Patient characteristics and clinical outcomes

This study included 81 elderly patients diagnosed with community-acquired pneumonia, of whom 67 (82.7%) survived and 14 (17.3%) did not. As shown in Table 1, baseline demographic characteristics showed a higher proportion of males (78.6% vs. 68.7%) and slightly older age (86.00 ± 7.78 years vs. 82.21 ± 14.38 years) in the non-survivor group compared to survivors, consistent with known risk factors for poor outcomes in pneumonia.

Table 1 Patient characteristics and clinical outcomesComorbidities and disease severity

The burden of comorbidities, as assessed by the Charlson Comorbidity Index, showed a median score of 6.00 in both groups. However, non-survivors had higher prevalence of several important comorbidities, including chronic obstructive pulmonary disease (28.6% vs. 15.2%), coronary artery disease (14.3% vs. 3.0%), and diabetes mellitus (35.7% vs. 26.9%). These patterns align with the clinical understanding that certain comorbidities contribute to poorer outcomes in pneumonia patients (Table 1).

Disease severity scores were notably higher in the non-survivor group, with median Pneumonia Severity Index of 151.50 compared to 137.00 in survivors, and median CURB-65 score of 3.00 versus 2.00. These elevated severity scores reflect the greater disease burden at presentation in patients who ultimately did not survive.

Clinical outcomes

Clear differences were observed in the hospital course metrics between groups. The median length of hospital stay was substantially shorter in the non-survivor group compared with survivors (5.00 days vs. 15.00 days; p = 0.038), as was the duration of critical care unit stay (4.50 days vs. 14.50 days; p = 0.032), reflecting the rapid clinical deterioration leading to earlier mortality in these patients.

Non-survivors demonstrated a more severe clinical course, with notably higher intensive care unit admission rates (71.4% vs. 41.8%) and greater need for ventilatory support (53.8% vs. 27.4%), reflecting the overall greater disease burden in this group (Table 1).

Baseline clinical, biochemical, and cytokine profiles

Table 2 summarizes the baseline characteristics of elderly patients with community-acquired pneumonia at the time of emergency department presentation, comparing traditional severity markers with a comprehensive cytokine profile between survival and mortality groups. Among conventional clinical parameters, respiratory rate was notably elevated in the mortality group (28.00 [26.00–30.00] vs. 24.00 [20.00–30.00] breaths per minute; p = 0.017), reflecting greater respiratory distress. Other vital signs showed clinically relevant patterns, including relatively higher heart rate (107.00 vs. 102.00 beats/min) and lower mean arterial pressure (88.82 vs. 95.17 mmHg) in non-survivors, consistent with the physiological stress response typically observed in more severely ill patients.

Table 2 Baseline clinical, biochemical, and cytokine profiles

At baseline, traditional inflammatory markers showed patterns consistent with more pronounced inflammation in the mortality group, with elevated white blood cell counts (12.565 [9.79–20.275] vs. 10.48 [8.40–14.495] × 109/L) and C-reactive protein levels (11.10 [5.31–14.80] vs. 8.71 [3.83–15.06] mg/dL). In the subset of patients with procalcitonin measurements available (n = 46), baseline levels were also higher in the mortality group (0.66 [IQR: 1.47] vs. 0.2 [IQR: 0.37] ng/mL, p = 0.126). These differences in inflammatory response underscored the importance of examining dynamic changes rather than single time-point measurements in elderly CAP patients.

Baseline cytokine analysis revealed differential inflammatory patterns between groups. IL-8 levels were considerably higher in the mortality group (56.30 [31.23–213.92] vs. 39.37 [21.00–84.55] pg/mL), consistent with its role in neutrophil recruitment during severe infections. Similarly, IL-6 baseline levels showed variability between groups (mortality: 53.25 [16.63, 251.21] vs. survival: 79.93 [31.81, 242.16] pg/mL), highlighting the heterogeneity of inflammatory responses in elderly patients and the limitations of single time-point cytokine measurements for prognostication.

Dynamic changes in cytokines and inflammatory markers

Table 3 summarizes the dynamic changes in inflammatory markers and cytokine profiles observed between the initial presentation and follow-up assessments. The mean time interval for follow-up measurements was comparable between the mortality and survival groups (15.46 [9.88, 17.40] hours vs. 14.78 [11.37, 20.69] hours, p = 0.520), ensuring valid temporal comparisons.

Table 3 Dynamic changes in cytokines and inflammatory markers

Among all measured inflammatory parameters, IL-6 was the only marker demonstrating significant dynamic changes predictive of mortality. In the mortality group, IL-6 levels markedly increased (ratio change: 0.88 [− 0.35, 1.49]), contrasting with a decrease observed in the survival group (ratio change: − 0.49 [− 0.76, − 0.10], p = 0.040). This distinct pattern was not seen in traditional inflammatory markers. C-reactive protein levels (ratio change: 0.24 [− 0.04, 0.71] vs. 0.10 [− 0.04, 0.44], p = 0.606) and white blood cell counts (ratio change: − 0.04 [− 0.26, 0.16] vs. − 0.18 [− 0.30, 0.00], p = 0.446) did not show significant discriminatory changes. PCT levels exhibited numerical differences between survivors (0.14 [− 0.16, 0.78]) and non-survivors (0.38 [− 0.30, 1.48]); however, the limited availability of data precludes definitive conclusions.

Other cytokines, such as IL-8 (ratio change: 0.32 vs. − 0.02) and IL-10 (ratio change: 0.28 vs. − 0.25), tended to be numerically elevated among non-survivors compared to survivors, although these differences did not reach statistical significance. IL-8 levels tended to rise in non-survivors (ratio change: 0.32 [− 0.30, 1.34] vs. − 0.02 [− 0.41, 0.25], p = 0.183), and IL-10 levels exhibited an upward trend (ratio change: 0.28 [− 0.44, 1.07] vs. − 0.25 [− 0.57, 0.11], p = 0.095). Changes in additional inflammatory mediators, including GM-CSF, TNF-α, IFN-γ, and other interleukins, remained comparable between the two groups.

These findings underscored the potential of dynamic IL-6 changes within the first 48 h as a sensitive indicator of disease progression in elderly patients with community-acquired pneumonia, surpassing the predictive utility of traditional inflammatory markers or other cytokines.

Prognostic value of cytokine dynamics in elderly CAP patients

To assess whether cytokine dynamics enhance the prognostic accuracy of traditional severity scores in elderly patients with community-acquired pneumonia, we performed both individual and combined analyses of risk factors.

In the individual analysis (Table 4, Section A), IL-6 dynamics demonstrated significant prognostic value. Patients with increased IL-6 levels exhibited a markedly higher mortality rate (36.00%) compared to those with decreased levels (8.93%), yielding an odds ratio of 5.39 (95% CI 1.69–18.83, p = 0.004).

Table 4 Integration of IL-6 changes with clinical severity scores for mortality risk assessment

The combined analysis (Table 4, Sections B and C) indicated that integrating IL-6 dynamics with conventional scoring systems significantly improved risk stratification. Among patients with very high Pneumonia Severity Index (PSI) scores (≥ 130), those with increased IL-6 levels had a significantly higher mortality rate (46.67%) compared to those with decreased levels (12.00%), corresponding to an odds ratio of 22.06 (95% CI 2.18–3007.59, p = 0.005). Similarly, in the moderate-risk CURB-65 group (score 2), increased IL-6 levels were associated with significantly higher mortality (41.67%) compared to decreased levels (9.52%), with an odds ratio of 21.27 (95% CI 1.98–2936.97, p = 0.008).

The comprehensive predictive model assessment using ROC curve analysis was presented in Table 5 and Fig. 1. The conventional severity scoring systems alone showed modest discriminatory power, with PSI achieving an AUC of 0.6631 (95% CI 0.5258–0.8004, p = 0.056) and CURB-65 reaching an AUC of 0.6231 (95% CI 0.4826–0.7637, p = 0.1491). Dynamic changes in IL-6 levels alone demonstrated improved predictive capability, with an AUC of 0.7020 (95% CI 0.5435–0.8606, p = 0.0179). When integrating IL-6 dynamics with traditional scoring systems, the predictive accuracy was further enhanced. The combination of PSI with IL-6 changes yielded the highest discriminatory power (AUC = 0.7676, 95% CI 0.6326–0.9026, p = 0.0017), followed by the combination of CURB-65 with IL-6 changes (AUC = 0.7564, 95% CI 0.6288–0.8840, p = 0.0027). These findings suggest that incorporating biomarker dynamics, particularly IL-6 changes, could potentially address some limitations of conventional severity scores in elderly CAP patients. To further contextualize the prognostic value of IL-6 dynamics, we evaluated models combining PSI and CURB-65 with CRP and PCT dynamics (Supplementary Table S4). These combinations yielded inferior or borderline AUC values, except for PSI plus PCT (AUC = 0.7527, p = 0.0348), which was limited by incomplete data (n = 46).

Table 5 Comparison of predictive models for mortality risk assessmentFig. 1

Receiver operating characteristic (ROC) curves for mortality prediction using PSI, CURB-65, and dynamic IL-6 changes A ROC curves comparing the prognostic performance of PSI alone (AUC = 0.6631, p = 0.056), IL-6 dynamics alone (AUC = 0.7020, p = 0.0179), and combinations of PSI with different inflammatory marker dynamics: PSI + IL-6 (AUC = 0.7676, p = 0.0017), PSI + PCT (AUC = 0.7527, p = 0.0348, n = 46), and PSI + CRP (AUC = 0.6471, p = 0.1152). The integration of IL-6 dynamics with PSI demonstrated strong discriminatory power in the full cohort. B ROC curves comparing the prognostic performance of CURB-65 alone (AUC = 0.6231, p = 0.1491), IL-6 dynamics alone (AUC = 0.7020, p = 0.0179), and combinations of CURB-65 with different inflammatory marker dynamics: CURB-65 + IL-6 (AUC = 0.7564, p = 0.0027), CURB-65 + PCT (AUC = 0.6740, p = 0.1463, n = 46), and CURB-65 + CRP (AUC = 0.6152, p = 0.2172). Among all combinations, CURB-65 + IL-6 demonstrated substantial improvement in mortality prediction (AUC = 0.7564, p = 0.0027). Note that PCT data were only available for a subset of patients (n = 46). Detailed statistical analyses are provided in Table 5 and Supplementary Table S4.