Group-Based Trajectory Modeling of Serum Sodium and Mortality in Critically Ill ICU Patients with Community-Acquired Pneumonia: An Analysis of the MIMIC-IV Database

Group-Based Trajectory Modeling of Serum Sodium and Mortality in Critically Ill ICU Patients with Community-Acquired Pneumonia: An Analysis of the MIMIC-IV Database | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Group-Based Trajectory Modeling of Serum Sodium and Mortality in Critically Ill ICU Patients with Community-Acquired Pneumonia: An Analysis of the MIMIC-IV Database Jingjing Cai, Bo Chen, Jun Lyu, Zhigang Wang This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7386815/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 5 You are reading this latest preprint version Abstract Purpose Community-acquired pneumonia (CAP) is a major cause of prolonged intensive care unit (ICU) admissions and is associated with substantial mortality. Abnormalities in serum sodium are common in critically ill patients and may carry prognostic significance. This study examined early serum sodium trajectories in ICU patients with CAP and their associations with 28-day, 90-day, and 365-day mortality. Methods Using the MIMIC-IV database, we retrospectively analyzed adult patients with a primary diagnosis of CAP, an ICU stay of more than 7 days, and at least four sodium measurements within the first 7 days. Group-based trajectory modeling was applied to identify distinct patterns of sodium change. Associations between trajectory groups and mortality were evaluated using multivariable Cox proportional hazards models, adjusting for demographics, comorbidities, illness severity, and laboratory parameters. Results A total of 3,146 patients were classified into five distinct trajectories: normal level with slow increase (trajectory 1), rapid increase followed by slow decrease (trajectory 2), low level with slow increase (trajectory 3), low level with rapid increase (trajectory 4), and normal stable (trajectory 5). Using trajectory 2 as the reference group, trajectory 5 was associated with significantly lower mortality at 28 days (adjusted HR 0.50, 95% CI 0.38–0.65, p < 0.001), 90 days (HR 0.63, 95% CI 0.50–0.79, p < 0.001), and 365 days (HR 0.68, 95% CI 0.55–0.83, p < 0.001). Similarly, trajectories 1 (28-day HR 0.58, 95% CI 0.45–0.76, p < 0.001) and 3 (28-day HR 0.51, 95% CI 0.36–0.73, p < 0.001) were associated with lower risk. Trajectory 4 showed no significant difference in 28-day mortality compared to trajectory 2 (HR 0.80, 95% CI 0.55–1.18, p = 0.263). Trajectory 2 was consistently associated with the highest mortality risk. Kaplan-Meier analysis demonstrated significant survival differences among the trajectory groups at 28 days (global log-rank test p = 0.019). Conclusions Distinct early sodium trajectories are independently associated with mortality in ICU patients with CAP and prolonged stays. The rapid increase followed by slow decrease pattern (trajectory 2), indicative of impaired physiological resilience, was linked to the poorest outcomes, whereas the normal stable pattern (trajectory 5) predicted optimal survival. Trajectory-based sodium monitoring may have relevance for early risk stratification and for consideration of individualized electrolyte management. Community-acquired pneumonia sodium trajectory intensive care unit mortality group-based trajectory modeling Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Introduction Community-acquired pneumonia (CAP) is among the most prevalent infectious diseases globally and remains a leading cause of hospital admission and death in adults. In the United States, it is estimated to result in roughly 1.4 million emergency department visits, 740 000 hospitalizations, and about 41 000 deaths each year [ 1 ] . Although most cases are treated successfully in outpatient settings, approximately one in ten patients require hospitalization, and of these, 20–25% are admitted to an intensive care unit (ICU). Severe CAP carries a particularly grave prognosis, with short- and long-term mortality reported between 27% and 50%, despite modern advances in antimicrobial therapy and critical care support [ 2 ] . This substantial disease burden is driven by the complex pathophysiology of CAP, encompassing widespread inflammatory responses, multiple organ dysfunction, and significant disruptions in physiological homeostasis. Most prior studies have relied on single sodium measurements at ICU admission, overlooking the dynamic fluctuations that occur during early critical illness. In reality, sodium levels change rapidly with disease progression, therapeutic interventions, and fluid balance shifts. Persistent abnormalities may indicate unresolved pathology, whereas normalization may signal recovery. These temporal patterns may offer stronger prognostic value than static values, especially for long-term outcomes, as excess mortality often extends beyond hospitalization due to persistent inflammation and organ injury [ 3 ] . Group-based trajectory modeling (GBTM) offers a robust method to identify distinct longitudinal patterns in repeated biomarker measurements, accommodating unbalanced or missing data. This approach has been widely applied in critical care research, revealing prognostically distinct subgroups in contexts [ 4 ] . However, studies specifically employing GBTM to characterize early serum sodium trajectories in critically ill patients with severe CAP remain scarce. Using the MIMIC-IV database, our objectives were threefold: (1) to characterize distinct patterns of serum sodium changes over the first seven days of ICU admission in patients with CAP; (2) to compare clinical characteristics among the identified trajectory groups; and (3) to examine how these patterns relate to mortality at 28 days, 90 days, and 365 days, using the identified highest-risk trajectory as the primary reference for comparison. Incorporating temporal sodium profiles into prognostic evaluation may improve risk stratification, help identify vulnerable patient subgroups, and guide individualized management—such as optimized fluid strategies and extended monitoring—aimed at improving outcomes in critically ill patients with CAP. Materials and Methods Data source The Medical Information Mart for Intensive Care (MIMIC) database, established in 2003, is a collaborative initiative supported by the Beth Israel Deaconess Medical Center (BIDMC), the National Institutes of Health, Massachusetts General Hospital, emergency physicians, intensivists, computer science experts, and various other professional critical care medicine networks (Yang et al., 2020). This observational study utilized the MIMIC-IV database (version 3.1; PhysioNet. RRID:SCR_007345; https://doi.org/10.13026/kpb9-mt58 ; covering ICU admissions from 2008 to 2022) for patients admitted to Beth Israel Deaconess Medical Center (BIDMC) in Boston, Massachusetts [ 5 , 6 ] . This includes a wide range of patient information, such as demographic data, vital signs, comorbidities, laboratory results, and mortality outcomes. The ethical approval statement and informed consent were not required for this study because all personal information was replaced with random codes and anonymized. Author Jingjing Cai completed a training program facilitated by the PhysioNet team and secured official approval to use the MIMIC-IV database (certification ID: 70520306). The study's methodology and reporting adhered to the guidelines set forth by the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) initiative [ 7 ] . Patient Selection Patients were eligible for screening if they were diagnosed with CAP during hospitalization, as determined by the International Classification of Diseases, Ninth or Tenth Revision (ICD-9/10) codes. For patients with more than one ICU admission, only the first was included in the analysis. The inclusion criteria were: (1) age ≥ 18 years at the time of ICU admission; and (2) pneumonia listed among the top ten principal discharge diagnoses. Exclusion criteria were: (1) age < 18 years; (2) pneumonia not listed among the top ten principal discharge diagnoses; (3) diagnosis of ventilator-associated pneumonia; or (4) ICU stay of fewer than seven consecutive days. After applying these criteria, 3,146 patients were included in the final analysis ( Fig. 1 ). For each patient, all available serum sodium measurements from the first 7 days of ICU stay were extracted to construct longitudinal sodium trajectories using GBTM. Baseline demographic characteristics, comorbidities, illness severity scores, laboratory parameters, and outcome variables were also collected for subsequent statistical analyses. Variable collection We collected baseline demographic characteristics (age, sex, race, and marital status), vital signs (heart rate, mean arterial pressure, respiratory rate, temperature, and SpO₂), and laboratory parameters (glucose, chloride, potassium, hematocrit, red cell distribution width(RDW), hemoglobin, platelet count, white blood cell count, blood urea nitrogen(BUN), total bilirubin, albumin, lactate, arterial PO₂, arterial PCO₂, pH, base excess, anion gap, and bicarbonate), as well as enteral and parenteral nutrition status. Disease severity was assessed using the Acute Kidney Injury (AKI) stage, Sequential Organ Failure Assessment (SOFA) score, Acute Physiology and Chronic Health Evaluation II (APACHE II), and Charlson Comorbidity Index (CCI). Additional treatment and clinical history variables included the use of vasopressors, mechanical ventilation, and renal replacement therapy (RRT), along with comorbidities such as hypertension, congestive heart failure, myocardial infarction, cerebrovascular accident, chronic kidney disease, chronic obstructive pulmonary disease, liver cirrhosis, diabetes, and cancer. If multiple measurements of the same variable were available within the first 24 hours after ICU admission, the value most deviated from the normal range was selected for baseline characterization and covariate adjustment. Variables with more than 20% missing data were excluded, and for the remaining variables, missing values were imputed using the random forest algorithm implemented in the mice package. For serum sodium, the observation period was defined as days 1 to 7 following ICU admission. For each day, only the first available sodium measurement was used, providing up to seven time points per patient. To ensure reliable trajectory estimation, patients were required to have at least four daily sodium values within this interval. The primary outcomes were all-cause mortality at 28, 90, and 365 days. Trajectory grouping of serum sodium content We applied GBTM to identify distinct longitudinal patterns of serum sodium concentration over the first 7 days after ICU admission. The time scale was defined in hours since ICU admission, with each patient contributing one sodium value per 24 hours as described above. Trajectory models were fitted using the lcmm package in R, initially testing linear, quadratic, and cubic polynomial functions of time without covariates. We compared models with different numbers of trajectories and polynomial orders, selecting the optimal solution based on the Akaike information criterion (AIC), Bayesian information criterion (BIC), sample size–adjusted BIC (SABIC), entropy, the proportion of patients in each trajectory group, and the average posterior probability of group membership (AvePP) [ 4 ] . A five-trajectory solution was pre-specified to enhance clinical interpretability and ensure adequate representation of each subgroup. The final model required that each trajectory group include at least 5% of the cohort and have an AvePP of 0.70 or higher. Each patient was then assigned to the trajectory group corresponding to the highest posterior probability, and these assignments were subsequently used for survival analyses and subgroup interaction testing. Statistical Analyses Baseline characteristics were summarized for the overall cohort and by sodium-trajectory groups. Continuous variables were tested for normality using the Shapiro–Wilk test. Normally distributed variables are presented as mean ± standard deviation (SD) and compared using one-way ANOVA. Categorical variables are presented as counts (percentages) and compared using the χ² test or Fisher’s exact test. The primary exposure was the latent trajectory of serum sodium during the first 7 days after ICU admission. Group-based trajectory modeling (GBTM) with a censored normal distribution was used to identify distinct sodium-trajectory patterns. The optimal model was selected using the Bayesian Information Criterion (BIC), average posterior probability of assignment (APPA ≥ 0.70 for all classes), entropy ≥ 0.70, odds of correct classification (OCC > 5), and clinical interpretability. Patients with missing data for key variables, including serum sodium, were excluded from analysis. Outcomes included all-cause mortality at 28, 90, and 365 days after admission. Kaplan–Meier curves were plotted for each trajectory group, and survival differences were compared using the log-rank test. Cox proportional hazards models were fitted to estimate hazard ratios (HRs) for the association between trajectory group and mortality at each follow-up. Because preliminary analyses showed the highest risk in Trajectory 2, this group was used as the reference to facilitate clinical interpretation. Models were specified as follows: the crude model was unadjusted; Model I adjusted for age, sex, and race; Model II further adjusted for ICU length of stay, hospital length of stay, Charlson Comorbidity Index (CCI), SOFA, and APACHE II scores; and Model III additionally adjusted for vital signs, laboratory parameters, nutrition status, additional treatments, clinical history variables, and comorbidities. An HR 1 indicated higher risk. All analyses were conducted in R (version 4.5.0; R Foundation for Statistical Computing, Vienna, Austria). Two-sided p-values < 0.05 were considered statistically significant. Results In total, 3,146 patients with CAP who fulfilled the inclusion criteria were analyzed. Using group-based trajectory modeling (GBTM), we delineated five distinct patterns of serum sodium evolution during the first seven days of ICU stay ( Fig. 2 ): Trajectory 1 (“normal level with slow increase”), Trajectory 2 (“rapid increase followed by slow decrease”), Trajectory 3 (“low level with slow increase”), Trajectory 4 (“low level with rapid increase”), and Trajectory 5 (“normal level, stable”). Table 1 presents the fit statistics for the group-based trajectory model. The five-class model met the selection criteria, with all trajectory groups containing more than 5% of participants and average posterior probabilities exceeding 0.7, indicating adequate classification accuracy. Additionally, the model showed favorable fit according to key selection criteria, including the Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), and sample size–adjusted BIC (SABIC), further supporting its optimality in modeling the data. Notably, baseline demographic and clinical profiles varied significantly among the five trajectory groups ( Table 2 ), suggesting potential heterogeneity in underlying pathophysiology and disease progression. Significant differences were observed in demographic factors, comorbidities, illness severity scores, and clinical parameters among the groups. Variables such as age, sex distribution, race, Charlson comorbidity index, SOFA, APACHE II scores, ICU and hospital length of stay, as well as laboratory values at admission, varied notably between trajectories, indicating distinct patient profiles associated with each sodium trend pattern. Kaplan-Meier curves demonstrated significant survival differences across trajectories at 28 days (global log-rank p = 0.019) ( Fig. 3A ). Kaplan-Meier survival curves comparing individual trajectories to the reference trajectory 2 confirmed significantly higher survival probabilities for trajectory 5 ( Fig. 3B–D ) and trajectory 1 ( Fig. 3E–G ) at 28, 90, and 365 days. In multivariable Cox proportional hazards analyses using trajectory 2 as the reference, distinct serum sodium trajectories were associated with differential mortality risks ( Table 3 ). For 28-day mortality, trajectories 1 (HR 0.58, 95% CI 0.45–0.76, P < 0.001), 3 (HR 0.51, 95% CI 0.36–0.73, P < 0.001), and 5 (HR 0.50, 95% CI 0.38–0.65, P < 0.001) were significantly associated with lower risk, whereas trajectory 4 showed no significant difference (HR 0.80, 95% CI 0.55–1.18, p = 0.263). Similar patterns were observed for 90-day mortality, with trajectories 1 (HR 0.74, 95% CI 0.60–0.92, p = 0.006), 3 (HR 0.68, 95% CI 0.51–0.91, p = 0.009), and 5 (HR 0.63, 95% CI 0.50–0.79, P < 0.001) remaining protective. For 365-day mortality, trajectories 1 (HR 0.75, 95% CI 0.62–0.91, p = 0.004) and 5 (HR 0.68, 95% CI 0.55–0.83, P < 0.001) retained a significant survival advantage, while trajectories 3 and 4 were not significant. Overall, trajectories 1, 3, and 5 were consistently linked to reduced mortality risk compared to the high-risk trajectory 2, with trajectories 1 and 5 also showing favorable long-term outcomes. In the age-stratified Cox proportional hazards analysis ( Fig. 4 ), using trajectory 2 as the reference, distinct patterns of mortality risk were observed across different age groups. For patients aged < 65 years, trajectory 1 was consistently associated with significantly lower mortality risk compared to trajectory 2 across all time points and models, with HRs ranging from 0.35 (95% CI: 0.22–0.55, p < 0.001) in Model I for 28-day mortality to 0.55 (95% CI: 0.40–0.76, p < 0.001) in Model II for 365-day mortality. Similarly, trajectory 5 showed a significant survival advantage over Trajectory 2, with HRs between 0.38 (95% CI: 0.24–0.60, p < 0.001) and 0.72 (95% CI: 0.53–0.97, p = 0.032). In contrast, Trajectories 3 and 4 demonstrated inconsistent associations, with several HRs close to 1.0 and non-significant p-values (e.g., Trajectory 3 vs 2: HR = 0.94, 95% CI: 0.58–1.54, p = 0.822 for 28-day mortality in Model I). For patients aged ≥ 65 years, the same general pattern was observed, though effect sizes were attenuated. Trajectory 1 remained protective in most models (e.g., HR = 0.66, 95% CI: 0.50–0.88, p = 0.005 in Model III for 28-day mortality; HR = 0.78, 95% CI: 0.63–0.96, p = 0.017 in Model II for 365-day mortality). Trajectory 5 also showed favorable outcomes, with HRs between 0.56 (95% CI: 0.41–0.75, p < 0.001) and 0.76 (95% CI: 0.61–0.94, p = 0.013). However, the differences for Trajectories 3 and 4 compared to Trajectory 2 were generally not statistically significant in older patients. Age-stratified analysis confirmed that the survival advantage of Trajectories 1 and 5 over the high-risk Trajectory 2 was evident in both younger (< 65 years) and older (≥ 65 years) patients, although the magnitude of risk reduction was generally greater in the younger group. In the CKD-stratified Cox proportional hazards analysis ( Fig. 5 ), using trajectory 2 (initial increase followed by decrease) as the reference, distinct patterns of mortality risk emerged across trajectory groups and follow-up periods (28-day, 90-day, and 365-day). For non-CKD patients, trajectory 1 (normal level with slow increase) demonstrated consistently lower mortality risk across all time points, with HRs ranging from 0.54 (95% CI: 0.41–0.70, p < 0.001) for 28-day mortality to 0.32 (95% CI: 0.19–0.55, p < 0.001) for 365-day mortality in Model III. Trajectory 5 (normal stable) also showed a significant survival benefit, with HRs between 0.38 (95% CI: 0.24–0.60, p < 0.001) and 0.56 (95% CI: 0.40–0.75, p < 0.001). In contrast, trajectories 3 and 4 exhibited inconsistent associations, with HRs close to 1.0 and non-significant p-values, suggesting weaker associations. Stratification by CKD status revealed that the significant reduction in mortality risk associated with trajectories 1 and 5, compared to trajectory 2, was substantially stronger and more consistent in patients without pre-existing CKD than in those with CKD. Discussion In this large retrospective cohort of critically ill patients with CAP, we identified five distinct early ICU serum sodium trajectories using GBTM. Our analysis indicates that dynamic changes in sodium levels rather than a single baseline measurement—carry important prognostic value. Our analysis underscores that dynamic changes in sodium levels, rather than a single baseline measurement, provide significant prognostic value for both short- and long-term mortality. Specifically, using the highest-risk trajectory (trajectory 2) as the reference, our analysis revealed that trajectory 5 and trajectory 1 were consistently associated with significantly lower mortality risk at all time points. trajectory 3 also showed reduced risk in the short term. trajectory 2 itself was confirmed as the pattern associated with the poorest prognosis. These results suggest that the sodium trajectory pattern in trajectory 2 may serve as a strong prognostic marker for both short- and long-term survival. In the age-stratified Cox proportional hazards analysis, distinct patterns of mortality risk were observed based on serum sodium trajectory groups, with consistently lower mortality risks for trajectory 1 and trajectory 5 across all follow-up periods. However, the strength of these associations was more pronounced in non-CKD patients, suggesting that the impact of sodium trajectory on mortality may be more robust in this group. In non-CKD patients, trajectories 1 and 5 consistently showed survival advantages versus trajectory 2, whereas trajectories 3 and 4 showed weaker and often non-significant associations, while in the CKD group, the associations were weaker, and the differences between trajectories 3 and 4 compared to trajectory 2 were not statistically significant. This difference in findings between CKD and non-CKD groups may highlight the altered physiological response to sodium fluctuations in patients with chronic kidney disease. The weaker associations in CKD patients suggest that other factors, such as renal function, fluid balance, and comorbidities, may attenuate the relationship between sodium trajectory and mortality. Therefore, the findings emphasize the need for tailored clinical management strategies for non-CKD patients, where sodium levels may not be as directly linked to mortality as in CKD patients. The diminished association in CKD patients may reflect their impaired renal capacity to regulate sodium and water homeostasis, potentially blunting the impact of acute fluctuations observed in the trajectories. Furthermore, the high burden of competing risks (e.g., cardiovascular disease, chronic inflammation) in CKD may obscure the specific contribution of short-term sodium dynamics to mortality. Future research should further explore the underlying mechanisms driving these differential associations and consider other biomarkers or clinical features that might explain these variations. These findings align with previous evidence that hypernatremia is independent predictors of mortality in critically ill patients [ 8 , 9 ] . Our results build on this knowledge by showing that early sodium fluctuations, particularly rapid rises, offer additional prognostic value [ 10 ] . ICU-acquired hypernatremia, often driven by excessive sodium administration, osmotic shifts, or impaired water regulation, has been associated with nearly doubled in-hospital mortality risk. In our cohort, trajectory 2 likely reflects acute hypernatremia precipitated by aggressive diuresis or sodium loading during initial resuscitation, followed by partial correction, whereas trajectory 4 may represent severe hyponatremia at admission with rapid overcorrection. Both patterns indicate marked instability in electrolyte homeostasis and significant physiological stress, which may partly explain the observed increase in adverse outcomes. From a pathophysiological perspective, sodium trajectories may serve as a surrogate marker of the host’s overall systemic homeostasis, analogous to the role of dynamic immune biomarkers in differentiating sepsis phenotypes. Abrupt sodium elevations in high-risk trajectories may signify a proinflammatory, hypercatabolic phase accompanied by osmotic stress at the cellular level, potentially leading to neurologic injury, impaired cardiac function, or further immune dysregulation [ 11 ]. In contrast, maintenance of normonatremia suggests preserved electrolytes balance and is associated with improved survival [ 12 ] , resembling the “normal, stable” immune profile described in other trajectory-based investigations. Clinically, these findings suggest that early monitoring of sodium trajectories may aid in stratifying ICU patients with CAP into distinct risk categories. Patients with rising sodium levels indicative of impending hypernatremia (trajectory 2) may benefit from tighter fluid balance, judicious diuretic use, and proactive sodium restriction. In contrast, those with severe hyponatremia undergoing rapid correction (trajectory 4) require adherence to recommended correction rates, timely administration of hypotonic fluids when appropriate, and close neurological surveillance to reduce iatrogenic risk. Patients maintaining stable normonatremia (trajectory 5) likely represent a lower-risk group for whom standard CAP management remains appropriate. Our study builds upon the findings of previous study [ 13 ] , which also demonstrated that both stable and fluctuating sodium levels were associated with mortality. However, we extend their work by offering a more nuanced understanding of the prognostic significance of sodium trajectory patterns, particularly emphasizing that the trajectory 2 pattern (rapid increase followed by slow decrease) independently predicted poor prognosis even after adjusting for chronic comorbidities and acute severity scores. Our findings underscore that rapid sodium fluctuations, likely induced by aggressive diuresis or sodium loading followed by partial correction, represent a significant risk and may lead to adverse outcomes due to the associated physiological stress. Our findings show that changes in serum sodium over time offer important prognostic information beyond single measurements [ 14 , 15 ] . Both rising and falling trajectories were independently linked to higher mortality compared with stable normonatremia, even after full adjustment. These results support the use of trajectory-based electrolyte monitoring as a simple and widely available tool in critical care. Recognizing sustained deviations early may prompt timely review of fluid management, renal function, and disease progression, enabling more individualized interventions and potentially improving outcomes. Our analysis revealed a clinically significant paradox in the baseline characteristics of trajectory groups. Despite exhibiting the highest 28-day mortality (34.8%) among all trajectories, patients in trajectory 2 demonstrated lower acute illness severity scores (median SOFA: 5.0; APACHE II: 19.0) compared to trajectories 3 and 4 (SOFA: 7.0/8.0; APACHE II: 22.0/22.0) ( Table 2 ). Paradoxically, this group carried the heaviest chronic disease burden, as evidenced by the highest Charlson Comorbidity Index (median CCI: 6.0). This dissociation suggests that: traditional severity scores may underestimate risk in comorbid patients While SOFA/APACHE II quantify acute physiological derangement, they may fail to capture the diminished physiological reserve in patients with high CCI. The aberrant sodium trajectory (rapid rise/slow decline) likely reflects this vulnerability—exposing an impaired capacity to maintain electrolyte homeostasis during acute stress. Sodium dynamics as a novel marker of physiological fragility. The trajectory pattern itself may serve as a biomarker of "homeostatic incompetence". Rapid sodium fluctuations could signify dysregulated neurohormonal responses (e.g., inappropriate ADH/RAAS activation) or subclinical organ dysfunction not captured by conventional scores—particularly relevant in patients with chronic comorbidities. Our models adjusted for CCI, SOFA, and APACHE II still identified trajectory 2 as an independent mortality predictor (adjusted HRs 1.48–2.01 vs. trajectory 5). This underscores that dynamic sodium monitoring provides prognostic information beyond static scores and comorbidity indices. In patients with high chronic disease burden (CCI ≥ 5), sodium trajectory analysis—particularly the detection of trajectory 2 patterns—may identify high-risk physiological fragility earlier than conventional metrics, enabling preemptive interventions to mitigate decompensation risk. Several limitations should be noted. As with any observational study, residual confounding cannot be fully excluded despite adjustment for a wide range of covariates [ 16 ] . Certain unmeasured factors, including the composition of administered fluids, the osmotic load from medications, and individual neuroendocrine responses, may still have influenced both sodium trajectories and outcomes. The frequency and timing of serum sodium measurements were determined by routine clinical practice rather than a standardized protocol, which may have introduced measurement bias. Although GBTM is designed to handle irregular and unbalanced data, variation in sampling intensity could still have affected trajectory classification. The consistency of associations observed across different adjustment models and follow-up periods lends support to the robustness of our findings, but cannot completely rule out bias from unmeasured variables. As repeated measurements were required to form trajectories, a degree of survivorship requirement is unavoidable and could influence estimates (akin to an immortal-time component). We therefore advise cautious interpretation. Finally, this single center retrospective analysis, based on the MIMIC-IV database, reflects the characteristics and practice patterns of one institution; replication in multicenter and prospective cohorts is necessary to confirm the reproducibility and broader applicability of these results. Despite these limitations, the use of a large, well-characterized ICU cohort and a trajectory-based analytical framework allowed for a detailed assessment of dynamic sodium changes and their prognostic relevance in critically ill patients with CAP. While hypernatremia is recognized as a marker of physiological derangement and increased mortality in general critical care populations, its prognostic role in this specific group has not been well established. In our analysis, early sodium trajectories identified clinically distinct prognostic subgroups, with a rapid rise in sodium—irrespective of baseline level—associated with markedly higher mortality. These results highlight the value of close sodium surveillance during the early ICU course and support further investigation into timely, targeted interventions for patients at greatest risk. Well-designed prospective studies, ideally incorporating randomized strategies, are needed to determine whether trajectory-guided management can improve outcomes in this high-risk population. Declarations Ethics approval and consent to participate. This retrospective study analyzed the publicly available, de-identified MIMIC-IV database (version 3.1). The creation and distribution of MIMIC-IV were approved by the Institutional Review Boards of the Massachusetts Institute of Technology and Beth Israel Deaconess Medical Center, which granted a waiver of individual informed consent because the database contains only de-identified data. Access to the database was granted to the authors after completion of the required training and Data Use Agreement (CITI certification ID: 70520306). No additional patient contact or intervention occurred. (Repository: PhysioNet; version DOI: 10.13026/kpb9-mt58.) Consent for publication. Not applicable. Funding. This research received no external funding. Clinical trial registration/number. Not applicable; this work is an observational secondary analysis of an existing database and not a registered interventional trial. Competing interests. The authors declare no competing interests. Data availability. The MIMIC-IV database is publicly available to credentialed researchers at physionet.org after completing the required training and data use agreement. (optional but useful) References Vaughn, V.M., et al., Community-Acquired Pneumonia: A Review. JAMA, 2024. 332(15): p. 1282-1295. Cillóniz, C., A. Torres and M.S. Niederman, Management of pneumonia in critically ill patients. BMJ, 2021. 375: p. e065871. Grim, C.C.A., et al., Association Between an Increase in Serum Sodium and In-Hospital Mortality in Critically Ill Patients. Crit Care Med, 2021. 49(12): p. 2070-2079. Nagin, D.S., B.L. Jones and J. Elmer, Recent Advances in Group-Based Trajectory Modeling for Clinical Research. Annu Rev Clin Psychol, 2024. 20(1): p. 285-305. Johnson, A.E.W., et al., MIMIC-IV, a freely accessible electronic health record dataset. Sci Data, 2023. 10(1): p. 1. Johnson A, Bulgarelli L, Pollard T, Gow B, Moody B, Horng S, Celi LA, Mark R (2024) MIMIC-IV (version 3.1). PhysioNet. 10.13026/kpb9-mt58.. von Elm, E., et al., The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. J Clin Epidemiol, 2008. 61(4): p. 344-9. Funk, G., et al., Incidence and prognosis of dysnatremias present on ICU admission. Intensive Care Medicine, 2010. 36(2): p. 304-311. Nasser, A., et al., ICU-acquired hypernatremia: Prevalence, patient characteristics, trajectory, risk factors, and outcomes. Critical Care and Resuscitation, 2024. 26(4): p. 303-310. Sakr, Y., et al., Fluctuations in serum sodium level are associated with an increased risk of death in surgical ICU patients. Crit Care Med, 2013. 41(1): p. 133-42. Müller, D.N., et al., Sodium in the microenvironment regulates immune responses and tissue homeostasis. Nat Rev Immunol, 2019. 19(4): p. 243-254. Kitisin, N., et al., Systematic review and meta-analysis of the treatment of hypernatremia in adult hospitalized patients: impact on mortality, morbidity, and treatment-related side effects. J Crit Care, 2025. 87: p. 155012. Chewcharat, A., et al., Trajectories of Serum Sodium on In-Hospital and 1-Year Survival among Hospitalized Patients. Clinical Journal of the American Society of Nephrology, 2020. 15(5): p. 600-607. Chewcharat, A., et al., Trajectories of Serum Sodium on In-Hospital and 1-Year Survival among Hospitalized Patients. Clin J Am Soc Nephrol, 2020. 15(5): p. 600-607. Huang, S., et al., Association between serum sodium trajectory and mortality in patients with acute kidney injury: a retrospective cohort study. BMC Nephrol, 2024. 25(1): p. 152. Gao, Y., et al., Confounder adjustment in observational studies investigating multiple risk factors: a methodological study. BMC Med, 2025. 23(1): p. 132. Tables Tables are available in the Supplementary Files section. Additional Declarations No competing interests reported. Supplementary Files Tables.docx Cite Share Download PDF Status: Under Review Version 1 posted Reviewers invited by journal 08 Oct, 2025 Editor assigned by journal 23 Sep, 2025 Editor invited by journal 03 Sep, 2025 Submission checks completed at journal 02 Sep, 2025 First submitted to journal 02 Sep, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7386815","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":531492904,"identity":"2986f9f5-e628-4490-96b1-fe6b4104e190","order_by":0,"name":"Jingjing Cai","email":"","orcid":"","institution":"The First Affiliated Hospital of Jinan University","correspondingAuthor":false,"prefix":"","firstName":"Jingjing","middleName":"","lastName":"Cai","suffix":""},{"id":531492905,"identity":"8bee5a9e-9eb7-4090-9cb0-32e13ef4cafd","order_by":1,"name":"Bo Chen","email":"","orcid":"","institution":"The First Affiliated Hospital of Jinan University","correspondingAuthor":false,"prefix":"","firstName":"Bo","middleName":"","lastName":"Chen","suffix":""},{"id":531492906,"identity":"2d80ae80-9402-4740-baa5-6c4748266a9a","order_by":2,"name":"Jun Lyu","email":"","orcid":"","institution":"The First Affiliated Hospital of Jinan University","correspondingAuthor":false,"prefix":"","firstName":"Jun","middleName":"","lastName":"Lyu","suffix":""},{"id":531492907,"identity":"f61aa717-96d7-4e2b-ad56-dba586d87cb0","order_by":3,"name":"Zhigang Wang","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA30lEQVRIiWNgGAWjYDCCA1CaXwJMScgQr0VyBgNjA1ALD/FaDG6AtTAQ1sJ3+4zh54JfdXabbzcff3SjxoKHgf3w0Q34tEieyzGWntnHlrztzrHE5pxjQIfxpKXdwKfF4AzvBmneHp5ksxs5hs05bEAtEjxmhLRs/s3bI5FsPAOk5R9xWrZJ8/wwsDOQAGrJbSNCi+QZ/m/WvA0JCRI30hJn5/ZJ8LAR8gvfGbbk2zx/6uz5ZyQf+JzzrU6On/3wMbxawICxjSGxAcZhI6gcDP4w2BOncBSMglEwCkYkAABQ/kgmhgj4igAAAABJRU5ErkJggg==","orcid":"","institution":"The First Affiliated Hospital of Jinan University","correspondingAuthor":true,"prefix":"","firstName":"Zhigang","middleName":"","lastName":"Wang","suffix":""}],"badges":[],"createdAt":"2025-08-16 10:23:17","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7386815/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7386815/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":93947351,"identity":"a22b5906-c230-41af-8835-be0acf38e522","added_by":"auto","created_at":"2025-10-20 14:26:38","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":2684520,"visible":true,"origin":"","legend":"","description":"","filename":"GroupBasedTrajectoryModelingofSerumSodiumandMortalityinCriticallyIllICUPatientswithCommunityAcquiredPneumoniaAnAnalysisoftheMIMICIVDatabase1.docx","url":"https://assets-eu.researchsquare.com/files/rs-7386815/v1/450ce7547b34d54f7871228a.docx"},{"id":93947347,"identity":"3c7317a9-6511-4656-a97f-360c34e265a8","added_by":"auto","created_at":"2025-10-20 14:26:38","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":7750,"visible":true,"origin":"","legend":"","description":"","filename":"a42fbcc54b9f41eb81a423733f758ffd.json","url":"https://assets-eu.researchsquare.com/files/rs-7386815/v1/48d4c0c3032f790741242f3f.json"},{"id":93947348,"identity":"58ba883a-9084-4466-ae7f-84b958b92cde","added_by":"auto","created_at":"2025-10-20 14:26:38","extension":"xml","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":64456,"visible":true,"origin":"","legend":"","description":"","filename":"a42fbcc54b9f41eb81a423733f758ffd1enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-7386815/v1/ee0b135c702734c790da9e92.xml"},{"id":93949342,"identity":"570ff1bb-85ad-41c6-8828-d7293f8f54f1","added_by":"auto","created_at":"2025-10-20 14:42:38","extension":"png","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":187595,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7386815/v1/673cf16015aefa7727c0d60d.png"},{"id":93947354,"identity":"add724d3-1593-4b8f-a7ee-c3a6351e5631","added_by":"auto","created_at":"2025-10-20 14:26:38","extension":"jpeg","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":562527,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7386815/v1/1d7b1e5089b95d0c6c25e60c.jpeg"},{"id":93947363,"identity":"0200e3fb-4d82-4902-82b0-347d889ed1f4","added_by":"auto","created_at":"2025-10-20 14:26:38","extension":"png","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":184390,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7386815/v1/070b7e490ea2398bcd9383f1.png"},{"id":93947357,"identity":"3cee8013-9e37-4626-a569-3d83cbb1f686","added_by":"auto","created_at":"2025-10-20 14:26:38","extension":"png","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":687582,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7386815/v1/e100b0ef778788c375935305.png"},{"id":93947356,"identity":"9c2acd96-ea45-4af0-a773-5dc673f33cf4","added_by":"auto","created_at":"2025-10-20 14:26:38","extension":"png","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":429093,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-7386815/v1/88964eafcb61afe0d727d45e.png"},{"id":93947977,"identity":"b7a85361-18ca-4d41-9b2a-eca697dd7d35","added_by":"auto","created_at":"2025-10-20 14:34:38","extension":"png","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":464585,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-7386815/v1/12c91b2bc53b4977a3133db0.png"},{"id":93947364,"identity":"ec03bfa6-7f5d-4bad-89d4-44fa20373d72","added_by":"auto","created_at":"2025-10-20 14:26:39","extension":"png","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":339519,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-7386815/v1/10225fb99cfcfa2d2215e285.png"},{"id":93947359,"identity":"18a3e76c-2b02-47e0-a951-77d93a29205d","added_by":"auto","created_at":"2025-10-20 14:26:38","extension":"png","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":42206,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7386815/v1/e47c35e5a231a8eafbe50158.png"},{"id":93947362,"identity":"09eeb35b-c7eb-478e-b0b1-0ff5b484b23e","added_by":"auto","created_at":"2025-10-20 14:26:38","extension":"png","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":101917,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7386815/v1/35d3dafbfd08228e7b5ef5fb.png"},{"id":93947368,"identity":"3f2d4137-0c63-4044-8245-d85423c98a99","added_by":"auto","created_at":"2025-10-20 14:26:39","extension":"png","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":95617,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7386815/v1/0cdc9897d08e8b66104b83f3.png"},{"id":93947980,"identity":"ecc61ca2-ed72-4eca-859d-178a0bdcd2ed","added_by":"auto","created_at":"2025-10-20 14:34:39","extension":"png","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":147743,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7386815/v1/9f7c8374f933547b90ff9551.png"},{"id":93949343,"identity":"97dca21f-5db4-4956-a38a-7f891829161b","added_by":"auto","created_at":"2025-10-20 14:42:38","extension":"png","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":78170,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-7386815/v1/989c0c91ace98803b98509a6.png"},{"id":93947365,"identity":"2893fa4c-4bc7-42a2-a624-b21bc55334b4","added_by":"auto","created_at":"2025-10-20 14:26:39","extension":"png","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":93645,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-7386815/v1/7cf7d003d8c511ad54d82f24.png"},{"id":93947979,"identity":"ed431836-3902-4a65-ae2f-10a9ffcb03d6","added_by":"auto","created_at":"2025-10-20 14:34:39","extension":"png","order_by":16,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":71721,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-7386815/v1/b4ef3a7e41838a5dafcc1c16.png"},{"id":93947369,"identity":"bff8d7d9-6627-4789-a2f0-69669db3830d","added_by":"auto","created_at":"2025-10-20 14:26:39","extension":"xml","order_by":17,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":60847,"visible":true,"origin":"","legend":"","description":"","filename":"a42fbcc54b9f41eb81a423733f758ffd1structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-7386815/v1/3749f6099500fbc756c0e91c.xml"},{"id":93949344,"identity":"bca45f90-a303-452d-9555-4ce01a193af1","added_by":"auto","created_at":"2025-10-20 14:42:38","extension":"html","order_by":18,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":70428,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-7386815/v1/987b9afc9c8645c7d3d53ccc.html"},{"id":93947973,"identity":"9987fc85-e532-4a61-aa09-a21acf71b8cc","added_by":"auto","created_at":"2025-10-20 14:34:38","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":275601,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend\u0026nbsp;\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-7386815/v1/263772dd214e836779962254.png"},{"id":93947974,"identity":"d3a6cbe6-66cd-4f2f-bf18-7d45535c2379","added_by":"auto","created_at":"2025-10-20 14:34:38","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":246015,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend\u0026nbsp;\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-7386815/v1/c4c1fd961aa6ee3456c85220.png"},{"id":93947350,"identity":"0cb418a2-eba1-4053-b7fa-06c0d9a3bd5f","added_by":"auto","created_at":"2025-10-20 14:26:38","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":532327,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend\u0026nbsp;\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-7386815/v1/aa29cc2d1dacf9d85a2e5b2b.png"},{"id":93947975,"identity":"b32cc7eb-6e0e-478e-ad9d-1036814bf234","added_by":"auto","created_at":"2025-10-20 14:34:38","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":830529,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend\u0026nbsp;\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-7386815/v1/4179e628bdea6e0e8dd08447.png"},{"id":93947976,"identity":"05451648-9844-45cc-a360-9757363e0a53","added_by":"auto","created_at":"2025-10-20 14:34:38","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":681229,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend\u0026nbsp;\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-7386815/v1/24ab503adc98b40ddf88ad55.png"},{"id":93949354,"identity":"78a84f96-4f0c-40b7-b979-4361d4642b2d","added_by":"auto","created_at":"2025-10-20 14:42:44","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":3040233,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7386815/v1/c4102919-bfe3-48a9-b19b-4f27b8354b2b.pdf"},{"id":93947345,"identity":"4755cd26-a09c-4a21-9d95-eeea4fa7e9d7","added_by":"auto","created_at":"2025-10-20 14:26:38","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":749576,"visible":true,"origin":"","legend":"","description":"","filename":"Tables.docx","url":"https://assets-eu.researchsquare.com/files/rs-7386815/v1/caa567bdc820001930773c0d.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Group-Based Trajectory Modeling of Serum Sodium and Mortality in Critically Ill ICU Patients with Community-Acquired Pneumonia: An Analysis of the MIMIC-IV Database","fulltext":[{"header":"Introduction","content":"\u003cp\u003eCommunity-acquired pneumonia (CAP) is among the most prevalent infectious diseases globally and remains a leading cause of hospital admission and death in adults. In the United States, it is estimated to result in roughly 1.4\u0026nbsp;million emergency department visits, 740 000 hospitalizations, and about 41 000 deaths each year\u003csup\u003e[\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]\u003c/sup\u003e. Although most cases are treated successfully in outpatient settings, approximately one in ten patients require hospitalization, and of these, 20\u0026ndash;25% are admitted to an intensive care unit (ICU). Severe CAP carries a particularly grave prognosis, with short- and long-term mortality reported between 27% and 50%, despite modern advances in antimicrobial therapy and critical care support\u003csup\u003e[\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]\u003c/sup\u003e. This substantial disease burden is driven by the complex pathophysiology of CAP, encompassing widespread inflammatory responses, multiple organ dysfunction, and significant disruptions in physiological homeostasis.\u003c/p\u003e\u003cp\u003eMost prior studies have relied on single sodium measurements at ICU admission, overlooking the dynamic fluctuations that occur during early critical illness. In reality, sodium levels change rapidly with disease progression, therapeutic interventions, and fluid balance shifts. Persistent abnormalities may indicate unresolved pathology, whereas normalization may signal recovery. These temporal patterns may offer stronger prognostic value than static values, especially for long-term outcomes, as excess mortality often extends beyond hospitalization due to persistent inflammation and organ injury\u003csup\u003e[\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eGroup-based trajectory modeling (GBTM) offers a robust method to identify distinct longitudinal patterns in repeated biomarker measurements, accommodating unbalanced or missing data. This approach has been widely applied in critical care research, revealing prognostically distinct subgroups in contexts\u003csup\u003e[\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]\u003c/sup\u003e. However, studies specifically employing GBTM to characterize early serum sodium trajectories in critically ill patients with severe CAP remain scarce.\u003c/p\u003e\u003cp\u003eUsing the MIMIC-IV database, our objectives were threefold: (1) to characterize distinct patterns of serum sodium changes over the first seven days of ICU admission in patients with CAP; (2) to compare clinical characteristics among the identified trajectory groups; and (3) to examine how these patterns relate to mortality at 28 days, 90 days, and 365 days, using the identified highest-risk trajectory as the primary reference for comparison. Incorporating temporal sodium profiles into prognostic evaluation may improve risk stratification, help identify vulnerable patient subgroups, and guide individualized management\u0026mdash;such as optimized fluid strategies and extended monitoring\u0026mdash;aimed at improving outcomes in critically ill patients with CAP.\u003c/p\u003e"},{"header":"Materials and Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003eData source\u003c/h2\u003e\u003cp\u003eThe Medical Information Mart for Intensive Care (MIMIC) database, established in 2003, is a collaborative initiative supported by the Beth Israel Deaconess Medical Center (BIDMC), the National Institutes of Health, Massachusetts General Hospital, emergency physicians, intensivists, computer science experts, and various other professional critical care medicine networks (Yang et al., 2020). This observational study utilized the MIMIC-IV database (version 3.1; PhysioNet. RRID:SCR_007345; \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.13026/kpb9-mt58\u003c/span\u003e\u003cspan address=\"10.13026/kpb9-mt58\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e; covering ICU admissions from 2008 to 2022) for patients admitted to Beth Israel Deaconess Medical Center (BIDMC) in Boston, Massachusetts\u003csup\u003e[\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]\u003c/sup\u003e. This includes a wide range of patient information, such as demographic data, vital signs, comorbidities, laboratory results, and mortality outcomes. The ethical approval statement and informed consent were not required for this study because all personal information was replaced with random codes and anonymized. Author Jingjing Cai completed a training program facilitated by the PhysioNet team and secured official approval to use the MIMIC-IV database (certification ID: 70520306). The study's methodology and reporting adhered to the guidelines set forth by the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) initiative\u003csup\u003e[\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]\u003c/sup\u003e.\u003c/p\u003e\u003c/div\u003e\n\u003ch3\u003ePatient Selection\u003c/h3\u003e\n\u003cp\u003ePatients were eligible for screening if they were diagnosed with CAP during hospitalization, as determined by the International Classification of Diseases, Ninth or Tenth Revision (ICD-9/10) codes. For patients with more than one ICU admission, only the first was included in the analysis. The inclusion criteria were: (1) age\u0026thinsp;\u0026ge;\u0026thinsp;18 years at the time of ICU admission; and (2) pneumonia listed among the top ten principal discharge diagnoses. Exclusion criteria were: (1) age\u0026thinsp;\u0026lt;\u0026thinsp;18 years; (2) pneumonia not listed among the top ten principal discharge diagnoses; (3) diagnosis of ventilator-associated pneumonia; or (4) ICU stay of fewer than seven consecutive days.\u003c/p\u003e\u003cp\u003eAfter applying these criteria, 3,146 patients were included in the final analysis (\u003cb\u003eFig.\u0026nbsp;1\u003c/b\u003e). For each patient, all available serum sodium measurements from the first 7 days of ICU stay were extracted to construct longitudinal sodium trajectories using GBTM. Baseline demographic characteristics, comorbidities, illness severity scores, laboratory parameters, and outcome variables were also collected for subsequent statistical analyses.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\n\u003ch3\u003eVariable collection\u003c/h3\u003e\n\u003cp\u003eWe collected baseline demographic characteristics (age, sex, race, and marital status), vital signs (heart rate, mean arterial pressure, respiratory rate, temperature, and SpO₂), and laboratory parameters (glucose, chloride, potassium, hematocrit, red cell distribution width(RDW), hemoglobin, platelet count, white blood cell count, blood urea nitrogen(BUN), total bilirubin, albumin, lactate, arterial PO₂, arterial PCO₂, pH, base excess, anion gap, and bicarbonate), as well as enteral and parenteral nutrition status. Disease severity was assessed using the Acute Kidney Injury (AKI) stage, Sequential Organ Failure Assessment (SOFA) score, Acute Physiology and Chronic Health Evaluation II (APACHE II), and Charlson Comorbidity Index (CCI). Additional treatment and clinical history variables included the use of vasopressors, mechanical ventilation, and renal replacement therapy (RRT), along with comorbidities such as hypertension, congestive heart failure, myocardial infarction, cerebrovascular accident, chronic kidney disease, chronic obstructive pulmonary disease, liver cirrhosis, diabetes, and cancer.\u003c/p\u003e\u003cp\u003eIf multiple measurements of the same variable were available within the first 24 hours after ICU admission, the value most deviated from the normal range was selected for baseline characterization and covariate adjustment. Variables with more than 20% missing data were excluded, and for the remaining variables, missing values were imputed using the random forest algorithm implemented in the \u003cem\u003emice\u003c/em\u003e package.\u003c/p\u003e\u003cp\u003eFor serum sodium, the observation period was defined as days 1 to 7 following ICU admission. For each day, only the first available sodium measurement was used, providing up to seven time points per patient. To ensure reliable trajectory estimation, patients were required to have at least four daily sodium values within this interval. The primary outcomes were all-cause mortality at 28, 90, and 365 days.\u003c/p\u003e\n\u003ch3\u003eTrajectory grouping of serum sodium content\u003c/h3\u003e\n\u003cp\u003eWe applied GBTM to identify distinct longitudinal patterns of serum sodium concentration over the first 7 days after ICU admission. The time scale was defined in hours since ICU admission, with each patient contributing one sodium value per 24 hours as described above.\u003c/p\u003e\u003cp\u003eTrajectory models were fitted using the \u003cem\u003elcmm\u003c/em\u003e package in R, initially testing linear, quadratic, and cubic polynomial functions of time without covariates. We compared models with different numbers of trajectories and polynomial orders, selecting the optimal solution based on the Akaike information criterion (AIC), Bayesian information criterion (BIC), sample size\u0026ndash;adjusted BIC (SABIC), entropy, the proportion of patients in each trajectory group, and the average posterior probability of group membership (AvePP)\u003csup\u003e[\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]\u003c/sup\u003e. A five-trajectory solution was pre-specified to enhance clinical interpretability and ensure adequate representation of each subgroup. The final model required that each trajectory group include at least 5% of the cohort and have an AvePP of 0.70 or higher. Each patient was then assigned to the trajectory group corresponding to the highest posterior probability, and these assignments were subsequently used for survival analyses and subgroup interaction testing.\u003c/p\u003e\n\u003ch3\u003eStatistical Analyses\u003c/h3\u003e\n\u003cp\u003eBaseline characteristics were summarized for the overall cohort and by sodium-trajectory groups. Continuous variables were tested for normality using the Shapiro\u0026ndash;Wilk test. Normally distributed variables are presented as mean\u0026thinsp;\u0026plusmn;\u0026thinsp;standard deviation (SD) and compared using one-way ANOVA. Categorical variables are presented as counts (percentages) and compared using the χ\u0026sup2; test or Fisher\u0026rsquo;s exact test. The primary exposure was the latent trajectory of serum sodium during the first 7 days after ICU admission. Group-based trajectory modeling (GBTM) with a censored normal distribution was used to identify distinct sodium-trajectory patterns. The optimal model was selected using the Bayesian Information Criterion (BIC), average posterior probability of assignment (APPA\u0026thinsp;\u0026ge;\u0026thinsp;0.70 for all classes), entropy\u0026thinsp;\u0026ge;\u0026thinsp;0.70, odds of correct classification (OCC\u0026thinsp;\u0026gt;\u0026thinsp;5), and clinical interpretability. Patients with missing data for key variables, including serum sodium, were excluded from analysis.\u003c/p\u003e\u003cp\u003eOutcomes included all-cause mortality at 28, 90, and 365 days after admission. Kaplan\u0026ndash;Meier curves were plotted for each trajectory group, and survival differences were compared using the log-rank test. Cox proportional hazards models were fitted to estimate hazard ratios (HRs) for the association between trajectory group and mortality at each follow-up. Because preliminary analyses showed the highest risk in Trajectory 2, this group was used as the reference to facilitate clinical interpretation. Models were specified as follows: the crude model was unadjusted; Model I adjusted for age, sex, and race; Model II further adjusted for ICU length of stay, hospital length of stay, Charlson Comorbidity Index (CCI), SOFA, and APACHE II scores; and Model III additionally adjusted for vital signs, laboratory parameters, nutrition status, additional treatments, clinical history variables, and comorbidities. An HR\u0026thinsp;\u0026lt;\u0026thinsp;1 indicated lower mortality risk relative to Trajectory 2, and an HR\u0026thinsp;\u0026gt;\u0026thinsp;1 indicated higher risk. All analyses were conducted in R (version 4.5.0; R Foundation for Statistical Computing, Vienna, Austria). Two-sided p-values\u0026thinsp;\u0026lt;\u0026thinsp;0.05 were considered statistically significant.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003eIn total, 3,146 patients with CAP who fulfilled the inclusion criteria were analyzed. Using group-based trajectory modeling (GBTM), we delineated five distinct patterns of serum sodium evolution during the first seven days of ICU stay (\u003cb\u003eFig.\u0026nbsp;2\u003c/b\u003e): Trajectory 1 (\u0026ldquo;normal level with slow increase\u0026rdquo;), Trajectory 2 (\u0026ldquo;rapid increase followed by slow decrease\u0026rdquo;), Trajectory 3 (\u0026ldquo;low level with slow increase\u0026rdquo;), Trajectory 4 (\u0026ldquo;low level with rapid increase\u0026rdquo;), and Trajectory 5 (\u0026ldquo;normal level, stable\u0026rdquo;). \u003cb\u003eTable\u0026nbsp;1\u003c/b\u003e presents the fit statistics for the group-based trajectory model. The five-class model met the selection criteria, with all trajectory groups containing more than 5% of participants and average posterior probabilities exceeding 0.7, indicating adequate classification accuracy. Additionally, the model showed favorable fit according to key selection criteria, including the Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), and sample size\u0026ndash;adjusted BIC (SABIC), further supporting its optimality in modeling the data.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eNotably, baseline demographic and clinical profiles varied significantly among the five trajectory groups (\u003cb\u003eTable\u0026nbsp;2\u003c/b\u003e), suggesting potential heterogeneity in underlying pathophysiology and disease progression. Significant differences were observed in demographic factors, comorbidities, illness severity scores, and clinical parameters among the groups. Variables such as age, sex distribution, race, Charlson comorbidity index, SOFA, APACHE II scores, ICU and hospital length of stay, as well as laboratory values at admission, varied notably between trajectories, indicating distinct patient profiles associated with each sodium trend pattern.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eKaplan-Meier curves demonstrated significant survival differences across trajectories at 28 days (global log-rank p\u0026thinsp;=\u0026thinsp;0.019) (\u003cb\u003eFig.\u0026nbsp;3A\u003c/b\u003e). Kaplan-Meier survival curves comparing individual trajectories to the reference trajectory 2 confirmed significantly higher survival probabilities for trajectory 5 (\u003cb\u003eFig.\u0026nbsp;3B\u0026ndash;D\u003c/b\u003e) and trajectory 1 (\u003cb\u003eFig.\u0026nbsp;3E\u0026ndash;G\u003c/b\u003e) at 28, 90, and 365 days.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eIn multivariable Cox proportional hazards analyses using trajectory 2 as the reference, distinct serum sodium trajectories were associated with differential mortality risks (\u003cb\u003eTable\u0026nbsp;3\u003c/b\u003e). For 28-day mortality, trajectories 1 (HR 0.58, 95% CI 0.45\u0026ndash;0.76, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001), 3 (HR 0.51, 95% CI 0.36\u0026ndash;0.73, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001), and 5 (HR 0.50, 95% CI 0.38\u0026ndash;0.65, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001) were significantly associated with lower risk, whereas trajectory 4 showed no significant difference (HR 0.80, 95% CI 0.55\u0026ndash;1.18, p\u0026thinsp;=\u0026thinsp;0.263). Similar patterns were observed for 90-day mortality, with trajectories 1 (HR 0.74, 95% CI 0.60\u0026ndash;0.92, p\u0026thinsp;=\u0026thinsp;0.006), 3 (HR 0.68, 95% CI 0.51\u0026ndash;0.91, p\u0026thinsp;=\u0026thinsp;0.009), and 5 (HR 0.63, 95% CI 0.50\u0026ndash;0.79, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001) remaining protective. For 365-day mortality, trajectories 1 (HR 0.75, 95% CI 0.62\u0026ndash;0.91, p\u0026thinsp;=\u0026thinsp;0.004) and 5 (HR 0.68, 95% CI 0.55\u0026ndash;0.83, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001) retained a significant survival advantage, while trajectories 3 and 4 were not significant. Overall, trajectories 1, 3, and 5 were consistently linked to reduced mortality risk compared to the high-risk trajectory 2, with trajectories 1 and 5 also showing favorable long-term outcomes.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eIn the age-stratified Cox proportional hazards analysis (\u003cb\u003eFig.\u0026nbsp;4\u003c/b\u003e), using trajectory 2 as the reference, distinct patterns of mortality risk were observed across different age groups. For patients aged\u0026thinsp;\u0026lt;\u0026thinsp;65 years, trajectory 1 was consistently associated with significantly lower mortality risk compared to trajectory 2 across all time points and models, with HRs ranging from 0.35 (95% CI: 0.22\u0026ndash;0.55, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001) in Model I for 28-day mortality to 0.55 (95% CI: 0.40\u0026ndash;0.76, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001) in Model II for 365-day mortality. Similarly, trajectory 5 showed a significant survival advantage over Trajectory 2, with HRs between 0.38 (95% CI: 0.24\u0026ndash;0.60, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001) and 0.72 (95% CI: 0.53\u0026ndash;0.97, p\u0026thinsp;=\u0026thinsp;0.032). In contrast, Trajectories 3 and 4 demonstrated inconsistent associations, with several HRs close to 1.0 and non-significant p-values (e.g., Trajectory 3 vs 2: HR\u0026thinsp;=\u0026thinsp;0.94, 95% CI: 0.58\u0026ndash;1.54, p\u0026thinsp;=\u0026thinsp;0.822 for 28-day mortality in Model I). For patients aged\u0026thinsp;\u0026ge;\u0026thinsp;65 years, the same general pattern was observed, though effect sizes were attenuated. Trajectory 1 remained protective in most models (e.g., HR\u0026thinsp;=\u0026thinsp;0.66, 95% CI: 0.50\u0026ndash;0.88, p\u0026thinsp;=\u0026thinsp;0.005 in Model III for 28-day mortality; HR\u0026thinsp;=\u0026thinsp;0.78, 95% CI: 0.63\u0026ndash;0.96, p\u0026thinsp;=\u0026thinsp;0.017 in Model II for 365-day mortality). Trajectory 5 also showed favorable outcomes, with HRs between 0.56 (95% CI: 0.41\u0026ndash;0.75, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001) and 0.76 (95% CI: 0.61\u0026ndash;0.94, p\u0026thinsp;=\u0026thinsp;0.013). However, the differences for Trajectories 3 and 4 compared to Trajectory 2 were generally not statistically significant in older patients. Age-stratified analysis confirmed that the survival advantage of Trajectories 1 and 5 over the high-risk Trajectory 2 was evident in both younger (\u0026lt;\u0026thinsp;65 years) and older (\u0026ge;\u0026thinsp;65 years) patients, although the magnitude of risk reduction was generally greater in the younger group.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eIn the CKD-stratified Cox proportional hazards analysis (\u003cb\u003eFig.\u0026nbsp;5\u003c/b\u003e), using trajectory 2 (initial increase followed by decrease) as the reference, distinct patterns of mortality risk emerged across trajectory groups and follow-up periods (28-day, 90-day, and 365-day). For non-CKD patients, trajectory 1 (normal level with slow increase) demonstrated consistently lower mortality risk across all time points, with HRs ranging from 0.54 (95% CI: 0.41\u0026ndash;0.70, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001) for 28-day mortality to 0.32 (95% CI: 0.19\u0026ndash;0.55, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001) for 365-day mortality in Model III. Trajectory 5 (normal stable) also showed a significant survival benefit, with HRs between 0.38 (95% CI: 0.24\u0026ndash;0.60, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001) and 0.56 (95% CI: 0.40\u0026ndash;0.75, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). In contrast, trajectories 3 and 4 exhibited inconsistent associations, with HRs close to 1.0 and non-significant p-values, suggesting weaker associations. Stratification by CKD status revealed that the significant reduction in mortality risk associated with trajectories 1 and 5, compared to trajectory 2, was substantially stronger and more consistent in patients without pre-existing CKD than in those with CKD.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eIn this large retrospective cohort of critically ill patients with CAP, we identified five distinct early ICU serum sodium trajectories using GBTM. Our analysis indicates that dynamic changes in sodium levels rather than a single baseline measurement\u0026mdash;carry important prognostic value. Our analysis underscores that dynamic changes in sodium levels, rather than a single baseline measurement, provide significant prognostic value for both short- and long-term mortality. Specifically, using the highest-risk trajectory (trajectory 2) as the reference, our analysis revealed that trajectory 5 and trajectory 1 were consistently associated with significantly lower mortality risk at all time points. trajectory 3 also showed reduced risk in the short term. trajectory 2 itself was confirmed as the pattern associated with the poorest prognosis. These results suggest that the sodium trajectory pattern in trajectory 2 may serve as a strong prognostic marker for both short- and long-term survival.\u003c/p\u003e\u003cp\u003eIn the age-stratified Cox proportional hazards analysis, distinct patterns of mortality risk were observed based on serum sodium trajectory groups, with consistently lower mortality risks for trajectory 1 and trajectory 5 across all follow-up periods. However, the strength of these associations was more pronounced in non-CKD patients, suggesting that the impact of sodium trajectory on mortality may be more robust in this group. In non-CKD patients, trajectories 1 and 5 consistently showed survival advantages versus trajectory 2, whereas trajectories 3 and 4 showed weaker and often non-significant associations, while in the CKD group, the associations were weaker, and the differences between trajectories 3 and 4 compared to trajectory 2 were not statistically significant.\u003c/p\u003e\u003cp\u003eThis difference in findings between CKD and non-CKD groups may highlight the altered physiological response to sodium fluctuations in patients with chronic kidney disease. The weaker associations in CKD patients suggest that other factors, such as renal function, fluid balance, and comorbidities, may attenuate the relationship between sodium trajectory and mortality. Therefore, the findings emphasize the need for tailored clinical management strategies for non-CKD patients, where sodium levels may not be as directly linked to mortality as in CKD patients. The diminished association in CKD patients may reflect their impaired renal capacity to regulate sodium and water homeostasis, potentially blunting the impact of acute fluctuations observed in the trajectories. Furthermore, the high burden of competing risks (e.g., cardiovascular disease, chronic inflammation) in CKD may obscure the specific contribution of short-term sodium dynamics to mortality. Future research should further explore the underlying mechanisms driving these differential associations and consider other biomarkers or clinical features that might explain these variations.\u003c/p\u003e\u003cp\u003eThese findings align with previous evidence that hypernatremia is independent predictors of mortality in critically ill patients\u003csup\u003e[\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]\u003c/sup\u003e. Our results build on this knowledge by showing that early sodium fluctuations, particularly rapid rises, offer additional prognostic value\u003csup\u003e[\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]\u003c/sup\u003e. ICU-acquired hypernatremia, often driven by excessive sodium administration, osmotic shifts, or impaired water regulation, has been associated with nearly doubled in-hospital mortality risk. In our cohort, trajectory 2 likely reflects acute hypernatremia precipitated by aggressive diuresis or sodium loading during initial resuscitation, followed by partial correction, whereas trajectory 4 may represent severe hyponatremia at admission with rapid overcorrection. Both patterns indicate marked instability in electrolyte homeostasis and significant physiological stress, which may partly explain the observed increase in adverse outcomes. From a pathophysiological perspective, sodium trajectories may serve as a surrogate marker of the host\u0026rsquo;s overall systemic homeostasis, analogous to the role of dynamic immune biomarkers in differentiating sepsis phenotypes. Abrupt sodium elevations in high-risk trajectories may signify a proinflammatory, hypercatabolic phase accompanied by osmotic stress at the cellular level, potentially leading to neurologic injury, impaired cardiac function, or further immune dysregulation\u003csup\u003e[\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e]. In contrast, maintenance of normonatremia suggests preserved electrolytes balance and is associated with improved survival\u003csup\u003e[\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]\u003c/sup\u003e, resembling the \u0026ldquo;normal, stable\u0026rdquo; immune profile described in other trajectory-based investigations.\u003c/p\u003e\u003cp\u003eClinically, these findings suggest that early monitoring of sodium trajectories may aid in stratifying ICU patients with CAP into distinct risk categories. Patients with rising sodium levels indicative of impending hypernatremia (trajectory 2) may benefit from tighter fluid balance, judicious diuretic use, and proactive sodium restriction. In contrast, those with severe hyponatremia undergoing rapid correction (trajectory 4) require adherence to recommended correction rates, timely administration of hypotonic fluids when appropriate, and close neurological surveillance to reduce iatrogenic risk. Patients maintaining stable normonatremia (trajectory 5) likely represent a lower-risk group for whom standard CAP management remains appropriate.\u003c/p\u003e\u003cp\u003eOur study builds upon the findings of previous study\u003csup\u003e[\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]\u003c/sup\u003e, which also demonstrated that both stable and fluctuating sodium levels were associated with mortality. However, we extend their work by offering a more nuanced understanding of the prognostic significance of sodium trajectory patterns, particularly emphasizing that the trajectory 2 pattern (rapid increase followed by slow decrease) independently predicted poor prognosis even after adjusting for chronic comorbidities and acute severity scores. Our findings underscore that rapid sodium fluctuations, likely induced by aggressive diuresis or sodium loading followed by partial correction, represent a significant risk and may lead to adverse outcomes due to the associated physiological stress.\u003c/p\u003e\u003cp\u003eOur findings show that changes in serum sodium over time offer important prognostic information beyond single measurements\u003csup\u003e[\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]\u003c/sup\u003e. Both rising and falling trajectories were independently linked to higher mortality compared with stable normonatremia, even after full adjustment. These results support the use of trajectory-based electrolyte monitoring as a simple and widely available tool in critical care. Recognizing sustained deviations early may prompt timely review of fluid management, renal function, and disease progression, enabling more individualized interventions and potentially improving outcomes.\u003c/p\u003e\u003cp\u003eOur analysis revealed a clinically significant paradox in the baseline characteristics of trajectory groups. Despite exhibiting the highest 28-day mortality (34.8%) among all trajectories, patients in trajectory 2 demonstrated lower acute illness severity scores (median SOFA: 5.0; APACHE II: 19.0) compared to trajectories 3 and 4 (SOFA: 7.0/8.0; APACHE II: 22.0/22.0) (\u003cb\u003eTable\u0026nbsp;2\u003c/b\u003e). Paradoxically, this group carried the heaviest chronic disease burden, as evidenced by the highest Charlson Comorbidity Index (median CCI: 6.0).\u003c/p\u003e\u003cp\u003eThis dissociation suggests that: traditional severity scores may underestimate risk in comorbid patients While SOFA/APACHE II quantify acute physiological derangement, they may fail to capture the diminished physiological reserve in patients with high CCI. The aberrant sodium trajectory (rapid rise/slow decline) likely reflects this vulnerability\u0026mdash;exposing an impaired capacity to maintain electrolyte homeostasis during acute stress. Sodium dynamics as a novel marker of physiological fragility. The trajectory pattern itself may serve as a biomarker of \"homeostatic incompetence\". Rapid sodium fluctuations could signify dysregulated neurohormonal responses (e.g., inappropriate ADH/RAAS activation) or subclinical organ dysfunction not captured by conventional scores\u0026mdash;particularly relevant in patients with chronic comorbidities. Our models adjusted for CCI, SOFA, and APACHE II still identified trajectory 2 as an independent mortality predictor (adjusted HRs 1.48\u0026ndash;2.01 vs. trajectory 5). This underscores that dynamic sodium monitoring provides prognostic information beyond static scores and comorbidity indices. In patients with high chronic disease burden (CCI\u0026thinsp;\u0026ge;\u0026thinsp;5), sodium trajectory analysis\u0026mdash;particularly the detection of trajectory 2 patterns\u0026mdash;may identify high-risk physiological fragility earlier than conventional metrics, enabling preemptive interventions to mitigate decompensation risk.\u003c/p\u003e\u003cp\u003eSeveral limitations should be noted. As with any observational study, residual confounding cannot be fully excluded despite adjustment for a wide range of covariates\u003csup\u003e[\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]\u003c/sup\u003e. Certain unmeasured factors, including the composition of administered fluids, the osmotic load from medications, and individual neuroendocrine responses, may still have influenced both sodium trajectories and outcomes. The frequency and timing of serum sodium measurements were determined by routine clinical practice rather than a standardized protocol, which may have introduced measurement bias. Although GBTM is designed to handle irregular and unbalanced data, variation in sampling intensity could still have affected trajectory classification. The consistency of associations observed across different adjustment models and follow-up periods lends support to the robustness of our findings, but cannot completely rule out bias from unmeasured variables. As repeated measurements were required to form trajectories, a degree of survivorship requirement is unavoidable and could influence estimates (akin to an immortal-time component). We therefore advise cautious interpretation. Finally, this single center retrospective analysis, based on the MIMIC-IV database, reflects the characteristics and practice patterns of one institution; replication in multicenter and prospective cohorts is necessary to confirm the reproducibility and broader applicability of these results.\u003c/p\u003e\u003cp\u003eDespite these limitations, the use of a large, well-characterized ICU cohort and a trajectory-based analytical framework allowed for a detailed assessment of dynamic sodium changes and their prognostic relevance in critically ill patients with CAP. While hypernatremia is recognized as a marker of physiological derangement and increased mortality in general critical care populations, its prognostic role in this specific group has not been well established. In our analysis, early sodium trajectories identified clinically distinct prognostic subgroups, with a rapid rise in sodium\u0026mdash;irrespective of baseline level\u0026mdash;associated with markedly higher mortality. These results highlight the value of close sodium surveillance during the early ICU course and support further investigation into timely, targeted interventions for patients at greatest risk. Well-designed prospective studies, ideally incorporating randomized strategies, are needed to determine whether trajectory-guided management can improve outcomes in this high-risk population.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis retrospective study analyzed the publicly available, de-identified MIMIC-IV database (version 3.1). The creation and distribution of MIMIC-IV were approved by the Institutional Review Boards of the Massachusetts Institute of Technology and Beth Israel Deaconess Medical Center, which granted a waiver of individual informed consent because the database contains only de-identified data. Access to the database was granted to the authors after completion of the required training and Data Use Agreement (CITI certification ID: 70520306). No additional patient contact or intervention occurred. (Repository: PhysioNet; version DOI: 10.13026/kpb9-mt58.)\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication.\u003c/strong\u003e Not applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding.\u003c/strong\u003e This research received no external funding.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eClinical trial registration/number.\u003c/strong\u003e Not applicable; this work is an observational secondary analysis of an existing database and not a registered interventional trial.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests.\u003c/strong\u003e The authors declare no competing interests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData availability.\u003c/strong\u003e The MIMIC-IV database is publicly available to credentialed researchers at physionet.org after completing the required training and data use agreement. \u003cem\u003e(optional but useful)\u003c/em\u003e\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eVaughn, V.M., et al., Community-Acquired Pneumonia: A Review. JAMA, 2024. 332(15): p. 1282-1295.\u003c/li\u003e\n\u003cli\u003eCill\u0026oacute;niz, C., A. Torres and M.S. Niederman, Management of pneumonia in critically ill patients. BMJ, 2021. 375: p. e065871.\u003c/li\u003e\n\u003cli\u003eGrim, C.C.A., et al., Association Between an Increase in Serum Sodium and In-Hospital Mortality in Critically Ill Patients. Crit Care Med, 2021. 49(12): p. 2070-2079.\u003c/li\u003e\n\u003cli\u003eNagin, D.S., B.L. Jones and J. Elmer, Recent Advances in Group-Based Trajectory Modeling for Clinical Research. Annu Rev Clin Psychol, 2024. 20(1): p. 285-305.\u003c/li\u003e\n\u003cli\u003eJohnson, A.E.W., et al., MIMIC-IV, a freely accessible electronic health record dataset. Sci Data, 2023. 10(1): p. 1.\u003c/li\u003e\n\u003cli\u003eJohnson A, Bulgarelli L, Pollard T, Gow B, Moody B, Horng S, Celi LA, Mark R (2024) MIMIC-IV (version 3.1). PhysioNet. 10.13026/kpb9-mt58..\u003c/li\u003e\n\u003cli\u003evon Elm, E., et al., The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. J Clin Epidemiol, 2008. 61(4): p. 344-9.\u003c/li\u003e\n\u003cli\u003eFunk, G., et al., Incidence and prognosis of dysnatremias present on ICU admission. Intensive Care Medicine, 2010. 36(2): p. 304-311.\u003c/li\u003e\n\u003cli\u003eNasser, A., et al., ICU-acquired hypernatremia: Prevalence, patient characteristics, trajectory, risk factors, and outcomes. Critical Care and Resuscitation, 2024. 26(4): p. 303-310.\u003c/li\u003e\n\u003cli\u003eSakr, Y., et al., Fluctuations in serum sodium level are associated with an increased risk of death in surgical ICU patients. Crit Care Med, 2013. 41(1): p. 133-42.\u003c/li\u003e\n\u003cli\u003eM\u0026uuml;ller, D.N., et al., Sodium in the microenvironment regulates immune responses and tissue homeostasis. Nat Rev Immunol, 2019. 19(4): p. 243-254.\u003c/li\u003e\n\u003cli\u003eKitisin, N., et al., Systematic review and meta-analysis of the treatment of hypernatremia in adult hospitalized patients: impact on mortality, morbidity, and treatment-related side effects. J Crit Care, 2025. 87: p. 155012.\u003c/li\u003e\n\u003cli\u003eChewcharat, A., et al., Trajectories of Serum Sodium on In-Hospital and 1-Year Survival among Hospitalized Patients. Clinical Journal of the American Society of Nephrology, 2020. 15(5): p. 600-607.\u003c/li\u003e\n\u003cli\u003eChewcharat, A., et al., Trajectories of Serum Sodium on In-Hospital and 1-Year Survival among Hospitalized Patients. Clin J Am Soc Nephrol, 2020. 15(5): p. 600-607.\u003c/li\u003e\n\u003cli\u003eHuang, S., et al., Association between serum sodium trajectory and mortality in patients with acute kidney injury: a retrospective cohort study. BMC Nephrol, 2024. 25(1): p. 152.\u003c/li\u003e\n\u003cli\u003eGao, Y., et al., Confounder adjustment in observational studies investigating multiple risk factors: a methodological study. BMC Med, 2025. 23(1): p. 132.\u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Tables","content":"\u003cp\u003eTables are available in the Supplementary Files section.\u003c/p\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"bmc-pulmonary-medicine","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"pulm","sideBox":"Learn more about [BMC Pulmonary Medicine](http://bmcpulmmed.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/pulm/default.aspx","title":"BMC Pulmonary Medicine","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Community-acquired pneumonia, sodium trajectory, intensive care unit, mortality, group-based trajectory modeling","lastPublishedDoi":"10.21203/rs.3.rs-7386815/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7386815/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003ePurpose\u003c/h2\u003e\u003cp\u003eCommunity-acquired pneumonia (CAP) is a major cause of prolonged intensive care unit (ICU) admissions and is associated with substantial mortality. Abnormalities in serum sodium are common in critically ill patients and may carry prognostic significance. This study examined early serum sodium trajectories in ICU patients with CAP and their associations with 28-day, 90-day, and 365-day mortality.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e\u003cp\u003eUsing the MIMIC-IV database, we retrospectively analyzed adult patients with a primary diagnosis of CAP, an ICU stay of more than 7 days, and at least four sodium measurements within the first 7 days. Group-based trajectory modeling was applied to identify distinct patterns of sodium change. Associations between trajectory groups and mortality were evaluated using multivariable Cox proportional hazards models, adjusting for demographics, comorbidities, illness severity, and laboratory parameters.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e\u003cp\u003eA total of 3,146 patients were classified into five distinct trajectories: normal level with slow increase (trajectory 1), rapid increase followed by slow decrease (trajectory 2), low level with slow increase (trajectory 3), low level with rapid increase (trajectory 4), and normal stable (trajectory 5). Using trajectory 2 as the reference group, trajectory 5 was associated with significantly lower mortality at 28 days (adjusted HR 0.50, 95% CI 0.38\u0026ndash;0.65, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), 90 days (HR 0.63, 95% CI 0.50\u0026ndash;0.79, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), and 365 days (HR 0.68, 95% CI 0.55\u0026ndash;0.83, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). Similarly, trajectories 1 (28-day HR 0.58, 95% CI 0.45\u0026ndash;0.76, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001) and 3 (28-day HR 0.51, 95% CI 0.36\u0026ndash;0.73, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001) were associated with lower risk. Trajectory 4 showed no significant difference in 28-day mortality compared to trajectory 2 (HR 0.80, 95% CI 0.55\u0026ndash;1.18, p\u0026thinsp;=\u0026thinsp;0.263). Trajectory 2 was consistently associated with the highest mortality risk. Kaplan-Meier analysis demonstrated significant survival differences among the trajectory groups at 28 days (global log-rank test p\u0026thinsp;=\u0026thinsp;0.019).\u003c/p\u003e\u003ch2\u003eConclusions\u003c/h2\u003e\u003cp\u003eDistinct early sodium trajectories are independently associated with mortality in ICU patients with CAP and prolonged stays. The rapid increase followed by slow decrease pattern (trajectory 2), indicative of impaired physiological resilience, was linked to the poorest outcomes, whereas the normal stable pattern (trajectory 5) predicted optimal survival. Trajectory-based sodium monitoring may have relevance for early risk stratification and for consideration of individualized electrolyte management.\u003c/p\u003e","manuscriptTitle":"Group-Based Trajectory Modeling of Serum Sodium and Mortality in Critically Ill ICU Patients with Community-Acquired Pneumonia: An Analysis of the MIMIC-IV Database","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-10-20 14:26:33","doi":"10.21203/rs.3.rs-7386815/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"reviewersInvited","content":"","date":"2025-10-08T05:08:51+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-09-23T09:05:40+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2025-09-03T11:40:54+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-09-02T18:09:41+00:00","index":"","fulltext":""},{"type":"submitted","content":"BMC Pulmonary Medicine","date":"2025-09-02T18:06:39+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"bmc-pulmonary-medicine","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"pulm","sideBox":"Learn more about [BMC Pulmonary Medicine](http://bmcpulmmed.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/pulm/default.aspx","title":"BMC Pulmonary Medicine","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"97931193-95ea-49d2-a6c4-48a59ac54321","owner":[],"postedDate":"October 20th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2025-10-20T14:26:33+00:00","versionOfRecord":[],"versionCreatedAt":"2025-10-20 14:26:33","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7386815","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7386815","identity":"rs-7386815","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}