The absolute event rates differed between the databases, underscoring the importance of distinguishing a reproducible relative association from a universal risk estimate. In MIMIC-IV, invasive reintubation within 48 hours occurred in 13.6% of 1,124 extubations with calculated osmolality above 300 mOsm/kg and 10.8% of 1,515 extubations at or below that level, an absolute difference of 2.8 percentage points. In eICU-CRD, the respective rates were 10.5% of 486 extubations and 4.6% of 904, a difference of 5.8 percentage points. The association also persisted when the outcome window was extended to 72 hours, with adjusted risk ratios of 1.25 (1.02–1.52) in MIMIC-IV and 1.53 (1.19–1.97) in eICU-CRD. The replication therefore involved the same exposure definition, outcome definitions, adjustment variables, transformations and analytic plan, rather than a threshold or model selected from the replication dataset.

Calculated serum osmolality was derived from the most recent complete set of sodium, glucose and blood urea nitrogen measurements available within 24 hours before extubation, with no more than two hours separating component collection times. The formula was 2 × sodium + glucose/18 + blood urea nitrogen/2.8, with sodium in millimoles per liter and glucose and blood urea nitrogen in milligrams per deciliter. Median calculated osmolality was 297.5 mOsm/kg in MIMIC-IV and 295.2 mOsm/kg in eICU-CRD; 42.6% and 35.0% of the respective cohorts exceeded 300 mOsm/kg. The investigators emphasized that this calculated value is not interchangeable with directly measured serum osmolality or effective osmolality, which excludes urea. It functions instead as a practical summary combining water–solute balance, glycemic stress and renal or catabolic burden using routinely available laboratory data.

The retrospective cohort included the first eligible extubation in each hospital admission after adults had received invasive mechanical ventilation. Exclusions covered tracheostomy conversion, terminal or palliative extubation, death-related cessation of ventilation and boundaries that could not be classified reliably. Noninvasive ventilation, continuous or bilevel positive airway pressure, high-flow nasal oxygen and standby ventilator modes were not treated as invasive ventilation. Invasive reintubation was tracked throughout the same hospital admission, and noninvasive respiratory support alone did not count as reintubation. Sensitivity analyses addressed planned versus unplanned extubation, potentially procedure-related reintubation, repeated admissions, narrower exposure windows, respiratory severity, fluid balance, renal status and identifiable hypertonic treatment.

Adjustment covered age, sex, ventilation duration, pre-extubation respiratory rate, oxygen saturation, Glasgow Coma Scale score, serum creatinine and vasopressor use during the preceding 24 hours. Missing covariates were handled separately within each database with 30 multiply imputed datasets, and complete-case analyses served as sensitivity checks. Additional respiratory adjustment had little effect within the subsets where PaO2/FiO2 was recorded. For example, in 1,329 MIMIC-IV extubations with 169 reintubations, the risk ratio was 1.47 both before and after adjustment for the oxygenation ratio. In 771 eICU-CRD extubations with 52 events, the corresponding estimates were 1.21 and 1.23, although their confidence intervals included no association. Adjustment for recorded fluid balance, loop-diuretic exposure and creatinine-based acute kidney injury likewise produced little change in the main pattern, but several analyses contained few events and were imprecise.

Exposure shape and biochemical-component analyses suggested that the result was not explained by a single laboratory value. Neighboring thresholds of above 295, 300 and 305 mOsm/kg produced associations in the same direction, although precision varied. An exploratory MIMIC-IV-derived threshold of 301.964 mOsm/kg, applied unchanged to eICU-CRD, corresponded to a risk ratio of 2.04 (1.41–2.96), but bootstrap-derived thresholds had substantial dispersion. Continuous associations differed between databases: each 10-mOsm/kg increase had an adjusted risk ratio of 1.03 (0.96–1.11) in MIMIC-IV and 1.21 (1.06–1.39) in eICU-CRD. Neither restricted cubic spline analysis showed evidence of nonlinearity. Separately modeled sodium, glucose and blood urea nitrogen did not reproduce the categorical association across both databases, while effective osmolality, which excludes urea, showed weaker estimates. The authors consequently interpret 300 mOsm/kg as a pragmatic stratification point rather than a sharply defined biological threshold.

Competing-event analyses showed that hyperosmolarity was also associated with death before reintubation within 48 hours. Cause-specific hazard ratios were 1.29 (1.02–1.63) in MIMIC-IV and 1.96 (1.34–2.89) in eICU-CRD, while Fine–Gray subdistribution hazard ratios were 1.34 (1.06–1.69) and 2.03 (1.45–2.86), respectively. For a composite endpoint of reintubation or death, however, the adjusted risk ratios attenuated to 1.09 (0.97–1.23) and 1.42 (1.13–1.78). This pattern indicates that some of the association with reintubation overlaps with a broader association involving mortality. It also reinforces the need to interpret calculated hyperosmolarity as a marker of overall illness severity rather than as an isolated mechanism specific to airway failure.

Secondary prognostic analyses found that calculated hyperosmolarity added little to the transportable accuracy of existing clinical models. In external validation with MIMIC-IV coefficients fixed before transfer to eICU-CRD, the area under the receiver operating characteristic curve increased from 0.5875 for the base model to 0.6074 after adding calculated hyperosmolarity. The numerical difference was 0.0199, but its 95% confidence interval ranged from −0.0327 to 0.0434 and included no improvement. Differences in precision–recall area and Brier score were also small. Original external validation indicated overprediction of absolute risk in eICU-CRD, with observed-to-expected ratios of 0.45 for both the base and osmolality-augmented models. The results support the marker’s value as an adjunctive risk signal, not as a replacement for respiratory mechanics, airway protection, secretion management, neurologic function or other established elements of extubation assessment.

The study has important limitations inherent to retrospective electronic health record research. Residual confounding is possible because cough strength, secretion burden and the rapid shallow breathing index could not be harmonized across databases. The eICU-CRD lacked a comparably reliable field for distinguishing planned from unplanned extubation, while treatment documentation was incomplete, particularly for medication names and high-flow nasal oxygen. Oxygenation and fluid data were available only in selected subsets, acute kidney injury was defined from serum creatinine rather than complete urine-output and kidney-replacement-therapy criteria, and calculated osmolality was not a direct laboratory measurement. Explicit causes of reintubation were recorded for too few events, and post-extubation laboratory sampling was influenced by the opportunity to obtain measurements. With only 93 primary reintubation events in eICU-CRD, some subgroup estimates were especially unstable. Both databases also represent US critical care practice, so replication in contemporary non-US cohorts would be needed to assess broader generalizability. The findings support prospective evaluation of calculated osmolality as a complementary component of extubation assessment, rather than a validated treatment threshold.

Subject of Research: Association of pre-extubation calculated serum hyperosmolarity with early invasive reintubation in critically ill adults

Article Title: Pre-extubation calculated serum osmolality and early invasive reintubation in critically ill adults: a dual-database study with multicenter external replication

Article References: Huang, W., Bian, J., Chen, Y., Huang, B., Liu, C., & Chen, C. (2026). Pre-extubation calculated serum osmolality and early invasive reintubation in critically ill adults: a dual-database study with multicenter external replication. Frontiers in Medicine, 13, Article 1984164. https://doi.org/10.3389/fmed.2026.1984164

Image Credits: AI Generated

DOI: 10.3389/fmed.2026.1984164

Keywords: calculated serum osmolality, extubation, invasive reintubation, critical care, mechanical ventilation, hyperosmolarity, metabolic–renal status, retrospective cohort study, MIMIC-IV, eICU-CRD, Pre-extubation, calculated

Because the study included no comparison group, the results show feasibility only and do not establish that the combination is safer or more effective than other anticoagulation strategies. The cohort comprised seven men and four women with a median age of 58 years. Eight patients had acute-on-chronic liver failure, one had acute liver failure following gastrointestinal bleeding, and two had subacute liver failure.

Sepsis or septic shock was the precipitating factor in five patients with acute-on-chronic liver failure, while two had cardiogenic or ischemic injury and one had malignant disease. The group was severely ill: seven of the 11 patients had grade 3 acute kidney injury, seven required vasoactive drugs, and the median Acute Physiology and Chronic Health Evaluation II (APACHE II) score was 23.5. Median platelet count was 44 × 10 9 /L, and liver-related coagulopathy was common.

The small, selected group came from the Department of Critical Care Medicine at Zhongnan Hospital of Wuhan University, where the investigators reviewed patients treated in 2025. The clinical problem addressed by the study is the narrow balance required for CRRT anticoagulation in liver failure. Reduced hepatic synthesis and coagulation dysfunction make systemic anticoagulation more hazardous, while impaired citrate metabolism can make conventional-dose regional citrate anticoagulation prone to accumulation, hypocalcemia and metabolic alkalosis.

Nafamostat can provide anticoagulation without citrate, but a dose sufficient to maintain filter patency may produce excessive systemic coagulation inhibition and increase bleeding concerns. The investigators therefore used a reduced citrate dose together with nafamostat, with the stated aim of limiting citrate exposure while maintaining anticoagulation of the extracorporeal circuit. Whether that combination improves the balance between filter patency and systemic safety cannot be determined from an uncontrolled series.

All patients were treated with Fresenius CRRT machines equipped with a citrate-calcium module and large-area filters. Blood was drawn at 80 mL/min and circulated at 120 mL/min, with post-dilution used and replacement-fluid and dialysate flow each set to 1,000 mL/h. The anticoagulant regimen consisted of 4% sodium citrate infused pre-filter at 150 mL/h, nafamostat mesylate at 20 mg/h before the filter, 5% sodium bicarbonate at 100 mL/h before the filter, and calcium gluconate at the blood-return line.

Post-dilution was selected to protect the venous drip chamber, while the calcium-containing replacement fluid made pre-dilution unnecessary. The investigators described the setup as a standardized bedside approach intended to be reproducible, although the protocol was not prospectively validated against another anticoagulation strategy. Calcium replacement was stratified according to peripheral ionized calcium.

The initial 10% calcium gluconate infusion was 6 mL/h when blood calcium exceeded 1.1 mmol/L, 8 mL/h at 1.0–1.1 mmol/L, 10 mL/h at 0.9–1.0 mmol/L, and 12 mL/h below 0.9 mmol/L. Additional intravenous calcium could be added after physician evaluation. Post-filter ionized calcium was used to assess regional anticoagulation, with a target above 0.4 mmol/L, while the peripheral activated partial thromboplastin time (APTT) was maintained at 1.2–1.5 times its baseline value.

Peripheral blood gases and APTT were checked every two hours for the first six hours and every four hours thereafter, with infusion rates adjusted as needed. The protocol also included inspection of the filter and pipeline and monitoring for bleeding, hypocalcemia, acid-base imbalance, electrolyte abnormalities and drug allergy. Safety monitoring included systemic coagulation tests and surveillance for bleeding and metabolic complications.

Citrate accumulation was defined as a sustained total-to-ionized calcium ratio of at least 2.5 accompanied by metabolic alkalosis or hypocalcemia. Hypocalcemia was defined as peripheral ionized calcium below 0.90 mmol/L. Metabolic alkalosis required arterial pH above 7.45 and bicarbonate above 26 mmol/L, while metabolic acidosis required pH below 7.35 and bicarbonate below 22 mmol/L, excluding respiratory influences.

Coagulation measurements and platelet counts were obtained before CRRT and immediately after the circuit was disconnected, before anticoagulation or other therapy was changed. Filter life ranged from 23 to 72 hours, with a median of 48 hours and an interquartile range of 38 to 68.75 hours. No severe pipeline or filter coagulation was recorded.

PT, INR, APTT, fibrinogen, D-dimer and platelet count did not differ significantly between baseline and the first measurement after CRRT ended; thrombin time was lower at the end of treatment, with a reported P value of 0.05. One patient, representing 9.1% of the cohort, had mild oral mucosal bleeding, but no severe or massive bleeding occurred. Six patients met the study’s hypocalcemia threshold, and five had metabolic alkalosis at baseline, although no severe hypocalcemia or serious acid-base and electrolyte disorder was observed during treatment.

These findings describe what happened in this selected cohort; they do not establish comparative event rates for the regimen. Several clinical measurements changed over the course of intensive treatment. Median procalcitonin fell from 4.78 to 1.92 ng/mL, with P = 0.005, and the median APACHE II score decreased from 23.5 to 17, with P = 0.011.

The median SOFA score fell from 6.67 to 5, but that change was not statistically significant at P = 0.463. Interleukin-6, white blood cell count and neutrophil percentage did not change significantly, while ALT, AST, total bilirubin, creatinine and urea nitrogen also showed no significant before-and-after differences. The study also recorded that liver and kidney function showed no obvious deterioration, but that observation cannot be interpreted as evidence of organ protection.

The before-and-after design, the small sample and the simultaneous use of other critical-care therapies make causal attribution particularly limited. All 11 patients simultaneously received other intensive-care interventions, including antibiotics, vasopressors, source control and broader organ support. The lower procalcitonin and APACHE II score therefore cannot be attributed to the mixed anticoagulation regimen.

The investigators also noted that the pre-filter bicarbonate strategy was based on institutional experience rather than prospective validation; it was intended to support acid-base stability when endogenous citrate metabolism and bicarbonate generation were impaired, but its adequacy in this population remains unevaluated. Median intensive-care unit stay was 12 days, median hospital stay was 21 days, and eight patients restarted CRRT within seven days after it was stopped because of clinical need. One patient died from severe irreversible disease.

These outcomes reflect the illness course and concurrent care, not a measured effect of anticoagulation. The findings are chiefly an operational description of how the regimen was delivered. They show that low-dose citrate combined with standardized nafamostat, calcium stratification, blood-gas monitoring and APTT adjustment could be implemented in selected patients with liver failure receiving CRRT.

The small, selected, single-center sample, absence of subgroup analysis and short follow-up further limit interpretation. The authors characterize the results as preliminary and hypothesis-generating and call for validation in adequately powered, multicenter, prospective controlled trials.

Subject of Research: Feasibility of low-dose citrate-nafamostat anticoagulation during CRRT in patients with liver failure

Article Title: Successful application of low-dose regional citrate combined with nafamostat mixed anticoagulation in CRRT for patients with liver failure: a retrospective case series study of 11 patients

Article References: Zhang, Y., Ma, J., Xiao, L., Ye, J., Li, H., Li, Z., Xiao, M., Ding, X., & Zhang, P. (2026). Successful application of low-dose regional citrate combined with nafamostat mixed anticoagulation in CRRT for patients with liver failure: a retrospective case series study of 11 patients. Frontiers in Pharmacology, 17, 1896125. https://doi.org/10.3389/fphar.2026.1896125

Image Credits: AI Generated

DOI: 10.3389/fphar.2026.1896125

Keywords: liver failure, continuous renal replacement therapy, regional citrate anticoagulation, nafamostat, mixed anticoagulation, filter patency, citrate accumulation, critical care, bleeding, hypocalcemia, Successful, application

The authors used the concept analysis methodology proposed by Walker and Avant, based on a comprehensive review of literature in PubMed, Web of Science, Ovid and ScienceDirect. This was a conceptual synthesis rather than a clinical trial. Its findings therefore address how prevention is defined and which antecedents, consequences and frameworks are relevant, rather than establishing whether a particular program improves outcomes.

According to the abstract, the analysis identified increased survival following pediatric critical illness, growing evidence of persistent physical, cognitive, psychological and social sequelae, and the recognition of PICS-p as a clinical construct among the most relevant antecedents of prevention. Potential consequences associated with implementing preventive approaches include better functional recovery, fewer long-term impairments and greater support for children and their families. These were presented as potential consequences, not demonstrated intervention effects.

Pediatric adaptations of the ABCDEF bundle and family-centered care were identified as relevant conceptual and clinical frameworks. The broader approach was characterized as comprehensive, focused on development and marked by active family participation and continuous interdisciplinary work from admission through the transition home. However, the abstract states that evidence regarding the direct effectiveness of these frameworks in preventing PICS-p remains limited. The review consequently provides a conceptual definition and organizes proposed components of prevention; it does not establish the prevalence of PICS-p, demonstrate the effectiveness of a specific bundle or nursing intervention, or show that any individual outcome is reduced by these approaches.

Subject of Research: Conceptual review of strategies for preventing post-intensive care syndrome in children

Article Title: Understanding and preventing post-intensive care syndrome in children: a conceptual exploration

Article References: Acuña-Caicedo, S. A., & Fuentes-Ramírez, A. (2026). Understanding and preventing post-intensive care syndrome in children: a conceptual exploration. BMC Pediatrics. https://doi.org/10.1186/s12887-026-07617-5

Image Credits: AI Generated

DOI: 10.1186/s12887-026-07617-5

Keywords: post-intensive care syndrome, pediatrics, critical care, concept analysis, ABCDEF bundle, family-centered care, functional recovery, long-term sequelae, Understanding, preventing, post-intensive, care

The central methodological challenge is counterfactual reasoning. Observational clinical records contain outcomes only for the care actually received; the outcomes that would have occurred under an alternative strategy remain unobserved. Existing nursing care evaluations, the authors note, primarily characterize population-level associations or predict observed outcomes, but they do not estimate the potential outcomes of individual patients under standard and enhanced acute care while accounting for evolving clinical states and predictive uncertainty. The Counterfactual Event Planner addresses this gap by integrating five components: the Manifold-Based Decision Model, the Agent-Based Contextual Navigator, the Probabilistic Outcome Synthesizer, the Counterfactual Pacing Strategy, and Bayesian Outcome Updating. Together, these modules construct clinically meaningful latent representations of patient health, explore context-aware intervention trajectories, estimate potential outcome distributions, and refine those estimates as new clinical evidence arrives.

The clinical motivation is grounded in the growing field of cardio-oncology. Cancer treatment may interact with baseline cardiovascular conditions, acute physiological stress, and supportive care requirements, increasing the need for timely nursing assessment and coordinated risk management. The authors deliberately selected oncology patients with pre-existing cardiovascular risk factors as the initial validation group for both clinical and methodological reasons. Clinically, this population is particularly vulnerable because concurrent clinical needs may affect treatment tolerance, symptom burden, care coordination, and long-term outcomes. Methodologically, a population with a higher expected burden of cardiovascular events provides a clearer setting for estimating heterogeneous care effects and comparing alternative nursing strategies while reducing the influence of severe outcome sparsity. The authors emphasize that this targeted population should be interpreted as an initial validation group rather than the only population that may benefit, since cardiovascular toxicity may also occur among patients initially classified as low risk.

Technically, the framework begins with the Manifold-Based Decision Model, which encodes demographic, oncological, cardiovascular, physiological, treatment, and nursing variables recorded during the first 24 hours of intensive care admission into a clinically informed latent patient state. The encoder maps high-dimensional clinical features onto a 64-dimensional manifold, capturing nonlinear dependencies among laboratory measurements, vital signs, and demographic attributes while reducing noise and redundancy. In implementation, the encoder is trained on standardized structured variables from the MIMIC-IV training cohort, and the decoder reconstructs observed clinical variables under an observed feature mask, preventing imputed or unavailable measurements from dominating representation learning. The same fitted preprocessing statistics and encoder parameters are then applied unchanged to validation, internal test, and external cohorts, a design choice that guards against information leakage.

The Agent-Based Contextual Navigator then operates on this encoded representation. Four agents, each governed by a policy network conditioned on patient history and current clinical observations, explore feasible intervention trajectories under two predefined acute care strategies: standard acute care and enhanced acute care. Enhanced acute care is assigned when a composite score of early nursing intensity, physiological monitoring frequency, and supportive care reaches the median threshold calculated from the MIMIC-IV training cohort. Crucially, the navigator is not optimized through online trial and error with patients. The action space is restricted to clinically observed strategies, and policy and transition parameters are learned offline from observed care assignments, baseline covariates, temporal nursing records, and outcome labels. This preserves clinical feasibility and avoids simulated interventions outside the range of observed nursing practice.

The Probabilistic Outcome Synthesizer assigns strategy-specific outcome distributions to the candidate trajectories. The clinical outcome vector includes intensive care unit mortality, in-hospital mortality, cardiovascular adverse events, intensive care unit length of stay, and clinical recovery status. Binary endpoints such as mortality and cardiovascular adverse events are modeled using Bernoulli likelihoods, while continuous endpoints such as length of stay and recovery scores use Gaussian likelihoods with learned means and variances. The Counterfactual Pacing Strategy generates one patient trajectory under each acute care strategy from the same baseline state, imposing temporal and clinical constraints so that simulated trajectories remain physiologically plausible. Bayesian Outcome Updating maintains a diagonal Gaussian posterior over the latent health state and performs one posterior update after every hourly observation, continuously refining outcome estimates and predictive uncertainty as additional data becomes available.

The empirical evaluation drew on two public critical care electronic health record resources. MIMIC-IV, containing 299,712 patients, 431,231 hospital admissions, and 73,181 intensive care unit stays from a tertiary academic medical center, was used for model development, training, validation, and internal testing. The eICU Collaborative Research Database, containing 200,859 intensive care unit encounters from 139,367 patients treated at 208 hospitals across the United States, was reserved exclusively for external validation. The eligible MIMIC-IV oncology cohort contained 8,640 intensive care encounters, split at the patient level into 6,912 training, 864 validation, and 864 internal test records, while the eligible eICU cohort contained 9,760 encounters. Adult oncology patients aged at least 18 years were retained when cancer diagnosis, intensive care admission, cardiovascular risk information, and complete outcome records were available, with encounters shorter than 24 hours excluded.

The results show consistent advantages over six representative baselines spanning traditional statistics, conventional causal inference, deep learning, causal representation learning, and tree-based estimation: Logistic Regression, Propensity Score Matching, Multi-Layer Perceptron, TARNet, DragonNet, and Causal Forest. On the MIMIC-IV internal test cohort, the Counterfactual Event Planner achieved a Precision in Estimation of Heterogeneous Effect of 0.099 compared with 0.111 for DragonNet, the strongest baseline, a relative reduction of 10.8 percent. Average Treatment Effect Error fell from 0.022 to 0.018, an 18.2 percent relative reduction, and Policy Value Error dropped from 0.027 to 0.023, a 14.8 percent improvement. Factual outcome prediction also improved, with root mean squared error of 0.185 and mean absolute error of 0.134 versus 0.192 and 0.139 for DragonNet. On the external eICU cohort, evaluated without retraining or cohort-specific adjustment, the framework again achieved the lowest error on every metric, indicating robust cross-hospital generalization under changes in patient composition, clinical practice, and measurement patterns.

Two additional validation strategies strengthened the evidence. In semi-synthetic counterfactual experiments, where real clinical covariates were retained but treatment assignment and potential outcomes were generated through predefined nonlinear causal mechanisms, the framework reduced true individual treatment effect error relative to DragonNet by 14.1, 17.8, and 23.0 percent under low, moderate, and strong confounding on MIMIC-IV, with corresponding external reductions of 14.9, 16.8, and 20.8 percent. The increasing improvement under stronger confounding suggests greater robustness when treatment assignment is strongly associated with baseline severity. A longitudinal landmark analysis at 6, 12, 18, and 24 hours showed that performance improved as clinical information accumulated, with PEHE on MIMIC-IV decreasing from 0.121 at 6 hours to 0.099 at 24 hours, and temporal consistency of individual effect estimates stabilizing over the observation window. An ablation study confirmed that removing the Counterfactual Pacing Strategy produced the largest degradation, underscoring the value of explicitly generating alternative care trajectories from a shared baseline state.

Beyond computational performance, the study incorporated a participatory clinical acceptability assessment in which specialist oncology nurses reviewed care strategy definitions, cardiovascular risk indicators, outcome variables, and representative counterfactual outputs. The framework received favorable ratings across clinical relevance, interpretability, workflow compatibility, and perceived safety, with an overall acceptability score of 4.38 on a five-point Likert scale and an 85.4 percent agreement rate. The authors position the work within the T1 to T2 translational stage, meaning computational discovery and retrospective validation, and explicitly state that prospective T3 and T4 studies examining workflow integration, alert burden, equity, and population-level impact remain necessary. They caution that estimated counterfactual outcomes should be regarded as decision support evidence requiring clinical interpretation rather than definitive causal truth, and that the framework is intended to support nurse-led assessment, risk stratification, and interdisciplinary care planning while final care decisions remain under clinician supervision and institutional protocols. If prospective validation succeeds, the approach could mark a shift from population-level nursing care evaluation toward genuinely individualized assessment for cancer patients whose hearts are on the line.

Subject of Research: Counterfactual machine learning for individualized nursing care assessment in oncology patients with cardiovascular risk factors

Article Title: Translational assessment of nursing care models for oncology patients with cardiovascular risk factors

Article References: Du, H., Wang, J., & Xing, Y. (2026). Translational assessment of nursing care models for oncology patients with cardiovascular risk factors. Frontiers in Cardiovascular Medicine, 13, Article 1834086. https://doi.org/10.3389/fcvm.2026.1834086

Image Credits: AI Generated

DOI: 10.3389/fcvm.2026.1834086

Keywords: counterfactual event planner, oncology nursing, cardiovascular risk factors, individualized treatment effect, MIMIC-IV, eICU database, causal inference, electronic health records, cardio-oncology, Bayesian outcome updating, critical care, clinical decision support

The research team, led by investigators affiliated with Southern Illinois University School of Medicine, Wayne State University School of Medicine, and collaborating institutions, mined the TriNetX database, a federated network of de-identified electronic health records spanning the years 2006 to 2026. They identified adults with a stroke diagnosis who required invasive mechanical ventilation within 48 hours of stroke onset and who had documented pulse oximetry readings in that same early window. Patients were classified as receiving a restrictive oxygenation strategy if every recorded peripheral oxygen saturation, or SpO₂, value fell between 92 and 96 percent, and as receiving a liberal strategy if all values were at or above 97 percent. Anyone whose readings straddled both ranges or dipped below 92 percent was excluded, a deliberately strict rule designed to create two cleanly separated, non-overlapping exposure groups.

Of the 17,207 patients who met the inclusion criteria, 2,366 had restrictive oxygenation profiles and 14,841 had liberal profiles before any statistical adjustment. Because patients who end up on conservative oxygen protocols often differ systematically from those kept at higher saturations—older, sicker, with more cardiopulmonary disease—the researchers used propensity score matching, a statistical technique that pairs patients with similar baseline characteristics. The model accounted for age, sex, race and ethnicity, body mass index, a Glasgow Coma Scale score of nine or below, and a long list of comorbidities including hypertension, ischemic heart disease, heart failure, atrial fibrillation, chronic obstructive pulmonary disease, pneumonia, and shock. After matching, 2,350 patients remained in each group, and all measured covariates were well balanced, with standardized mean differences below 0.1.

The results were striking. In the matched cohort, restrictive oxygenation during the first 48 hours of ventilation was associated with a 13 percent higher hazard of death at one month, with a hazard ratio of 1.13 and a 95 percent confidence interval of 1.03 to 1.23, a statistically significant finding with a p-value of 0.008. The association persisted at three months, with a hazard ratio of 1.11 and a confidence interval of 1.02 to 1.20. Kaplan–Meier survival curves, which track survival over time, visibly separated between the two groups in the combined stroke population, indicating that the mortality gap emerged early and persisted through the follow-up period.

When the investigators split the population by stroke subtype, the picture became more nuanced. In the acute ischemic stroke subgroup, built from 1,214 restrictive and 7,463 liberal patients before matching, restrictive oxygenation was associated with a 26 percent higher hazard of death at one month (hazard ratio 1.26, 95 percent confidence interval 1.11 to 1.43) and a 22 percent higher hazard at three months (hazard ratio 1.22, 95 percent confidence interval 1.08 to 1.37). In the intracerebral hemorrhage subgroup, however, the corresponding hazard ratios—1.09 at one month and 1.11 at three months—had confidence intervals that crossed one, meaning no statistically significant difference could be detected. The authors note that the hemorrhagic subgroup was smaller, with 728 patients per arm after matching, so limited statistical power may have masked a real effect; notably, the point estimates trended in the same direction as those seen in ischemic stroke.

The biology behind these findings is grounded in the peculiar physiology of the acutely injured brain. In the first two days after an ischemic stroke, cerebral autoregulation—the mechanism that normally keeps blood flow constant across swings in blood pressure—is often impaired. Surrounding the core of dead tissue lies the penumbra, a zone of hypoperfused but still viable brain that depends precariously on whatever oxygen the circulation delivers. Even modest reductions in arterial oxygen content could, in theory, tip penumbral tissue into irreversible infarction, extending the damage and worsening outcomes. A restrictive saturation target of 92 to 96 percent, while safe for many critically ill patients, may leave less reserve for this vulnerable tissue than a liberal target of 97 percent or above.

Yet the push toward conservative oxygen in intensive care has its own logic. Excess oxygen generates reactive oxygen species, drives oxidative stress, and can cause cerebral vasoconstriction, and randomized trials in general ICU populations—including ICU-ROX and the Oxygen-ICU trial—have explored whether tighter oxygen control reduces harm. The new study suggests that extrapolating those protocols to stroke patients may be misguided. The authors point to post-hoc analyses of ICU-ROX as a telling parallel: among ventilated patients with acute brain pathologies other than hypoxic-ischemic encephalopathy, 180-day mortality was numerically higher under conservative oxygen (34.5 percent versus 26.9 percent), a heterogeneity considered plausible enough to be formalized as a prespecified hypothesis in the ongoing Mega-ROX Brains trial, which is randomizing brain-injured patients to conservative or liberal oxygenation targets.

Among the secondary outcomes, one result stood out as counterintuitive: patients in the restrictive group were significantly less likely to undergo tracheostomy within 30 days, at 8.5 percent versus 13.1 percent in the liberal group, a risk ratio of 0.653 with a p-value below 0.001. The authors caution that this pattern may reflect survivorship bias—patients who die early cannot receive a tracheostomy—or may be shaped by unmeasured factors such as the reasons for intubation, the trajectory of neurological recovery, and goals-of-care discussions that could not be captured in the database. Ventilator-associated pneumonia, by contrast, showed no significant difference between groups, with event rates of 0.6 percent and 0.89 percent, though the very low number of events limited statistical power.

The study’s observational design imposes important caveats. Because oxygenation was observed rather than allocated, residual confounding by unmeasured variables cannot be excluded, and the association cannot be read as proof of causation. Stroke severity measures such as infarct volume, hematoma size, and detailed neurological examinations were not available in the federated record network and could not be incorporated into the matching model. The exposure was defined by documented pulse oximetry readings rather than continuous arterial oxygen measurements, so the analysis reflects oxygenation strategy rather than precise quantification of hyperoxemia or hypoxemia. The two saturation thresholds are also adjacent, making the contrast between groups relatively narrow, and follow-up was limited to three months.

Even with those limitations, the sheer size of the cohort and the rigor of the matching make this one of the most pointed challenges yet to the assumption that conservative oxygen targets are universally safe. The authors frame their findings as hypothesis-generating rather than practice-changing, but the implication is hard to ignore: the first 48 hours of mechanical ventilation may represent a modifiable window in stroke care, and the optimal saturation target may differ between an ischemic brain starved of blood flow and a hemorrhagic brain compressed by expanding blood. As stroke remains a leading cause of death and disability worldwide, the authors argue that disease-specific randomized trials—rather than broad ICU-wide oxygen protocols—are urgently needed to determine whether liberal oxygenation genuinely protects the penumbra, or whether individualized titration can balance the competing risks of hypoxemia and oxygen toxicity in this uniquely vulnerable population.

Subject of Research: Oxygen saturation targets and short-term mortality in mechanically ventilated stroke patients

Article Title: Association of liberal versus restrictive oxygen saturation targets with mortality in mechanically ventilated stroke patients: a multicenter retrospective propensity-matched analysis

Article References: Ghafarian, A., Saad, J., Wajahath, M., Ghafarian, M., Saasouh, W., Saleh, K., Nasser, M., Badami, A., Shahait, A. D., & Loggini, A. (2026). Association of liberal versus restrictive oxygen saturation targets with mortality in mechanically ventilated stroke patients: a multicenter retrospective propensity-matched analysis. Frontiers in Neurology, 17, 1840183. https://doi.org/10.3389/fneur.2026.1840183

Image Credits: AI Generated

DOI: 10.3389/fneur.2026.1840183

Keywords: stroke, mechanical ventilation, oxygen saturation, acute ischemic stroke, intracerebral hemorrhage, mortality, propensity score matching, critical care, neurocritical care, TriNetX, hyperoxia, ventilator-associated pneumonia

The study focused on patients who had undergone elective curative hepatic resection, the surgical removal of part of the liver to treat hepatocellular carcinoma, the most common primary liver cancer in adults. Hepatocellular carcinoma is a major global health burden, and surgical resection remains one of the most effective curative options for early-stage disease. Yet the operation is a major physiological insult, and the hours and days that follow are typically spent in an intensive care unit, an environment notorious for disrupting sleep. Alarms, lighting, monitoring equipment, noise, and pain all conspire to fragment sleep, and poor sleep in critically ill patients has been linked in prior research to worse outcomes, including postoperative delirium.

The research team, led by nurses at Linkou Chang Gung Memorial Hospital in Taoyuan, Taiwan, together with colleagues at Chang Gung University of Science and Technology and Taipei Medical University, designed a prospective two-group quasi-experimental study. Between 2022 and 2023, they enrolled 52 adults who had undergone elective curative hepatic resection at the tertiary medical center in northern Taiwan. Rather than randomizing patients individually, the researchers enrolled participants sequentially according to the recruitment period: one group received usual postoperative ICU care alone, while the other received inhaled lavender aromatherapy as an adjunct to that same usual care. This design, while less rigorous than a randomized controlled trial, is a common and pragmatic approach in busy clinical settings where contamination between concurrent patients would otherwise be difficult to control.

The intervention itself was strikingly simple. On the evenings of postoperative days one and two, two drops of lavender essential oil were applied to a gauze pad that was held near the patient’s nose for ten breaths and then fixed to the patient’s collar from eight in the evening until eight the following morning. In other words, patients were continuously exposed to a gentle lavender scent throughout the first two nights of their ICU stay, a period when sleep disturbance and discomfort tend to peak. No additional drugs were given, no sedatives were changed, and the control group simply received the standard care that all patients would normally get.

To measure the effects, the team used three validated instruments. Sleep quality was assessed with the Richards-Campbell Sleep Questionnaire, a tool specifically developed to capture how well patients feel they slept in the noisy, interrupt-laden ICU environment. Pain was evaluated with the Critical Care Pain Observation Tool, a behavioral scale that is particularly valuable in intensive care because it relies on observable indicators such as facial expression and body movement rather than self-report, which is useful when patients are too unwell or too fatigued to answer questions reliably. Anxiety was measured with the Visual Analogue Scale for Anxiety, on which patients mark their level of anxiety along a continuous line. Assessments were performed at baseline before the intervention, and again at two postoperative follow-up points, allowing the researchers to track changes over time.

Because the same patients were measured repeatedly, the analysis used generalized estimating equations, a statistical framework designed for longitudinal data that can account for the correlation between repeated measurements within each person while adjusting for covariates. The key question was whether the two groups followed different trajectories over time, which shows up statistically as a significant group-by-time interaction. The results were consistent across all three outcomes. For pain, the interaction test yielded a Wald chi-square of 23.41 with a p-value below .001, indicating a highly significant divergence between the groups. For anxiety, the Wald chi-square was 9.38 with a p-value of .009, and for sleep quality it was 8.33 with a p-value of .016. All three are statistically robust findings, not marginal ones.

The magnitude of the differences at the final follow-up point is what makes the study clinically interesting. By the second assessment, patients who had received lavender aromatherapy had significantly lower pain scores on the Critical Care Pain Observation Tool and significantly lower anxiety scores on the Visual Analogue Scale for Anxiety than patients who had received usual care alone. Their sleep quality scores on the Richards-Campbell Sleep Questionnaire were significantly higher. Most notably, the adjusted between-group difference in sleep quality was 14.81 points, with a 95 percent confidence interval ranging from 4.82 to 24.80. Because that interval lies entirely above zero and its magnitude is large relative to the scale, the authors characterize it as a potentially clinically meaningful improvement in sleep quality, the kind of change a patient would genuinely feel rather than a statistical artifact.

Importantly, the two groups of 26 patients each were comparable at the outset, with no significant differences in demographic or clinical characteristics, which strengthens the argument that the aromatherapy itself, rather than some pre-existing imbalance, drove the observed differences. The study also has transparent limitations that the quasi-experimental design imposes. Because patients were enrolled by recruitment period rather than randomly assigned, unmeasured temporal factors, such as changes in staffing, seasonal variation, or shifts in unit practice, cannot be fully excluded as alternative explanations. The sample size of 52 is modest, and the findings apply specifically to adults recovering from curative resection for early-stage hepatocellular carcinoma at a single tertiary center, so generalization to other surgical populations or other ICUs requires caution. Self-reported sleep and anxiety measures are also inherently subjective, although the use of a behavioral pain tool partially mitigates this concern.

Why would a plant extract delivered through the nose influence pain, anxiety, and sleep at all? The biological plausibility rests on the anatomy of olfaction. Odorant molecules bind to receptors in the nasal epithelium that send signals directly to the olfactory bulb and from there to limbic brain structures, including the amygdala and hypothalamus, which are central to emotion, stress responses, and the regulation of arousal and sleep-wake cycles. Lavender oil contains compounds such as linalool and linalyl acetate, which have been studied for their calming and anxiolytic properties, and systematic reviews have reported anxiolytic effects of Lavandula angustifolia across multiple clinical contexts. Recent reviews of inhalation aromatherapy have also suggested benefits for postoperative pain management, possibly through reduction of anxiety-related amplification of pain perception, since pain and anxiety reinforce one another in a well-known feedback loop.

The practical implications could be considerable. Pharmacologic options for promoting sleep in the ICU are limited, and a rapid systematic review of pharmacologic sleep promotion modalities in intensive care has highlighted how weak the evidence base is for most drug-based approaches, along with concerns about side effects such as next-day sedation and delirium. Against that backdrop, a two-drop, low-cost, low-risk intervention that requires no prescription and no change to existing protocols is attractive, provided it works. This study stops short of proving causation with the certainty of a randomized trial, but its longitudinal analysis, validated measures, and clinically meaningful effect sizes make a credible case that lavender aromatherapy deserves testing in larger randomized studies. For now, the image of an ICU patient drifting to sleep with a lavender-scented gauze clipped to the collar is less a folk remedy than an emerging, evidence-informed candidate for supportive cancer care, one whose biggest virtues, simplicity and safety, may be exactly what critical care medicine needs.

Subject of Research: Effects of inhaled lavender essential oil aromatherapy on sleep, pain, and anxiety in postoperative ICU patients

Article Title: Effects of inhaled lavender essential oil on sleep quality, pain, and anxiety in ICU patients after curative hepatocellular carcinoma resection: a quasi-experimental study

Article References: Lu, Y.-Y., Tsai, P.-C., Hu, Y.-C., & Hsu, W.-C. (2026). Effects of inhaled lavender essential oil on sleep quality, pain, and anxiety in ICU patients after curative hepatocellular carcinoma resection: a quasi-experimental study. Supportive Care in Cancer, 34(10), Article 942. https://doi.org/10.1007/s00520-026-11184-x

Image Credits: AI Generated

DOI: 10.1007/s00520-026-11184-x

Keywords: lavender aromatherapy, hepatocellular carcinoma, intensive care unit, sleep quality, postoperative pain, anxiety, essential oils, non-pharmacological intervention, liver resection, supportive care, Richards-Campbell Sleep Questionnaire, critical care

Sepsis is among the most common and deadliest medical emergencies in the United States, affecting an estimated 1.7 million Americans each year. It arises when the body’s response to an infection turns against its own tissues and organs, triggering a cascade of inflammation that can lead to widespread organ dysfunction. Septic shock represents the most severe form of this syndrome. In these patients, the inflammatory response causes blood vessels to dilate excessively, and blood pressure falls so low that blood and oxygen can no longer reach vital organs such as the brain, kidneys, and heart in sufficient quantities. Without rapid intervention, the result is often multi-organ failure and death. Septic shock contributes to an estimated one in three hospital deaths nationwide, making it one of the most consequential conditions in modern medicine.

The standard approach to septic shock involves two parallel strategies. Clinicians administer antibiotics and control the source of the infection, while simultaneously using medications called vasopressors to constrict blood vessels and raise blood pressure to a level that keeps organs perfused. First-line vasopressors are agents related to adrenaline, such as norepinephrine, which act on receptors that tighten blood vessels and support cardiac output. But when these first-line drugs are not enough, physicians turn to a second agent: vasopressin, a naturally occurring hormone that raises blood pressure through a different mechanism, constricting vessels via vasopressin receptors rather than adrenergic ones. Vasopressin is also, notably, an expensive infusion, and its use consumes significant hospital resources.

What has never been clear is when to add it. Some critical care clinicians begin vasopressin relatively early, as soon as a patient’s blood pressure starts to become difficult to control on norepinephrine alone. Others wait, reserving the drug for patients who require higher doses of the first-line agents before escalation. Both strategies have coexisted in intensive care units around the world, guided by physiological reasoning rather than direct comparative evidence. “For years, clinicians have known what medications to use for septic shock, but there has been little evidence about when to add vasopressin after initial vasopressor therapy,” said Ithan D. Peltan, MD, principal investigator of the study and an Intermountain Health critical care physician and researcher.

The new findings come from the Vasopressin for Septic Shock Pragmatic (VASSPR) Trial, which evaluated the two treatment strategies across 13 Intermountain Health hospitals in Utah and Idaho between February 2024 and July 2025. The trial enrolled 3,959 adults with septic shock, making it one of the largest studies of its kind in the field. Rather than randomizing individual patients, the investigators used a hospital-level design, comparing periods in which hospitals recommended starting vasopressin earlier against periods in which they recommended waiting until a patient’s shock potentially worsened. Researchers then evaluated outcomes over the following 28 days.

The result was striking in its simplicity: no meaningful differences between the groups. Survival, the amount of time patients spent alive and out of the hospital, and other measures of recovery were similar regardless of when vasopressin treatment was initiated. The trial did observe slightly higher rates of cardiogenic shock in the early-treatment group, a finding that may reflect physiological differences in how the two strategies affect the heart, but the researchers determined that this difference had no meaningful impact on patient-centered outcomes such as survival and recovery.

“The question wasn’t whether vasopressin works. We already know it can help raise blood pressure in patients with septic shock. The question is when it should be added to other vasopressors,” Dr. Peltan explained. “This trial directly answers a common bedside question that affects thousands of patients every year: is it safe to wait to see if a patient’s sepsis gets worse before giving them costly infusions they may not need? Our study found that it is.”

The implications extend beyond the specific question of vasopressin timing. Sepsis is a leading cause of intensive care unit admissions and one of the most expensive conditions treated in U.S. hospitals, costing billions of dollars to treat annually. If hospitals can safely adopt a more conservative approach to escalation, waiting to confirm that a patient truly needs a second vasopressor before starting it, the savings in medication costs, infusion equipment, and nursing workload could be substantial, all without compromising patient outcomes. In an era of intense pressure on hospital resources, evidence that supports restraint is a rare and valuable commodity.

“In healthcare, findings that show a treatment doesn’t need to be used earlier are often just as important as findings that support a new therapy,” said Dr. Peltan. “Our study suggests that hospitals can take a more conservative approach, helping deliver effective care while using resources wisely.” Researchers say the findings will help inform future clinical practice guidelines and provide physicians with greater confidence when making treatment decisions for patients with septic shock, replacing habitual variation between institutions with a shared, evidence-based standard.

Beyond its clinical findings, the VASSPR trial is also an example of an important advancement in how health systems conduct research. By embedding large-scale clinical trials within routine patient care, researchers can answer important questions efficiently in real-world settings, enrolling broad patient populations at a fraction of the cost and time required by traditional trials. The infrastructure developed for the study is already being used to launch additional clinical trials within Intermountain Health, designed to address the challenging questions that physicians and nurses encounter daily at the bedside. “Trials like this one have the potential to make meaningful changes and guide us on difficult questions that are hard to answer with traditional clinical trials,” said Dr. Peltan. For the hundreds of thousands of patients who develop septic shock each year, that means the evidence base guiding their care is finally catching up with the urgency of their condition.

Subject of Research: Timing of vasopressin initiation for the treatment of septic shock

Article Title: New study finds escalating treatment earlier for patients with life-threatening septic shock doesn’t improve survival or recovery

Article References: New study finds escalating treatment earlier for patients with life-threatening septic shock doesn’t improve survival or recovery. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: septic shock, sepsis, vasopressin, vasopressors, critical care, randomized controlled trial, VASSPR trial, Intermountain Health, JAMA, blood pressure, intensive care, clinical guidelines

One of the headline studies examines a question that has long troubled asthma specialists: whether empiric treatment with chronic proton pump inhibitors, medications routinely prescribed to suppress stomach acid, might actually worsen outcomes in patients with difficult-to-treat asthma. Using a propensity-matched United States cohort design, investigators analyzed biologic-treated patients with difficult-to-treat asthma and found that chronic proton pump inhibitor use was associated with an increased rate of exacerbations. The study does not establish causation, but the association is striking enough to prompt a reevaluation of a common prescribing pattern, particularly in a population already burdened by frequent flare-ups and heavy medication use.

The mechanistic rationale for caution is plausible. Proton pump inhibitors alter gastric pH, which can influence airway inflammation through microaspiration and changes in the microbiome, and prior observational work has suggested links between acid suppression and respiratory infections. In patients whose asthma is already difficult to control and who are receiving biologic therapy, any additional factor that tips the balance toward exacerbation carries clinical weight. The propensity-matching approach used in the study helps control for confounding by indication, since patients prescribed acid suppression often differ systematically from those who are not, but the authors and meeting organizers emphasize that the findings are associative and warrant prospective testing.

A second asthma-adjacent study shifts attention to chronic obstructive pulmonary disease with type 2 inflammation. Mepolizumab, a monoclonal antibody that blocks interleukin-5 and thereby reduces eosinophilic airway inflammation, has an established role in severe eosinophilic asthma. Pooled results from phase 3 randomized trials presented at the meeting indicate that mepolizumab reduces systemic corticosteroid utilization in patients with COPD and evidence of type 2 inflammation. This matters because chronic oral corticosteroid exposure drives a cascade of harms, including osteoporosis, diabetes, weight gain, and increased infection risk, and clinicians have long sought alternatives for the subset of COPD patients with eosinophilic biology.

The pooled analysis aggregates data across multiple phase 3 randomized trials, strengthening statistical power and allowing subgroup patterns to emerge that single trials might miss. For a disease as heterogeneous as COPD, identifying the eosinophilic, type 2 inflammatory phenotype as a treatment-responsive subgroup represents a shift toward precision medicine in an area where broad-spectrum bronchodilation has historically dominated. The results do not suggest mepolizumab for all COPD patients; rather, they support biomarker-guided selection, with blood eosinophil counts and other markers of type 2 inflammation serving as the gatekeepers to biologic therapy.

Bronchiectasis research at the meeting introduces a deceptively simple clinical tool: the color of a patient’s sputum as reported by the patient. In a post hoc analysis from the ASPEN trial, investigators examined whether patient-reported sputum color correlates with objective measures of airway inflammation, specifically neutrophil serine proteases, and with exacerbation risk. The findings suggest that the yellow or green discoloration long noted by clinicians at the bedside reflects real biochemical activity in the airways, since neutrophil-derived enzymes such as myeloperoxidase and neutrophil elastase impart the characteristic green hue through oxidative chemistry involving heme groups.

The ASPEN trial post hoc analysis is notable because it links a subjective, zero-cost patient report to molecular markers and clinical outcomes in a large trial population. If sputum color tracks with neutrophil serine protease activity and predicts exacerbations, it could serve as a practical triage signal in clinics and telemedicine encounters, flagging patients who need intensified anti-inflammatory or antimicrobial treatment before objective laboratory results return. The analysis is post hoc, meaning the endpoint was not the trial’s primary design goal, so the association requires confirmation in prospective studies, but the clinical logic is compelling and the implications for low-resource settings are considerable.

Critical care presentations tackle two of the most consequential questions in intensive care. The first applies machine learning to the problem of extubation failure, the situation in which a patient is removed from mechanical ventilation only to deteriorate and require reintubation within hours or days. Traditional weaning parameters, such as the rapid shallow breathing index and spontaneous breathing trial performance, have limited predictive power, and extubation failure carries significant mortality and morbidity penalties. In an analysis of the MIMIC-IV database, a large publicly available repository of intensive care unit records from Beth Israel Deaconess Medical Center, investigators trained machine learning models to predict extubation failure using features that extend beyond conventional weaning parameters.

The MIMIC-IV analysis exemplifies the current wave of clinical machine learning, in which high-dimensional electronic health record data, including vital signs, laboratory values, medication exposures, and ventilation settings, are mined for patterns invisible to single-parameter rules. Models of this kind can integrate trajectory information, capturing how a patient’s respiratory mechanics evolve over hours rather than relying on a single snapshot. The critical caveats are familiar: retrospective models trained on a single institution’s data risk overfitting and degraded performance on external populations, and algorithmic predictions must be validated prospectively and integrated thoughtfully into clinician decision-making. Still, the direction is clear, and the meeting’s inclusion of this work signals growing institutional interest in augmenting, not replacing, bedside judgment.

The second critical care study addresses telemedicine in the intensive care unit, evaluating tele-ICU implementation and program maturation in urban safety net hospitals through an 18-month multisite interrupted time-series analysis. Safety net hospitals serve disproportionately disadvantaged populations and often operate with constrained staffing, making them ideal test beds for remote intensivist support. Interrupted time-series design is among the strongest quasi-experimental methods available for evaluating health system interventions when randomized assignment is impractical, comparing outcome trajectories before and after implementation while accounting for secular trends. The 18-month horizon also allows the investigators to distinguish early implementation effects from the maturation phase, when teams adapt workflows and trust in the remote system develops.

Tele-ICU programs connect bedside teams with remote critical care specialists who monitor patients, review data, and advise on management around the clock. Evidence on tele-ICU effectiveness has been mixed across prior studies, with some large implementations showing mortality benefits and others showing null results, often attributed to differences in how the programs intervene rather than whether they exist. The new multisite analysis in safety net hospitals adds an equity dimension to this literature, since these institutions care for patients who historically experience worse critical care outcomes. The maturation finding, if it shows improving trajectories over time, would echo prior observations that tele-ICU benefits accrue gradually as integration deepens.

Interstitial lung disease research at the meeting advances two complementary frontiers. The first is noninvasive sphingolipid profiling, which distinguishes idiopathic pulmonary fibrosis from fibrotic hypersensitivity pneumonitis. These two fibrotic lung diseases can be difficult to differentiate clinically and radiographically, yet they carry different prognoses and treatment pathways, including the eligibility landscape for antifibrotic therapy and the importance of antigen avoidance in hypersensitivity pneumonitis. Sphingolipids, a class of lipid signaling molecules involved in cell proliferation, apoptosis, and inflammatory regulation, have emerged as biomarker candidates in fibrotic disease, and a blood-based profiling approach could spare patients from invasive surgical lung biopsy while accelerating diagnostic clarity.

The second interstitial lung disease study connects telomere biology to imaging. Shorter telomere length, a marker of cellular aging and replicative stress, is a recognized risk factor for pulmonary fibrosis, and telomere-related gene mutations are implicated in familial forms of the disease. The new research reports that shorter telomere length is associated with automated detection of usual interstitial pneumonia on imaging in patients with pulmonary fibrosis. Usual interstitial pneumonia is the histopathologic and radiologic pattern characteristic of idiopathic pulmonary fibrosis, and automated detection algorithms are increasingly capable of quantifying this pattern on high-resolution computed tomography. The association suggests that telomere attrition and the radiologic fingerprint of fibrosis are linked, potentially refining risk stratification in a disease where early identification changes treatment eligibility.

Lung cancer screening, one of the most consequential public health interventions in thoracic medicine, is the subject of a national gap analysis quantifying deficits in awareness, communication, and uptake. Low-dose computed tomography screening reduces lung cancer mortality in high-risk individuals, yet uptake in the United States remains persistently low, with eligible individuals often unaware of their eligibility or never receiving a recommendation from a clinician. The new analysis quantifies these deficits nationally, dissecting where the screening pipeline fails, from public awareness through clinician communication to actual completion of imaging. The framing matters because each stage of the pipeline represents a distinct intervention target, from mass media campaigns to electronic health record prompts that flag eligible patients during routine visits.

Taken together, the CHEST 2026 abstracts portray a specialty in transition, moving from one-size-fits-all approaches toward phenotype-guided therapy, from subjective bedside impressions toward molecular and algorithmic validation, and from fragmented care toward system-level interventions that reach underserved populations. The biologic era is expanding from asthma into COPD and bronchiectasis, machine learning is pressing against the limits of traditional critical care parameters, and lipidomics and telomere biology are reshaping the diagnosis of fibrotic lung disease. Meeting organizers note that all research presented at the meeting is available on the journal CHEST website, and the society emphasizes that the abstracts represent preliminary findings that will inform, but not replace, the peer-reviewed literature and clinical guidelines that follow.

Subject of Research: Research highlights from CHEST 2026 Annual Meeting

Article Title: Research highlights from CHEST 2026 Annual Meeting

Article References: Research highlights from CHEST 2026 Annual Meeting. Original publication (Bibliographic details unavailable.)

Image Credits: AI Generated

DOI: Not provided

Keywords: CHEST 2026, pulmonary medicine, critical care, asthma, COPD, bronchiectasis, pulmonary fibrosis, lung cancer screening, tele-ICU, machine learning, Research, highlights

The study, published in the Journal of Cellular and Molecular Medicine, followed 462 adults with sepsis admitted to the intensive care unit of Beijing Hospital of Traditional Chinese Medicine, Capital Medical University, between June 2014 and March 2025. All patients met the Sepsis-3 criteria, the internationally accepted definition of life-threatening organ dysfunction caused by a dysregulated host response to infection. Rather than relying on a single blood draw at admission, the investigators tracked serum levels of C-reactive protein, or CRP, and albumin on days 1, 3, 5, 7 and 14 after ICU admission, capturing the dynamic choreography of the body’s inflammatory and metabolic responses.

CRP and albumin are two of the most familiar numbers on any critical care chart, and both are manufactured in the liver. They are biological mirror images. CRP is a prototypic positive acute-phase reactant, ramped up by pro-inflammatory cytokines, particularly interleukin-6, and serves as a robust readout of systemic inflammatory activity. Albumin, by contrast, is a negative acute-phase protein: as inflammation flares, the liver suppresses albumin synthesis, and serum levels fall. The degree of hypoalbuminemia in critically ill patients therefore tracks the intensity of the inflammatory storm, while also reflecting nutritional reserve, hepatic synthetic function and vascular permeability. Previous work has shown that composite indices combining the two, such as the CRP-to-albumin ratio and the CALLY index, can outperform either marker alone in diagnosing and prognosticating sepsis.

What has been missing, the researchers argue, is a way to capture the coordinated temporal behaviour of the two proteins rather than their static values. Most prior studies examined single time points, ratios, or separate trajectories of one marker at a time. To fill that gap, the team applied group-based dual-trajectory modelling, or GBDTM, a statistical technique that extends conventional group-based trajectory modelling by jointly modelling correlated longitudinal markers and identifying clinically interpretable subphenotypes. Using the lcmm package in R, they tested linear, quadratic and cubic time functions across solutions containing one to six latent classes, judging the models by log likelihood, the Akaike and Bayesian information criteria, entropy, class-specific average posterior probabilities, minimum class prevalence and clinical interpretability.

The winning model was a three-class solution with an entropy of 0.80 and average posterior probabilities of 0.84, 0.87 and 0.92, indicating clean separation between groups. It carved the cohort into three distinct CRP-albumin phenotypes. The vast majority, 368 patients or 79.65 percent, followed a slow-decreasing pattern in which inflammation gradually subsided. A second group of 48 patients, 10.39 percent, showed a rapid-decreasing pattern. The third and smallest group, 46 patients or 9.96 percent, displayed the ominous signature: CRP and albumin trajectories that increased over the observation window, a pattern suggesting persistent or worsening inflammatory activity coupled with deteriorating metabolic reserve.

The clinical stakes became clear in the survival analysis. The overall 28-day mortality rate in the cohort was 34.6 percent, but it was far from evenly distributed. Patients in the increasing group died within 28 days at a rate of 52.2 percent, compared with 32.6 percent in the slow-decreasing group and 33.3 percent in the rapid-decreasing group. Kaplan-Meier curves diverged significantly between the trajectory groups, with log-rank p values of 0.034 for 28-day mortality and 0.049 for 60-day mortality. The baseline characteristics of the groups also differed in telling ways: the increasing group had a lower prevalence of skin and soft-tissue infection, different distributions of acute gastrointestinal injury grade, blood pH and hemofiltration duration.

To determine whether the increasing phenotype carried independent prognostic weight, the researchers turned to least absolute shrinkage and selection operator, or LASSO, Cox regression, which selected 27 adjustment covariates from clinically relevant demographic, physiological, infection-related, organ-function, treatment and laboratory domains. In the fully adjusted model, membership in the increasing trajectory group remained associated with 28-day mortality with a hazard ratio of 1.943, a 95 percent confidence interval of 1.202 to 3.140 and a p value of 0.007, and with 60-day mortality with a hazard ratio of 1.946, a 95 percent confidence interval of 1.218 to 3.108 and a p value of 0.005, relative to the slow-decreasing reference group. The rapid-decreasing group showed no significant excess risk. Scaled Schoenfeld-residual tests found no violation of the proportional hazards assumption, and subgroup analyses detected no evidence that the association differed by sex, smoking, alcohol use or infection site.

Sensitivity analyses reinforced the signal while adding important nuance. Parsimonious models adjusted only for age, sex and either the Sequential Organ Failure Assessment or the Acute Physiology and Chronic Health Evaluation II score yielded similar estimates, with hazard ratios for the increasing group ranging from 1.564 to 1.808. Excluding red blood cell count to resolve multicollinearity left the association essentially unchanged. Yet the incremental value over established severity scores proved modest: adding trajectory membership to a SOFA-based model raised the C-index from 0.689 to 0.702 for 28-day mortality, a change of 0.013 that fell short of statistical significance, and the corresponding gains in APACHE II-based models were even smaller. The joint model also did not consistently outperform a CRP-only trajectory model, though it did edge out an albumin-only model in both Akaike information criterion and discrimination.

The authors are candid about these limits. The value of the joint trajectory, they conclude, lies in dynamic phenotyping rather than in replacing established severity scores. The increasing pattern describes a biologically coherent subgroup whose inflammatory and metabolic systems are moving in the wrong direction together, a picture that admission values alone cannot capture. But the high-risk class proved sensitive to the number of latent classes specified: a four-class solution split the adverse phenotype into a 30-patient high-increasing group with an attenuated association and an 18-patient moderate inverted-U group with a strong association but a prevalence below the 5 percent reliability threshold, hinting at heterogeneity that the present sample cannot stably resolve.

Limitations temper the excitement. The study was single-center and retrospective, so trajectory prevalence and effect sizes may shift across different health-care systems, sepsis etiologies and measurement schedules, and external validation in independent multicenter cohorts or public databases such as MIMIC-IV and eICU remains to be done. Imputation of missing baseline covariates introduces uncertainty, and the small size of the increasing group yields imprecise estimates. The researchers also did not establish the earliest time point at which class membership can be assigned reliably, nor did they test whether trajectory-guided treatment improves outcomes. For now, the message is hypothesis-generating rather than practice-changing: in the first two weeks of sepsis, the coupled dance of CRP rising while albumin fails to recover may identify a small but distinctly vulnerable population, one that prospective multicenter studies must now confirm before this dynamic signature can inform bedside decisions.

Subject of Research: Joint longitudinal trajectories of C-reactive protein and albumin as prognostic phenotypes in early sepsis

Article Title: Dynamic Co‐Evolution of C‐Reactive Protein and Albumin in Early Sepsis: A Group‐Based Dual‐Trajectory Analysis for Risk Stratification

Article References: Chen, T., Zhang, S., Ding, M., Yang, Y., Yang, Q., Guo, W., Xu, X., & Liu, Q. (2026). Dynamic Co‐Evolution of C‐Reactive Protein and Albumin in Early Sepsis: A Group‐Based Dual‐Trajectory Analysis for Risk Stratification. Journal of Cellular and Molecular Medicine, 30(20), Article e71402. https://doi.org/10.1111/jcmm.71402

Image Credits: AI Generated

DOI: 10.1111/jcmm.71402

Keywords: sepsis, C-reactive protein, albumin, group-based dual-trajectory modelling, ICU mortality, risk stratification, acute-phase response, prognosis, critical care, biomarkers, Cox regression, SOFA score

The study, led by Guillermo Gutierrez and colleagues, addresses a foundational problem in clinical measurement: the reference standards against which new devices are validated are themselves imperfect. The team developed EBLA—Evidence-Based Label Adjustment—an algorithm that treats disagreement between two measurement methods as evidence to be weighed rather than error to be assigned, using the joint patterns of agreement and disagreement to estimate which method is more likely at fault on a case-by-case basis.

“The fundamental problem is that when we validate a new monitor against a clinical standard, we assume the standard is correct,” the authors explain. “But clinical assessments are themselves measurements with their own error rates. When the two disagree, conventional analysis simply assigns the error to the new device, which may substantially overstate its true inaccuracy.”

The approach exploits a statistical property of paired imperfect measurements: the relationship between how often two methods agree and their individual error rates follows a predictable mathematical structure. By modeling this structure across a patient population, EBLA can estimate the latent error rates of both methods simultaneously, without requiring either to be treated as ground truth. The algorithm then uses these estimates to adjust the labels on individual cases where the methods diverge, producing a corrected reference standard against which the new device can be more fairly evaluated.

To test the approach, the team turned to a genuine clinical problem: assessing respiratory effort in patients on mechanical ventilators who cannot communicate. The clinical standard—the modified Borg scale, in which nurses rate a patient’s breathing distress on a zero-to-ten scale—is subjective and noisy. The new technology is a continuous, noninvasive monitor of inspiratory effort developed by the same group. In a validation dataset of 1,292 paired measurements from 121 ventilated patients, the two methods agreed only about 71 percent of the time, a level of disagreement typical of comparisons between subjective and continuous measures.

Applying EBLA to this dataset produced a striking result: after label adjustment, the estimated error rate of the continuous monitor fell substantially, while the apparent error rate of the clinical standard rose. The area under the receiver operating characteristic curve for the monitor’s ability to detect elevated respiratory effort improved from 0.79 against the raw clinical ratings to 0.92 against the adjusted labels. “The monitor was being penalized for disagreements that were, in many cases, the fault of the reference standard,” the authors report.

The team validated the algorithm’s logic through simulation, constructing synthetic datasets in which the true error rates of both methods were known by design. Across scenarios varying the prevalence of the target condition and the error rates of each method, EBLA recovered the true error rates with high fidelity and consistently outperformed conventional analyses that treated the clinical standard as correct. The simulations also clarified the method’s boundaries: it requires that the two methods’ errors be at least partially independent, and it performs best when agreement between the methods is neither near-perfect nor near-chance.

The implications extend to any clinical field where new monitoring technologies are judged against imperfect standards—sepsis scoring, pain assessment, delirium screening, and weaning readiness among them. In each case, conventional validation may systematically understate the value of continuous measurement simply because the continuous method detects episodes that intermittent clinical assessment misses. A method that can estimate the reliability of both sides of a comparison, rather than assuming one side is correct, offers a way to break that bias.

The authors caution that EBLA is a tool for specific situations, not a universal replacement for validation studies. It requires paired measurements on the same subjects, reasonable sample sizes, and attention to the independence assumptions underlying the model. When a true gold standard exists, it should be used. But in the many clinical domains where no such standard exists, the choice has traditionally been between accepting an imperfect reference or abandoning validation altogether. This work suggests a third option: treat the disagreement itself as data, and let the mathematics sort out who is likely wrong.

The statistical core of the EBLA approach rests on a simple but powerful observation: when two independently fallible instruments agree, the probability that both are simultaneously wrong is much lower than the probability that either is wrong alone. If each method carries an independent error rate, the expected contamination among concordant observations shrinks roughly with the square of that rate. For a per-method error rate of fifteen percent, for example, the expected error rate among agreeing pairs falls to approximately three percent—a fivefold reduction in label noise. This enrichment of correct labels is what allows the concordant subset to serve as a high-confidence training set, even though neither instrument can be treated as ground truth.

Once the concordant observations are identified, EBLA trains a committee of five deliberately diverse machine-learning classifiers: a random forest, a gradient-boosted tree ensemble, logistic regression, a support vector machine with a radial basis kernel, and k-nearest neighbors. The diversity is not incidental. Because the adjudication of disputed cases depends on the classifiers reaching conclusions independently, the authors deliberately excluded architecturally similar algorithms to minimize correlated errors. Each learner embodies a distinct inductive bias—linear separation, kernel-based mapping, bagged trees, boosting, and instance-based reasoning—so that any systematic weakness in one model is unlikely to be shared by the others. Discordant observations are then adjudicated by threshold voting, with majority rule requiring at least three of five votes, while stricter supermajority and unanimity thresholds leave unresolved cases designated as ambiguous and reported separately rather than forced into a decision.

Because the clinical dyspnea dataset lacks any independent ground truth, the team first stress-tested the method on the Wisconsin Breast Cancer Dataset, a benchmark of 569 cases with pathology-confirmed diagnoses. By randomly flipping class labels at controlled error rates—ranging from symmetric ten-percent noise in both sources to asymmetric combinations such as thirty-percent corruption in one label set and twenty in the other—the researchers could measure exactly how well EBLA recovered the true labels. The fifteen-percent symmetric scenario was designated the primary test case because it approximates both the roughly nineteen-percent discordance observed in the ventilator cohort and a plausible error range for behavioral observation scales. Notably, only the labels were perturbed; the underlying feature values remained intact, isolating the effect of label noise from measurement noise in the features themselves.

The validation protocol incorporated several safeguards against optimistic bias. Classifier performance was assessed with stratified five-fold cross-validation, preserving the proportion of positive and negative cases in each fold, and feature standardization was performed separately within each iteration using parameters estimated only from the training folds. This prevented information from the held-out data from leaking into preprocessing, a common source of inflated performance estimates in machine-learning studies of clinical data.

The authors are explicit about the conditions under which the method is appropriate. Errors between the two instruments must be at least partially independent; if both are influenced by a shared confounder—deep sedation, for instance, might simultaneously blunt behavioral signs of distress and alter the physiological signals—the concordant subset loses its purity and the entire adjudication logic weakens. The concordant subset must also be large enough to support reliable classifier training, with the authors suggesting at least eighty concordant observations in their experience, and informative features must be available that are independent of those used to define the original classification thresholds. Finally, the voting threshold should be specified in advance according to the relative clinical consequences of false-positive and false-negative classifications, rather than chosen after seeing the results.

These caveats matter because real-world measurement error is rarely as clean as the simulated random flips used in the benchmark tests. Systematic biases—an observer who consistently under-rates distress, a monitor whose calibration drifts in a particular direction—violate the independence assumption in ways that random corruption does not. The authors acknowledge this limitation directly, positioning EBLA as a tool for assessing the potential impact of reference-standard error on validation results rather than as a replacement for the primary analysis. The framework is best understood as a sensitivity analysis: it quantifies how much of the observed disagreement between a new measure and an established one can plausibly be attributed to imperfection in the reference itself.

The clinical motivation remains pressing. Dyspnea affects roughly half of mechanically ventilated patients and is associated with prolonged ventilation, increased mortality, and post-traumatic stress disorder, yet in patients who cannot communicate it can only be detected through behavioral observation. The Mechanical Ventilation Respiratory Distress Observation Scale, the most extensively validated instrument for this purpose, is applied intermittently and is susceptible to inter-rater variability—precisely the profile of an imperfect reference standard. A continuous estimate of inspiratory effort derived from routine ventilator airway pressure and flow signals offers the possibility of monitoring that does not depend on observer availability or interpretation, and methods like EBLA provide a principled way to evaluate such measures when the alternative comparator is itself noisy.

Subject of Research: Ensemble‐based label adjudication (EBLA): A method for validating continuous physiological measures against observer‐dependent reference standards

Article Title: Ensemble‐based label adjudication (EBLA): A method for validating continuous physiological measures against observer‐dependent reference standards

Article References: Gutierrez, G., & Papa, F. (2026). Ensemble‐based label adjudication ( EBLA ): A method for validating continuous physiological measures against observer‐dependent reference standards. Physiological Reports, 14(19), Article e71127. https://doi.org/10.14814/phy2.71127

Image Credits: AI Generated

DOI: 10.14814/phy2.71127

Keywords: diagnostics, clinical measurement, machine learning, respiratory monitoring, validation methods, medical devices, methodology, critical care, Ensemble, based, label, adjudication