Journal of Clinical Monitoring and Computing最新文献

Purpose: Preoperative anesthesia evaluation is a crucial step in ensuring patient safety and optimizing perioperative care. A heterogenous patient population requiring varying levels of assessment often leads to inefficiencies and additional resource allocation. This study proposes using pre-trained language models to assist in triaging the appropriate degree of preoperative anesthesia evaluation for surgical patients.

Methods: Retrospective institutional data were obtained from surgical patients evaluated at a single center preoperative anesthesia care clinic. The performance of four pre-trained language models (RoBERTa, BERT, ClinicalBERT, and PubMedBERT) in the classification of which patients would be appropriate for a nursing preoperative phone call versus in-person clinician evaluation was assessed using F1-score, area under the receiver operating characteristics curve (AUC), specificity, sensitivity, and average precision. For each pre-trained language model, three different data input combinations were assessed: (1) diagnosis codes (D); (2) clinical text data (N); and (3) diagnosis codes and clinical text (D + N). The data were split into training (75%) and test set (25%).

Results: There were 1,761 unique patients, with an average of 12 notes per patient and a total of 46,922 clinical documents, included in the analysis. The AUC range between the four language models was highest in the D + N analyses (0.70 - 0.74), lower in the N analyses (0.58 - 0.73) and lowest in the D analyses (0.57 - 0.62). RoBERTa had the highest score compared to the other language models for all data types.

Conclusions: Automating integrated analysis using pre-trained language models to aid in preoperative triaging could enhance accuracy and efficiency at scale, reducing manual review and provider burden.

Cardiac index (CI) is a key physiologic indicator correlated with end-organ perfusion in cardiac surgical patients, yet it is not routinely measured in all cases. This study evaluated the accuracy of estimating CI using routinely available physiologic monitor data, adjusted for relevant patient, physiologic, and procedural factors documented in perioperative anesthesia records. We analyzed anesthesia records from adult cardiac surgical patients with thermodilution-based CI measurements across seven US hospitals from 2014 to 2022. Four published formulas-based on intraoperative blood pressure and heart rate-were used to estimate CI in generalized linear models, with adjustment for perioperative patient and procedure characteristics. Bland-Altman analysis compared adjusted CI estimates to reference thermodilution CI values. The ability of each estimator to classify patients with low CI (< 2.2 L/min/m²) was assessed for concordance. In a cohort of 5,989 patients, the median (IQR = interquartile range) thermodilution-based CIs were 2.1 (1.8-2.6) and 2.4 (2.0-2.9) L/min/m² before and after cardiopulmonary bypass, respectively. The best-performing formula, Liljestrand and Zander, achieved mean absolute errors of 0.45 and 0.47 L/min/m² before and after bypass, respectively. However, its reliability in classifying low CI was limited (Cohen's kappa = 0.26 pre-bypass, 0.20 post-bypass). Routinely collected physiologic and patient data can be used to generate population-level cardiac index estimates in adult cardiac surgery patients when appropriately adjusted, though individual-level discrimination of low CI is limited. These findings inform future large-scale perioperative hemodynamic research.

Cerebrovascular autoregulation maintains stable cerebral blood flow by counteracting slow changes in cerebral perfusion pressure (termed "slow waves"). Conventional assessment involves invasive techniques using intracranial pressure (ICP) or technically challenging cerebral blood flow velocity (FV) measurements. Near-infrared spectroscopy (NIRS) has emerged as a non-invasive alternative; however, its ability to accurately capture the slow-wave oscillations fundamental to cerebrovascular autoregulation remains uncertain. 412 h of simultaneous ICP, FV, NIRS, and arterial blood pressure (ABP) monitoring from 35 traumatic brain injury patients were explored. Coherence, gain, and Granger causality analyses were employed to assess whether NIRS adequately reflects slow waves in ABP, FV, or ICP to investigate whether NIRS is a suitable alternative for assessing the state of cerebrovascular autoregulation In this single-centre observational cohort study, 89 recordings from 35 moderate to severe traumatic brain injury (TBI) patients (totalling 412 h of artefact-free data) were analysed. Simultaneous high-resolution recordings of NIRS, ICP, FV, and arterial blood pressure (ABP) were acquired. Coherence and gain were computed across defined frequency bands (0.001-0.5 Hz), with a focus on the range most relevant to cerebrovascular autoregulation (0.005-0.05 Hz). Granger causality was used to explore directional relationships between physiological inputs (ABP, FV, ICP) and NIRS outputs (rSO2 and haemoglobin metrics). Haemoglobin-based NIRS metrics (total, oxy-, deoxy-, and delta haemoglobin) demonstrated significantly higher coherence and Granger causality with FV and ICP compared to rSO2 (p < 0.001, large effect sizes) capturing the slow-wave oscillations central to cerebrovascular autoregulation. In contrast, rSO₂ exhibited poor coherence and low causality, especially with ABP, likely due to device-specific post-processing and resolution limitations. NIRS derived haemoglobin metrics reliably capture slow-wave dynamics reflective of cerebrovascular autoregulation and reactivity, offering a non-invasive alternative to traditional methods. Conversely, rSO2 lacks sufficient temporal fidelity to detect these fluctuations under routine clinical conditions, limiting its utility for cerebrovascular autoregulation assessment.

Critically ill patients often require complex extracorporeal treatments, such as extracorporeal blood purification (EBP). At the bedside, there can be reluctance or uncertainty about when to initiate EBP, and there is no standard agreement on which goals to pursue, what prescriptions to use to achieve those goals, or which recommendations to follow to prevent complications. Furthermore, an accurate analysis of why clinical goals are not achieved or how often the patient should be reassessed to readjust the EBP prescription is not currently standardized. This narrative review describes the main actions characterizing a quality improvement program for EBP in the ICU, which took place at the University of Florence and was subsequently adopted at the national level. The pillars of this program were: (1) definition, implementation, and dissemination of information and communication technology tools aimed at objectively measuring results at the bedside, supporting dynamic prescribing and precision medicine, and promoting advances in knowledge in this field; (2) creation of a national multi-professional network of clinical users and researchers in EBP; (3) promotion and maintenance of technical and non-technical skills in EBP based on the reformulation of advanced academic training in this field.

Postoperative delirium (POD) is a common and multifactorial complication following cardiac surgery, with cardiopulmonary bypass (CPB) playing a significant contributory role. Impaired cerebral autoregulation (CA) during CPB, particularly in older patients, may lead to cerebral hypo- or hyperperfusion. While several methods exist to assess CA and cerebral blood flow, many require specialized equipment not widely available. This prospective observational study aimed to investigate whether altered cerebral artery flow velocity, measured preoperatively by transcranial Doppler (TCD), is associated with the development of POD. We enrolled 41 patients undergoing elective cardiac surgery with CPB. Bilateral peak flow velocities of the middle cerebral arteries were measured preoperatively using TCD. The mean middle cerebral artery velocity (mMCAv_mean) was calculated for each patient. POD occurred in 21 patients (51%). A lower mMCAv_mean was significantly associated with an increased risk of POD. Specifically, each 1 cm/s decrease in mMCAv_mean increased the likelihood of POD by 9.2% (odds ratio 0.908; 95% confidence interval: 0.840-0.981; p = 0.015). Reduced cerebral blood flow velocity during CPB, as measured by TCD, is associated with a higher risk of POD. These findings highlight the potential utility of intraoperative TCD monitoring for early identification of at-risk patients and support further research into TCD-guided preventive strategies in cardiac surgery.

This study aimed to evaluate whether continuous axillary temperature monitoring using a wearable patch enables earlier detection of postoperative infections compared to conventional intermittent infrared thermometry. 103 surgical patients were included in this prospective, single-center study and monitored over an 11-month period. Continuous axillary temperature monitoring using the SteadyTemp^® patch was compared to routine infrared measurements performed as part of clinical routine. The primary outcome was fever detection rate (≥ 38.0 °C). Secondary outcomes included the correlation between fever detection and laboratory values as well as the frequency of clinical interventions. Out of 103 included patients, fever was detected in 33 cases. Continuous monitoring identified fever in 31 of these 33 patients (93.9%), whereas infrared thermometry detected fever in only 12 cases (36.4%). In 16 cases where antibiotic therapy was initiated or adjusted due to newly detected fever, the patch detected fever in 15 patients, compared to only 7 detections by infrared thermometry. Surgical interventions due to suspected infections were performed in 5 patients, and fever was detected by the patch in all cases, while infrared thermometry detected fever in only 2 of these patients. Due to the frequent failure of infrared thermometry to detect fever, a scoring system was developed to assess the clinical relevance of fever detection. Continuous temperature monitoring with the SteadyTemp^® patch demonstrated superior fever detection compared to infrared thermometry, leading to earlier identification of febrile events. This study suggests that continuous temperature monitoring may enhance infection surveillance in surgical patients, allowing for more timely clinical interventions.

We evaluated the accuracy and precision of zero-heat flux (ZHF) and dual sensor (DS) non-invasive temperature probes in intensive care unit (ICU) patients undergoing hypothermic temperature control, hypothesizing that both devices would accurately estimate core temperature. In a single-center prospective cohort study, we enrolled 35 ICU patients and applied continuous, non-invasive ZHF and/or DS probes to the lateral forehead or anterior chest to collect 358 observations. Conditions potentially influencing temperature estimation were recorded. Using Bland-Altman analysis with multiple paired observations per individual, we compared the bias between non-invasive probes and direct core temperature measurements. Lin's concordance coefficient (LCC) was computed to quantify precision. The mean bias between the ZHF probe and invasive temperature was + 0.98 °C; for the DS probe, it was - 2.19 °C. In hypothermic patients, the ZHF probe's accuracy improved (bias + 0.28 °C, LCC 0.86), while the DS probe remained inaccurate (bias - 2.52 °C, LCC 0.07). Clinical confounders like vasoactive agents or temperature control devices did not consistently affect bias, accuracy, or precision. Neither the ZHF nor DS non-invasive probes provided sufficient accuracy or precision to guide clinical decisions in the ICU. These results contrast with previous studies reporting biases within ± 0.5 °C. However, the ZHF probe showed promising limited deviation, especially in hypothermic patients.

Background: Assessing fluid responsiveness is crucial in managing critically ill patients. Echocardiography, particularly passive leg raising (PLR)-induced changes in the velocity-time integral of the left ventricular outflow tract (VTI_LVOT), is widely used for this purpose. We hypothesized that PLR-induced changes in the mitral valve velocity-time integral (VTI_Mi) could serve as a reliable alternative.

Methods: This prospective single-center study included septic ICU patients requiring fluid responsiveness assessment. VTI_LVOT and VTI_Mi were measured at baseline and after PLR. Fluid responsiveness was defined as a PLR-induced increase in VTI_LVOT ≥10%. The ability of PLR-induced VTI_Mi changes to predict fluid responsiveness was assessed via ROC curve and gray zone analyses.

Results: Fifty consecutive patients were included (median age 65 years [IQR: 57-73], APACHE II score 22 [IQR: 18-27]). Septic shock was present in 27 (54%), 21 (42%) were mechanically ventilated, and 23 (46%) were classified as responders. PLR-induced changes in VTI_Mi and VTI_LVOT were significantly correlated (ρ = 0.656, p < 0.001). The area under the ROC curve for VTI_Mi was 0.927 (95% CI: 0.849-1, p < 0.001). A 10% increase in VTI_Mi predicted fluid responsiveness with a sensitivity of 83% (95% CI: 61-95) and specificity of 96% (95% CI: 83-99). The gray zone ranged between 5% and 8%, encompassing 16% of the cohort.

Conclusion: PLR-induced changes in VTI_Mi reliably predict fluid responsiveness in critically ill patients. VTI_Mi represents a viable alternative to VTI_LVOT for fluid responsiveness assessment, contributing to individualized hemodynamic management.

Trial registration: NCT05538637.