Data source and study population
This cross-sectional study analyzed electronic medical records of all geriatric psychiatry inpatients admitted to Huzhou Third Municipal Hospital between July 2020 and December 2024 (n = 4,445). Data extraction excluded all personal identifiers, retaining only clinical variables. The Ethics Committee of Huzhou Third Municipal Hospital approved this low-risk retrospective observational study (Approval No. 2025–119). Clinical trial number: not applicable.
Eligibility required complete documentation of pre-admission medications and hepatorenal function biomarker panels (ALT, AST, BUN, Cr). Patients using hepatoprotective agents were excluded to avoid underestimation of liver injury biomarkers (ALT/AST ratio). Hepatoprotective agents may normalize ALT/AST ratio levels in individuals with subclinical hepatic impairment, thereby obscuring the true association between liver function biomarker (ALT/AST ratio) and depression risk. However, this restriction may limit generalizability to elderly populations actively managed with such medications. After excluding patients with hepatoprotectant exposure or incomplete data, the final analytical sample included 1,783 patients. Figure 1 illustrates a schematic of the selection methodology.
Flow chart of the study population inclusion
Hepatorenal function biomarkers
The primary independent variables were ALT/AST ratio, BUN (mmol/L), and BUN/Cr ratio. All biomarkers were measured within 24 h of admission using chemiluminescence microparticle immunoassay (CMIA) on an Abbott Architect i2000 analyzer (Abbott Diagnostics, Chicago, USA). The proteinuria was assessed by dipstick test. Ratios were calculated as: (1) ALT (U/L)/AST (U/L) [24], (2) BUN (mg/dL)/Cr (mg/dL) [25]. Variables were categorized into quartiles for analysis.
Assessment of depression
Depression diagnosis required meeting DSM-5 criteria [26]:
-
a.
MDD: ≥ 5 core symptoms (including depressed mood/anhedonia) over ≥ 2 weeks.
-
b.
Persistent Depressive Disorder: Depressive symptoms lasting ≥ 2 years with ≤ 2 months’ remission.
-
c.
Standardized assessments included MINI 7.0 [27] or SCID-5 [28], confirmed by psychiatrists.
Covariates
Demographic/clinical covariates: age, sex, education, personality, marital status, smoking, drinking, viral hepatitis (VH), fatty liver (FL), cerebral infarction (CI), anxiety, schizophrenia, bipolar disorder, blood pressure (systolic blood pressure [SBP]/diastolic blood pressure [DBP], mmHg), glucose (GLU, mmol/L), lipids (triglyceride [TG]/total cholesterol [TC], mmol/L), albumin (ALB, g/L), prothrombin time (PT, s), white blood cell (WBC, 10⁹/L), total bilirubin (TBIL, μmol/L), eGFR (mL/min/1.73m2), proteinuria, GGT/ALT/AST/alkaline phosphatase (ALP) (U/L), BUN (mmol/L), Cr (μmol/L) and antipsychotic drugs using.
Missing values in covariates were addressed using a multivariate single imputation method. This approach utilized an iterative imputer, with a Bayesian Ridge model serving as the estimator in each step of the round-robin imputation process. The validity of this method is supported by evidence that when covariates exhibit ≤ 10% missingness, single imputation negligibly impacts statistical power and effect estimation [29].
Statistical analysis
In our primary analysis of datasets, continuous variables with normal distribution were described using mean ± standard deviation (SD), skewed continuous variables using median (interquartile range, IQR), and categorical variables as frequencies (percentages). Between-group comparisons utilized: independent Student’s t-tests (normal continuous data), Mann–Whitney U tests (non-normal continuous data), and χ2 tests (categorical variables). Logistic regression models generated odds ratios (ORs) with 95% confidence intervals (CIs).
To investigate the association between biomarkers and depression risk, we categorized ALT/AST ratio, BUN and BUN/Cr ratio into quartiles (Q1-Q4) based on clinical thresholds. Three progressively adjusted models were developed: Model 1 controlled for age and sex; Model 2 added lifestyle factors (smoking, drinking) and sociodemographics (personality, marital status, education); Model 3 further incorporated biological markers (blood pressure, GLU, lipids, ALB, WBC, eGFR and proteinuria), chronic comorbidities (VH, FL, CI), psychiatric comorbidities (anxiety, schizophrenia, bipolar disorder) and antipsychotic drugs using, thus comprehensively accounting for confounders. The robustness of study outcomes was verified by comparing effect magnitudes and p-values across the three refined models. This approach ensured the reliability of conclusions through systematic evaluation of potential influences on findings. Subsequently, we applied restricted cubic spline (RCS) models with four knots (5th, 35th, 65th, 95th percentiles) to evaluate relationships between these biomarkers and depression risk, adjusting for covariates as specified in Model 3. Covariate selection adhered to predefined criteria: clinical relevance, prior literature evidence, statistical significance in univariate analyses (P < 0.05), or variables altering effect estimates by > 10%.
Interaction and stratified analyses were performed across prespecified subgroups defined by age (< 80 vs. ≥ 80 years), sex, education, personality, drinking status, FL and CI. Interaction effects between ALT/AST ratio, BUN and BUN/Cr ratio with subgroup variables were evaluated using likelihood ratio tests. Moreover, sensitivity analyses were conducted employing a complete-case analysis strategy.
The statistical analyses in the study were carried out using R Statistical Software (Version 4.2.2, The R Foundation) and the Free Statistics Analysis Platform (Version 1.9, A two-sided P < 0.05 was considered statistically significant.
link