Herasevich, Vitaly

AbstractObjective Video recording and video recognition (VR) with computer vision have become widely used in many aspects of modern life. Hospitals have employed VR technology for security purposes, however, despite the growing number of studies showing the feasibility of VR software for physiologic monitoring or detection of patient movement, its use in the intensive care unit (ICU) in real-time is sparse and the perception of this novel technology is unknown. The objective of this study is to understand the attitudes of providers, patients, and patient's families toward using VR in the ICU.Design A 10-question survey instrument was used and distributed into two groups of participants: clinicians (MDs, advance practice providers, registered nurses), patients and families (adult patients and patients' relatives). Questions were specifically worded and section for free text-comments created to elicit respondents' thoughts and attitudes on potential issues and barriers toward implementation of VR in the ICU.Setting The survey was conducted at Mayo Clinic in Minnesota and Florida.Results A total of 233 clinicians' and 50 patients' surveys were collected. Both cohorts favored VR under specific circumstances (e.g., invasive intervention and diagnostic manipulation). Acceptable reasons for VR usage according to clinicians were anticipated positive impact on patient safety (70%), and diagnostic suggestions and decision support (51%). A minority of providers was concerned that artificial intelligence (AI) would replace their job (14%) or erode professional skills (28%). The potential use of VR in lawsuits (81% clinicians) and privacy breaches (59% patients) were major areas of concern. Further identified barriers were lack of trust for AI, deterioration of the patient–clinician rapport. Patients agreed with VR unless it does not reduce nursing care or record sensitive scenarios.Conclusion The survey provides valuable information on the acceptance of VR cameras in the critical care setting including an overview of real concerns and attitudes toward the use of VR technology in the ICU.

Abstract Objective Access to palliative care (PC) is important for many patients with uncontrolled symptom burden from serious or complex illness. However, many patients who could benefit from PC do not receive it early enough or at all. We sought to address this problem by building a predictive model into a comprehensive clinical framework with the aims to (i) identify in-hospital patients likely to benefit from a PC consult, and (ii) intervene on such patients by contacting their care team. Materials and Methods Electronic health record data for 68 349 inpatient encounters in 2017 at a large hospital were used to train a model to predict the need for PC consult. This model was published as a web service, connected to institutional data pipelines, and consumed by a downstream display application monitored by the PC team. For those patients that the PC team deems appropriate, a team member then contacts the patient’s corresponding care team. Results Training performance AUC based on a 20% holdout validation set was 0.90. The most influential variables were previous palliative care, hospital unit, Albumin, Troponin, and metastatic cancer. The model has been successfully integrated into the clinical workflow making real-time predictions on hundreds of patients per day. The model had an “in-production” AUC of 0.91. A clinical trial is currently underway to assess the effect on clinical outcomes. Conclusions A machine learning model can effectively predict the need for an inpatient PC consult and has been successfully integrated into practice to refer new patients to PC.

ABSTRACT Introduction The emergence of more complex Prolonged Field Care in austere settings and the need to assist inexperienced providers’ ability to treat patients create an urgent need for effective tools to support care. We report on a project to develop a phone-/tablet-based decision support system for prehospital tactical combat casualty care that collects physiologic and other clinical data and uses machine learning to detect and differentiate shock manifestation. Materials and Methods Software interface development methods included literature review, rapid prototyping, and subject matter expert design requirements reviews. Machine learning algorithm methods included development of a model trained on publicly available Medical Information Mart for Intensive Care data, then on de-identified data from Mayo Clinic Intensive Care Unit. Results The project team interviewed 17 Army, Air Force, and Navy medical subject matter experts during design requirements review sessions. They had an average of 17 years of service in military medicine and an average of 4 deployments apiece and all had performed tactical combat casualty care on live patients during deployment. Comments provided requirements for shock identification and management in prehospital settings, including support for indication of shock probability and shock differentiation. The machine learning algorithm based on logistic regression performed best among other algorithms we tested and was able to predict shock onset 90 minutes before it occurred with better than 75% accuracy in the test dataset. Conclusions We expect the Trauma Triage, Treatment, and Training Decision Support system will augment a medic’s ability to make informed decisions based on salient patient data and to diagnose multiple types of shock through remotely trained, field deployed ML models.

Abstract BackgroundAcute respiratory failure associated with sepsis contributes to higher in-hospital mortality. Intubation and invasive mechanical ventilation is a common rescue procedure. However, the 2016 International Guidelines for Management of Sepsis and Septic Shock does not provide any recommendation on indication nor timing of intubation. Timely intubation may improve outcome. The decision to intubate those patients is often hampered by the fear of further hemodynamic deterioration following intubation. MethodsThis study aimed at evaluating the impact of timely intubation on outcome in sepsis associated respiratory failure. We conducted an ancillary analysis of a prospective registry od adult ICU patients with septic shock admitted to the medical ICU in a tertiary medical center, between April 30th, 2014 and December 31st, 2017. All cases of sepsis with lactate >4 mmol/L, mean arterial pressure <65 mmHg, or vasopressor use after 30 mL/kg fluid boluses and suspected or confirmed infection. Patients who remained hospitalized at 24 hours following sepsis onset were separated into intubated and non-intubated groups. The primary outcome was hospital mortality. Univariate and multivariable analyses were used, adjusted for admission characteristics and stabilization of shock within 6 hours. In a secondary analysis, time-dependent propensity score matching was used to match intubated and non-intubated patients.ResultsWe identified 345 (33%) patients intubated within 24 hours and 707 (67%) not intubated. Intubated patients were younger, transferred more often from an outside facility, had higher severity of illness scores, more lung infection, achieved blood pressure goals more often but less often lactate normalization within 6 hours. The crude in-hospital mortality was higher, 89 (26%) vs. 82 (12%), p<0.001, as were ICU mortality, and ICU and hospital length of stay. After adjustment, intubation showed no effect on hospital mortality but fewer hospital-free days through day 28. After 1:1 propensity score matching, there was no difference in hospital mortality, but fewer hospital-free days in the intubated group. ConclusionsIntubation within 24 hours of sepsis onset was safe and not associated with hospital mortality, but was associated with less 28-day hospital-free days. Intubation should not be discouraged in appropriate patients with septic shock.

Objectives: To develop and verify a digital twin model of critically ill patient using the causal artificial intelligence approach to predict the response to specific treatment during the first 24 hours of sepsis. Design: Directed acyclic graphs were used to define explicitly the causal relationship among organ systems and specific treatments used. A hybrid approach of agent-based modeling, discrete-event simulation, and Bayesian network was used to simulate treatment effect across multiple stages and interactions of major organ systems (cardiovascular, neurologic, renal, respiratory, gastrointestinal, inflammatory, and hematology). Organ systems were visualized using relevant clinical markers. The application was iteratively revised and debugged by clinical experts and engineers. Agreement statistics was used to test the performance of the model by comparing the observed patient response versus the expected response (primary and secondary) predicted by digital twin. Setting: Medical ICU of a large quaternary- care academic medical center in the United States. Patients or Subjects: Adult (> 18 year yr old), medical ICU patients were included in the study. Interventions: No additional interventions were made beyond the standard of care for this study. Measurements and Main Results: During the verification phase, model performance was prospectively tested on 145 observations in a convenience sample of 29 patients. Median age was 60 years (54–66 d) with a median Sequential Organ Failure Assessment score of 9.5 (interquartile range, 5.0–14.0). The most common source of sepsis was pneumonia, followed by hepatobiliary. The observations were made during the first 24 hours of the ICU admission with one-step interventions, comparing the output in the digital twin with the real patient response. The agreement between the observed versus and the expected response ranged from fair (kappa coefficient of 0.41) for primary response to good (kappa coefficient of 0.65) for secondary response to the intervention. The most common error detected was coding error in 50 observations (35%), followed by expert rule error in 29 observations (20%) and timing error in seven observations (5%). Conclusions: We confirmed the feasibility of development and prospective testing of causal artificial intelligence model to predict the response to treatment in early stages of critical illness. The availability of qualitative and quantitative data and a relatively short turnaround time makes the ICU an ideal environment for development and testing of digital twin patient models. An accurate digital twin model will allow the effect of an intervention to be tested in a virtual environment prior to use on real patients.

Abstract Background Although diagnostic error (DE) is a significant problem, it remains challenging for clinicians to identify it reliably and to recognize its contribution to the clinical trajectory of their patients. The purpose of this work was to evaluate the reliability of real-time electronic health record (EHR) reviews using a search strategy for the identification of DE as a contributor to the rapid response team (RRT) activation. Objectives Early and accurate recognition of critical illness is of paramount importance. The objective of this study was to test the feasibility and reliability of prospective, real-time EHR reviews as a means of identification of DE. Methods We conducted this prospective observational study in June 2019 and included consecutive adult patients experiencing their first RRT activation. An EHR search strategy and a standard operating procedure were refined based on the literature and expert clinician inputs. Two physician-investigators independently reviewed eligible patient EHRs for the evidence of DE within 24 hours after RRT activation. In cases of disagreement, a secondary review of the EHR using a taxonomy approach was applied. The reviewers categorized patient experience of DE as Yes/No/Uncertain. Results We reviewed 112 patient records. DE was identified in 15% of cases by both reviewers. Kappa agreement with the initial review was 0.23 and with the secondary review 0.65. No evidence of DE was detected in 60% of patients. In 25% of cases, the reviewers could not determine whether DE was present or absent. Conclusion EHR review is of limited value in the real-time identification of DE in hospitalized patients. Alternative approaches are needed for research and quality improvement efforts in this field.