step 1, screening for patients at nutritional risk using the NRS 2002 score (score ≥ 3); Step 2, making the final diagnosis of malnutrition according to the phenotypic and etiological criteria recognized by the GLIM consensus. The phenotypic criteria specifically included the following three items: (1) Weight loss: involuntary weight loss exceeding 5% within 6 months, or weight loss exceeding 10% beyond 6 months; (2) Low BMI: according to the GLIM committee’s recommendations for Asian populations [14], BMI < 18.5 kg/m² for patients under 70 years old, and BMI < 20 kg/m² for patients aged 70 and above; (3) Reduced muscle mass: given the lack of body composition detection data in this dataset, this study used two surrogate criteria for judgment — Criterion 1: reduced CC (men < 34 cm, women < 33 cm); Criterion 2: reduced handgrip strength (age ≥ 65 years: men < 28 kg, women < 18 kg; age < 65 years: men < 34 kg, women < 20 kg). Reduced muscle mass was determined only when both criteria were simultaneously positive. This determination method is consistent with previous studies and the 2025 consensus of the Asian Working Group for Sarcopenia [15]. The etiological criteria included the following two items: (1) Reduced food intake or impaired digestion/absorption function: a reduction in energy intake exceeding 50% for more than one week, or any sustained energy intake deficit lasting more than two weeks, or the presence of chronic gastrointestinal diseases affecting food intake and nutrient absorption; (2) Disease burden and inflammatory state: the presence of acute or chronic inflammation-related diseases or trauma, comprehensively determined based on the patient’s acute and chronic disease history at admission. This study used reduced ALB level (< 35 g/L) as a surrogate indicator of systemic inflammation [16–17].
Statistical analysis
All statistical analyses were performed using R software (version 4.4.3). Continuous variables were reported as mean ± standard deviation for normally distributed data, or as median with interquartile range for skewed distributions. Categorical variables were presented as frequencies and percentages. Group comparisons for continuous variables were conducted using one-way analysis of variance or the Kruskal-Wallis H test, as appropriate. Categorical variables were compared using the Chi-square test or Fisher’s exact test. Given the skewed distribution of the original TCBI values, a natural logarithmic transformation was applied to achieve approximate normality (TCBI-LN). Pearson correlation coefficients or Spearman’s rank correlation coefficients were calculated to assess associations between baseline parameters and TCBI-LN.
Univariate analysis was performed between TCBI and other clinical features and HAM at discharge in patients without malnutrition. Covariate selection in this study used a two-step method: first, integrating univariate analysis results to identify potentially relevant confounding factors. Key confounding factors included demographic characteristics, anthropometric indicators, nutritional status indicators, disease characteristics, and comorbidities. Subsequently, we constructed directed acyclic graphs (DAGs) incorporating these potential confounding factors to determine the necessary adjustment variables. Thereafter, multivariable logistic regression analysis was used to assess the relationship between TCBI and HAM. Sequential multivariable models were constructed as follows. Model 1 was unadjusted. Model 2 was adjusted for sex and age. Model 3 was additionally adjusted for BMI, marital status, education level, and payment method. Model 4 was further adjusted for hospital department and primary diagnosis. Model 5, the fully adjusted model, additionally included diabetes mellitus, hypertension, coronary heart disease, cerebral infarction, and NRS 2002 status. Results were expressed as odds ratios (OR) and their corresponding 95% confidence intervals (CI) per unit increase, thereby constructing both crude and adjusted models. Collinearity was checked using Variance Inflation Factor (VIF). In addition, restricted cubic splines (RCS) were used to display the potential nonlinear relationship between TCBI and HAM. All patients were grouped according to the inflection point values of the RCS curve, and the occurrence of HAM was compared among groups. RCS analysis was performed using four knots placed at the 5th, 35th, 65th, and 95th percentiles of TCBI-LN, with the nadir of the fitted curve set as the reference value; bootstrap resampling with 1,000 repetitions was further conducted to estimate percentile-based 95% confidence intervals for the exploratory lower turning point, nadir, and upper turning point. Finally, stratified logistic regression analysis was performed according to sex, age, BMI (18.5–23.0 kg/m² vs. <18.5 or ≥ 23.0 kg/m²), hospital department type, history of hypertension, history of coronary heart disease, history of diabetes mellitus, and history of cerebral infarction to explore the relationship between TCBI values and HAM in different subgroups. The likelihood ratio test was used to test interactions between subgroups. RCS curves were also drawn and compared for patients with different hypertension histories and hospital departments. All tests were two-sided, and P < 0.05 was considered statistically significant.
Results
Descriptive comparison between Non-HAM and HAM groups at discharge
A total of 3,382 inpatients without malnutrition at admission were included, of whom 731 (21.6%) were classified as having HAM at discharge. In descriptive comparisons, patients with HAM had higher observed proportions of infection-related complications (5.3% vs. 1.6%, P < 0.001) and other complications (4.9% vs. 1.5%, P < 0.001), longer hospital stays (14.93 ± 5.78 days vs. 13.53 ± 5.65 days, P < 0.001), and higher total medical costs (median: 48,980 RMB vs. 20,640 RMB, P < 0.001) than those without HAM (Table 1).
Table 1.
Descriptive comparison of clinical characteristics and outcomes by HAM status at discharge
| Characteristics | Overall(n = 3,382) | Non-HAM (n = 2,651) |
HAM(n = 731) | P value | Statistic |
|---|---|---|---|---|---|
| Infection-related complications, n (%) | <0.001 | 33.393 | |||
| No | 3300 (97.6%) | 2608 (98.4%) | 692 (94.7%) | ||
| Yes | 82 (2.4%) | 43 (1.6%) | 39 (5.3%) | ||
| Other complications, n (%) | <0.001 | 29.390 | |||
| No | 3305 (97.7%) | 2610 (98.5%) | 695 (95.1%) | ||
| Yes | 77 (2.3%) | 41 (1.5%) | 36 (4.9%) | ||
| Total length of hospital stay(days), mean ± sd | 13.84 ± 5.71 | 13.53 ± 5.65 | 14.93 ± 5.78 | <0.001 | -5.890 |
| ICU admission, n (%) | 0.110 | 2.557 | |||
| No | 3042 (89.9%) | 2396 (90.4%) | 646 (88.4%) | ||
| Yes | 340 (10.1%) | 255 (9.6%) | 85 (11.6%) | ||
| Total healthcare costs(CNY), median (IQR) | 30,120 (14050, 58010) | 20,640 (13040, 51630) | 48,980 (21050, 71580) | <0.001 | -11.553 |
Statistics: t (independent t-test); W (Wilcoxon rank-sum test); χ² (chi-square test)
Clinical characteristics based on TCBI-LN quartiles
Among the 3,382 patients without malnutrition at admission, because the original TCBI variable exhibited a skewed distribution, we performed natural logarithmic transformation on it to obtain the variable TCBI-LN, which approximated a normal distribution (Supplementary Fig. 1). According to the quartile cutoff values of TCBI-LN, the patients were divided into four groups (Q1–Q4) (Table 2). As the TCBI-LN level increased, patient age gradually decreased, and the proportion of males, BMI, BW, CC, ALB, TP, and other nutrition-related indicators showed a significant upward trend (P < 0.001). Simultaneously, the proportion of patients with NRS 2002 ≥ 3 in the high TCBI-LN group decreased significantly, TG, TC, and TyG index increased, while inflammation-related indicators (such as PLR) decreased. Marked baseline heterogeneity was observed across TCBI-LN quartiles, including differences in hospital department, primary diagnosis, anthropometric measurements, lipid profiles, and nutritional risk status, indicating that quartile-based comparisons should be interpreted descriptively and with consideration of potential confounding. The incidence of HAM also varied across disease types within different TCBI-LN intervals (Fig. 2).
Table 2.
Baseline clinical characteristics of the study population stratified by TCBI-LN quartiles
| Characteristics | Overall(n = 3382) | Quartile 1 (n = 846) |
Quartile 2 (n = 846) |
Quartile 3 (n = 845) |
Quartile 4 (n = 845) |
P value | Statistic |
|---|---|---|---|---|---|---|---|
| Sex, n (%) | < 0.001 | 25.420 | |||||
| Male | 1950 (57.7%) | 450 (53.2%) | 476 (56.3%) | 477 (56.4%) | 547 (64.7%) | ||
| Female | 1432 (42.3%) | 396 (46.8%) | 370 (43.7%) | 368 (43.6%) | 298 (35.3%) | ||
| Age (year), mean ± SD | 58.78 ± 14.66 | 60.94 ± 15.79 | 60.94 ± 14.13 | 58.99 ± 13.71 | 54.26 ± 13.92 | < 0.001 | 40.424 |
| Marital status, n (%) | 0.279 | 3.838 | |||||
| Married | 3149 (93.1%) | 787 (93.0%) | 799 (94.4%) | 785 (92.9%) | 778 (92.1%) | ||
| Single or divorced | 233 (6.9%) | 59 (7.0%) | 47 (5.6%) | 60 (7.1%) | 67 (7.9%) | ||
| Weight (kg), mean ± SD | 66.19 ± 10.57 | 62.17 ± 9.79 | 64.73 ± 9.47 | 67.39 ± 9.89 | 70.48 ± 11.21 | < 0.001 | 105.157 |
| Education Level, n (%) | 0.277 | 7.500 | |||||
| Junior high school or below | 614 (18.2%) | 148 (17.5%) | 150 (17.7%) | 160 (18.9%) | 156 (18.5%) | ||
| Senior high school | 570 (16.9%) | 152 (18.0%) | 158 (18.7%) | 140 (16.6%) | 120 (14.2%) | ||
| Unknown | 2198 (65.0%) | 546 (64.5%) | 538 (63.6%) | 545 (64.5%) | 569 (67.3%) | ||
| Payment Method, n (%) | < 0.001 | 74.878 | |||||
| Out-of-pocket | 562 (16.6%) | 132 (15.6%) | 118 (13.9%) | 135 (16.0%) | 177 (20.9%) | ||
| Basic medical insurance | 2442 (72.2%) | 610 (72.1%) | 636 (75.2%) | 621 (73.5%) | 575 (68.0%) | ||
| Public funding (government) | 52 (1.5%) | 6 (0.7%) | 7 (0.8%) | 5 (0.6%) | 34 (4.0%) | ||
| Others | 149 (4.4%) | 49 (5.8%) | 34 (4.0%) | 41 (4.9%) | 25 (3.0%) | ||
| Unknown | 177 (5.2%) | 49 (5.8%) | 51 (6.0%) | 43 (5.1%) | 34 (4.0%) | ||
| Department, n (%) | < 0.001 | 143.290 | |||||
| Medical department | 2027 (59.9%) | 524 (61.9%) | 560 (66.2%) | 580 (68.6%) | 363 (43.0%) | ||
| Surgical department | 1355 (40.1%) | 322 (38.1%) | 286 (33.8%) | 265 (31.4%) | 482 (57.0%) | ||
| Primary diagnosis, n (%) | < 0.001 | 843.299 | |||||
| Endocrine and Rheumatic Immune Diseases | 58 (1.7%) | 21 (2.5%) | 14 (1.7%) | 13 (1.5%) | 10 (1.2%) | ||
| Neurological and Psychiatric Diseases | 406 (12.0%) | 28 (3.3%) | 21 (2.5%) | 23 (2.7%) | 334 (39.5%) | ||
| Bone and Joint Diseases | 149 (4.4%) | 32 (3.8%) | 34 (4.0%) | 47 (5.6%) | 36 (4.3%) | ||
| Benign Digestive System Diseases | 805 (23.8%) | 221 (26.1%) | 208 (24.6%) | 206 (24.4%) | 170 (20.1%) | ||
| Others | 86 (2.5%) | 31 (3.7%) | 21 (2.5%) | 23 (2.7%) | 11 (1.3%) | ||
| Benign Respiratory System Diseases | 179 (5.3%) | 61 (7.2%) | 50 (5.9%) | 44 (5.2%) | 24 (2.8%) | ||
| Cardiovascular Diseases | 107 (3.2%) | 31 (3.7%) | 28 (3.3%) | 25 (3.0%) | 23 (2.7%) | ||
| Malignant Tumors | 1592 (47.1%) | 421 (49.8%) | 470 (55.6%) | 464 (54.9%) | 237 (28.0%) | ||
| Diabetes mellitus, n (%) | 0.136 | 5.541 | |||||
| No | 3147 (93.1%) | 777 (91.8%) | 784 (92.7%) | 786 (93.0%) | 800 (94.7%) | ||
| Yes | 235 (6.9%) | 69 (8.2%) | 62 (7.3%) | 59 (7.0%) | 45 (5.3%) | ||
| Hypertension, n (%) | < 0.001 | 19.082 | |||||
| No | 2758 (81.5%) | 687 (81.2%) | 663 (78.4%) | 679 (80.4%) | 729 (86.3%) | ||
| Yes | 624 (18.5%) | 159 (18.8%) | 183 (21.6%) | 166 (19.6%) | 116 (13.7%) | ||
| Coronary heart disease, n (%) | < 0.001 | 17.905 | |||||
| No | 3217 (95.1%) | 795 (94.0%) | 791 (93.5%) | 807 (95.5%) | 824 (97.5%) | ||
| Yes | 165 (4.9%) | 51 (6.0%) | 55 (6.5%) | 38 (4.5%) | 21 (2.5%) | ||
| Cerebral infarction, n (%) | 0.111 | 6.010 | |||||
| No | 3293 (97.4%) | 816 (96.5%) | 821 (97.0%) | 826 (97.8%) | 830 (98.2%) | ||
| Yes | 89 (2.6%) | 30 (3.5%) | 25 (3.0%) | 19 (2.2%) | 15 (1.8%) | ||
| NRS2002 ≥ 3, n (%) | < 0.001 | 72.324 | |||||
| NRS2002 < 3 | 2498 (73.9%) | 572 (67.6%) | 589 (69.6%) | 625 (74.0%) | 712 (84.3%) | ||
| NRS2002 ≥ 3 | 884 (26.1%) | 274 (32.4%) | 257 (30.4%) | 220 (26.0%) | 133 (15.7%) | ||
| Calf circumference (cm), mean ± SD | 33.69 ± 3.90 | 33.20 ± 3.66 | 33.74 ± 3.65 | 34.27 ± 3.53 | 33.55 ± 4.59 | < 0.001 | 11.252 |
| Handgrip strength (kg), mean ± SD | 28.24 ± 10.41 | 26.77 ± 9.82 | 27.86 ± 10.11 | 27.99 ± 10.11 | 30.34 ± 11.22 | < 0.001 | 17.817 |
| BMI (kg/m²), mean ± SD | 24.00 ± 3.00 | 22.99 ± 2.84 | 23.72 ± 2.87 | 24.44 ± 2.93 | 24.87 ± 3.01 | < 0.001 | 67.641 |
| WBC (×10⁹/L), median [IQR] | 6.20 [4.90, 7.96] | 5.66 [4.52, 7.10] | 5.90 [4.70, 7.20] | 6.15 [5.00, 7.70] | 7.40 [5.63, 10.30] | < 0.001 | 234.431 |
| Lymphocytes (×10⁹/L), median [IQR] | 1.61 [1.20, 2.15] | 1.50 [1.10, 1.92] | 1.59 [1.20, 2.04] | 1.66 [1.22, 2.10] | 2.00 [1.35, 3.00] | < 0.001 | 124.179 |
| Hemoglobin (g/L), mean ± SD | 127.01 ± 20.21 | 121.73 ± 21.29 | 125.76 ± 20.37 | 128.20 ± 20.03 | 132.36 ± 17.52 | < 0.001 | 42.561 |
| Platelets (×10⁹/L), mean ± SD | 213.29 ± 73.32 | 206.03 ± 79.36 | 211.45 ± 70.13 | 224.92 ± 73.66 | 210.78 ± 68.41 | < 0.001 | 10.449 |
| PLR, median [IQR] | 117.32 [78.61, 166.27] | 125.32 [88.73, 173.15] | 127.54 [96.25, 171.13] | 126.55 [95.76, 181.88] | 74.29 [16.67, 128.71] | < 0.001 | 352.742 |
| Glucose (mmol/L), median [IQR] | 5.30 [4.82, 6.30] | 5.10 [4.64, 5.80] | 5.18 [4.80, 5.83] | 5.32 [4.84, 6.04] | 6.11 [5.14, 7.80] | < 0.001 | 274.341 |
| TyG index, mean ± SD | 8.99 ± 0.87 | 8.27 ± 0.54 | 8.65 ± 0.36 | 8.99 ± 0.39 | 10.06 ± 0.81 | < 0.001 | 1634.597 |
| ALT (U/L), median [IQR] | 19.00 [13.90, 30.00] | 16.90 [12.00, 24.93] | 18.61 [13.03, 27.15] | 19.00 [14.00, 29.75] | 24.00 [16.00, 36.00] | < 0.001 | 142.715 |
| Total bilirubin (µmol/L), median [IQR] | 11.70 [8.60, 15.60] | 11.92 [8.91, 15.90] | 11.80 [9.10, 15.55] | 11.77 [8.50, 16.26] | 11.40 [7.90, 14.70] | 0.009 | 11.685 |
| BUN (mmol/L), median [IQR] | 5.30 [4.32, 6.33] | 5.20 [4.10, 6.38] | 5.34 [4.40, 6.25] | 5.18 [4.21, 6.21] | 5.40 [4.50, 6.45] | 0.005 | 12.899 |
| Creatinine (µmol/L), median [IQR] | 69.00 [57.00, 82.00] | 67.91 [58.00, 81.00] | 71.95 [61.00, 82.92] | 72.00 [61.00, 83.30] | 65.00 [47.00, 81.00] | < 0.001 | 90.960 |
| Triglycerides (mmol/L), median [IQR] | 1.53 [1.12, 2.36] | 0.87 [0.70, 1.10] | 1.29 [1.16, 1.46] | 1.73 [1.53, 2.01] | 4.19 [2.53, 8.40] | < 0.001 | 2370.864 |
| Total cholesterol (mmol/L), mean ± SD | 4.31 ± 1.18 | 3.50 ± 1.14 | 4.20 ± 0.85 | 4.49 ± 0.94 | 5.04 ± 1.21 | < 0.001 | 320.071 |
| TCBI-LN, mean ± SD | 7.40 ± 0.87 | 6.43 ± 0.38 | 7.08 ± 0.12 | 7.49 ± 0.14 | 8.58 ± 0.66 | < 0.001 | 4490.072 |
| Albumin (g/L), mean ± SD | 40.08 ± 5.53 | 37.81 ± 5.37 | 38.91 ± 4.88 | 40.24 ± 4.87 | 43.38 ± 5.36 | < 0.001 | 186.943 |
| PNI, median [IQR] | 48.45 [43.96, 53.15] | 46.00 [41.79, 50.21] | 47.38 [42.98, 51.49] | 49.10 [45.25, 52.80] | 53.55 [48.40, 59.10] | < 0.001 | 369.249 |
| Total protein (g/L), mean ± SD | 66.79 ± 7.00 | 64.76 ± 7.21 | 65.39 ± 6.87 | 66.77 ± 6.74 | 70.26 ± 5.78 | < 0.001 | 114.967 |
| Nutritional support, n (%) | 0.137 | 13.619 | |||||
| None | 2003 (59.2%) | 519 (61.3%) | 492 (58.2%) | 481 (56.9%) | 511 (60.5%) | ||
| Enteral nutrition | 710 (21.0%) | 183 (21.6%) | 164 (19.4%) | 185 (21.9%) | 178 (21.1%) | ||
| Parenteral nutrition | 315 (9.3%) | 72 (8.5%) | 95 (11.2%) | 78 (9.2%) | 70 (8.3%) | ||
| Both | 354 (10.5%) | 72 (8.5%) | 95 (11.2%) | 101 (12.0%) | 86 (10.2%) | ||
| Hospital-acquired malnutrition, n (%) | < 0.001 | 46.780 | |||||
| No | 2651 (78.4%) | 604 (71.4%) | 646 (76.4%) | 700 (82.8%) | 701 (83.0%) | ||
| Yes | 731 (21.6%) | 242 (28.6%) | 200 (23.6%) | 145 (17.2%) | 144 (17.0%) |
Statistics: F (one-way ANOVA); H (Kruskal-Wallis test); χ² (chi-square test)
Fig. 2.
HAM Rates Across TCBI-LN Quartiles by Disease Type. Note: group A: Endocrine and rheumatic immune diseases; group B: Neurological and psychiatric diseases; group C: Bone and joint diseases; group D: Benign digestive system diseases; group E: Others; group F: Benign respiratory system diseases; group G: Cardiovascular diseases; group H: Malignant tumors.
Correlation analysis between TCBI-LN and clinical characteristics
The results of the correlation analysis showed that TCBI-LN was significantly correlated with multiple indicators related to lipid metabolism, nutrition, and inflammation. Specifically, TCBI-LN exhibited strong positive correlations with TG (r = 0.868, P < 0.001), TC (r = 0.540, P < 0.001), and TyG (r = 0.820, P < 0.001), suggesting that TCBI-LN can effectively reflect the body’s lipid metabolism status (Fig. 3). Regarding nutrition-related indicators, TCBI-LN showed significant positive correlations with PNI (r = 0.450, P < 0.001), ALB (r = 0.400, P < 0.001), TP (r = 0.311, P < 0.001), and LYM (r = 0.339, P < 0.001), and was also positively correlated with BMI (r = 0.248, P < 0.001), HGS (r = 0.115, P < 0.001), and CC (r = 0.048, P = 0.005). In addition, TCBI-LN was negatively correlated with PLR (r = -0.261, P < 0.001), age (r = -0.179, P < 0.001), and various comorbidities. Meanwhile, TCBI-LN also showed significant negative correlations with HAM (r = -0.118, P < 0.001) and TBIL (r = -0.060, P < 0.001). All of the above correlations were statistically significant, indicating that TCBI-LN is an effective indicator that comprehensively reflects a patient’s lipid metabolism, nutritional status, inflammation level, and clinical prognosis.
Fig. 3.
Correlation analysis between TCBI-LN and clinical parameters. Note: CC: Calf Circumference; HGS: Handgrip Strength; LYM: Lymphocyte Count; Hb: Hemoglobin; PLT: Platelet Count; ALT: Alanine Aminotransferase; GLU: Glucose; BUN: Blood Urea Nitrogen; TBIL: Total Bilirubin; TP: Total Protein; DM: Diabetes; HTN: Hypertension; CHD: Coronary Heart Disease; CI: Cerebral Infarction. *p<0.05 ; **p<0.01 ; ***p<0.001
Association of TCBI-LN for HAM at discharge
As shown in Table 3, the results of univariate analysis indicated that patients who developed HAM at discharge had significantly lower TCBI-LN levels compared with those who did not develop HAM (P < 0.001). In addition, patients in the HAM group also exhibited significant differences in age, sex ratio, hospital department, and inflammatory and nutritional indicators (all P < 0.05).
Table 3.
Univariate analysis of clinical indicators and HAM
| Characteristics | Overall (n = 3382) |
Without HAM (n = 2651) |
With HAM (n = 731) |
P value | Statistic |
|---|---|---|---|---|---|
| Sex, n (%) | 0.013 | 6.229 | |||
| Male | 1950 (57.7%) | 1499 (56.5%) | 451 (61.7%) | ||
| Female | 1432 (42.3%) | 1152 (43.5%) | 280 (38.3%) | ||
| Age (year), mean ± SD | 58.78 ± 14.66 | 57.75 ± 14.75 | 62.53 ± 13.71 | < 0.001 | -7.878 |
| Marital status, n (%) | 0.472 | 0.518 | |||
| Married | 3149 (93.1%) | 2464 (92.9%) | 685 (93.7%) | ||
| Single or divorced | 233 (6.9%) | 187 (7.1%) | 46 (6.3%) | ||
| Weight (kg), mean ± SD | 66.19 ± 10.57 | 66.81 ± 10.44 | 63.92 ± 10.71 | < 0.001 | 6.594 |
| Education Level, n (%) | 0.341 | 2.153 | |||
| Junior high school or below | 614 (18.2%) | 468 (17.7%) | 146 (20.0%) | ||
| Senior high school | 570 (16.9%) | 452 (17.1%) | 118 (16.1%) | ||
| Unknown | 2198 (65.0%) | 1731 (65.3%) | 467 (63.9%) | ||
| Payment Method, n (%) | 0.113 | 7.464 | |||
| Public funding (government) | 52 (1.5%) | 44 (1.7%) | 8 (1.1%) | ||
| Basic medical insurance | 2442 (72.2%) | 1886 (71.1%) | 556 (76.1%) | ||
| Out-of-pocket | 562 (16.6%) | 454 (17.1%) | 108 (14.8%) | ||
| Others | 149 (4.4%) | 123 (4.6%) | 26 (3.6%) | ||
| Unknown | 177 (5.2%) | 144 (5.4%) | 33 (4.5%) | ||
| Hospital Department, n (%) | < 0.001 | 26.055 | |||
| Medical department | 2027 (59.9%) | 1529 (57.7%) | 498 (68.1%) | ||
| Surgical department | 1355 (40.1%) | 1122 (42.3%) | 233 (31.9%) | ||
| Primary Diagnosis, n (%) | < 0.001 | 116.066 | |||
| Endocrine and Rheumatic Immune Diseases | 58 (1.7%) | 56 (2.1%) | 2 (0.3%) | ||
| Neurological and Psychiatric Diseases | 406 (12.0%) | 339 (12.8%) | 67 (9.2%) | ||
| Bone and Joint Diseases | 149 (4.4%) | 131 (4.9%) | 18 (2.5%) | ||
| Benign Digestive System Diseases | 805 (23.8%) | 683 (25.8%) | 122 (16.7%) | ||
| Benign Genitourinary Diseases | 86 (2.5%) | 79 (3.0%) | 7 (1.0%) | ||
| Benign Respiratory System Diseases | 179 (5.3%) | 149 (5.6%) | 30 (4.1%) | ||
| Cardiovascular Diseases | 107 (3.2%) | 90 (3.4%) | 17 (2.3%) | ||
| Malignant Tumors | 1592 (47.1%) | 1124 (42.4%) | 468 (64.0%) | ||
| Diabetes mellitus, n (%) | 0.003 | 8.947 | |||
| No | 3147 (93.1%) | 2485 (93.7%) | 662 (90.6%) | ||
| Yes | 235 (6.9%) | 166 (6.3%) | 69 (9.4%) | ||
| Hypertension, n (%) | < 0.001 | 13.508 | |||
| No | 2758 (81.5%) | 2196 (82.8%) | 562 (76.9%) | ||
| Yes | 624 (18.5%) | 455 (17.2%) | 169 (23.1%) | ||
| Coronary heart disease, n (%) | 0.400 | 0.707 | |||
| No | 3217 (95.1%) | 2526 (95.3%) | 691 (94.5%) | ||
| Yes | 165 (4.9%) | 125 (4.7%) | 40 (5.5%) | ||
| Cerebral infarction, n (%) | 0.214 | 1.545 | |||
| No | 3293 (97.4%) | 2586 (97.5%) | 707 (96.7%) | ||
| Yes | 89 (2.6%) | 65 (2.5%) | 24 (3.3%) | ||
| NRS2002 ≥ 3, n (%) | < 0.001 | 53.496 | |||
| NRS2002 < 3 | 2498 (73.9%) | 2035 (76.8%) | 463 (63.3%) | ||
| NRS2002 ≥ 3 | 884 (26.1%) | 616 (23.2%) | 268 (36.7%) | ||
| Calf circumference (cm), mean ± SD | 33.69 ± 3.90 | 33.99 ± 3.89 | 32.60 ± 3.75 | < 0.001 | 8.604 |
| Handgrip strength (kg), mean ± SD | 28.24 ± 10.41 | 28.65 ± 10.42 | 26.74 ± 10.22 | < 0.001 | 4.394 |
| BMI (kg/m²), mean ± SD | 24.00 ± 3.00 | 24.19 ± 2.95 | 23.33 ± 3.10 | < 0.001 | 6.912 |
| WBC (×10⁹/L), median [IQR] | 6.20 [4.90, 7.96] | 6.23 [4.96, 8.10] | 5.96 [4.68, 7.62] | < 0.001 | 1063361.000 |
| Lymphocytes (×10⁹/L), median [IQR] | 1.73 [1.26, 2.47] | 1.75 [1.27, 2.50] | 1.69 [1.23, 2.34] | 0.031 | 1019423.000 |
| Hemoglobin (g/L), mean ± SD | 127.01 ± 20.21 | 128.18 ± 19.98 | 122.77 ± 20.50 | < 0.001 | 6.440 |
| Platelets (×10⁹/L), mean ± SD | 213.29 ± 73.32 | 211.64 ± 72.45 | 219.29 ± 76.13 | 0.012 | -2.501 |
| PLR, median [IQR] | 117.32 [78.61, 166.27] | 114.71 [76.30, 163.95] | 125.40 [87.73, 175.15] | < 0.001 | 881821.000 |
| Glucose (mmol/L), median [IQR] | 5.30 [4.82, 6.30] | 5.31 [4.80, 6.39] | 5.30 [4.87, 6.12] | 0.540 | 983260.500 |
| TyG index, mean ± SD | 8.99 ± 0.87 | 9.04 ± 0.87 | 8.84 ± 0.85 | < 0.001 | 5.428 |
| ALT (U/L), median [IQR] | 19.00 [13.90, 30.00] | 19.00 [13.38, 31.00] | 19.00 [14.00, 28.00] | 0.462 | 986123.500 |
| Total bilirubin (µmol/L), median [IQR] | 11.70 [8.60, 15.60] | 11.60 [8.40, 15.60] | 12.20 [9.20, 15.58] | 0.036 | 920020.500 |
| BUN (mmol/L), median [IQR] | 5.30 [4.32, 6.33] | 5.20 [4.26, 6.29] | 5.52 [4.66, 6.43] | < 0.001 | 864989.500 |
| Creatinine (µmol/L), median [IQR] | 69.00 [57.00, 82.00] | 68.00 [56.00, 81.40] | 73.00 [62.00, 83.39] | < 0.001 | 847647.500 |
| Triglycerides (mmol/L), median [IQR] | 1.53 [1.12, 2.36] | 1.58 [1.16, 2.48] | 1.32 [1.01, 2.02] | < 0.001 | 1125232.000 |
| Total cholesterol (mmol/L), mean ± SD | 4.31 ± 1.18 | 4.30 ± 1.17 | 4.33 ± 1.22 | 0.507 | -0.663 |
| TCBI-LN, mean ± SD | 7.40 ± 0.87 | 7.44 ± 0.87 | 7.23 ± 0.86 | < 0.001 | 5.990 |
| Albumin (g/L), mean ± SD | 40.08 ± 5.53 | 40.37 ± 5.56 | 39.06 ± 5.33 | < 0.001 | 5.697 |
| PNI, median [IQR] | 49.40 [44.50, 55.45] | 49.90 [44.90, 56.00] | 47.64 [43.12, 53.72] | < 0.001 | 1097146.000 |
| Total protein (g/L), mean ± SD | 66.79 ± 7.00 | 67.16 ± 6.92 | 65.45 ± 7.14 | < 0.001 | 5.884 |
| Nutritional support, n (%) | < 0.001 | 170.369 | |||
| None | 2003 (59.2%) | 1697 (64.0%) | 306 (41.9%) | ||
| Enteral nutrition | 710 (21.0%) | 539 (20.3%) | 171 (23.4%) | ||
| Parenteral nutrition | 315 (9.3%) | 175 (6.6%) | 140 (19.2%) | ||
| Both | 354 (10.5%) | 240 (9.1%) | 114 (15.6%) |
Statistics: t (independent t-test); W (Wilcoxon rank-sum); χ² (chi-square test)
Based on the results of univariate analysis and DAGs (Supplementary Fig. 2), we identified the necessary adjustment variables for multivariable logistic regression analysis to further confirm the independent association between TCBI-LN and HAM at discharge. Multivariable logistic regression analysis further confirmed the independent association between TCBI-LN and HAM (Table 4). After sequentially adjusting for potential confounding factors, this association remained significant. When analyzed as a continuous variable, in the unadjusted model (Model 1), each 1-unit increase in TCBI-LN was associated with an approximately 26.2% reduction in the risk of HAM occurrence (OR = 0.738, 95% CI: 0.668–0.817). After sequentially incorporating sex and age (Model 2); BMI, marital status, educational level, and medical payment method (Model 3); hospital department and primary diagnosis (Model 4); and history of diabetes mellitus, hypertension, coronary heart disease, cerebral infarction, and NRS 2002 score (Model 5, fully adjusted model), this protective association remained consistently significant (OR = 0.853, 95% CI: 0.746–0.975). When TCBI-LN was analyzed by quartiles, participants in Q2 and Q3 had significantly lower odds of HAM compared with those in Q1 in the fully adjusted model. However, the association in Q4 was attenuated and did not reach statistical significance. Therefore, the quartile-based findings should not be interpreted as evidence of a simple monotonic protective trend. Instead, they suggest a non-monotonic pattern, with the lowest HAM risk observed in the intermediate TCBI-LN range. This interpretation is consistent with the restricted cubic spline analysis, which demonstrated a significant nonlinear association between TCBI-LN and HAM. The variance inflation factor (VIF) values for all models were < 3.0, ruling out significant multicollinearity.
Table 4.
Predictive Value of TCBI-LN for HAM
| Variables | Model 1 | Model 2 | Model 3 | Model 4 | Model 5 |
|---|---|---|---|---|---|
| OR (95% CI) | OR (95% CI) | OR (95% CI) | OR (95% CI) | OR (95% CI) | |
| TCBI-LN |
0.738 (0.668–0.817)*** |
0.773 (0.696–0.857)*** |
0.844 (0.758–0.940)** |
0.848 (0.744–0.967)* |
0.853 (0.746–0.975)* |
|
TCBI-LN quartiles |
|||||
|
Quartile 1 (n = 846) |
1 | 1 | 1 | 1 | 1 |
|
Quartile 2 (n = 846) |
0.774 (0.623–0.962)* |
0.770 (0.618–0.959)* |
0.810 (0.648–1.012) |
0.760 (0.605–0.955)* |
0.758 (0.602–0.954)* |
|
Quartile 3 (n = 845) |
0.517 (0.410–0.653)*** |
0.534 (0.422–0.675)*** |
0.612 (0.481–0.779)*** |
0.573 (0.448–0.733)*** |
0.573 (0.447–0.733)*** |
|
Quartile 4 (n = 845) |
0.512 (0.405–0.646)*** |
0.574 (0.452–0.729)*** |
0.703 (0.548–0.903)** |
0.786 (0.594–1.040) |
0.806 (0.606–1.070) |
| P for trend | < 0.001 | < 0.001 | < 0.001 | 0.003 | 0.004 |
OR = Odds Ratio; CI = Confidence Interval. Model 1: Unadjusted. Model 2: + sex, age. Model 3: + BMI, marital status, education, payment. Model 4: + department, primary diagnosis. Model 5: + diabetes, hypertension, CHD, cerebral infarction, NRS 2002
*p<0.05 ; **p<0.01 ; ***p<0.001. P for trend: quartile number entered as continuous variable
RCS analysis, after adjusting for all covariates in Model 5, revealed a statistically significant association between TCBI-LN and HAM risk (P for overall < 0.001), and the test for nonlinearity was also statistically significant (P for nonlinearity < 0.001), indicating that the relationship between the two is not a simple linear one (Fig. 4). The RCS curve exhibited a nonlinear pattern characterized by an initial rapid decline followed by a plateau or slight fluctuation, suggesting that the protective effect of TCBI-LN against HAM is more sensitive at lower levels and gradually saturates at higher levels. This finding further supports the complex dose-response characteristics of TCBI-LN as a continuous variable in evaluating nutritional risk among hospitalized patients. Bootstrap analysis showed that the RCS-derived nadir was relatively stable at 7.85, whereas the lower and upper turning points showed wider uncertainty intervals, supporting their interpretation as exploratory data-derived points rather than definitive clinical thresholds.
Fig. 4.
Restricted cubic spline plot. (Adjusted for variables in Model 5)
Risk stratification analysis based on RCS inflection points
Based on the two TCBI-LN inflection points identified from the RCS curve (Fig. 5), all 3,382 patients were divided into three risk strata (Table 5): a low-TCBI high-risk group, an optimal low-risk group, and a high-TCBI high-risk group. The HAM incidence rates in the three groups were 26.1%, 17.0%, and 18.3%, respectively, and the differences among the groups were highly statistically significant (P < 0.001). The HAM incidence in the low-TCBI high-risk group was significantly higher than that in the other two groups, whereas the difference between the optimal low-risk group and the high-TCBI high-risk group was relatively small, further corroborating the nonlinear saturation effect revealed by the RCS analysis.
Fig. 5.
Forest plot for interaction analysis (Adjusted for variables in Model 5)
Table 5.
HAM incidence across RCS-derived TCBI-LN risk strata
| Variables | Low-TCBI.high-risk.group | Low-risk.optimal.group | High-TCBI.high-risk.group | Total | P | Statistic |
|---|---|---|---|---|---|---|
| TCBI-LN range | < 7.27 | 7.27 to < 8.50 | ≥ 8.50 | < 0.001 | 40.910 | |
| With HAM | 442(26.1%) | 278(17.0%) | 11(18.3%) | 731 | ||
| Without HAM | 1249(73.9%) | 1353(83.0%) | 49(81.7%) | 2651 | ||
| Total | 1691 | 1631 | 60 | 3382 |
Statistics: χ² (chi-square test)
Comparison of predictive performance across different models
To evaluate whether TCBI-LN provides incremental predictive information beyond its individual components, we compared models incorporating TCBI-LN versus TG, TC, body weight, ln(TC×TG), BMI, or NRS2002 separately (Supplementary Table 1). The TCBI-LN model achieved the highest AUC (0.695, 95% CI: 0.675–0.717), lowest AIC (3294.3), and lowest BIC (3435.2) among all models. TG alone (AUC = 0.667) and TC alone (AUC = 0.667) provided virtually no improvement over the base model (AUC = 0.666), while body weight alone (AUC = 0.675) and ln(TC×TG) (AUC = 0.670) showed only modest gains. DeLong tests confirmed that TCBI-LN significantly outperformed all comparator models (all P < 0.05), indicating that the multiplicative integration of TC, TG, and body weight captures synergistic metabolic-anthropometric information beyond any individual component. ROC analysis further showed that the TCBI-LN model outperformed models incorporating its individual components or established nutritional indicators, supporting its incremental discriminative value (Supplementary Fig. 3).
Adjusted subgroup analysis and interaction testing
To evaluate the cross-subgroup robustness of the association between TCBI-LN and HAM, after adjusting for all variables in Model 5, we performed stratified logistic regression and interaction testing according to sex, age (≥ 65 years vs. <65 years), BMI (normal vs. abnormal), hospital department, history of hypertension, history of diabetes mellitus, history of coronary heart disease, and history of cerebral infarction (Fig. 5). In the majority of subgroups, the protective effect of TCBI-LN was consistent in direction. However, the interaction test identified two significant effect modifications: hospital department (P for interaction = 0.003) and history of hypertension (P for interaction = 0.004). With regard to hospital department, the protective effect of TCBI-LN was more prominent among surgical patients (OR = 0.723, 95% CI: 0.605–0.864), whereas it did not reach statistical significance among medical patients (OR = 0.981, 95% CI: 0.847–1.138). In the hypertension history subgroup, the protective effect of TCBI-LN was clearly evident in patients without hypertension (OR = 0.845, 95% CI: 0.745–0.957), whereas the effect was in the opposite direction among patients with a history of hypertension (OR = 1.220, 95% CI: 0.933–1.595). The interaction tests for the remaining subgroups did not reach statistical significance, indicating that the association between TCBI-LN and HAM is relatively stable across the above-mentioned population characteristics. The RCS curves stratified by hypertension history and hospital department further visually illustrate the dose-response differences underlying the above effect modifications (Fig. 6). Although the associations between hospital department and hypertension history remained significant at a Bonferroni-corrected P < 0.006 level, these findings should be interpreted with caution, as they are exploratory in nature and based on hypothetical relationships. This is due to the issue of multiple testing and limited statistical power in some subgroups.
Fig. 6.
Subgroup RCS analysis of history of hypertension and hospitalization department (Adjusted for variables in Model 5)
Discussion
Based on 3,382 inpatients without malnutrition at admission from 34 tertiary referral hospitals in China, this study systematically evaluated the independent association between TCBI and HAM at discharge. To our knowledge, this is the largest multicenter real-world study of its kind in a similar population to date. Multivariable logistic regression analysis demonstrated that after fully adjusting for demographic characteristics, anthropometric measurements, nutritional risk scores, and major comorbidities, TCBI-LN remained independently and inversely associated with the risk of HAM occurrence. Restricted cubic spline analysis further revealed a nonlinear relationship between the two, identifying two key inflection points at original TCBI-LN values of 7.27 and 8.50. Based on these RCS-derived exploratory cut-points, the HAM incidence rates in the low-TCBI high-risk group, optimal low-risk group, and high-TCBI high-risk group were 26.1%, 17.0%, and 18.3%, respectively (P < 0.001), suggesting a possible nonlinear pattern, with lower risk in the intermediate TCBI-LN range. These findings may inform future validation studies but should not be regarded as established clinical thresholds.
As a composite nutritional index integrating TG, TC, and BW, TCBI has previously been shown to be associated with clinical outcomes in various cardiovascular, cerebrovascular, and critical care settings. Wang et al. [18] first reported in patients with dilated cardiomyopathy that the highest TCBI tertile had significantly lower all-cause mortality than the lowest tertile, establishing TCBI as an independent predictor of long-term survival. Liu et al. [19] found that low TCBI was closely associated with the occurrence of stroke-associated pneumonia. In patients with acute heart failure, low TCBI similarly indicates more severe cardiac impairment and poorer prognosis [20]. These findings are highly consistent in direction with those of the present study. Notably, a recently published study on acute ischemic stroke reported a nonlinear relationship between TCBI and 90-day adverse functional outcomes, with an inflection point of approximately 1227.3, which is relatively close to the first inflection point of 1436.6 (TCBI-LN represents the corresponding TCBI value at 7.27) identified in our study. This suggests that the saturation threshold of the protective effect of TCBI may possess a certain degree of cross-disease generalizability [21]. However, all of the above studies were conducted in specific disease cohorts. The present study is the first to demonstrate the independent predictive value of TCBI for HAM among inpatients with diverse disease conditions, thereby expanding the clinical application scope of TCBI.
Several possible mechanisms may explain the observed nonlinear association between TCBI-LN and HAM, although these explanations should be considered hypothesis-generating rather than direct mechanistic evidence. TCBI integrates triglycerides, total cholesterol, and body weight, and may therefore reflect both lipid-related metabolic reserve and body-size-related energy reserve [11]. Patients with lower TCBI-LN may have limited energy reserves and greater vulnerability to nutritional deterioration during hospitalization. However, our study did not longitudinally assess lipid metabolism pathways, hepatic lipoprotein synthesis, medication use, inflammatory trajectories, or detailed nutritional intake. Therefore, we cannot determine whether lipid metabolism itself mediates the association between TCBI-LN and HAM.
Notably, this study observed a “high-TCBI high-risk” phenomenon, which contradicts the hypothesis of a simple linear protective effect of TCBI. This pattern should not be interpreted as a confirmed U-shaped relationship; rather, it suggests that HAM risk was higher at low TCBI-LN levels, whereas the association became attenuated and uncertain at very high TCBI-LN levels. One possible explanation is that extremely high TCBI values may reflect heterogeneous metabolic conditions, such as metabolic syndrome, rather than uniformly adequate nutritional reserve. Under hospitalization-related stress, such patients may have impaired metabolic flexibility and may remain susceptible to nutritional deterioration. However, the high-TCBI high-risk group comprised only 60 patients with 11 HAM events, yielding wide confidence intervals. Therefore, this stratum should be regarded as hypothesis-generating rather than a validated clinical risk category. Independent validation in larger cohorts enriched for high TCBI values is warranted. In addition, the observed association may be partly related to the so-called lipid paradox, in which lower lipid levels in hospitalized or chronically ill patients may reflect inflammation, disease burden, or catabolic status rather than favorable metabolic health [22]. These potential mechanisms remain plausible hypotheses and require validation in future studies with repeated metabolic, inflammatory, pharmacological, and nutritional assessments.
Subgroup analysis identified two significant effect modifiers. First, the hospital department in which the patient was admitted significantly modified the association between TCBI-LN and HAM (P for interaction = 0.003). The protective effect was prominent in surgical patients (OR = 0.723), whereas no statistical significance was observed in medical patients (OR = 0.981). Surgical trauma, perioperative fasting, and stress-induced metabolic consumption can rapidly deplete the nutritional reserves that patients possess upon admission within a short period. This renders the baseline nutritional-metabolic level quantified by TCBI more sensitive in predicting HAM occurrence among surgical patients. This finding is consistent with previous reports indicating that poor preoperative nutritional status significantly increases the risk of postoperative complications [23], suggesting that systematic TCBI assessment and nutritional intervention should be implemented in surgical patients at the time of admission. Second, a history of hypertension was another important effect modifier (P for interaction = 0.004). The protective effect of TCBI-LN was clearly evident in patients without a history of hypertension (OR = 0.845), whereas the effect was reversed in patients with a history of hypertension (OR = 1.220). The interaction by hypertension status should be interpreted with caution. Patients with hypertension may differ from those without hypertension in triglyceride and cholesterol levels, body weight, adiposity, diabetes burden, chronic inflammatory status, and medication exposure. Because TG, TC, and body weight are direct components of TCBI, the weaker or reversed association observed among hypertensive patients may reflect differences in the distribution and clinical meaning of TCBI in this subgroup rather than a true difference in nutritional susceptibility. In hypertensive patients, a higher TCBI value may be more likely to represent an adverse metabolic phenotype rather than uniformly favorable nutritional reserve. In addition, medication data were unavailable, and residual confounding by lipid-lowering, antihypertensive, antidiabetic, or other therapies could not be assessed. Given the limited size of the hypertensive subgroup, this interaction finding should be regarded as exploratory and requires validation in cohorts with detailed metabolic and medication data.
Nevertheless, several limitations should be acknowledged. First, although this was a multicenter study, all hospitals were tertiary institutions in China, which may limit the generalizability of the findings. The analysis was also based on complete cases, and patients with missing handgrip strength, calf circumference, weight-change data, laboratory values, or discharge assessments were excluded. Because these variables are closely related to frailty, nutritional status, and disease severity, selection bias cannot be ruled out. Second, hospital-level clustering was not explicitly modeled. Differences in laboratory testing, nutritional care, and discharge assessment across centers may have affected the precision of the estimates. Third, the RCS-derived turning points were data-driven and should not be interpreted as established clinical thresholds. This is especially important for the high-TCBI range, where the high-TCBI high-risk group included only 60 patients and 11 HAM events; therefore, this subgroup should be regarded as exploratory. In particular, we could not exclude or separately analyze patients receiving lipid-lowering therapy or those with severe dyslipidemia, although these conditions may alter TG and TC levels independently of nutritional status. Fourth, partial overlap exists between the exposure and outcome definitions. TCBI includes body weight, whereas GLIM-defined malnutrition partly relies on body-size- and muscle-related criteria, such as low BMI, weight loss, and reduced muscle mass surrogates. Although BMI and other covariates were adjusted for, this structural overlap cannot be fully eliminated. Fifth, detailed medication data were unavailable. Lipid-lowering therapy, antihypertensive treatment, antidiabetic medications, corticosteroids, and nutrition-related interventions may influence lipid levels, body weight, inflammation, and HAM risk, leading to residual confounding. Therefore, subgroup differences, particularly those related to hypertension or diabetes, should be interpreted cautiously and not attributed to medication-related mechanisms. Finally, muscle mass was assessed using calf circumference and handgrip strength rather than direct body-composition methods, and the study did not include post-discharge outcomes or external predictive validation. Thus, TCBI-LN should be viewed as a metabolic-anthropometric marker associated with HAM risk, rather than a validated standalone prediction tool.
Conclusion
Based on 3,382 inpatients without malnutrition at admission from 34 tertiary referral hospitals in China, this study confirms that admission TCBI is independently associated with the risk of HAM, exhibiting a significant nonlinear dose-response relationship. These findings support the integration of TCBI into routine admission assessment protocols to aid in the identification of patients at high risk for HAM and to guide early metabolic-nutritional interventions.
Supplementary Information
Below is the link to the electronic supplementary material.
Author contributions
Yonghao Li: conceptualization, methodology, investigation, data curation, writing (original draft), and visualization; Xiaoyi Luan: methodology, formal analysis, writing (review and editing); Yilin Li: methodology, data curation, writing (review and editing); Liru Chen: conceptualization, writing (review and editing); Mingwei Zhu: conceptualization, methodology, formal analysis, investigation, writing (review and editing), and supervision.
Funding
National High Level Hospital Clinical Research Funding Project (BJ-2025-204).
Data availability
The dataset supporting the conclusions of this article is included within the article.
Declarations
Human ethics and consent to participate declarations
Each patient agreed and signed the informed consent form. This study was approved by the Ethics Committee of Beijing Hospital (Approval No. 2014BJYYEC-022-02) and was conducted in strict accordance with the Declaration of Helsinki. The study was registered with the Chinese Clinical Trial Registry (Registration No.: ChiCTR-EPC14005253). All patients signed informed consent forms and agreed to participate in the study.
Declaration of generative AI and AI-assisted technologies in the writing process
The authors declare that no generative AI and AI-assisted technologies were used in the writing process.
Competing interests
The authors declare no competing interests.
Footnotes
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Yonghao Li, Xiaoyi Luan and Yilin Li contributed equally to this work.
References
- 1.Bellanti F, Lo Buglio A, Quiete S, Vendemiale G. Malnutrition in Hospitalized Old Patients: Screening and Diagnosis, Clinical Outcomes, and Management. Nutrients. 2022;14(4):910. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Meyer F, Valentini L. Disease-Related Malnutrition and Sarcopenia as Determinants of Clinical Outcome. Visc Med. 2019;35(5):282–91. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Braunschweig C, Gomez S, Sheean PM. Impact of declines in nutritional status on outcomes in adult patients hospitalized for more than 7 days. J Am Diet Assoc. 2000;100(11):1316–22. quiz 1323-4. [DOI] [PubMed] [Google Scholar]
- 4.Allard JP, Keller H, Jeejeebhoy KN, Laporte M, Duerksen DR, Gramlich L, Payette H, Bernier P, Davidson B, Teterina A, Lou W. Decline in nutritional status is associated with prolonged length of stay in hospitalized patients admitted for 7 days or more: A prospective cohort study. Clin Nutr. 2016;35(1):144–52. [DOI] [PubMed] [Google Scholar]
- 5.Lima J, Teixeira PP, Eckert IDC, Burgel CF, Silva FM. Decline of nutritional status in the first week of hospitalisation predicts longer length of stay and hospital readmission during 6-month follow-up. Br J Nutr. 2021;125(10):1132–9. [DOI] [PubMed] [Google Scholar]
- 6.White JV, Guenter P, Jensen G, Malone A, Schofield M. Academy of Nutrition and Dietetics Malnutrition Work Group; A.S.P.E.N. Malnutrition Task Force; A.S.P.E.N. Board of Directors. Consensus statement of the Academy of Nutrition and Dietetics/American Society for Parenteral and Enteral Nutrition: characteristics recommended for the identification and documentation of adult malnutrition (undernutrition). J Acad Nutr Diet. 2012;112(5):730–8. [DOI] [PubMed] [Google Scholar]
- 7.Botero L, Banks MD, Gordon EH, Bauer J, Young AM. Incidence and outcomes of in-hospital nutritional decline: A prospective observational cohort study in adult patients. Clin Nutr. 2024;43(5):1057–64. [DOI] [PubMed] [Google Scholar]
- 8.Botero L, Young AM, Banks MD, Bauer J. Incidence and criteria used in the diagnosis of hospital-acquired malnutrition in adults: a systematic review and pooled incidence analysis. Eur J Clin Nutr. 2023;77(1):23–35. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Jensen GL, Cederholm T, Correia MITD, Gonzalez MC, Fukushima R, Pisprasert V, Blaauw R, Braz DC, Carrasco F, Cruz Jentoft AJ, Cuerda C, Evans DC, Fuchs-Tarlovsky V, Gramlich L, Shi HP, Hasse JM, Hiesmayr M, Hiki N, Jager-Wittenaar H, Jahit S, Jáquez A, Keller H, Klek S, Malone A, Mogensen KM, Mori N, Mundi M, Muscaritoli M, Ng D, Nyulasi I, Pirlich M, Schneider S, de van der Schueren M, Siltharm S, Singer P, Steiber A, Tappenden KA, Yu J, van Gossum A, Wang JY, Winkler MF, Compher C, Barazzoni R. GLIM consensus approach to diagnosis of malnutrition: A 5-year update. JPEN J Parenter Enter Nutr. 2025;49(4):414–27. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Mayneris-Perxachs J, Swann JR. Metabolic phenotyping of malnutrition during the first 1000 days of life. Eur J Nutr. 2019;58(3):909–30. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Li Y, Li Y, Luan X, Chen L, Liu P, Cui H, Zhu M. Lower triglyceride-cholesterol-body weight index is independently associated with increased in-hospital complication risk: a large multicenter real-world study. Lipids Health Dis. 2026;25(1):132. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Yang-Giuliano X, Osborn EA. Paradoxically simple: A new nutritional index for predicting coronary risk. Int J Cardiol. 2018;262:106–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Kondrup J, Allison SP, Elia M, Vellas B, Plauth M, Educational, Committee CP. European Society of Parenteral and Enteral Nutrition (ESPEN). ESPEN guidelines for nutrition screening 2002. Clin Nutr. 2003;22(4):415–21. [DOI] [PubMed] [Google Scholar]
- 14.Cederholm T, Jensen GL, Correia MITD, Gonzalez MC, Fukushima R, Higashiguchi T, Baptista G, Barazzoni R, Blaauw R, Coats AJS, Crivelli AN, Evans DC, Gramlich L, Fuchs-Tarlovsky V, Keller H, Llido L, Malone A, Mogensen KM, Morley JE, Muscaritoli M, Nyulasi I, Pirlich M, Pisprasert V, de van der Schueren MAE, Siltharm S, Singer P, Tappenden K, Velasco N, Waitzberg D, Yamwong P, Yu J, Van Gossum A, Compher C, GLIM Core Leadership Committee, GLIM Working Group. GLIM criteria for the diagnosis of malnutrition - A consensus report from the global clinical nutrition community. J Cachexia Sarcopenia Muscle. 2019;10(1):207–17. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Chen LK, Hsiao FY, Akishita M, Assantachai P, Lee WJ, Lim WS, Muangpaisan W, Kim M, Merchant RA, Peng LN, Tan MP, Won CW, Yamada M, Woo J, Arai H. A focus shift from sarcopenia to muscle health in the Asian Working Group for Sarcopenia 2025 Consensus Update. Nat Aging. 2025;5(11):2164–75. [DOI] [PubMed] [Google Scholar]
- 16.Cederholm T, Bosaeus I, Barazzoni R, Bauer J, Van Gossum A, Klek S, Muscaritoli M, Nyulasi I, Ockenga J, Schneider SM, de van der Schueren MA, Singer P. Diagnostic criteria for malnutrition - An ESPEN Consensus Statement. Clin Nutr. 2015;34(3):335–40. [DOI] [PubMed] [Google Scholar]
- 17.Li Y, Chen L, Yang X, Cui H, Li Z, Chen W, Shi H, Zhu M. Dynamic association of serum albumin changes with inflammation, nutritional status and clinical outcomes: a secondary analysis of a large prospective observational cohort study. Eur J Med Res. 2025;30(1):679. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Wang C, Qing Y, Chen W, Li G. A novel nutritional index as a predictor of mortality in dilated cardiomyopathy: a retrospective study. PeerJ. 2022;10:e12704. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Liu Y, Chen Y, Zhi Z, Wang P, Wang M, Li Q, Wang Y, Zhao L, Chen C, Association Between TCBI, Triglycerides. Total Cholesterol, and Body Weight Index) and Stroke-Associated Pneumonia in Acute Ischemic Stroke Patients. Clin Interv Aging. 2024;19:1091–101. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Pedersen MH, Davidovski FS, Espersen C, Adam LM, Khoraizat A, Stanchev A, Durukan E, Skaarup KG, Dons M, Rastoder E, Al-Rubai AH, Sengeløv M, Bernholm KF, Lassen MCH, Pareek M, Jørgensen PG, Wolsk E, Schou M, Biering-Sørensen T. The Triglycerides, Total Cholesterol and Body Weight Index (TCBI) in acute heart failure - a post-hoc analysis of a prospective cohort study. Nutr Metab Cardiovasc Dis. 2026;36(5):104532. [DOI] [PubMed] [Google Scholar]
- 21.Xiao B, Zhu J, Han Y. Association of a novel nutritional marker, the triglyceride-cholesterol-body weight index, with 90-day unfavorable outcomes in acute ischemic stroke: a prospective cohort study. Front Nutr. 2026;12:1707231. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Thurmann KE, Mukherjee TG, White MD. Inflammation-Driven Lipid Suppression in Hospitalized Patients: Insights Into the Inflammatory Lipid Paradox From a Retrospective Study. Cureus. 2025;17(8):e89488. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Keerio RB, Ali M, Shah KA, Iqbal A, Mehmood A, Iqbal S. Evaluating the Impact of Preoperative Nutritional Status on Surgical Outcomes and Public Health Implications in General Surgery Patients. Cureus. 2024;16(12):e76633. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The dataset supporting the conclusions of this article is included within the article.