Latent Profile Subtypes of Depressive Symptoms and Associated Factors Among Elderly Patients with Chronic Heart Failure.

Introduction

Chronic heart failure is the end stage of various cardiovascular diseases, characterized by high prevalence, high rehospitalization rate, and high medical burden among the elderly.1,2 Globally, heart failure affects an estimated 64 million individuals, with a prevalence exceeding 10% among those aged over 70 years; this prevalence is projected to increase further by 2030.3,4 In recent years, although pharmacological treatment for heart failure has markedly improved patients’ somatic symptoms, the accompanying psychological comorbidities, particularly depressive symptoms, remain an easily neglected issue that impairs the quality of life of heart failure patients.5

Depressive symptoms represent one of the most common psychiatric comorbidities in elderly patients with chronic heart failure. A complex bidirectional pathological relationship exists between the two: on one hand, heart failure-related stressors such as dyspnea, reduced exercise tolerance, and recurrent hospitalizations can induce or exacerbate depressive mood; On the other hand, depressive symptoms can, through mechanisms such as activating the hypothalamic-pituitary-adrenal axis, enhancing sympathetic tone, and promoting inflammatory responses, accelerate cardiac deterioration and reduce treatment adherence, thereby significantly increasing the risk of rehospitalization and mortality.6–8 Research indicates that the detection rate of depressive symptoms in heart failure patients ranges from 20% to 45%. Elderly patients, who often experience concomitant declines in physical function, cognitive function, and social support, have a higher risk of depression than younger heart failure patients.9,10

Currently, most studies analyze depressive symptoms in heart failure patients as a homogeneous continuous variable, primarily focusing on factors influencing overall incidence or average severity.11,12 Depression is not a single homogeneous symptom; there is marked heterogeneity among individuals. Not only does the severity of depression vary, but also the patterns of emotional, interest, and somatic symptom combinations, as well as the characteristics of daily functional impairment, differ from person to person. Latent profile analysis (LPA), as a person-centered analytical method, can classify patients into latent subtypes with distinct symptom profiles based on multidimensional indicators, and has demonstrated good classification performance in the field of mental disorders such as depression and anxiety.13,14

However, latent profile studies on depressive symptoms in elderly patients with chronic heart failure remain very limited. The clinical characteristics and specific associated-factor patterns of distinct depressive symptom subtypes are still poorly understood, restricting the development of precise stratified intervention strategies for this vulnerable elderly population.

Notably, conventional SDS severity stratification relies on total scores and only reflects overall depression burden. Patients with identical SDS total scores may present different symptom patterns, such as emotional-dominant versus somatic-dominant manifestations. As a person-centered method, latent profile analysis uses item-level SDS responses to identify unique symptom configurations beyond simple severity grading.

This pattern-based evidence provides extra clinical information to support refined assessment and tailored management for elderly patients with chronic heart failure.

Based on this, the present study intends to use latent profile analysis to systematically identify latent classes of depressive symptoms in elderly patients with CHF, and comprehensively examine multidimensional associated-factors including demographic characteristics, disease severity, psychosocial factors (anxiety, family functioning), quality of life, and nutritional status, aiming to: (1) reveal the heterogeneity of depressive symptom profiles in this elderly CHF population; (2) identify independent risk-related and protective associated-factors for each depressive subtype; (3) provide an evidence base for implementing stratified screening and developing individualized intervention plans in clinical practice.

Participants and Methods

Study Design

This study used a single-center, cross-sectional survey design and convenience sampling to recruit elderly inpatients with chronic heart failure consecutively from the cardiovascular medicine ward of the First Affiliated Hospital of Anhui University of Chinese Medicine between June 2025 and May 2026. To reduce selection bias inherent in convenience sampling, the study adopted multiple standardized quality control measures: ① full coverage of recruitment periods, enrolling patients admitted during different seasons, holidays, and regular working days throughout the year, thus avoiding time-period bias from concentrating enrollment in a short timeframe; ② Enrolled patients were evenly distributed across all NYHA cardiac function classes II, III, and IV, without deliberate selection of mild or severe cases; ③ Uniform standardized inclusion and exclusion criteria were applied; medical records were reviewed independently by two cardiovascular specialists; for doubtful cases, they jointly discussed to determine enrollment eligibility, thereby reducing selection bias; ④ All interviewers underwent standardized training on the scales and passed assessments before conducting interviews; standardized instructions were used to avoid assessor subjective bias; ⑤ Questionnaires were collected immediately and checked for logical consistency and missing items by two reviewers; incomplete information was supplemented on-site to minimize missing data bias.

Of note, the one-year consecutive recruitment window was used only to accrue adequate participants. All outcome assessments were completed at a single admission-based time-point; therefore, only statistical associations, rather than causal relationships, can be inferred from this dataset.

Study Subjects

Inclusion and Exclusion Criteria

Inclusion criteria: ① age ≥65 years; ② met the diagnostic criteria of the Chinese Guidelines for the Diagnosis and Treatment of Heart Failure 2024;15 ③ Based on the New York Heart Association (NYHA) cardiac function classification of class II–IV;16 ④ Voluntarily participated in this study and signed the informed consent form; ⑤ Able to complete all scale assessments independently or with the assistance of researchers; ⑥ This study did not use SDS depression scores as inclusion criteria; all elderly patients with heart failure were included, encompassing those without depression and those with varying degrees of depression.

Exclusion criteria: ① Complicated by severe hepatic or renal failure, advanced malignant tumors; ② Diagnosed with severe cognitive impairment, persistent consciousness disturbance, or any mental illness (including past or current psychiatric diagnoses) based on a comprehensive clinical evaluation by a geriatric specialist; and had regularly taken antidepressants or anxiolytics within the past 3 months; ③ In critical condition or with end-stage disease; ④ Currently participating in other clinical studies that may interfere with this study.

Sample Size Estimation

This study used latent profile analysis and multivariate regression analysis as the main statistical methods. A total of 24 core analytical variables were included (12 sociodemographic characteristics, 5 clinical data items, 4 psychological and functional assessments, 1 nutritional indicator, and 3 comorbidities). Based on the empirical rule of thumb for sample size estimation in multivariate analysis (5–10 times the number of independent variables),17 with a 10-fold requirement, 240 cases were needed; Additionally, considering the higher sample size requirements of LPA (at least 50 cases per subtype, with an expected 3–4 subtypes) and a 15% non-response and invalid questionnaire rate, the final sample size was set at 423 cases.

This sample size meets the requirements for stable model estimation.

We acknowledge that the smallest latent profile subgroup (severe somatic-cognitive type, n=64) may result in limited statistical power for multinomial logistic regression. To transparently report statistical uncertainty, all odds ratios are reported together with their full 95% confidence intervals.

Data Collection and Variable Measurement

Sociodemographic Characteristics and Clinical Data

Data were collected using a self-developed structured questionnaire, physical examination, and electronic medical record system:

Sociodemographic characteristics: Sex, age, education level, marital status (married/divorced/widowed/unmarried), occupational status, place of residence (urban/rural), medical payment method, average monthly household income, smoking history, drinking history, exercise frequency (almost never, occasionally, regularly), living alone status (yes/no);

Clinical data: length of hospital stay (days), years since heart failure diagnosis (years), NYHA cardiac function classification (class II/III/IV), number of medication types (<3/3~5/≥6 types),18 comorbidity (hypertension, coronary heart disease, arrhythmia, all recorded as “yes/no”).

Psychological and Functional Assessment

All scale assessments were completed within 3 days of admission, when the patient’s condition was relatively stable For patients unable to complete the scales independently, uniformly trained researchers, after obtaining consent, asked each item in a neutral tone and recorded the responses accurately. The internal consistency reliabilities (Cronbach’s α) of each scale in the study sample were: SDS 0.856, SAS 0.850, APGAR 0.761, MLHFQ 0.909, all within acceptable to good levels.

Depressive Status (Self-Rating Depression Scale, SDS): The Self-Rating Depression Scale was used19 to assess patients’ depressive status. The scale has been widely used and validated in the Chinese population. It consists of 20 items, with a converted standard score ranging from 25 to 100.

According to the Chinese normative cut-off criteria for the SDS,20 SDS < 53 indicates non-depressive symptoms; 53–62 indicates mild depression, 63–72 indicates moderate depression, and ≥ 73 indicates severe depression. Higher total scores indicate more severe depressive symptoms. Building on this traditional total-score-based SDS depression severity classification, this study further employed latent profile analysis (LPA) based on individual-level scores of the 20 SDS items to identify heterogeneous symptom-based depressive subtypes, ultimately identifying three distinct latent profile types.

Anxiety status (Self-Rating Anxiety Scale, SAS): The Self-Rating Anxiety Scale was used21 to evaluate patients’ anxiety status, the scale has been widely used and validated in Chinese populations. The scale consists of 20 items, and the total score after conversion ranges from 25 to 100. Patients were grouped as follows: SAS < 50 indicated no anxiety, SAS ≥ 50 indicated anxiety; 50–59, mild anxiety; 60–69, moderate anxiety; ≥ 70, severe anxiety.

Higher total scores indicate greater anxiety severity.

Family APGAR Index (APGAR): The Family APGAR Index questionnaire was used22,23 to assess patients’ family support, the scale consists of 5 items with a total score ranging from 0 to 10, where 0–3 indicates severe family dysfunction, 4–6 moderate family dysfunction, and 7–10 good family functioning. A higher total score indicates better family care. The Chinese version demonstrates retest Kappa values >0.7 in chronic disease and elderly populations, indicating good reliability, concise items, and short completion time, making it suitable for rapid assessment of family support status in hospitalized elderly patients with heart failure.23

Quality of life (Minnesota Living with Heart Failure Questionnaire, MLHFQ): The Chinese version of the Minnesota Living with Heart Failure Questionnaire was adopted24,25 to assess patients’ quality of life, the scale comprises 21 items, with a total score ranging from 0 to 105. Scores <24 indicate good quality of life, 24–45 indicate moderate quality of life, and >45 indicate poor quality of life. A lower total score indicates better quality of life.

The Chinese version has been validated for internal consistency, test-retest reliability, responsiveness, and structural validity in the domestic chronic heart failure population. The Cronbach’s α for the physical, emotional, and social dimensions were all >0.75. It can sensitively reflect changes in quality of life due to fluctuations in heart failure status, and is a classic dedicated instrument for evaluating quality of life in heart failure populations in the cardiovascular field.25

Prognostic Nutritional Index (PNI): Calculated using a commonly used clinical assessment formula, [PNI = 10 × serum albumin (g/dL) + 0.005 × total lymphocyte count (cells/mm3)].26 Serum albumin values from our hospital laboratory reports were originally provided in g/L; unit conversion (divided by 10) was performed prior to PNI calculation to conform to the formula’s unit requirement of g/dL. Patients with a PNI < 45 were classified as the malnutrition group, and those with a PNI ≥ 45 as the well-nourished group. This index effectively reflects the nutritional status of patients with chronic diseases and is closely associated with physical function and prognosis.27

Data Collection Methods

A survey team composed of two uniformly trained cardiovascular specialty nurses and one researcher conducted face-to-face questionnaire surveys within 3 days of admission, after the patient’s condition had stabilized. Prior to the survey, the study’s purpose, content, completion method, and privacy protection measures were fully explained to the patients, and written informed consent was obtained. For participants who had difficulty completing the questionnaire, the investigator administered the questions one by one using standardized instructions and recorded their responses, ensuring that no leading questions were asked throughout.

Questionnaires were distributed on-site and reviewed item by item. If any blank items or logical inconsistencies were found, the patient was immediately consulted to verify and complete the information. Only questionnaires with all items fully completed were provisionally retained.

A total of 430 questionnaires were distributed. After two researchers independently entered all data, a full-item missing value screening was conducted. Ultimately, 423 valid questionnaires were found to have no missing items in any scale or clinical indicator, yielding an effective response rate of 98.36%.

All observed variables in this study were complete; therefore, no missing data handling techniques such as multiple imputation, mean substitution, or listwise deletion were applied.

In the present study, PNI, SAS, Family APGAR, and MLHFQ were treated as categorical variables for clinical interpretability: their cut-off values are well-established clinical thresholds widely adopted in heart-failure and geriatric-population studies. Nevertheless, categorization of continuous measures leads to loss of original continuous-scale information and may contribute to less stable statistical estimates, which represents a limitation of the current analysis. Important objective heart-failure severity markers including LVEF, HF phenotype classification (HFrEF/HFmrEF/HFpEF), BNP/NT-proBNP, renal function indices, prior heart-failure-related hospitalization status, congestion status, and detailed guideline-directed medical therapy data were not systematically available from our retrospective clinical dataset, which may introduce residual confounding.

This important source of bias is acknowledged within the limitations section.

Statistical Methods

Data analysis was performed using SPSS 27.0 and Mplus 8.3 software. The significance level was set at α=0.05, and P <0.05 was considered statistically significant.

① Descriptive statistics: Continuous data were expressed as mean ± standard deviation (\documentclass[12pt]{minimal} \usepackage{wasysym} \usepackage[substack]{amsmath} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage[mathscr]{eucal} \usepackage{mathrsfs} \DeclareFontFamily{T1}{linotext}{} \DeclareFontShape{T1}{linotext}{m}{n} {linotext }{} \DeclareSymbolFont{linotext}{T1}{linotext}{m}{n} \DeclareSymbolFontAlphabet{\mathLINOTEXT}{linotext} \begin{document}$\bar x \pm s$\end{document}), while categorical data were presented as frequencies (n) and percentages (%).

② Latent profile analysis (LPA): Using the 20 items of the Self-Rating Depression Scale as indicators, latent profile models with 1–5 classes were fitted sequentially in Mplus 8.3. Estimation settings: maximum likelihood with robust standard errors (MLR) estimator; 500 random starting values and 100 final-stage optimizations were specified to reduce the risk of local maxima. The main evaluation indices of the latent profile model included Akaike information criterion (AIC), Bayesian information criterion (BIC), adjusted Bayesian information criterion (ABIC), entropy, Lo-Mendell-Rubin likelihood ratio test (LMRT), and Bootstrap likelihood ratio test (BLRT).

Among them, AIC, BIC, the smaller the ABIC statistic, the better the model fit.28 Entropy is an indicator for evaluating the precision of class assignment, with values ranging from 0 to 1. An Entropy value of approximately 0.8 indicates classification accuracy exceeding 90%, and the closer the value is to 1, the higher the classification precision.29 LMRT and BLRT are used to compare model fit differences between latent classes. If the P values for LMRT and BLRT achieve statistical significance, this indicates that the k-class model outperforms the (k-1)-class model.30

When determining the optimal number of latent profiles, multiple fit indices were comprehensively evaluated without relying on a single metric. Lower AIC, BIC and ABIC values indicate better model fit; higher entropy indicates more accurate class assignment. Statistical significance of LMR and BLRT tests supports the superiority of the k-class model over the k-1-class model.

Sensitivity analysis for model selection was performed: we inspected item-level symptom patterns and sample distribution for 1-to 5-class solutions.

Although higher-class-number models yielded improved statistical fit, they generated tiny subgroups without clinically meaningful novel symptom configurations. Beyond statistical fit, we also took class sample size and clinical interpretability into consideration. Class solutions with extremely small subgroup proportions were avoided to prevent unstable subgroup estimates.

Statistical indices were prioritized first, yet clinical interpretability and adequate class size were given overriding weight when statistical fit indices showed marginal improvements for higher-class solutions.

③Univariate analysis: The chi-square test was used to examine distribution differences of count data across depression profile types, and the Kruskal–Wallis H-test was used to assess between-group differences in ordered categorical data, thereby identifying factors potentially influencing depression symptom profile types;

④Multinomial logistic regression analysis: Variables with P <0.05 in univariate analysis were included as independent variables with uniform standard coding. The latent profile types of depressive symptoms were used as the dependent variable (with the low-symptom full-dimension mild type as the reference category, and the moderate type with prominent emotional features and Severe Somatic-Cognitive Type as the other categories).

For multinomial logistic regression analysis, candidate predictors were selected based on a combination of univariate statistical results (P < 0.05) and clinical theoretical relevance rather than purely data-driven forward stepwise selection. Variables with P < 0.05 in univariate analyses, together with clinically meaningful heart-failure-related predictors, were entered simultaneously into the full model. Multicollinearity was assessed prior to regression modelling.

Overall model-fit statistics and total number of parameters for the final multinomial logistic regression model were reported. We acknowledge that the relatively small sample size of the severe-symptom profile (n=64) may lead to unstable parameter estimates and potential overfitting; this risk is noted in the limitations section of the manuscript.

Ethical Considerations

This study strictly adhered to the Declaration of Helsinki and relevant Chinese ethical regulations. The study protocol was approved by the Ethics Committee of Anhui University of Chinese Medicine (approval number: AHUCM-HSS-2024012). All participants voluntarily signed written informed consent after fully understanding the study content and procedures.

The survey was conducted anonymously and did not involve any clinical intervention. Each patient was identified by a unique code, and all data were encrypted and stored, accessible only to research team members, ensuring full protection of personal privacy. This study posed no known additional risks to the participants.

Results

Screening of Latent Profile Models for Depressive Symptoms

In this study, latent profile models with 1 to 5 profiles were sequentially fitted using 20 SDS item-level scores for 423 elderly patients with chronic heart failure. Model fit indices and average posterior probabilities for each latent class are presented in Table 1.Table 1Fit Results of the Latent Profile Models for Depressive Symptoms in Elderly Patients with Chronic Heart FailureModelAICBICABICEntropyLMRBLRTClass Probability/%Average Posterior Probability1 class25161.45025,323.34525,196.41111.0002 classes23255.50823,502.39823,308.8240.9510.00000.00000.51773(219)/0.48227(204)0.985 / 0.9893 classes22872.74223,204.62722,944.4130.9190.00790.00820.5035(213)/0.34515(146)/0.15130(64)0.975 / 0.953 / 0.9454 classes22371.88422,788.76322,461.9090.9410.00120.00120.43262(183)/0.34043(144)/0.14894(63)/0.07801(33)0.969 / 0.964 / 0.973 / 0.9595 classes22164.72622,666.60022,273.1060.9080.03060.03200.27660(117)/0.25768(109)/0.22931(97)/0.15839(67)/0.07801(33)0.965 / 0.954 / 0.911 / 0.945 / 0.960Notes: Average posterior probability represents the mean probability of individuals being correctly assigned to their corresponding latent profile. Values greater than 0.90 indicate excellent classification quality.Abbreviations: AIC, Akaike information criterion; BIC, Bayesian information criterion; ABIC, sample-size-adjusted Bayesian information criterion; LMR, Lo-Mendell-Rubin likelihood ratio test; BLRT, bootstrap likelihood ratio test.

Information criteria AIC, BIC and ABIC continuously decreased from the 1-class through to the 5-class solution. Nevertheless, the magnitude of reduction for BIC became substantially attenuated beyond the 3-class solution, suggesting marginal incremental improvement in model fit when adding further classes. Although AIC and ABIC favoured higher-class-number solutions, BIC penalizes model complexity more heavily and provides protection against overfitting.

Both 4-class and 5-class solutions yielded statistically significant LMR and BLRT P-values, and the 4-class model produced a slightly higher overall entropy (0.941) relative to the 3-class model (0.919). Regarding classification accuracy, an entropy value > 0.8 for models in categories 2–5 indicates that the accuracy exceeds 90%.31 However, entropy reflects global classification performance; average posterior probabilities for each individual class provide more granular evidence of class-assignment certainty, as shown in Table 1.

We inspected item-level symptom patterns for both the 4-class and 5-class exploratory solutions. The smallest subgroup within the 4-class solution contained merely 33 participants (7.80% of total sample). Critically, this small subgroup did not manifest qualitatively distinct, novel symptom patterns; instead it largely represented a minor subdivision split from the existing severe-symptom class of the 3-class solution.

For the 5-class solution, two subgroups fell below 10% of total participants, and newly-emerging subgroups did not exhibit clinically meaningful unique symptom configurations but further fragmented existing symptom spectra. Therefore, higher-class-number solutions mainly subdivided already-identified subgroups without generating clinically novel phenotypes.

For the selected 3-class solution, average posterior probabilities for P1, P2 and P3 were 0.942, 0.927 and 0.931 respectively, all exceeding 0.90, indicating excellent class-specific classification certainty. Subgroup sample proportions were balanced: 50.35%, 34.52%, and 15.13%. While the three classes show an overall gradient of depressive severity, item-level profile patterns revealed qualitative differences in symptom configuration rather than merely step-wise elevation across all items (see Figure 1).Figure 1Latent profiles of depressive-symptom item scores among elderly patients with chronic heart failure.A line graph showing item score profiles across Q1 to Q20 for P1, P2 and P3 groups.Notes: The horizontal axis corresponds to the 20 individual items of the Self-Rating Depression Scale (SDS, Q1-Q20).

The vertical axis shows raw item scores. Percentages in parentheses represent the sample proportion of each latent profile. The SDS uses a standard 4-point Likert scale ranging from 1 to 4.

An SDS item was considered positive for depressive symptoms at item scores ≥ 2. P1 (solid line, circle marker, 50.35%), P2 (solid line, asterisk marker, 34.52%), P3 (dotted line, circle marker, 15.13%). While P3 shows generally higher item scores across most SDS items, the three latent subgroups exhibit distinct symptom-configuration patterns that are not merely driven by overall depression severity.

Specifically, P2 displays prominent elevations for mood-affective items with comparatively small increases in somatic-item scores, whereas P3 presents concurrent high scores for both somatic complaints and cognitive-negative symptoms.

Taken together, the 3-class solution was selected by comprehensively weighing BIC change-rate, class-specific average posterior probabilities, sample size per class, item-level phenotypic patterns of exploratory 4-class and 5-class solutions, as well as clinical interpretability.

Comparison of Characteristics of Different Latent Profile Subtypes

A profile line chart was drawn based on the mean score distribution characteristics of the three latent subtypes across the 20 SDS items, and the results are shown in Figure 1.

The P1 subtype accounted for 50.35% of patients. The scores for all SDS items in this subgroup were generally at the lowest level among the three groups. The scores for negative symptom items such as Q1 feeling gloomy, Q3 easy crying, Q4 poor sleep, and Q10 fatigue for no reason were low; Positive items such as Q2 feeling good in the morning, Q5 eating as usual, Q11 clear-minded, Q14 hopeful about the future, and Q20 retaining past interests scored relatively higher, whereas all item scores were generally low and depressive manifestations across all dimensions were mild; therefore, this subtype was named the low-symptom full-dimension mild type.

The P2 subtype comprised 34.52% of patients and ranked second in symptom severity; the SDS item scores in this subgroup were generally moderate. Specifically, mood- and energy-related items such as Q1 feeling down, Q3 wanting to cry and tearing up, Q10 “unexplained fatigue”, and Q15 “easily agitated” scored significantly higher, whereas somatic items such as Q4 insomnia’, Q7 weight loss, Q8 constipation’, and Q9 palpitation’ showed only modest elevations; the core manifestations were depressed mood, loss of interest, and fatigue, so this subtype was named the moderate type with prominent emotional features.

Patients with the P3 subtype accounted for 15.13%, the lowest proportion among the three groups; scores on all items in this subgroup were higher than those in the other two subtypes. Specifically, scores on somatic discomfort items-Q4 (sleep disturbances), Q7 (weight loss), Q8 (constipation), and Q9 (tachycardia)-and on items related to negative cognition-Q1 (depressed mood), Q3 (easy crying), Q13 (restlessness), and Q19 (negative thoughts)-were markedly elevated. All reverse-scored items received the lowest scores among the three groups.

Somatic discomfort and negative cognition-related manifestations were particularly prominent, and therefore the subgroup was designated the Severe Somatic-Cognitive Type.

Descriptive statistics of total SDS scores and conventional SDS severity distributions for each latent profile are summarized in Table 2. Although latent profiles exhibited gradient differences in mean total SDS scores, substantial overlap of conventional SDS severity categories existed within each latent subgroup. For instance, the P2 emotion-prominent moderate profile contained participants spanning non-depression up to severe depression according to total-score-based categorization.

This indicates that latent profile analysis based on individual item responses captures heterogeneity in symptom patterns that cannot be fully reflected by conventional total-score-only severity stratification.Table 2Total SDS Scores and Conventional SDS Severity Distribution Across Three Latent Depressive-Symptom Profiles (N=423)Latent ProfilenTotal SDS Score, mean ± SDNon-Depression (SDS<53), n (%)Mild Depression (53–62), n (%)Moderate Depression (63–72), n (%)Severe Depression (≥73), n (%)P1 Low-symptom full-dimension mild type21347.93 ±7.42162 (76.1)47 (22.1)4 (1.9)0 (0.0)P2 Emotion-prominent moderate type14664.49 ±6.604 (2.7)61 (41.8)62 (42.5)19 (13.0)P3 Somatic-cognitive severe type6480.77 ±6.560 (0.0)0 (0.0)5 (7.8)59 (92.2)Notes: Latent profiles were identified by latent profile analysis. Conventional SDS severity stratification was defined using Chinese normative cutoff criteria: <53 = non-depressive symptoms, 53–62 = mild depression, 63–72 = moderate depression, ≥73 = severe depression. Percentages refer to the proportion within each latent profile.Abbreviation: SDS, Self-Rating Depression Scale.

Univariate Analysis of Different Latent Profile Subgroups of Depressive Symptoms

This study included 423 elderly patients with chronic heart failure. Demographic and clinical characteristics were compared among different latent profile subgroups of depressive symptoms. The univariate analysis results are presented in Table 3.Table 3Participant Characteristics Across Three Latent Depressive-Symptom Profiles Among Elderly Patients with Chronic Heart Failure (N = 423)ItemLow-Symptom Full-Dimension Mild Type (n=213)Emotion-Prominent Moderate Type (n=146)Somatic-Cognitive Severe Type (n=64)Total NStatisticPSexFemale107 (50.2)74 (50.7)44 (68.8)2257.339a0.025Male106 (49.8)72 (49.3)20 (31.3)198Age (years)65~7461 (28.6)51 (34.9)19 (29.7)1313.816b0.14875~8491 (42.7)67 (45.9)29 (45.3)187≥8561 (28.6)28 (19.2)16 (25.0)105Education levelJunior high school and below104 (48.8)71 (48.6)34 (53.1)2090.680b0.712Technical secondary school and high school58 (27.2)48 (32.9)18 (28.1)124College and above51 (23.9)27 (18.5)12 (18.8)90Marital statusDivorced/widowed31 (14.6)13 (8.9)35 (54.7)7966.208a<0.001Married182 (85.4)133 (91.1)29 (45.3)344Occupation statusRetired98 (46.0)54 (37.0)36 (56.3)1887.356a0.118Farmer87 (40.8)68 (46.6)22 (34.4)177Other28 (13.1)24 (16.4)6 (9.4)58Place of residenceUrban142 (66.7)95 (65.1)37 (57.8)2740.534b0.766County town9 (4.2)11 (7.5)9 (14.1)29Township24 (11.3)11 (7.5)9 (14.1)44Rural38 (17.8)29 (19.9)9 (14.1)76Medical payment methodUrban employee medical insurance106 (49.8)63 (43.2)34 (53.1)2032.315a0.314Urban and rural resident medical insurance107 (50.2)83 (56.8)30 (46.9)220Average monthly household incomeBelow 2000 yuan79 (37.1)53 (36.3)21 (32.8)1530.302b0.8602000–5000 yuan71 (33.3)56 (38.4)25 (39.1)152Above 5000 yuan63 (29.6)37 (25.3)18 (28.1)118Smoking historyNever smoked171 (80.3)100 (68.5)51 (79.7)3226.364b0.041Currently smoking24 (11.3)30 (20.5)7 (10.9)61Former smoker18 (8.5)16 (11.0)6 (9.4)40History of alcohol consumptionNever consumed alcohol163 (76.5)94 (64.4)45 (70.3)3025.692b0.058Currently drinking alcohol30 (14.1)34 (23.3)10 (15.6)74Quit drinking20 (9.4)18 (12.3)9 (14.1)47Exercise frequencyAlmost never exercise78 (36.6)51 (34.9)32 (50.0)1614.250b0.119Occasional exercise81 (38.0)47 (32.2)17 (26.6)145Regular exercise54 (25.4)48 (32.9)15 (23.4)117Length of hospital stay (days)≥1652 (24.4)51 (34.9)40 (62.5)14320.360b<0.0018~15121 (56.8)81 (55.5)12 (18.8)214<840 (18.8)14 (9.6)12 (18.8)66Years since heart failure diagnosis<5136 (63.8)95 (65.1)37 (57.8)2680.627b0.7315~946 (21.6)24 (16.4)17 (26.6)87≥1031 (14.6)27 (18.5)10 (15.6)68NYHA classificationIV24 (11.3)37 (25.3)39 (60.9)10063.841b<0.001III114 (53.5)90 (61.6)18 (28.1)222II75 (35.2)19 (13.0)7 (10.9)101Number of medication types≥644 (20.7)21 (14.4)35 (54.7)10027.787b<0.0013~5119 (55.9)104 (71.2)22 (34.4)245<350 (23.5)21 (14.4)7 (10.9)78HypertensionNo59 (27.7)38 (26.0)17 (26.6)1140.129a0.938Yes154 (72.3)108 (74.0)47 (73.4)309Coronary heart diseaseNo81 (38.0)46 (31.5)18 (28.1)1452.903a0.234Yes132 (62.0)100 (68.5)46 (71.9)278ArrhythmiaNo126 (59.2)93 (63.7)39 (60.9)2580.752a0.687Yes87 (40.8)53 (36.3)25 (39.1)165Living alone statusYes37 (17.4)63 (43.2)41 (64.1)14157.950a<0.001No176 (82.6)83 (56.8)23 (35.9)282PNI total scoreMalnutrition group63 (29.6)59 (40.4)49 (76.6)17145.111a<0.001Well-nourished group150 (70.4)87 (59.6)15 (23.4)252SAS total scoreSevere Anxiety21 (9.9)8 (5.5)34 (53.1)6347.768b<0.001Moderate anxiety26 (12.2)15 (10.3)9 (14.1)50Mild Anxiety60 (28.2)55 (37.7)7 (10.9)122Normal106 (49.8)68 (46.6)14 (21.9)188APGAR total scoreSevere family dysfunction15 (7.0)17 (11.6)25 (39.1)5786.317b<0.001Moderate family dysfunction34 (16.0)75 (51.4)24 (37.5)133Good family functioning164 (77.0)54 (37.0)15 (23.4)233MLHFQ total scorePoor quality of life121 (56.8)96 (65.8)50 (78.1)2676.318b0.042Moderate quality of life85 (39.9)38 (26.0)7 (10.9)130Good quality of life7 (3.3)12 (8.2)7 (10.9)26Notes: Values are n (%). aChi-square test; bKruskal–Wallis H-test.

Medication types mainly include diuretics, β-blockers, statins, calcium channel blockers, digitalis preparations, etc. (excluding psychotropic medications). PNI, Prognostic Nutritional Index: Malnutrition <45 points, well-nourished ≥45 points.

SAS, Self-Rating Anxiety Scale: Normal <50 points; mild anxiety 50–59 points; moderate anxiety 60–69 points; severe anxiety ≥70 points. APGAR, Family APGAR index: Good family functioning 7–10 points; moderate dysfunction 4–6 points; severe dysfunction 0–3 points. MLHFQ, Minnesota Living with Heart Failure Questionnaire: Good quality of life <24 points; moderate quality of life 24–45 points; poor quality of life >45 points.

Univariate analysis showed that there were statistically significant differences among the three groups in sex, marital status, smoking history, length of hospital stay (days), NYHA cardiac function classification, number of medication types, living alone status, nutritional status (PNI total score), anxiety level (SAS total score), family functioning (APGAR total score), and quality of life (MLHFQ total score) (all P <0.05); while age, education level, occupational status, place of residence, medical payment method, monthly household income, drinking history, exercise frequency, years since heart failure diagnosis, and comorbidities such as hypertension, coronary heart disease, and arrhythmia showed no statistically significant inter-group differences (all P >0.05).

In the severe depression-related subgroup, the proportions of individuals who were female, divorced or widowed, had a longer length of hospital stay, had NYHA functional class IV, took a higher number of medication types, lived alone, had malnutrition, had severe anxiety, had severe family dysfunction, and had poor quality of life were significantly higher; In the mild depression subgroup, the proportion of married individuals was higher, and overall family functioning and quality of life were relatively better.

Multivariate Logistic Regression Analysis

Variables with P <0.05 in the univariate analysis [sex, marital status, smoking history, length of hospital stay (days), living alone status, PNI classification, SAS classification, APGAR score, MLHFQ score] were included in the model. Combined with clinical expertise, NYHA cardiac function classification and number of medication types were also included as independent variables. Standardized variable assignments are shown in Table 4.

With latent profile types of depressive symptoms as the dependent variable, P1 low-symptom full-dimension mild type was set as the reference group. P2 moderate type with prominent emotional features and P3 severe somatic-cognitive type were set as comparison groups for multinomial logistic regression analysis. Statistical results are presented in Table 5.Table 4Assignment of Independent VariablesIndependent VariableAssignment MethodSexFemale=1, Male=2 (reference)Marital statusDivorced/widowed=1, Married=2 (reference)Smoking historyNever smoked=1, Currently smoking=2, Former smoker=3 (reference)Length of hospital stay (days)≥16=1, 8–15=2, <8=3 (reference)NYHA classificationClass IV=0, Class III=1, Class II=2 (reference)Number of medication types≥6=1, 3–5=2, <3=3 (reference)Living alone statusYes=0, No=1 (reference)PNIMalnutrition group=0, Well-nourished group=1 (reference)SASSevere anxiety=1, Moderate anxiety=2, Mild anxiety=3, Normal=4 (reference)APGARSevere family dysfunction=1, Moderate family dysfunction=2, Good family functioning=3 (reference)MLHFQPoor quality of life=1, Moderate quality of life=2, Good quality of life=3 (reference)Notes: The category marked as (reference) was set as the reference group for multinomial logistic regression analyses.

All categorical predictors were dummy-coded prior to being entered into the regression model.Abbreviations: PNI, Prognostic Nutritional Index; SAS, Self-Rating Anxiety Scale; APGAR, Family APGAR index; MLHFQ, Minnesota Living with Heart Failure Questionnaire.Table 5Multivariate Analysis of Different Latent Profile Types of Depressive Symptoms in Elderly Patients with Chronic Heart Failure (N= 423)VariableP2 vs P1(Emotion-Predominant Moderate Profile vs Low-Symptom Mild Profile)P3 vs P1(Somatic-Cognitive Severe Profile vs Low-Symptom Mild Profile)BSEχ2POR95% CIBSEχ2POR95% CIIntercept−2.1040.8845.6670.017−6.4001.89011.4650.001SexFemale0.2620.2970.7790.3771.2990.727~2.3231.1080.6363.0330.0823.0280.87~10.531MaleMarital statusDivorced/ widowed−0.6210.4232.1610.1420.5370.235~1.232.2220.62812.524< 0.0019.2262.695~31.586MarriedSmoking historyNever smoked−0.7110.4632.3590.1250.4910.198~1.217−0.3961.0390.1450.7030.6730.088~5.159Currently smoking−0.0060.5270.0000.9900.9940.354~2.7930.4751.2450.1460.7031.6090.14~18.476Former smokerLength of hospital stay (days)≥160.6840.4482.3330.1271.9830.824~4.7711.3970.8702.5800.1084.0420.735~22.2278–150.4730.4071.350.2451.6050.723~3.563−0.3160.8370.1420.7060.7290.141~3.762<8NYHA classificationIV class1.4590.44010.9710.0014.3011.814~10.1962.6840.81510.8600.00114.6462.968~72.287III class0.8940.3476.6370.0102.4441.238~4.8230.1990.7480.0710.7901.2200.282~5.281II classNumber of medication types≥6−0.3130.4420.5020.4790.7310.307~1.7391.7670.7705.2720.0225.8561.295~26.4723–50.5290.3542.2340.1351.6970.848~3.397−0.3680.8630.1820.6700.6920.128~3.755<3Living alone statusYes1.0600.29213.213< 0.0012.8871.63~5.1131.3100.5365.9670.0153.7061.296~10.601NoPNIMalnutrition group0.5740.2824.1520.0421.7751.022~3.0822.5820.61017.939< 0.00113.2234.003~43.672Well-nourished groupSASSevere Anxiety−0.6850.5251.7030.1920.5040.18~1.411.5910.6885.3480.0214.9111.275~18.921Moderate anxiety−0.4180.4440.8850.3470.6590.276–1.5720.3290.8550.1480.701.3900.26–7.421Mild Anxiety0.2790.3070.8240.3641.3210.724–2.411−1.6960.8294.1810.0410.1830.036–0.932NormalAPGARFamily functioningSevere impairment1.3520.4528.9660.0033.8651.595~9.3653.0880.80114.859< 0.00121.9234.561~105.368Family functioningModerate impairment1.8340.29438.996< 0.0016.2593.52~11.1311.8560.6468.2430.0046.3961.802~22.701Family functioningGoodMLHFQPoor quality of life−0.4480.5630.6320.4270.6390.212~1.927−1.1020.9841.2550.2630.3320.048~2.284Moderate quality of life−0.9130.5902.3910.1220.4010.126~1.277−2.6671.2204.7830.0290.0690.006~0.758Good quality of lifeNotes: P1 (low-symptom full-dimension mild profile) served as the reference category for multinomial logistic regression analyses. P2 = emotion-predominant moderate profile; P3 = severe somatic-cognitive profile. B, regression coefficient; SE, standard error; χ2, Wald chi-square; OR, odds ratio; 95% CI, 95% confidence interval.

Multicollinearity diagnostics were performed; no substantial multicollinearity was detected among predictors (all VIF <5). Overall model fit: −2 log-likelihood = 456.686, Model χ2 = 370.484, df = 38, P < 0.001; total number of parameters in the final model = 38. Pseudo-R2: Cox-Snell = 0.583, Nagelkerke = 0.675, McFadden = 0.439.Abbreviations: PNI, Prognostic Nutritional Index; SAS, Self-Rating Anxiety Scale; APGAR, Family APGAR index; MLHFQ, Minnesota Living with Heart Failure Questionnaire; VIF, variance inflation factor.

P2 Moderate Depression compared to P1 Mild Depression: NYHA cardiac function classification IV (OR=4.301, 95% CI: 1.814–10.196), NYHA cardiac function classification III (OR=2.444, 95% CI: 1.238–4.823), living alone (OR=2.887, 95% CI: 1.630–5.113), PNI Malnutrition group (OR=1.775, 95% CI: 1.022–3.082), severe family dysfunction (OR=3.865, 95% CI: 1.595–9.365), and moderate family dysfunction (OR=6.259, 95% CI: 3.520–11.131) were independently associated with higher odds of membership in the moderate emotion-predominant profile. The differences between groups were all statistically significant (all P <0.05).

When comparing P3 severe depression to P1 mild depression, variables independently associated with higher odds of membership in the severe somatic-cognitive profile included: divorced/widowed (OR=9.226, 95% CI: 2.695–31.586), NYHA cardiac function classification IV (OR=14.646, 95% CI: 2.968–72.287), number of medication types ≥6 (OR=5.856, 95% CI: 1.295–26.472), living alone (OR=3.706, 95% CI: 1.296–10.601), PNI malnutrition group (OR=13.223, 95% CI: 4.003–43.672), severe anxiety (OR=4.911, 95% CI: 1.275–18.921), severe family dysfunction (OR=21.923, 95% CI: 4.561–105.368), and moderate family dysfunction (OR=6.396, 95% CI: 1.802–22.701) (all P<0.05); Mild anxiety (OR=0.183, 95% CI: 0.036–0.932) and moderate quality of life (OR=0.069, 95% CI: 0.006–0.758) were independently associated with lower odds of belonging to the severe somatic-cognitive profile (all P <0.05).

Sex, smoking history, length of hospital stay, SAS moderate/normal anxiety, MLHFQ poor/good quality of life, and other indicators were associated with the latent profile types of depressive symptoms in univariate analysis. After adjusting for confounding factors via multinomial logistic regression, the between-group differences for these variables were not statistically significant (all P >0.05), indicating that they are not independent influencing factors for worsening depressive severity in elderly patients with chronic heart failure.

It should be noted that several odds ratios comparing the smallest severe somatic-cognitive subtype (P3, n=64) against the reference P1 group exhibited relatively wide 95% confidence intervals. This is attributable to the limited sample size of this subgroup. Even for statistically significant results, these point estimates need to be interpreted cautiously considering the inherent statistical uncertainty.

Discussion

This study identified, through latent profile analysis, that depressive symptoms in elderly patients with chronic heart failure exhibit significant heterogeneity and can be classified into three latent subtypes: low-symptom full-dimension mild type (50.35%), moderate type with prominent emotional features (34.52%), and severe somatic-cognitive type (15.13%). This classification reveals differences in symptom dimensions that are difficult to discern from a traditional homogeneous perspective, thereby providing an empirical basis for clinically implementing stratified screening and precise interventions.

Latent Profile Characteristics of Depressive Symptoms in Elderly Patients with Chronic Heart Failure

Previous studies on heart failure complicated by depression have predominantly treated the total depression score as a continuous variable in overall correlation analyses. Only a few investigations have examined depression subtypes specifically within the heart failure population, and existing typologies largely focus on young and middle-aged cohorts. Targeted, stratified research on elderly hospitalized patients with chronic heart failure remains notably limited.11,12

This study conducted latent profile analysis based on all SDS item scores and identified three depressive subtypes with distinct characteristics. The first subtype was a low-symptom, all-dimension mild subtype, characterized by generally low scores on depressive items across all dimensions, concealed depressive manifestations, and a high clinical risk of missed diagnosis due to being masked by the somatic symptoms of heart failure. The second subtype was an emotional-motivational prominent moderate subtype, with core features of depressed mood, loss of interest, and fatigue; somatic depressive symptoms were relatively mild, and this group exhibited a higher potential for response to psychological and behavioral interventions.

The third type is a severe subtype combining somatic and cognitive symptoms, with superimposed severe somatic discomfort and negative cognition. This group often has concomitant grade IV cardiac insufficiency, malnutrition, and severe family functioning deficit, and constitutes the key intervention population with the highest clinical psychological risk. The symptom distribution patterns of the three subtypes align with the multimorbidity and somatic symptom predominance of heart failure in the elderly, overcoming the limitation of a single total depression score in distinguishing symptom heterogeneity.

One important methodological consideration for the present study is symptom overlap between intrinsic somatic manifestations of chronic heart failure and somatic items within the SDS scale. SDS items evaluating sleep disturbance, palpitations, fatigue, and constipation may partly reflect physical symptoms of heart failure itself rather than pure depressive psychopathology. Such symptom overlap may inflate scores on somatic-related SDS items, particularly for participants assigned to the severe somatic-cognitive subtype identified in this study.

Analysis of Independent Associated Factors of Depressive Symptoms in Elderly Patients with Chronic Heart Failure

Clinical Condition-Related Risk Factors: Cardiac Function Classification and Polypharmacy

NYHA cardiac function classification is closely associated with depression subtypes; the higher the classification, the stronger the association with moderate and severe depression. The NYHA class IV group has the highest proportion of the severe somatic-cognitive type. In NYHA class IV patients, myocardial contractile function is severely impaired, and somatic discomforts such as dyspnea, generalized edema, and activity limitation persist chronically.

Somatic discomfort and negative emotions often coexist;32,33 However, as this study was cross-sectional, the temporal sequence of these effects could not be determined. Furthermore, severe cardiac insufficiency can reduce cerebral perfusion and cause neuroendocrine dysregulation, disrupting the secretion of mood-regulating neurotransmitters such as 5-hydroxytryptamine and dopamine, thereby establishing a physiological link characterized by the coexistence of somatic discomfort and emotional disturbances.34

Length of hospital stay showed between-group differences in univariate analysis, but after adjusting for confounders including NYHA classification and living alone, this difference was no longer significant, suggesting that length of hospital stay is merely a concomitant marker of cardiac function and social support, rather than an independent factor for depression.

Number of medication types ≥6 was positively associated with severe depression. Polypharmacy populations are more prone to adverse drug reactions: β-blockers cause fatigue, diuretics lead to nocturia, and superimposed somatic symptoms such as dizziness and gastrointestinal discomfort; somatic discomfort is positively associated with depressive symptoms.35,36

Risk Factors Related to Living Status: Marital Status, Living Alone Status

Divorced/widowed status was independently and positively associated with severe depression. Spouses are the core source of emotional and life care for elderly patients. Patients lacking spousal support are chronically affected by the course of heart failure, and negative emotions such as loneliness and helplessness often co-occur with severe depression, showing a significant correlation.

This result is consistent with previous studies.32

Living alone is a common associated factor for both moderate and severe depression, with a stronger association with severe depression. Elderly patients living alone lack daily interpersonal interactions, find it difficult to quickly obtain external help when sudden illness strikes, and experience more prominent worries about their condition; living alone is positively correlated with depression severity.37,38

Risk Factors Related to Physical Condition: Nutritional Status

Malnutrition is significantly associated with both moderate and severe depression, with the strength of association increasing with depression severity, and it is most prominent in the severe somatic-cognitive type. Existing research confirms that decreased appetite, gastrointestinal congestion, and metabolic disorders induced by heart failure increase the risk of malnutrition; Insufficient intake of protein and micronutrients can disrupt the synthesis of dopamine and 5-hydroxytryptamine, thereby explaining the physiological link between nutrition and mood;39–41 This cross-sectional data reveals that patients with malnutrition as assessed by PNI scored significantly higher on the depression scale. Heart failure, somatic discomfort, nutritional deficiency, and depression often coexist synchronously, indicating only an association rather than causality.

The population characteristics are consistent with those reported in previous cross-sectional studies of heart failure.42

Social Support-Related Risk Factors: Family Functioning

Family dysfunction displayed a gradient association: compared with individuals with good family functioning, moderate and severe family dysfunction were both significantly positively correlated with moderate-to-severe depression, and those with severe family dysfunction exhibited a higher risk of severe depression. Individuals with intact family functioning can receive adequate emotional, practical, and financial support, thereby alleviating the psychological stress of the disease; Patients lacking family support experience greater helplessness when confronting illness, and disease fluctuations and depressive symptoms are more likely to co-occur.43–45

Protective Association of Mood and Quality of Life with Depression

Mild anxiety showed a negative statistical association with membership in the severe somatic-cognitive depressive subtype in this cross-sectional analysis. We cannot infer causal protective effects from this observational association; this correlation may reflect illness-related adaptive worry or unmeasured confounding variables. Moderate quality of life also exhibited an inverse statistical association with severe somatic-cognitive subtype membership.

Alleviation of somatic discomfort and maintenance of basic self-care ability may reduce disease-related psychological stress and negative emotions, which is consistent with existing research findings.46

These seemingly protective-like associations for mild anxiety and moderate quality of life should both be interpreted cautiously. The subgroup with good quality of life only contained 26 participants, and the number of participants with mild anxiety within the severe-symptom subgroup was limited; thus, effect-size estimates may be unstable.

Nursing Intervention Strategies

Based on the characteristics of the three depression subtypes, as well as independent associated factors observed in this cross-sectional study, the following stratified intervention proposals are put forward for future prospective evaluation. These proposals have not been empirically tested within the present investigation. Clinical considerations for potential stratified precision nursing are as follows:

Corresponding interventions targeting cardiac function classification and polypharmacy: NYHA class III/IV and taking ≥6 medications are common risk factors for moderate-to-severe depression. In clinical practice, a systematic review of all oral medications should be conducted at admission and follow-up visits, in collaboration with physicians to deprescribe non-essential adjuvant medications; For patients on long-term β-blockers and diuretics, provide early education about adverse reactions such as fatigue and nocturia, and implement fatigue management and nighttime toileting assistance to alleviate negative emotions induced by somatic discomfort; Routine assessment of somatic distress in patients with severe NYHA class IV disease, along with timely alleviation of symptoms such as shortness of breath and edema, may reduce the risk of depression in this population.

Psychological intervention for individuals living alone, divorced, or widowed: Divorced, widowed, and living alone are high-risk groups for severe depression. Upon admission, screen and flag high-risk patients and establish a dedicated follow-up file; during hospitalization, set fixed one-on-one communication periods to alleviate feelings of loneliness and helplessness; upon discharge, coordinate with community healthcare providers for regular home visits to reduce prolonged solitude. For those without family caregivers, coordinate short-term companionship from community volunteers to fill the gaps in emotional and daily care.

Standardized whole-course nutritional intervention: Malnutrition is an independent risk factor for depression. The PNI nutritional screening should be incorporated into the routine admission assessment for elderly patients with heart failure, with dynamic monitoring of albumin and lymphocyte counts. For malnourished patients, formulate an individualized high-protein dietary plan to supplement raw materials for neurotransmitter synthesis, thereby concurrently improving the coexisting state of physical frailty and depression.

Family function optimization care: Moderate to severe family dysfunction can significantly elevate the risk of depression. During hospitalization, provide concurrent health education to patients and their families, teaching heart failure care and emotional soothing skills; guide family members to offer daily positive encouragement. For families with conflicts or insufficient care willingness, provide individual communication and counseling to optimize the family support atmosphere.

Protective interventions targeting emotional well-being and quality of life: Mild anxiety does not require excessive intervention; patients can be positively guided to actively monitor their condition and seek standardized medical care. For patients with limited physical function and low quality of life, prioritizing the optimization of symptomatic treatment for heart failure and maintaining basic self-care ability can help reduce the risk of severe depression.

Stratified and differentiated interventions based on depression subtypes: (1) Low-symptom full-dimension mild type (50.35%): Focus on early screening and prevention. All elderly heart failure patients aged 65 years and older should complete the SDS psychological screening upon admission and discharge. Routine health education on heart failure and emotional health should be conducted, and regular brief one-on-one psychological interviews should be carried out to delay the progression of negative emotions.47–49 (2) Moderate type with prominent emotional features (34.52%): This represents the golden window for psychological intervention, with behavioral activation nursing as the core.

A low-intensity exercise program is formulated based on cardiac function tolerance, and structured emotional counseling is implemented simultaneously to enhance psychological adjustment and self-management abilities.50–52 (3) Severe Somatic-Cognitive Type (15.13%): Implement multidisciplinary collaborative intervention, integrating cardiovascular, nutritional, and psychological specialties; Simultaneously optimize heart failure treatment and individualized nutritional intervention, supplemented with cognitive behavioral modification, to alleviate the coexistence of somatic discomfort and negative cognition.53,54

In summary, high NYHA classification, polypharmacy, living alone and widowed, malnutrition, and family functioning deficits are all modifiable and controllable related factors. Clinically, psychological screening can be routinely implemented, with focused follow-up for high-risk groups, and individualized care can be provided based on depression subtypes, shifting from homogeneous care to stratified precision intervention, which is expected to improve the psychosomatic comorbidity manifestations in this population.

Research Limitations and Future Directions

This study has several limitations. First, this study employed a single-center convenience sampling design, with the sample drawn from a single medical institution, which may limit the generalizability of the findings to populations in other regions or medical institutions of different levels. Second, this study used a cross-sectional design, which can only reveal statistical associations between variables, cannot infer causal direction, and cannot reflect the dynamic trajectory of depressive symptoms.

Third, living alone was included in the analysis only as a binary variable, without incorporating refined indicators such as the intensity of social support and the frequency of interpersonal contact, which may underestimate the buffering effect of social contact on depressive symptoms. Fourth, the latent profile analysis was based solely on items from the SDS, without incorporating psychosocial variables such as psychological resilience and coping styles; thus, the classification system requires further refinement.

In addition, participants were assigned to latent classes via maximum posterior probability, which cannot fully account for uncertainty in latent-class membership. Although average posterior probabilities exceeded 0.90 for all subgroups and indicated good classification certainty, this method remains less optimal than the three-step approach. The severe somatic-cognitive profile contained a relatively small sample size (n=64); despite variable selection guided by both clinical theory and univariate findings, unstable parameter estimates and potential overfitting cannot be completely ruled out in the multinomial logistic regression model.

Furthermore, several key objective markers of heart failure severity, including LVEF, HF phenotype (HFrEF/HFmrEF/HFpEF), BNP/NT-proBNP, renal function, prior HF hospitalization, congestion status, and detailed guideline-directed medical therapy information, were not systematically available in our dataset, which may lead to residual confounding. Finally, continuous indicators including PNI, SAS, Family APGAR, and MLHFQ were converted to categorical variables using established clinical cut-offs to enhance clinical interpretability. However, categorization inevitably results in loss of continuous-scale information and may increase the risk of unstable statistical estimates.

Moreover, somatic items of the SDS may overlap with intrinsic physical symptoms of chronic heart failure, which could confound the measurement of depressive symptom profiles. In addition, all nursing-related intervention suggestions presented in the discussion are hypothesis-generating proposals and have not been validated in the current study.

Future research could consider: (1) conducting multicenter, large-sample longitudinal cohort studies to verify the stability of the depression subtype classification identified in this study and to explore the dynamic patterns of depressive symptoms and their causal pathways; (2) expanding the assessment dimensions by including multi-level indicators such as physical function, psychological resources, and social support, refining assessments related to living alone and social contact, and constructing a more comprehensive framework for classifying depression heterogeneity; (3) Based on the depression subtypes identified in this study and their characteristic influencing factors, design and validate the effectiveness of stratified intervention strategies to provide a high-quality evidence base for the precise prevention and control of depressive symptoms and individualized care in elderly patients with chronic heart failure.

Conclusion

Depressive symptoms in elderly patients with chronic heart failure exhibit significant heterogeneity and can be classified into three latent profile subtypes based on the present three-profile LPA model and PNI-based nutritional assessment. The low-symptom mild subtype accounts for the highest proportion. In this cross-sectional study, cardiac dysfunction, polypharmacy, being divorced or widowed, living alone, malnutrition, and family dysfunction were associated with higher odds of membership in more severe depressive symptom profiles.

By contrast, mild anxiety and moderate quality of life were associated with lower odds of belonging to the severe somatic-cognitive profile. These represent observed statistical associations derived from the current analytical scheme; causal or temporal relationships cannot be inferred, and the latent profile classification together with PNI-based nutritional stratification await further external validation in independent study populations.

In clinical practice, recognition of these distinct depressive symptom subtypes may inform subtype-oriented screening and assessment. Targeted interventions addressing these associated factors are only hypothetical considerations. The present study provides no empirical evidence for intervention effectiveness, and the clinical benefits of these strategies remain to be verified in future prospective investigations.