Impact of post-stroke depression on the prognosis of stroke-associated pneumonia and analysis of predictive factors: A retrospective study of 186 patients

Abstract

BACKGROUND

Stroke-associated pneumonia (SAP) is the most common infectious complication after acute cerebral infarction, occurring in 7%-38% of hospitalized stroke patients and independently associated with higher mortality rates and extended hospital stays. Post-stroke depression (PSD) is present in approximately 30%-50% of stroke survivors and may additionally impair immunity, decrease compliance with rehabilitation, and worsen the clinical trajectory of SAP. However, the association of concomitant PSD with SAP outcomes and the determinants of poor prognosis in this population are not well defined.

AIM

To investigate the impact of PSD on outcomes in patients with SAP and identify independent predictors of poor prognosis in patients with SAP and comorbid depression.

METHODS

This study included 186 patients with SAP after acute cerebral infarction who were admitted to the Department of Neurological Intensive Rehabilitation, Cangzhou Hospital of Integrated Traditional Chinese and Western Medicine, Hebei Province, China, between December 2014 and December 2025. Patients with a recorded Hamilton Depression Rating Scale score 2 weeks after stroke were divided into a depression group (n = 92) and non-depression group (n = 94). Overall data were gathered, including demographics, comorbidities, stroke characteristics, SAP severity indices, treatment modalities (including traditional Chinese medicine intervention like warm moxibustion therapy), laboratory biomarkers [procalcitonin, C-reactive protein, albumin, lymphocyte count, serum brain-derived neurotrophic factor (BDNF), interleukin-6 and cortisol] and outcome measures. Pulmonary infection resolution time, total hospital stays, mechanical ventilation duration, 28-day mortality, and functional outcomes measured by the modified Rankin Scale (mRS) were primary endpoints. Univariate and multivariable logistic regression analyses were performed to identify independent predictors of an adverse prognosis.

RESULTS

Comorbid PSD was present in 92 (49.5%) of the 186 SAP patients. Patients in the depression group had significantly worse pulmonary infection resolution time (16.8 ± 4.5 days vs 11.2 ± 3.2 days, P < 0.001), longer hospitalization stay (28.5 ± 7.8 days vs 19.3 ± 5.6 days, P < 0.001), longer duration on mechanical ventilation (8.5 vs 5.2 days, P < 0.001), higher mortality within the following 28-day timeframe (18.5% vs 7.4%, P = 0.024), and worse functional outcomes defined by an mRS ≥ 3 at discharge (P = 0.001). Seven independent adverse prognosis predictors were identified by multivariable logistic regression modeling: Comorbid PSD [odds ratio (OR) = 3.12, 95% confidence interval (95%CI): 1.85-5.26, P < 0.001], Acute Physiology and Chronic Health Evaluation II score ≥ 18 (OR = 3.45, 95%CI: 1.92-6.20, P < 0.001), serum albumin < 30 g/L (OR = 2.78, 95%CI: 1.55-4.98, P = 0.001), mechanical ventilation > 7 days (OR = 2.65, 95%CI: 1.48-4.75, P = 0.001), multi-drug resistant organism infection (OR = 2.52, 95%CI: 1.42-4.47, P = 0.002), BDNF < 14.5 ng/mL (OR = 2.35, 95%CI: 1.28-4.32, P = 0.006), and age ≥ 75 years (OR = 1.98, 95%CI: 1.15-3.41, P = 0.014). The area under the curve was 0.91 (95%CI: 0.86-0.96), and the combined predictive model had a sensitivity of 84.6% and a specificity of 87.3%. Patients that received integrated traditional Chinese and Western medicine treatment, including warm moxibustion, had a faster time of infection resolution (13.2 ± 3.8 days vs 15.5 ± 4.5 days, P = 0.012) and improved functional outcomes compared with controls.

CONCLUSION

Comorbid PSD further worsened the clinical outcomes of patients with SAP with a longer time to infection resolution, longer length of stay, increased mortality, and poor functional recovery. A predictive model based on the combination of depression status, disease severity, nutritional indicators and serum biomarkers had high diagnostic accuracy in identifying at-risk patients. Warm moxibustion, which is the simple application of heat and acupuncture point stimulation, improved time of infection resolution and functional outcomes. However, approaches that integrate traditional Chinese medicine and similar Western medicine concordantly may be more efficacious. These findings support the rationale for early identification and management of SAD in affected patients with SAP as a means to improve outcomes.

Key Words: Stroke-associated pneumonia; Post-stroke depression; Prognosis; Predictive factors; Warm moxibustion; Brain-derived neurotrophic factor

Core Tip: This retrospective study of 186 patients with stroke-associated pneumonia (SAP) demonstrates that co-morbid post-stroke depression is highly prevalent (approximately 50% of SAP) and an independent predictor of adverse prognosis, including markedly longer time to resolution of infection, prolonged hospital stays, worse functional recovery at discharge, and higher 28-day mortality. A robust seven-variable predictor model incorporating depression status, disease severity indices (Acute Physiology and Chronic Health Evaluation II), nutritional biomarkers (albumin), infectious parameters (multi-drug resistant organism status), neurotrophic factors (brain-derived neurotrophic factor), mechanical ventilation duration, and age, yielded excellent discriminative power with an area under the curve of 0.91, representing a clinically actionable framework for early identification of high-risk SAP patients. These findings suggest that routine depression screening may be important in SAP management, and integrated traditional Chinese and Western medicine approaches, especially warm moxibustion, may have beneficial effects.

INTRODUCTION

Stroke-associated pneumonia (SAP), or pneumonia that occurs in the first 7 days after stroke onset, is one of the most common and serious medical complications for patients with acute cerebral infarction. Epidemiological data show that SAP occurs in 7%-38% of hospitalized stroke patients, with highly variable incidence rates depending on the stroke severity, patient age, and the use of diagnostic criteria[1-4]. SAP alone is independently correlated with a three-fold increase in 30-day mortality, increased intensive care unit (ICU) admission and hospital length of stay (LOS), and markedly elevated healthcare costs, as well as significantly impaired functional outcomes at discharge and during long-term follow-up[5-7]. Stroke-induced immunosuppression, aspiration due to dysphagia, decreased cough reflex and the central nervous system-mediated changes in pulmonary defense mechanisms contribute to SAP pathogenesis[8-10].

Approximately one third of patients suffer from depression after a stroke; this increases to approximately 50% within 1 year after the event[11-13]. The harmful effects of post-stroke depression (PSD) on recovery after a stroke are already known, including impaired functional rehabilitation[14], poor treatment adherence[15], increased recurrence of cerebral vascular disease[16] and increased mortality. From a pathophysiological perspective, PSD is characterized by severe dysregulation of the hypothalamic-pituitary-adrenal (HPA) axis, increased pro-inflammatory cytokines, decreased brain-derived neurotrophic factor (BDNF) levels and impaired cellular immunity[17-19], all having direct relevance to susceptibility toward infection and in its resolution.

The relationship between PSD and SAP is clinically relevant but under-investigated. In theory, the PSD-related immune dysregulation [high cortisol levels, increased interleukin-6 (IL-6), decreased lymphocyte activity, and lower natural killer cell function], can compromise host defense against pulmonary infection[20-22]. Moreover, behavioral adaptations related to depression, including cessation of exercise, malnourishment, cumulative swallowing failure, and decrease in cough strength, could enhance aspiration and slow recovery from pneumonia[23,24]. However, there is a limited number of studies and they have not systematically assessed the relevant prognostic effect of comorbid PSD on SAP outcomes or identified factors that predict poor prognosis among this double-burden population.

Traditional Chinese medicine (TCM), especially warm moxibustion (Wen He Jiu) therapy, has been gradually integrated as part of rehabilitation protocols for stroke patients in China. According to initial hypotheses, warm moxibustion may have qi- and blood-promoting circulation effect[25], anti-inflammatory activity[26], an immunomodulatory effect[27] and increased gastrointestinal motility, which is fitting as an SAP management mechanism. However, the evidence is still lacking that its use as an adjunctive treatment for SAP, particularly in patients with comorbid depression, would be effective.

The present study aimed to: (1) Evaluate the clinical outcomes of SAP patients with or without comorbid PSD; (2) Explore the independent predictors of adverse prognosis using multivariable analysis; and (3) Assess the performance of serum biomarkers for prognostic risk stratification and traditional Chinese and Western medicine treatment, including warm moxibustion, that may be beneficial in this population. These findings aim to provide an evidence-based basis for a multidisciplinary approach to SAP patients suffering from co-morbid depression.

MATERIALS AND METHODS

Study design and participants

This retrospective study retrieved clinical data from SAP patients who were admitted to the Department of Neurological Intensive Rehabilitation, Cangzhou Hospital of Integrated Traditional Chinese and Western Medicine, Hebei Province, China, from December 2014 to December 2025 after ischemic stroke. The study protocol was approved by the Ethics Committee of Cangzhou Hospital of Integrated Traditional Chinese and Western Medicine, Hebei Province, China (Approval No. CZX2024-KY052-02), and written informed consent was obtained from all participants or their legal representatives.

Inclusion criteria: (1) Age ≥ 18 years; (2) Acute ischemic stroke confirmed by cranial computed tomography and/or magnetic resonance imaging within 72 hours of admission; (3) Diagnosis of SAP according to the modified Centers for Disease Control and Prevention criteria, with pneumonia onset occurring within 7 days following stroke; (4) Availability of complete clinical, imaging, laboratory, and follow-up data; and (5) Ability to undergo depression screening 2 weeks after stroke.

Exclusion criteria: (1) Hemorrhagic stroke or subarachnoid hemorrhage; (2) Community-acquired pneumonia before and at stroke onset; (3) Existing diagnosis of major depressive disorder or other severe psychiatric illness that required continuous medication; (4) Disturbance of consciousness was severely impaired as assessed by Glasgow Coma Scale (GCS) ≤ 8, making depression assessment impossible; (5) Active pulmonary tuberculosis or pulmonary malignancy; (6) Immunosuppressive treatment within 3 months before admission; or (7) Death expected less than 48 hours after admission.

In total, 238 patients were screened (Figure 1). The final cohort included 186 patients after 52 were excluded based on the above exclusion criteria: 15 for failure to complete depression assessment, 11 for history of psychiatric disorders prior to hospitalization, 8 for diagnosis of community acquired pneumonia, 7 with missing follow up data, 5 with associated pulmonary malignancy, 4 on immunosuppressive therapy, and 2 expected to die within 48 hours.

Figure 1 Receiver operating characteristic curves comparing the diagnostic performance of the combined predictive model and individual predictors for adverse prognosis in stroke-associated pneumonia patients. The combined model integrating clinical, biochemical, and microbiological parameters achieved the highest area under the curve (AUC; 0.91), followed by Acute Physiology and Chronic Health Evaluation II score alone (AUC = 0.82). All curves were significantly superior to the reference line (all P < 0.001). AUC: Area under the curve; APACHE II: Acute Physiology and Chronic Health Evaluation II; PSD: Post-stroke depression; BDNF: Brain-derived neurotrophic factor; MV: Mechanical ventilation.

Assessment and grouping

Depression [Hamilton Depression Rating Scale (HAMD-17)] was assessed 2 weeks (± 3 days) post-stroke in all patients blinded to clinical data and SAP outcomes by a trained psychiatrist. Depression was defined as a HAMD-17 score ≥ 7, consistent with standard definitions used in PSD research[11,12]. As a supplementary self-report measure, the Patient Health Questionnaire 9 (PHQ-9) was administered to evaluate concordance with HAMD-17-based group classification. PHQ-9 scores showed substantial agreement with the HAMD-17-derived groupings (κ = 0.74, indicating good concordance); however, formal cross-validation results are not reported in this retrospective analysis because PHQ-9 data were unavailable for 23 patients (12.4%), precluding a complete-case concordance analysis. Therefore, the HAMD-17 clinician-rated scale was therefore used as the sole instrument for group assignment. Patients were divided into two groups: Depression (n = 92; 49.5%) or non-depression (n = 94; 50.5%). Depression severity was subclassified as mild (HAMD-17: 7-16; n = 42, 45.7%), moderate (17-23; n = 33, 35.9%) and severe (≥ 24; n = 17, 18.5%).

The primary outcome was a composite adverse prognosis, defined as not having reached at least one of the following: (1) 28-day all-cause mortality; (2) Time to pulmonary infection resolution > 21 days; (3) Need for prolonged mechanical ventilation (> 14 days); or (4) Modified Rankin Scale (mRS) score ≥ 4 at hospital discharge. Among the 186 patients, adverse prognosis criteria were fulfilled in 78 cases (41.9%).

Data collection

Demographic and clinical information upon admission included: Age, sex, body mass index (BMI), education level, smoking history, alcohol consumption history, medical comorbidities (hypertension, diabetes mellitus, coronary heart disease, chronic obstructive pulmonary disease, chronic kidney disease, and previous stroke), stroke severity as measured by the National Institutes of Health Stroke Scale (NIHSS) score, GCS dysphagia status using the standard swallowing assessment, and nutritional risk using the Nutritional Risk Screening 2002 tool. Social support was assessed using the Social Support Rating Scale (SSRS).

SAP-specific data: Time from stroke onset to pneumonia diagnosis, pneumonia severity determined by the CURB-65 score, and the Acute Physiology and Chronic Health Evaluation II (APACHE II) score; sputum culture results and pathogenic organism identification; patterns of antimicrobial resistance; requirement for and duration of mechanical ventilation; ICU admission and length-of-stay data; as well as having undergone tracheostomy.

Treatment-related data recorded: Antibiotic regimen and duration, corticosteroid use, proton pump inhibitor use, enteral nutrition timing and route (oral or gastrointestinal tract), physical rehabilitation interventions, and TCM treatments. Specifically, a binary variable was created to indicate whether patients were treated with warm moxibustion therapy [moxibustion applied on Zusanli (ST36), Feishu (BL13), and Dazhui (GV14) acupoints 30 minutes per session, once a day for 14 consecutive days].

Laboratory biomarkers

Venous blood was drawn between 7:00 AM and 8:00 AM after overnight fasting at admission and on days 3, 7, and 14 after SAP diagnosis. Serum was separated within 30 minutes and stored at -80 °C until bulk analysis. The following biomarkers were measured: (1) Infection markers: Procalcitonin (PCT), C-reactive protein (CRP) and white blood cell count; inflammatory cytokines-IL-6, tumor necrosis factor-alpha (TNF-α), interleukin-10 (IL-10) by ELISA; (2) Neurotrophic factors: BDNF by ELISA; (3) Neuroendocrine markers: Morning serum cortisol; (4) Immune function-lavanthocyte counts, CD4⁺ T-cell counts and CD4/CD8 ratio by flow cytometry; and (5) Nutritional markers-serum albumin, prealbumin and transferrin, as well as routine biochemical parameters like hepatic and renal function, fasting glucose, and glycated hemoglobin.

Statistical analysis

SPSS 26.0 (IBM Corp.) and R 4.2.1 software were used for statistical analyses. Normal variables were reported as mean ± SD and compared by independent-samples t test; non-normally distributed were reported as median (interquartile range) and compared with a Mann-Whitney U test. Categorical variables were presented as n (%) and compared with χ² test or Fisher's exact test. Variables with P < 0.10 in univariate analysis were selected for multivariable logistic regression by forward stepwise selection. The results were expressed as odds ratios (ORs) and 95% confidence intervals (95%CIs). The diagnostic value of each predictor and the model was assessed via receiver operating characteristic (ROC) curve analysis, with area under the curve (AUC), sensitivity, specificity and optimal cutoff values calculated by the Youden index. Model calibration was evaluated with Hosmer-Lemeshow goodness-of-fit test. Kaplan-Meier survival curves were created to compare 28-day mortality among the groups, and log-rank test was used for significance. All tests were two sided; P < 0.05 was considered statistically significant.

RESULTS

Demographic and clinical characteristics

For the 186 enrolled patients with SAP, 92 (49.5%) were classified into the depression group and 94 (50.5%) were classified into the non-depression group. The demographic and clinical characteristics of each group are shown in Table 1. Compared with patients in the non-depression group, patients in the depression group exhibited a higher proportion of females (58.7% vs 36.2%, P = 0.002), higher NIHSS at admission (10.2 ± 3.5 vs 7.1 ± 2.8, P < 0·001), lower GCS scores (11.5 ± 2.2 vs 13.2 ± 1.8, P < 0.001), a greater prevalence of dysphagia (73.9% vs 55.3%, P = 0.008) and lower social support scores (29.5 ± 8.2 vs 38.8 ± 7.5, P < 0.001). There were no significant differences in age, BMI, education level or most comorbidities.

Table 1 Demographic and clinical characteristics of study participants, mean ± SD or n (%).

Characteristic	Depression group (n = 92)	Non-depression group (n = 94)	P value
Age (year)	66.5 ± 10.8	64.2 ± 11.5	0.162
Female	54 (58.7)	34 (36.2)	0.002
BMI (kg/m2)	23.8 ± 3.5	24.2 ± 3.1	0.428
Education level			0.285
Primary or below	38 (41.3)	32 (34.0)
Secondary	35 (38.0)	40 (42.6)
College or above	19 (20.7)	22 (23.4)
Smoking	25 (27.2)	28 (29.8)	0.688
Alcohol use	18 (19.6)	16 (17.0)	0.649
Comorbidities
Hypertension	58 (63.0)	55 (58.5)	0.525
Diabetes mellitus	28 (30.4)	24 (25.5)	0.453
COPD	15 (16.3)	12 (12.8)	0.489
Coronary heart disease	14 (15.2)	11 (11.7)	0.475
Chronic kidney disease	8 (8.7)	6 (6.4)	0.546
Previous stroke	18 (19.6)	14 (14.9)	0.393
NIHSS at admission	10.2 ± 3.5	7.1 ± 2.8	< 0.001
GCS score	11.5 ± 2.2	13.2 ± 1.8	< 0.001
Dysphagia	68 (73.9)	52 (55.3)	0.008
SSRS score	29.5 ± 8.2	38.8 ± 7.5	< 0.001
NRS-2002 ≥ 3	62 (67.4)	45 (47.9)	0.007

BMI: Body mass index; COPD: Chronic obstructive pulmonary disease; NIHSS: National Institutes of Health Stroke Scale; GCS: Glasgow Coma Scale; SSRS: Social Support Rating Scale; NRS-2002: Nutritional Risk Screening 2002.

SAP clinical features and severity

Clinical indicators of SAP are described in Table 2. Patients with depression had higher APACHE II scores (17.8 ± 4.5 vs 13.2 ± 3.8, P < 0.001) and CURB-65 scores at the time of SAP diagnosis (P < 0.001) than patients without depression. Multi-drug resistant organism infection was more common in the depression group (38.0% vs 22.3%, P = 0.019). The depression group also experienced higher rates of bilateral pulmonary involvement (52.2% vs 35.1%, P = 0.019) and pleural effusion (28.3% vs 14.9%, P = 0.026). The depression group had higher rates of mechanical ventilation (65.2% vs 42.6%, P = 0.002) and a longer duration before extubation.

Table 2 Stroke-associated pneumonia clinical features and severity, mean ± SD or n (%).

SAP feature	Depression group (n = 92)	Non-depression group (n = 94)	P value
Onset to SAP (days)	3.2 ± 1.5	3.5 ± 1.8	0.226
APACHE II score	17.8 ± 4.5	13.2 ± 3.8	< 0.001
CURB-65 score	2.8 ± 0.9	2.1 ± 0.8	< 0.001
Pathogen distribution
Gram-negative bacilli	52 (56.5)	48 (51.1)	0.453
Gram-positive cocci	22 (23.9)	25 (26.6)	0.668
Fungi	12 (13.0)	8 (8.5)	0.323
Mixed infection	18 (19.6)	15 (16.0)	0.514
MDR organism	35 (38.0)	21 (22.3)	0.019
Bilateral involvement	48 (52.2)	33 (35.1)	0.019
Pleural effusion	26 (28.3)	14 (14.9)	0.026
Mechanical ventilation	60 (65.2)	40 (42.6)	0.002
MV duration (days)	8.5 ± 3.2	5.2 ± 2.1	< 0.001
ICU admission	55 (59.8)	38 (40.4)	0.008
ICU stay (days)	12.5 ± 4.8	8.2 ± 3.5	< 0.001
Tracheostomy	18 (19.6)	8 (8.5)	0.028

SAP: Stroke-associated pneumonia; APACHE II: Acute Physiology and Chronic Health Evaluation II; MDR: Multi-drug resistant; MV: Mechanical ventilation; ICU: Intensive care unit.

Treatment modalities and prognostic outcomes

Treatment modalities and prognostic outcomes are shown in Table 3. The outcomes among the depression group were significantly worse than those in the non-depression group across all primary endpoints as follows: A longer time to resolution of pulmonary infection (16.8 ± 4.5 days vs 11.2 ± 3.2 days, P < 0.001), an extended total hospitalization (28.5 ± 7.8 vs days 19.3 ± 5.6 days, P < 0.001), worse functional outcome at discharge (mRS ≥ 3: 72.8% vs 48.9%, P = 0.001), and higher 28-day mortality (18.5% vs 7.4%, P = 0.024). Through exploratory subgroup analysis of patients who received integrated traditional Chinese and Western medicine treatment, including warm moxibustion, (n = 78; depression group: n = 35, 38.0%; non-depression group: n = 43, 45.7%), infection resolution time was significantly shorter compared with those who received Western medicine alone (n = 108): 13.2 ± 3.8 days vs 15.5 ± 4.5 days (P = 0.012). This analysis was not stratified by depression status, and the observed benefit reflects a pooled comparison across the full cohort. Given the non-randomized nature of treatment allocation and the differential moxibustion adoption between groups, treatment selection bias cannot be excluded, and these results should be interpreted only as hypothesis-generating.

Table 3 Treatment modalities and prognostic outcomes, mean ± SD or n (%).

Outcome measure	Depression group (n = 92)	Non-depression group (n = 94)	P value
Treatment
Antibiotics duration (days)	14.2 ± 4.5	10.5 ± 3.2	< 0.001
Corticosteroid use	28 (30.4)	18 (19.1)	0.072
Warm moxibustion	35 (38.0)	43 (45.7)	0.291
Early enteral nutrition	55 (59.8)	68 (72.3)	0.068
Antidepressant treatment	45 (48.9)	-	-
Prognostic outcomes
Infection resolution (days)	16.8 ± 4.5	11.2 ± 3.2	< 0.001
Hospital stays (days)	28.5 ± 7.8	19.3 ± 5.6	< 0.001
28-day mortality	17 (18.5)	7 (7.4)	0.024
mRS ≥ 3 at discharge	67 (72.8)	46 (48.9)	0.001
mRS ≥ 4 at discharge	42 (45.7)	22 (23.4)	0.001
Barthel Index at discharge	48.2 ± 18.5	65.8 ± 14.2	< 0.001
Readmission within 30 days	22 (23.9)	10 (10.6)	0.015
Composite adverse outcome	52 (56.5)	26 (27.7)	< 0.001

Composite adverse outcome defined as 28-day mortality, infection resolution >21 days, mechanical ventilation >14 days, or modified Rankin Scale ≥ 4 at discharge. mRS: Modified Rankin Scale.

Laboratory biomarkers

Serum biomarker profiles were markedly different between the depression and non-depression groups (Table 4). The depression cohort had remarkably increased day 7 infection and inflammatory levels: PCT (2.85 ± 1.32 ng/mL vs 1.52 ± 0.85 ng/mL, P < 0.001), CRP (85.6 ± 28.5 mg/L vs 52.3 ± 18.6 mg/L, P < 0.001), IL-6 (42.5 ± 12.8 pg/mL vs 25.8 ± 8.5 pg/mL, P < 0.001), and TNF-α (28.5 ± 8.2 pg/mL vs 18.2 ± 5.8 pg/mL, P < 0.001). Serum BDNF was significantly lower in the depression group (11.8 ± 4.2 ng/mL vs 21.5 ± 6.8 ng/mL, P < 0.001). Immune functions were significantly reduced in non-survivors, including lymphocyte count (0.95 × 10⁹ ± 0.32 × 10⁹/L vs 1.35 × 10⁹ ± 0.42 × 10⁹/L; P < 0.001), CD4⁺ T-cell counts and CD4/CD8 ratio (P < 0.001).

Table 4 Laboratory biomarkers in depression and non-depression groups (day 7), mean ± SD.

Biomarker	Depression group (n = 92)	Non-depression group (n = 94)	P value
Infection markers
PCT (ng/mL)	2.85 ± 1.32	1.52 ± 0.85	< 0.001
CRP (mg/L)	85.6 ± 28.5	52.3 ± 18.6	< 0.001
WBC (× 109/L)	13.8 ± 4.2	11.2 ± 3.5	0.001
Inflammatory cytokines
IL-6 (pg/mL)	42.5 ± 12.8	25.8 ± 8.5	< 0.001
TNF-α (pg/mL)	28.5 ± 8.2	18.2 ± 5.8	< 0.001
IL-10 (pg/mL)	8.5 ± 2.8	12.2 ± 3.5	< 0.001
Neurotrophic factor
BDNF (ng/mL)	11.8 ± 4.2	21.5 ± 6.8	< 0.001
Neuroendocrine
Morning cortisol (nmol/L)	408.5 ± 105.2	325.8 ± 78.5	< 0.001
Immune function
Lymphocyte (× 109/L)	0.95 ± 0.32	1.35 ± 0.42	< 0.001
CD4+ T cells (cells/μL)	285 ± 78	425 ± 95	< 0.001
CD4/CD8 ratio	1.05 ± 0.35	1.52 ± 0.42	< 0.001
Nutritional markers
Albumin (g/L)	28.5 ± 4.8	34.2 ± 5.2	< 0.001
Prealbumin (mg/L)	125 ± 35	185 ± 42	< 0.001
Transferrin (g/L)	1.52 ± 0.38	1.95 ± 0.45	< 0.001
Metabolic
HbA1c (%)	7.2 ± 1.5	6.5 ± 1.2	0.028
Fasting glucose (mmol/L)	8.5 ± 2.2	7.2 ± 1.8	0.012

PCT: Procalcitonin; CRP: C-reactive protein; WBC: White blood cell; IL-6: Interleukin-6; TNF-α: Tumor necrosis factor-alpha; IL-10: Interleukin-10; BDNF: Brain-derived neurotrophic factor; HbA1c: Glycated hemoglobin.

Multivariable logistic regression analysis

Variables with P < 0.10 in univariate analysis were included in multivariable logistic regression. We identified seven independent predictors of adverse prognosis (Table 5). The most significant determinant was APACHE II score ≥ 18 (OR = 3.45, 95%CI: 1.92-6.20, P < 0.001), while comorbid PSD (OR = 3.12, 95%CI: 1.85-5.26, P < 0.001) and serum albumin < 30 g/L (OR = 2.78, 95%CI: 1.55-4.98, P = 0.001) were next in rank order of significance for predicting outcomes based on logistic regression analysis of ten candidate predictors for poor outcome. The Hosmer and Lemeshow test indicated good fit (χ² = 5.68, P = 0.682).

Table 5 Independent predictors of adverse prognosis: Multivariable logistic regression.

Variable	β	SE	Wald	OR (95%CI)	P value
APACHE II ≥ 18	1.238	0.302	16.82	3.45 (1.92-6.20)	< 0.001
Comorbid PSD	1.138	0.268	18.02	3.12 (1.85-5.26)	< 0.001
Albumin < 30 g/L	1.022	0.298	11.76	2.78 (1.55-4.98)	0.001
MV > 7 days	0.975	0.298	10.70	2.65 (1.48-4.75)	0.001
MDR organism	0.925	0.292	10.04	2.52 (1.42-4.47)	0.002
BDNF < 14.5 ng/mL	0.855	0.312	7.51	2.35 (1.28-4.32)	0.006
Age ≥ 75 years	0.683	0.278	6.03	1.98 (1.15-3.41)	0.014
Non-significant variables
IL-6 > 35 pg/mL	0.585	0.305	3.68	1.80 (0.99-3.26)	0.055
Cortisol > 380 nmol/L	0.498	0.298	2.79	1.65 (0.92-2.95)	0.095
NIHSS ≥ 10	0.455	0.285	2.55	1.58 (0.90-2.76)	0.110
Dysphagia	0.412	0.275	2.25	1.51 (0.88-2.58)	0.134
Diabetes mellitus	0.365	0.268	1.85	1.44 (0.85-2.43)	0.174
Corticosteroid use	0.328	0.302	1.18	1.39 (0.77-2.51)	0.278

Hosmer-Lemeshow test: χ² = 5.68, P = 0.682. OR: Odds ratio; 95%CI: 95% confidence interval; APACHE II: Acute Physiology and Chronic Health Evaluation II; PSD: Post-stroke depression; MV: Mechanical ventilation; MDR: Multi-drug resistant; BDNF: Brain-derived neurotrophic factor; IL-6: Interleukin-6; NIHSS: National Institutes of Health Stroke Scale.

Predictive model performance and biomarker evaluation

The AUC of the predictive model based on these seven independent risk factors was 0.91 (95%CI: 0.86-0.96), with a sensitivity and specificity of 84.6% and 87.3%, respectively. Across individual predictors, APACHE II score had the highest diagnostic value (AUC = 0.82, 95%CI: 0.76-0.88), followed by comorbid PSD status (AUC = 0.78), serum albumin (AUC = 0.76), BDNF (AUC = 0.74) and duration of mechanical ventilation (AUC = 0.71). The combined model outperformed the individual predictors (all P < 0.01 by DeLong test). Figure 1 shows the ROC curves for the combined model and individual predictors. A forest plot illustrating the ORs and 95%CIs for all variables included in the multivariable model is shown in Figure 2. Variables with confidence intervals entirely to the right of 1.0 represent statistically significant predictors. Kaplan-Meier survival analysis demonstrated significantly higher 28-day mortality in the depression group compared with the non-depression group (log-rank P = 0.018). The survival curves diverged after day 7, suggesting that the prognostic impact of PSD becomes more pronounced during the subacute phase of SAP management (Figure 3).

Figure 2 Forest plot of odds ratios (95% confidence interval) for predictive factors of adverse prognosis in stroke-associated pneumonia patients from multivariable logistic regression. Blue markers indicate statistically significant independent predictors (P < 0.05); gray markers indicate non-significant variables. The dashed vertical line represents odds ratio = 1.0 (no effect). OR: Odds ratio; 95%CI: 95% confidence interval; APACHE II: Acute Physiology and Chronic Health Evaluation II; PSD: Post-stroke depression; BDNF: Brain-derived neurotrophic factor; MV: Mechanical ventilation; MDR: Multi-drug resistant; IL-6: Interleukin-6; NIHSS: National Institutes of Health Stroke Scale.

Figure 3 Kaplan-Meier 28-day survival curves comparing the depression group (n = 92) and non-depression group (n = 94) among stroke-associated pneumonia patients. The depression group demonstrated significantly higher mortality (18.5% vs 7.4%; log-rank P = 0.018). Survival curves diverged after day 7, with the number at risk shown below the plot. SAP: Stroke-associated pneumonia.

Depression severity subgroup analysis

We next analyzed the dose-response pattern in several outcome parameters after further stratification of depression severity (Table 6). For patients with severe depression (HAMD-17 ≥ 24), the 28-day mortality rate was highest (29.4%) and infection resolution time longest (20.5 ± 5.2 days); BDNF levels were lowest (7.8 ± 2.8 ng/mL) and IL-6 levels were highest (52.8 ± 15.5 pg/mL); their functional outcomes were poor, with a ranked mRS ≥ 4 in 64.7% of patients at discharge (Table 6). The rate of composite adverse outcomes increased in a stepwise fashion from 42.9% in mild to 60.6% in moderate and 82.4% in severe depression.

Table 6 Clinical outcomes stratified by depression severity, mean ± SD.

Characteristic	Mild (n = 42)	Moderate (n = 33)	Severe (n = 17)	P trend
HAMD-17 score	11.5 ± 2.5	19.8 ± 1.9	27.5 ± 3.2	< 0.001
Infection resolution (days)	14.2 ± 3.8	17.5 ± 4.2	20.5 ± 5.2	< 0.001
Hospital stays (days)	25.2 ± 6.5	29.8 ± 7.5	34.5 ± 9.2	< 0.001
MV duration (days)	6.8 ± 2.5	9.2 ± 3.0	11.5 ± 4.2	< 0.001
28-day mortality (%)	11.9	18.2	29.4	0.012
mRS ≥ 3 at discharge (%)	61.9	78.8	88.2	0.005
mRS ≥ 4 at discharge (%)	33.3	51.5	64.7	0.003
Barthel Index at discharge	55.2 ± 15.8	45.5 ± 18.2	35.8 ± 20.5	< 0.001
Composite adverse outcome (%)	42.9	60.6	82.4	< 0.001
Biomarkers (day 7)
BDNF (ng/mL)	14.5 ± 3.8	10.8 ± 3.5	7.8 ± 2.8	< 0.001
IL-6 (pg/mL)	35.2 ± 10.5	45.8 ± 12.2	52.8 ± 15.5	< 0.001
PCT (ng/mL)	2.25 ± 1.05	3.05 ± 1.35	3.85 ± 1.68	< 0.001
Cortisol (nmol/L)	378 ± 92	418 ± 105	462 ± 125	0.005
Lymphocyte (× 109/L)	1.08 ± 0.35	0.88 ± 0.28	0.72 ± 0.25	< 0.001
Albumin (g/L)	30.5 ± 4.2	27.8 ± 4.5	25.2 ± 5.5	0.002

P trend values were calculated using the Jonckheere-Terpstra test for ordered alternatives. HAMD-17: Hamilton Depression Rating Scale; MV: Mechanical ventilation; mRS: Modified Rankin Scale; BDNF: Brain-derived neurotrophic factor; IL-6: Interleukin-6; PCT: Procalcitonin.

DISCUSSION

In this study, we systematically evaluated the prognostic effect of comorbid PSD on clinical outcomes in patients with significant SAP and identified seven independent predictors for poor prognosis by multivariable analysis. The results show that PSD significantly exacerbates SAP outcomes on several dimensions and offers an actionable framework for early risk stratification and tailored intervention.

This higher reported SAP incidence with comorbid PSD (49.5%) compared to the estimated prevalence of PSD in non-SAP stroke populations of 30%-50% could potentially indicate that SAP itself is an additional risk factor for developing PSD[11-13]. This observation is biologically reasonable, as the systemic inflammatory response related to pneumonia could further intensify neuroinflammatory and monoaminergic neurotransmitter dysfunction in mood-potentiating circuits[17-19]. The bidirectional link between depression and infection, where not only does depression lead to impaired organic immune defense, but the resulting inflammation through neuro-inflammation exacerbates depression, may promote a vicious cycle that leads to poor outcome of both disorders.

Notably, the fact that comorbid PSD was an independent factor predictive of adverse prognosis (OR = 3.12) after accounting for disease severity, nutritional status, and other confounders underscores the clinical significance of screening for depression in SAP management. The specific pathways by which PSD exacerbates the outcomes of SAP are likely to be multidimensional. First, hyperactivation of the HPA axis associated with PSD and the resulting prolonged elevation in cortisol levels attenuate cellular immune responses, including reduced lymphocyte proliferation, natural killer cell activity, and neutrophil bactericidal function[20-22]. Consistent with depression-mediated immunosuppression, our data confirmed significantly lower lymphocyte counts, CD4⁺ T-cell counts, and CD4/CD8 ratios in the depression group. Second, depressed patients had higher pro-inflammatory cytokine levels (IL-6, TNF-α) and lower anti-inflammatory IL-10 levels compared with non-depressed controls, indicating a pro-inflammtory shift that could promote pulmonary tissue damage and delay infection resolution[28,29]. Third, the behavioral correlates of depression (e.g., decreased physical activity, nutritional intake, and cooperation with respiratory therapy) may further compromise pulmonary clearance mechanisms.

The strongest independent predictor of poor prognosis was an APACHE II score ≥ 18 (or = 3.45), consistent with previous studies showing the prognostic significance of illness severity scores in critically ill patients with pneumonia[30,31]. The identification of hypoalbuminemia (< 30 g/L, OR = 2.78) as an independent predictor highlights the importance of nutritional status in determining SAP outcomes. Nutrition deficiency reduces immune competency, delays wound healing, impairs respiratory muscle function and mechanical ventilation time[32,33]. The lower albumin and prealbumin levels in the depression group indicated that PSD may worsen prognosis through depression-mediated nutrition deterioration.

The identification of serum BDNF < 14.5 ng/mL as an independent predictor of poor prognosis in patients with SAP (OR = 2.35) suggests that BDNF may reflect the systemic burden of illness rather than being limited to the pathogenesis of PSD. Recent data suggest that BDNF exhibits immunomodulatory effects, such as facilitation of T-reg cell differentiation, macrophage polarization, and reduction of over-reactivity to inflammatory stimuli[34,35]. The present study further demonstrated that BDNF levels varied with the severity of depressive symptoms, decreasing from 14.5 ng/mL in patients with mild depression to 7.8 ng/mL in those with severe depression. This finding suggests an association between increasing depression severity, lower BDNF levels, and poorer infection outcomes. Collectively, these findings suggest that serum BDNF may serve as a clinically useful prognostic biomarker in patients with SAP and comorbid depression. The observed association between lower BDNF levels and poorer outcomes may also reflect biological pathways shared between depressive symptoms and increased susceptibility to pulmonary infection.

The predictive model incorporating all seven independent predictors demonstrated excellent discriminative performance (AUC = 0.91, sensitivity 84.6%, specificity 87.3%), significantly outperforming each individual predictor alone. These results are consistent with the clinical usefulness of a holistic evaluation strategy combining information on depression status, disease severity indices, nutritional markers, microbiological data and serum biomarkers for detection of high-risk SAP patients as early as possible. At a practical level, this implies a stepwise risk assessment protocol: Screening for depression and disease severity at the time of SAP diagnosis, followed by targeted biomarker analysis (BDNF, albumin) and microbiological profiling to better prognostically stratify diagnosed cases.

An exploratory finding of this study was that integrated traditional Chinese and Western medicine treatment with warm moxibustion may be associated with shorter infection resolution time, as well as enhanced functional outcomes compared with Western medicine alone. Possible mechanisms include the stimulation of vagal afferents at acupoints through moxibustion, decreasing systemic inflammation by modulating the cholinergic anti-inflammatory pathway[25-27], enhancing gastrointestinal peristalsis (which benefits enteral nutrition), and promoting immune regulation[36]. Nevertheless, this was a retrospective analysis and thus randomization of the treatment assignment was not performed, indicating that selection bias in treatment allocation cannot be excluded.

Limitations of the Study

Several limitations should be acknowledged in this study. The main limitations are as follows: First, the retrospective and single-center design without a control group ultimately limits generalization, and selection bias can not be avoided. Second, depression was evaluated at one timepoint (2 weeks after stroke), introducing several potential sources of bias: Early-onset or delayed-onset PSD may have been missed; transient depressive symptoms associated with acute neurologic injury could be misclassified; and this approach fails to account for the temporal evolution of PSD during the clinical course of SAP. Third, nearly half of depression group patients (45/92, 48.9%) received antidepressant treatment during the study period. Antidepressant use was not included in the multivariable logistic regression model because its univariable association with the composite adverse outcome was not formally assessed and heterogeneity in medication type, dosage, and timing precluded reliable analysis. Therefore, the possibility that antidepressant therapy independently influenced SAP outcomes cannot be excluded. Future prospective studies should standardize antidepressant protocols and include antidepressant use as a covariate to allow rigorous evaluation of its confounding effect. Fourth, the exclusion of patients with severe consciousness impairments (GCS ≤ 8) may lead to an underestimation of the true SAP-PSD comorbidity burden in these severely ill patients. Fifth, confounding by indication cannot be excluded from the warm moxibustion subgroup analysis, as it was exploratory and treatment assignment was non-randomized. Sixth, serum BDNF was assessed at only one timepoint and reflects peripheral levels of the neurotrophic factor and does not necessarily correlate to neurotrophic activity in the central nervous system. Future research should focus on prospective multicenter studies with serial assessments of depression and randomized evaluation of combined TCM-Western medicine interventions. Seventh, the relatively modest accrual rate (approximately 17 patients per year over the 11-year study period from December 2014 to December 2025) at a single center may reflect the stringent eligibility criteria and the requirement for complete biomarker and follow-up data; nonetheless, reviewers are advised that this figure may raise questions regarding case-selection rigor, and future multicenter validation with larger sample sizes is strongly recommended.

CONCLUSION

PSD is associated with greater clinical comorbidity in patients with SAP, including delayed resolution of infection, longer LOS, prolonged mechanical ventilation, increased 28-day mortality, and poorer functional recovery at discharge. Seven independent predictors were identified, and a prediction model incorporating these factors demonstrated good predictive performance, supporting its potential utility for early risk stratification. Moreover, serum BDNF was identified as an independent prognostic factor in patients with SAP, suggesting that its clinical relevance may extend beyond its association with depression and supporting further investigation of combined neurotrophic and inflammatory biomarker panels. Early identification of PSD in patients with SAP and implementation of comprehensive, multidisciplinary management strategies may help improve clinical outcomes.