Left ventricular (LV) dysfunction is associated with poor clinical outcomes in acute respiratory failure (ARF). This study evaluates the efficacy of LV strain in detecting LV dysfunction in ARF patients requiring invasive mechanical ventilation (IMV) compared to conventionally measured left ventricular ejection fraction (LVEF). ARF patients requiring IMV who had echocardiography performed during MICU admission were included. LV global longitudinal strain (LVGLS) and LVEF were measured retrospectively using speckle tracking (STE) and traditional transthoracic echocardiography (TTE), respectively, by investigators blinded to the status of IMV and clinical data. The cohort was divided into three groups: TTE during IMV (TTE-IMV), before IMV (TTE-bIMV), and after IMV (TTE-aIMV). Multivariable regression models, adjusted for illness severity score, chronic cardiac disease, acute respiratory failure etiology, body mass index, chronic obstructive pulmonary disease, and obstructive sleep apnea, evaluated associations between LV function parameters and the presence of IMV. Among 376 patients, TTE-IMV, TTE-bIMV, and TTE-aIMV groups constituted 223, 68, and 85 patients, respectively. The median age was 65 years (IQR: 56–74), with 53.2% male participants. Adjusted models showed significantly higher LVGLS in groups not on IMV at the time of TTE (TTE-bIMV: β = 4.19, 95% CI 2.31 to 6.08, p < 0.001; TTE-aIMV: β = 3.79, 95% CI 2.03 to 5.55, p < 0.001), while no significant differences in LVEF were observed across groups. In a subgroup analysis of patients with LVEF ≥55%, the significant difference in LVGLS among the groups remained (TTE-bIMV: β = 4.18, 95% CI 2.22 to 6.15, p < 0.001; TTE-aIMV: β = 3.45, 95% CI 1.50 to 5.40, p < 0.001), but was no longer present in those with LVEF < 55%. This suggests an association between IMV and lower LVGLS in ARF patients requiring IMV, indicating that LVGLS may be a more sensitive marker for detecting subclinical LV dysfunction compared to LVEF in this population. Future studies should track and assess serial echocardiography data in the same cohort of patients pre-, during, and post-IMV in order to validate these findings and prognosticate STE-detected LV dysfunction in ARF patients requiring IMV.

Right ventricular (RV) dysfunction is associated with poor prognosis in acute respiratory failure (ARF). Our study evaluates the efficacy of RV strain in detecting RV dysfunction in ARF patients requiring invasive mechanical ventilation (IMV) compared to tricuspid annular plane systolic excursion (TAPSE). In this retrospective study involving 376 patients diagnosed with ARF and requiring IMV, we extracted clinical and outcome data from patient records. RV global longitudinal strain (RVGLS), free wall longitudinal strain (FWLS), and TAPSE were measured retrospectively using speckle tracking echocardiography (STE) and traditional echocardiography, respectively. We divided the cohort into three groups: TTE during IMV (TTE-IMV, 223 patients), before IMV (TTE-bIMV, 68 patients), and after IMV (TTE-aIMV, 85 patients). Multivariable regression analysis, adjusted for covariates, revealed significantly higher RVGLS and FWLS in the groups not on IMV at the time of TTE compared to the TTE-IMV group. Specifically, the TTE-bIMV group showed higher RVGLS (β = 7.28, 95% CI 5.07, 9.48) and FWLS (β = 5.83, 95% CI 3.36, 8.31), while the TTE-aIMV group exhibited higher RVGLS (β = 9.39, 95% CI 6.10, 12.69) and FWLS (β = 7.54, 95% CI 4.83, 10.24). TAPSE did not reveal any significant differences across the groups. Our study suggests an association between IMV and lower RVGLS and FWLS in ARF patients, indicating that IMV itself may contribute to RV dysfunction. RVGLS and FWLS appear to be more sensitive than TAPSE in detecting changes in RV function that were previously subclinical in patients on IMV. Prospective studies with TTE before, during, and after IMV are necessary to assess the primary driver of RV dysfunction and to prognosticate STE-detected RV dysfunction in this population.

BACKGROUND: At least 50% of individuals who suffer sudden cardiac arrest (SCA) have warning symptoms before their SCA, but these are not sufficient to predict imminent SCA (ISCA). We hypothesized that combining symptoms with clinical profile could effectively predict ISCA. METHODS: In 2 community-based studies of SCA in Oregon and California, survivors of SCA who had experienced warning symptoms and called 911 (n=364) were compared with control subjects (n=313) who also called 911 for similar symptoms but did not have SCA. Symptom information was obtained directly from interviews as well as prehospital emergency records. We used machine learning to identify predictive combinations of symptoms and clinical features in this case-control study. The area under the receiver operating characteristic curve (AUC) and 5-fold cross-validation were used to assess model performance and stability. RESULTS: Warning symptom-clinical profile combinations were successfully identified. The combination of chest pain and known heart failure predicted ISCA among males, while chest pain and known coronary artery disease predicted ISCA among females (AUC=0.788). With dyspnea, both history of coronary artery disease and heart failure were predictors (AUC=0.745). In the absence of chest pain or dyspnea, seizure-like symptoms, coronary artery disease, and heart failure were predictors (AUC, 0.775). 5-fold cross-validation produced consistent results. CONCLUSIONS: Combinations of warning symptoms and clinical features distinguished individuals with SCA from individuals without SCA with good accuracy (AUCs, 0.745–0.788). These findings have implications for clinical triage of ISCA warning symptoms in emergency settings and warrant investigation in larger studies.

Background: Identification of in-hospital cardiac arrest (IHCA) through manual chart abstraction is the gold standard, but its time-consuming and resource-intensive, which naturally limits its practicality. Diagnostic codes provide an accessible automated alternative, but prior work has shown this method of extraction has poor sensitivity and poor positive predictive value. More accurate and scalable automated methods are needed to support IHCA surveillance and research. Methods: We developed large language model (LLM)-based systems that incorporated physician-defined clinical knowledge through guideline-enriched prompting, supervised finetuning, and a self-correction procedure. The systems identified IHCA events and their locations from electronic health record notes. Performance was evaluated at the encounter level against physician-adjudicated chart review and compared with ICD-code–based identification. Results: Compared with an ICD-code–based algorithm (precision 0.68, recall 0.91, F1 score 0.78), the supervised fine-tuned LLM offered substantially better performance for IHCA identification on a curated evaluation corpus, reaching a precision of 0.88, a recall of 1.00, and an F1 score of 0.94. When tested on a larger, unselected validation cohort of 45,525 encounters, a separate LLM configuration—relying on guideline-enriched prompting and a self-correction procedure rather than fine-tuning—achieved a precision of 0.79 for encounter-level IHCA identification, meaning most encounters flagged by the model were genuine IHCA cases. The fine-tuned LLM underperformed on this cohort due to its inability to self-correct its decisions, an issue that can be mitigated by having a separate model perform the correction step. Conclusion: Explicit integration of clinical knowledge through prompting and supervised fine-tuning enabled accurate identification of IHCA events and locations from routine clinical documentation. In a two-pass workflow, these systems could prioritize encounters for clinician adjudication, reduce manual screening requirements, and support more scalable IHCA surveillance and registry development.