OBJECTIVES: Echocardiography is commonly used for hemodynamic assessment in sepsis, but data regarding its association with outcome are conflicting. The aim of this study was to evaluate the association between echocardiography and outcomes in patients with septic shock using the Medical Information Mart for Intensive Care IV database. DESIGN: Retrospective cohort study comparing patients who did or did not undergo transthoracic echocardiography within the first 5 days of admission for the primary outcome of 28-day mortality. SETTING: Admissions to the Beth Israel Deaconess Medical Center intensive care from 2008 to 2019. PATIENTS: Adults 16 years old or older with septic shock requiring vasopressor support within 48 hours of admission. Readmissions and patients admitted to the coronary care unit or cardiovascular intensive care were excluded, as well as patients with ST-elevation myocardial infarction or cardiac arrest. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: Echocardiography was performed in 1,515 (27%) of 5,697 eligible admissions. The primary outcome was analyzed using a marginal structural model and rolling entry matching to adjust for baseline and time-varying confounders. Patients who underwent echocardiography showed no significant difference in 28-day mortality (adjusted hazard ratio 1.09; 95% CI, 0.95–1.25; p = 0.24). This was consistent across multiple sensitivity analyses. Secondary outcomes were changes in management instituted within 4 hours of imaging. Treatment changes occurred in 493 patients (33%) compared with 431 matched controls (29%), with the most common intervention being the administration of a fluid bolus. CONCLUSIONS: Echocardiography in sepsis was not associated with a reduction in 28-day mortality based on observational data. These findings do not negate the utility of echo in cases of diagnostic uncertainty or inadequate response to initial treatment.

Introduction Small case series have reported that diabetic ketoacidosis is associated with an elevated osmolar gap, while no previous studies have assessed the accuracy of calculated osmolarity in the hyperosmolar hyperglycemic state. The aim of this study was to characterize the magnitude of the osmolar gap in these conditions and assess whether this changes over time. Methods In this retrospective cohort study, two publicly available intensive care datasets were used: Medical Information Mart of Intensive Care IV and the eICU Collaborative Research Database. We identified adult admissions with diabetic ketoacidosis and the hyperosmolar hyperglycemic state who had measured osmolality results available contemporaneously with sodium, urea and glucose values. Calculated osmolarity was derived using the formula 2Na + glucose + urea (all values in mmol/L). Results We identified 995 paired values for measured and calculated osmolarity from 547 admissions (321 diabetic ketoacidosis, 103 hyperosmolar hyperglycemic states and 123 mixed presentations). A wide variation in the osmolar gap was seen, including substantial elevations and low and negative values. There was a greater frequency of raised osmolar gaps at the start of the admission, which tends to normalize by around 12–24 h. Similar results were seen regardless of the admission diagnosis. Conclusions The osmolar gap varies widely in diabetic ketoacidosis and the hyperosmolar hyperglycemic state and may be highly elevated, especially at admission. Clinicians should be aware that measured and calculated osmolarity values are not interchangeable in this population. These findings should be confirmed in a prospective study.

Implements Sequential Doubly Robust (SDR) and infinite-dimensional Targeted Maximum Likelihood (iTMLE) estimators for longitudinal modified treatment policies in settings with transitioning states, such as ICU, ward, or emergency department care episodes. Treatment is permitted in active states and becomes structurally inapplicable after a state transition (e.g. discharge or death). Supports asymmetric g- and Q-model regularisation, k-fold cross-fitting, and pluggable SuperLearner ensembles. Includes specialised SuperLearner wrappers (SL.tgt.* and SL.tmle_* families) for the iTMLE targeting step, which pass the logit offset as a covariate column to preserve correct subsetting during SuperLearner cross-validation. Methods based on Diaz et al. (2021) and Luedtke et al. (2017) .