Perioperative mental health intervention for depression and anxiety symptoms in older adults study protocol: design and methods for three linked randomised controlled trials.

INTRODUCTION: Preoperative anxiety and depression symptoms among older surgical patients are associated with poor postoperative outcomes, yet evidence-based interventions for anxiety and depression have not been applied within this setting. We present a protocol for randomised controlled trials (RCTs) in three surgical cohorts: cardiac, oncological and orthopaedic, investigating whether a perioperative mental health intervention, with psychological and pharmacological components, reduces perioperative symptoms of depression and anxiety in older surgical patients. METHODS AND ANALYSIS: Adults ≥60 years undergoing cardiac, orthopaedic or oncological surgery will be enrolled in one of three-linked type 1 hybrid effectiveness/implementation RCTs that will be conducted in tandem with similar methods. In each trial, 100 participants will be randomised to a remotely delivered perioperative behavioural treatment incorporating principles of behavioural activation, compassion and care coordination, and medication optimisation, or enhanced usual care with mental health-related resources for this population. The primary outcome is change in depression and anxiety symptoms assessed with the Patient Health Questionnaire-Anxiety Depression Scale from baseline to 3 months post surgery. Other outcomes include quality of life, delirium, length of stay, falls, rehospitalisation, pain and implementation outcomes, including study and intervention reach, acceptability, feasibility and appropriateness, and patient experience with the intervention. ETHICS AND DISSEMINATION: The trials have received ethics approval from the Washington University School of Medicine Institutional Review Board. Informed consent is required for participation in the trials. The results will be submitted for publication in peer-reviewed journals, presented at clinical research conferences and disseminated via the Center for Perioperative Mental Health website. TRIAL REGISTRATION NUMBERS: NCT05575128, NCT05685511, NCT05697835, pre-results.

Predicting Primary Care Physician Burnout From Electronic Health Record Use Measures.

OBJECTIVE: To evaluate the ability of routinely collected electronic health record (EHR) use measures to predict clinical work units at increased risk of burnout and potentially most in need of targeted interventions. METHODS: In this observational study of primary care physicians, we compiled clinical workload and EHR efficiency measures, then linked these measures to 2 years of well-being surveys (using the Stanford Professional Fulfillment Index) conducted from April 1, 2019, through October 16, 2020. Physicians were grouped into training and confirmation data sets to develop predictive models for burnout. We used gradient boosting classifier and other prediction modeling algorithms to quantify the predictive performance by the area under the receiver operating characteristics curve (AUC). RESULTS: Of 278 invited physicians from across 60 clinics, 233 (84%) completed 396 surveys. Physicians were 67% women with a median age category of 45 to 49 years. Aggregate burnout score was in the high range (≥3.325/10) on 111 of 396 (28%) surveys. Gradient boosting classifier of EHR use measures to predict burnout achieved an AUC of 0.59 (95% CI, 0.48 to 0.77) and an area under the precision-recall curve of 0.29 (95% CI, 0.20 to 0.66). Other models' confirmation set AUCs ranged from 0.56 (random forest) to 0.66 (penalized linear regression followed by dichotomization). Among the most predictive features were physician age, team member contributions to notes, and orders placed with user-defined preferences. Clinic-level aggregate measures identified the top quartile of clinics with 56% sensitivity and 85% specificity. CONCLUSION: In a sample of primary care physicians, routinely collected EHR use measures demonstrated limited ability to predict individual burnout and moderate ability to identify high-risk clinics.

National Multi-Institutional Validation of a Surgical Transfusion Risk Prediction Model.

BACKGROUND: Accurate estimation of surgical transfusion risk is important for many aspects of surgical planning, yet few methods for estimating are available for estimating such risk. There is a need for reliable validated methods for transfusion risk stratification to support effective perioperative planning and resource stewardship. STUDY DESIGN: This study was conducted using the American College of Surgeons NSQIP datafile from 2019. S-PATH performance was evaluated at each contributing hospital, with and without hospital-specific model tuning. Linear regression was used to assess the relationship between hospital characteristics and area under the receiver operating characteristic (AUROC) curve. RESULTS: A total of 1,000,927 surgical cases from 414 hospitals were evaluated. Aggregate AUROC was 0.910 (95% CI 0.904 to 0.916) without model tuning and 0.925 (95% CI 0.919 to 0.931) with model tuning. AUROC varied across individual hospitals (median 0.900, interquartile range 0.849 to 0.944), but no statistically significant relationships were found between hospital-level characteristics studied and model AUROC. CONCLUSIONS: S-PATH demonstrated excellent discriminative performance, although there was variation across hospitals that was not well-explained by hospital-level characteristics. These results highlight the S-PATH's viability as a generalizable surgical transfusion risk prediction tool.

A Perioperative Mental Health Intervention for Depressed and Anxious Older Surgical Patients: Results From a Feasibility Study.

OBJECTIVES: The perioperative period is challenging and stressful for older adults. Those with depression and/or anxiety have an increased risk of adverse surgical outcomes. We assessed the feasibility of a perioperative mental health intervention composed of medication optimization and a wellness program following principles of behavioral activation and care coordination for older surgical patients. METHODS: We included orthopedic, oncologic, and cardiac surgical patients aged 60 and older. Feasibility outcomes included study reach, the number of patients who agreed to participate out of the total eligible; and intervention reach, the number of patients who completed the intervention out of patients who agreed to participate. Intervention efficacy was assessed using the Patient Health Questionnaire for Anxiety and Depression (PHQ-ADS). Implementation potential and experiences were collected using patient surveys and qualitative interviews. Complementary caregiver feedback was also collected. RESULTS: Twenty-three out of 28 eligible older adults participated in this study (mean age 68.0 years, 65% women), achieving study reach of 82% and intervention reach of 83%. In qualitative interviews, patients (n = 15) and caregivers (complementary data, n = 5) described overwhelmingly positive experiences with both the intervention components and the interventionist, and reported improvement in managing depression and/or anxiety. Preliminary efficacy analysis indicated improvement in PHQ-ADS scores (F = 12.13, p <0.001). CONCLUSIONS: The study procedures were reported by participants as feasible and the perioperative mental health intervention to reduce anxiety and depression in older surgical patients showed strong implementation potential. Preliminary data suggest its efficacy for improving depression and/or anxiety symptoms. A randomized controlled trial assessing the intervention and implementation effectiveness is currently ongoing.

Guidance for reporting analyses of metadata on electronic health record use.

INTRODUCTION: Research on how people interact with electronic health records (EHRs) increasingly involves the analysis of metadata on EHR use. These metadata can be recorded unobtrusively and capture EHR use at a scale unattainable through direct observation or self-reports. However, there is substantial variation in how metadata on EHR use are recorded, analyzed and described, limiting understanding, replication, and synthesis across studies. RECOMMENDATIONS: In this perspective, we provide guidance to those working with EHR use metadata by describing 4 common types, how they are recorded, and how they can be aggregated into higher-level measures of EHR use. We also describe guidelines for reporting analyses of EHR use metadata-or measures of EHR use derived from them-to foster clarity, standardization, and reproducibility in this emerging and critical area of research.

Association Between User Interaction and Treatment Response of a Voice-Based Coach for Treating Depression and Anxiety: Secondary Analysis of a Pilot Randomized Controlled Trial.

BACKGROUND: The quality of user interaction with therapeutic tools has been positively associated with treatment response; however, no studies have investigated these relationships for voice-based digital tools. OBJECTIVE: This study evaluated the relationships between objective and subjective user interaction measures as well as treatment response on Lumen, a novel voice-based coach, delivering problem-solving treatment to patients with mild to moderate depression or anxiety or both. METHODS: In a pilot trial, 42 adults with clinically significant depression (Patient Health Questionnaire-9 [PHQ-9]) or anxiety (7-item Generalized Anxiety Disorder Scale [GAD-7]) symptoms or both received Lumen, a voice-based coach delivering 8 problem-solving treatment sessions. Objective (number of conversational breakdowns, ie, instances where a participant's voice input could not be interpreted by Lumen) and subjective user interaction measures (task-related workload, user experience, and treatment alliance) were obtained for each session. Changes in PHQ-9 and GAD-7 scores at each ensuing session after session 1 measured the treatment response. RESULTS: Participants were 38.9 (SD 12.9) years old, 28 (67%) were women, 8 (19%) were Black, 12 (29%) were Latino, 5 (12%) were Asian, and 28 (67%) had a high school or college education. Mean (SD) across sessions showed breakdowns (mean 6.5, SD 4.4 to mean 2.3, SD 1.8) decreasing over sessions, favorable task-related workload (mean 14.5, SD 5.6 to mean 17.6, SD 5.6) decreasing over sessions, neutral-to-positive user experience (mean 0.5, SD 1.4 to mean 1.1, SD 1.3), and high treatment alliance (mean 5.0, SD 1.4 to mean 5.3, SD 0.9). PHQ-9 (Ptrend=.001) and GAD-7 scores (Ptrend=.01) improved significantly over sessions. Treatment alliance correlated with improvements in PHQ-9 (Pearson r=-0.02 to -0.46) and GAD-7 (r=0.03 to -0.57) scores across sessions, whereas breakdowns and task-related workload did not. Mixed models showed that participants with higher individual mean treatment alliance had greater improvements in PHQ-9 (ß=-1.13, 95% CI -2.16 to -0.10) and GAD-7 (ß=-1.17, 95% CI -2.13 to -0.20) scores. CONCLUSIONS: The participants had fewer conversational breakdowns and largely favorable user interactions with Lumen across sessions. Conversational breakdowns were not associated with subjective user interaction measures or treatment responses, highlighting how participants adapted and effectively used Lumen. Individuals experiencing higher treatment alliance had greater improvements in depression and anxiety. Understanding treatment alliance can provide insights on improving treatment response for this new delivery modality, which provides accessibility, flexibility, comfort with disclosure, and cost-related advantages compared to conventional psychotherapy. TRIAL REGISTRATION: ClinicalTrials.gov NCT04524104; https://clinicaltrials.gov/study/NCT04524104.

Characterizing the Patterns of Electronic Health Record-Integrated Secure Messaging Use: Cross-Sectional Study.

BACKGROUND: Communication among health care professionals is essential for the delivery of safe clinical care. Secure messaging has rapidly emerged as a new mode of asynchronous communication. Despite its popularity, relatively little is known about how secure messaging is used and how such use contributes to communication burden. OBJECTIVE: This study aims to characterize the use of an electronic health record-integrated secure messaging platform across 14 hospitals and 263 outpatient clinics within a large health care system. METHODS: We collected metadata on the use of the Epic Systems Secure Chat platform for 6 months (July 2022 to January 2023). Information was retrieved on message volume, response times, message characteristics, messages sent and received by users, user roles, and work settings (inpatient vs outpatient). RESULTS: A total of 32,881 users sent 9,639,149 messages during the study. Median daily message volume was 53,951 during the first 2 weeks of the study and 69,526 during the last 2 weeks, resulting in an overall increase of 29% (P=.03). Nurses were the most frequent users of secure messaging (3,884,270/9,639,149, 40% messages), followed by physicians (2,387,634/9,639,149, 25% messages), and medical assistants (1,135,577/9,639,149, 12% messages). Daily message frequency varied across users; inpatient advanced practice providers and social workers interacted with the highest number of messages per day (median 19). Conversations were predominantly between 2 users (1,258,036/1,547,879, 81% conversations), with a median of 2 conversational turns and a median response time of 2.4 minutes. The largest proportion of inpatient messages was from nurses to physicians (972,243/4,749,186, 20% messages) and physicians to nurses (606,576/4,749,186, 13% messages), while the largest proportion of outpatient messages was from physicians to nurses (344,048/2,192,488, 16% messages) and medical assistants to other medical assistants (236,694/2,192,488, 11% messages). CONCLUSIONS: Secure messaging was widely used by a diverse range of health care professionals, with ongoing growth throughout the study and many users interacting with more than 20 messages per day. The short message response times and high messaging volume observed highlight the interruptive nature of secure messaging, raising questions about its potentially harmful effects on clinician workflow, cognition, and errors.

Identifying and Addressing Barriers to Implementing Core Electronic Health Record Use Metrics for Ambulatory Care: Virtual Consensus Conference Proceedings.

Precise, reliable, valid metrics that are cost-effective and require reasonable implementation time and effort are needed to drive electronic health record (EHR) improvements and decrease EHR burden. Differences exist between research and vendor definitions of metrics. PROCESS: We convened three stakeholder groups (health system informatics leaders, EHR vendor representatives, and researchers) in a virtual workshop series to achieve consensus on barriers, solutions, and next steps to implementing the core EHR use metrics in ambulatory care. CONCLUSION: Actionable solutions identified to address core categories of EHR metric implementation challenges include: (1) maintaining broad stakeholder engagement, (2) reaching agreement on standardized measure definitions across vendors, (3) integrating clinician perspectives, and (4) addressing cognitive and EHR burden. Building upon the momentum of this workshop's outputs offers promise for overcoming barriers to implementing EHR use metrics.

An Intraoperative Telemedicine Program to Improve Perioperative Quality Measures: The ACTFAST-3 Randomized Clinical Trial.

Importance: Telemedicine for clinical decision support has been adopted in many health care settings, but its utility in improving intraoperative care has not been assessed. Objective: To pilot the implementation of a real-time intraoperative telemedicine decision support program and evaluate whether it reduces postoperative hypothermia and hyperglycemia as well as other quality of care measures. Design, Setting, and Participants: This single-center pilot randomized clinical trial (Anesthesiology Control Tower-Feedback Alerts to Supplement Treatments [ACTFAST-3]) was conducted from April 3, 2017, to June 30, 2019, at a large academic medical center in the US. A total of 26 254 adult surgical patients were randomized to receive either usual intraoperative care (control group; n = 12 980) or usual care augmented by telemedicine decision support (intervention group; n = 13 274). Data were initially analyzed from April 22 to May 19, 2021, with updates in November 2022 and February 2023. Intervention: Patients received either usual care (medical direction from the anesthesia care team) or intraoperative anesthesia care monitored and augmented by decision support from the Anesthesiology Control Tower (ACT), a real-time, live telemedicine intervention. The ACT incorporated remote monitoring of operating rooms by a team of anesthesia clinicians with customized analysis software. The ACT reviewed alerts and electronic health record data to inform recommendations to operating room clinicians. Main Outcomes and Measures: The primary outcomes were avoidance of postoperative hypothermia (defined as the proportion of patients with a final recorded intraoperative core temperature >36 °C) and hyperglycemia (defined as the proportion of patients with diabetes who had a blood glucose level ≤180 mg/dL on arrival to the postanesthesia recovery area). Secondary outcomes included intraoperative hypotension, temperature monitoring, timely antibiotic redosing, intraoperative glucose evaluation and management, neuromuscular blockade documentation, ventilator management, and volatile anesthetic overuse. Results: Among 26 254 participants, 13 393 (51.0%) were female and 20 169 (76.8%) were White, with a median (IQR) age of 60 (47-69) years. There was no treatment effect on avoidance of hyperglycemia (7445 of 8676 patients [85.8%] in the intervention group vs 7559 of 8815 [85.8%] in the control group; rate ratio [RR], 1.00; 95% CI, 0.99-1.01) or hypothermia (7602 of 11 447 patients [66.4%] in the intervention group vs 7783 of 11 672 [66.7.%] in the control group; RR, 1.00; 95% CI, 0.97-1.02). Intraoperative glucose measurement was more common among patients with diabetes in the intervention group (RR, 1.07; 95% CI, 1.01-1.15), but other secondary outcomes were not significantly different. Conclusions and Relevance: In this randomized clinical trial, anesthesia care quality measures did not differ between groups, with high confidence in the findings. These results suggest that the intervention did not affect the targeted care practices. Further streamlining of clinical decision support and workflows may help the intraoperative telemedicine program achieve improvement in targeted clinical measures. Trial Registration: ClinicalTrials.gov Identifier: NCT02830126.

Anesthesia Clinical Workload Estimated From Electronic Health Record Documentation vs Billed Relative Value Units.

Importance: Accurate measurements of clinical workload are needed to inform health care policy. Existing methods for measuring clinical workload rely on surveys or time-motion studies, which are labor-intensive to collect and subject to biases. Objective: To compare anesthesia clinical workload estimated from electronic health record (EHR) audit log data vs billed relative value units. Design, Setting, and Participants: This cross-sectional study of anesthetic encounters occurring between August 26, 2019, and February 9, 2020, used data from 8 academic hospitals, community hospitals, and surgical centers across Missouri and Illinois. Clinicians who provided anesthetic services for at least 1 surgical encounter were included. Data were analyzed from January 2022 to January 2023. Exposure: Anesthetic encounters associated with a surgical procedure were included. Encounters associated with labor analgesia and endoscopy were excluded. Main Outcomes and Measures: For each encounter, EHR-derived clinical workload was estimated as the sum of all EHR actions recorded in the audit log by anesthesia clinicians who provided care. Billing-derived clinical workload was measured as the total number of units billed for the encounter. A linear mixed-effects model was used to estimate the relative contribution of patient complexity (American Society of Anesthesiology [ASA] physical status modifier), procedure complexity (ASA base unit value for the procedure), and anesthetic duration (time units) to EHR-derived and billing-derived workload. The resulting ß coefficients were interpreted as the expected effect of a 1-unit change in each independent variable on the standardized workload outcome. The analysis plan was developed after the data were obtained. Results: A total of 405 clinicians who provided anesthesia for 31 688 encounters were included in the study. A total of 8 288 132 audit log actions corresponding to 39 131 hours of EHR use were used to measure EHR-derived workload. The contributions of patient complexity, procedural complexity, and anesthesia duration to EHR-derived workload differed significantly from their contributions to billing-derived workload. The contribution of patient complexity toward EHR-derived workload (ß = 0.162; 95% CI, 0.153-0.171) was more than 50% greater than its contribution toward billing-derived workload (ß = 0.106; 95% CI, 0.097-0.116; P < .001). In contrast, the contribution of procedure complexity toward EHR-derived workload (ß = 0.033; 95% CI, 0.031-0.035) was approximately one-third its contribution toward billing-derived workload (ß = 0.106; 95% CI, 0.104-0.108; P < .001). Conclusions and Relevance: In this cross-sectional study of 8 hospitals, reimbursement for anesthesiology services overcompensated for procedural complexity and undercompensated for patient complexity. This method for measuring clinical workload could be used to improve reimbursement valuations for anesthesia and other specialties.

Mediating Effects of Neural Targets on Depression, Weight, and Anxiety Outcomes of an Integrated Collaborative Care Intervention: The ENGAGE-2 Mechanistic Pilot Randomized Clinical Trial.

Background: Integrated treatments for comorbid depression (often with anxiety) and obesity are lacking; mechanisms are poorly investigated. Methods: In a mechanistic pilot trial, adults with body mass index ≥30 and Patient Health Questionnaire-9 scores ≥10 were randomized to usual care (n = 35) or an integrated behavioral intervention (n = 71). Changes at 6 months in body mass index and Depression Symptom Checklist-20 scores were co-primary outcomes, and Generalized Anxiety Disorder Scale-7 score was a secondary outcome. Changes at 2 months in the activation and functional connectivity of regions of interest in the negative affect circuit were primary neural targets, and secondary targets were in the cognitive control, default mode, and positive affect circuits. Results: Participants were 47.0 years (SD = 11.9 years), 76% women, 55% Black, and 20% Latino. Depression Symptom Checklist-20 (between-group difference, -0.3 [95% CI: -0.6 to -0.1]) and Generalized Anxiety Disorder Scale-7 (-2.9 [-4.7 to -1.1]) scores, but not body mass index, decreased significantly at 6 months in the intervention versus usual care groups. Only Generalized Anxiety Disorder Scale-7 score changes at 6 months significantly correlated with neural target changes at 2 months in the negative affect (anterior insula, subgenual/pregenual anterior cingulate cortex, amygdala) and cognitive control circuits (dorsal lateral prefrontal cortex, dorsal anterior cingulate cortex). Effects were medium to large (0.41-1.18 SDs). Neural target changes at 2 months in the cognitive control circuit only differed by treatment group. Effects were medium (0.58-0.79 SDs). Conclusions: Compared with usual care, the study intervention led to significantly improved depression but not weight loss, and the results on neural targets were null for both outcomes. The significant intervention effect on anxiety might be mediated through changes in the cognitive control circuit, but this warrants replication.

Effects of a virtual voice-based coach delivering problem-solving treatment on emotional distress and brain function: a pilot RCT in depression and anxiety.

Consumer-based voice assistants have the ability to deliver evidence-based treatment, but their therapeutic potential is largely unknown. In a pilot trial of a virtual voice-based coach, Lumen, delivering problem-solving treatment, adults with mild-to-moderate depression and/or anxiety were randomized to the Lumen intervention (n = 42) or waitlist control (n = 21). The main outcomes included changes in neural measures of emotional reactivity and cognitive control, and Hospital Anxiety and Depression Scale [HADS] symptom scores over 16 weeks. Participants were 37.8 years (SD = 12.4), 68% women, 25% Black, 24% Latino, and 11% Asian. Activation of the right dlPFC (neural region of interest in cognitive control) decreased in the intervention group but increased in the control group, with an effect size meeting the prespecified threshold for a meaningful effect (Cohen's d = 0.3). Between-group differences in the change in activation of the left dlPFC and bilateral amygdala were observed, but were of smaller magnitude (d = 0.2). Change in right dlPFC activation was also meaningfully associated (r ≥ 0.4) with changes in self-reported problem-solving ability and avoidance in the intervention. Lumen intervention also led to decreased HADS depression, anxiety, and overall psychological distress scores, with medium effect sizes (Cohen's d = 0.49, 0.51, and 0.55, respectively), compared with the waitlist control group. This pilot trial showed promising effects of a novel digital mental health intervention on cognitive control using neuroimaging and depression and anxiety symptoms, providing foundational evidence for a future confirmatory study.

Associations between daily step count trajectories and clinical outcomes among adults with comorbid obesity and depression.

Purpose: To examine the relationship between features of daily measured step count trajectories and clinical outcomes among people with comorbid obesity and depression in the ENGAGE-2 Trial. Methods: This post hoc analysis used data from the ENGAGE-2 trial where adults (n=106) with comorbid obesity (BMI ≥30.0 or 27.0 if Asian) and depressive symptoms (Patient Health Questionnaire-9 score ≥10) were randomized (2:1) to receive the experimental intervention or usual care. Daily step count trajectories over the first 60 days (Fitbit Alta HR) were characterized using functional principal component analyses. 7-day and 30-day trajectories were also explored. Functional principal component scores that described features of step count trajectories were entered into linear mixed models to predict weight (kg), depression (Symptom Checklist-20), and anxiety (Generalized Anxiety Disorder Questionnaire-7) at 2-months (2M) and 6-months (6M). Results: Features of 60-day step count trajectories were interpreted as overall sustained high, continuous decline, and disrupted decline. Overall sustained high step count was associated with low anxiety (2M, ß=-0.78, p<.05; 6M, ß=-0.80, p<.05) and low depressive symptoms (6M, ß=-0.15, p<.05). Continuous decline in step count was associated with high weight (2M, ß=0.58, p<.05). Disrupted decline was not associated with clinical outcomes at 2M or 6M. Features of 30-day step count trajectories were also associated with weight (2M, 6M), depression (6M), and anxiety (2M, 6M); Features of 7-day step count trajectories were not associated with weight, depression, or anxiety at 2M or 6M. Conclusions: Features of step count trajectories identified using functional principal component analysis were associated with depression, anxiety, and weight outcomes among adults with comorbid obesity and depression. Functional principal component analysis may be a useful analytic method that leverages daily measured physical activity levels to allow for precise tailoring of future behavioral interventions.

Measuring the cognitive effort associated with task switching in routine EHR-based tasks.

OBJECTIVE: Clinical work involves performing overlapping, time-sensitive tasks that frequently require clinicians to switch their attention between multiple tasks. We developed a methodological approach using EHR-based audit logs to determine switch costs-the cognitive burden associated with task switching-and assessed its magnitude during routine EHR-based clinical tasks. METHOD: Physician trainees (N = 75) participated in a longitudinal study where they provided access to their EHR-based audit logs. Physicians' audit log actions were used to create a taxonomy of EHR tasks. These tasks were transformed into task sequences and the time spent on each task in a sequence was computed. Within these task sequences, instances of task switching (i.e., switching from one task to the next) and non-switching were identified. The primary outcome of interest was the time spent on a post-switch task. Using a mixed-effects regression model, we compared the durations of post-switch and non-switch tasks. RESULTS: 2,781,679 audit log events over 117,822 sessions from 75 physicians were analyzed. Physicians spent most time on chart review (Median (IQR) = 5,439 (2,492-8,336) seconds), note review (1,936 (827-3,321) seconds), and navigating the EHR interface (1,048 (365.5-2,006) seconds) daily. Post task switch activity times were greater for documentation (Median increase = 5 s), order entry (Median increase = 3 s) and results review (Median increase = 3 s). Mixed-effects regression showed that time spent on tasks were longer following a task switch (ß = 0.03; 95% CIlower = 0.027, CIupper = 0.034), with greater post-swtich task times for imaging, order entry, note review, handoff, note entry, chart review and best practice advisory tasks. DISCUSSION: Increased task switching time-an indicator of the cognitive burden associated with switching between tasks-is prevalent in routine EHR-based tasks. We discuss the cumulative impact of incremental switch costs have on overall EHR workload, wellness, and error rates. Relying on theoretical cognitive foundations, we suggest pragmatic design considerations for mitigating the effects of cognitive burden associated with task switching.

Classifying clinical work settings using EHR audit logs: a machine learning approach.

OBJECTIVES: We used electronic health record (EHR)-based raw audit logs to classify the work settings of anesthesiology physicians providing care in both surgical intensive care units (ICUs) and operating rooms. STUDY DESIGN: Observational study. METHODS: Attending anesthesiologists who worked at least 1 shift in 1 of 4 surgical ICUs in calendar year 2019 were included. Time-stamped EHR-based audit log events for each week were used to create event frequencies and represented as a term frequency-inverse document frequency matrix. Primary classification outcome of interest was a physician's clinical work setting. Performance of multiple supervised machine learning classifiers were evaluated. RESULTS: A total of 24 attending physicians were included; physicians performed a median (IQR) of 2545 (906-5071) EHR-based actions per week and worked a median (IQR) of 5 (3-7) weeks in a surgical ICU. A random forest classifier yielded the best discriminative performance (mean [SD] area under receiver operating characteristic curve, 0.88 [0.05]; mean [SD] area under precision-recall curve, 0.72 [0.13]). Model explanations illustrated that clinical activities related to signing of clinical notes, printing handoff data, and updating diagnosis information were associated with the positive prediction of working in a surgical ICU setting. CONCLUSIONS: A random forest classifier using a frequency-based feature engineering approach successfully predicted work settings of physicians with multiple clinical responsibilities with high accuracy. These findings highlight opportunities for using audit logs for automated assessment of clinician activities and their work settings, thereby affording the ability to accurately assess context-specific work characteristics (eg, workload).

Characterizing the macrostructure of electronic health record work using raw audit logs: an unsupervised action embeddings approach.

Raw audit logs provide a comprehensive record of clinicians' activities on an electronic health record (EHR) and have considerable potential for studying clinician behaviors. However, research using raw audit logs is limited because they lack context for clinical tasks, leading to difficulties in interpretation. We describe a novel unsupervised approach using the comparison and visualization of EHR action embeddings to learn context and structure from raw audit log activities. Using a dataset of 15 767 634 raw audit log actions performed by 88 intern physicians over 6 months of EHR use across inpatient and outpatient settings, we demonstrated that embeddings can be used to learn the situated context for EHR-based work activities, identify discrete clinical workflows, and discern activities typically performed across diverse contexts. Our approach represents an important methodological advance in raw audit log research, facilitating the future development of metrics and predictive models to measure clinician behaviors at the macroscale.

Multi-horizon predictive models for guiding extracorporeal resource allocation in critically ill COVID-19 patients.

OBJECTIVE: Extracorporeal membrane oxygenation (ECMO) resource allocation tools are currently lacking. We developed machine learning (ML) models for predicting COVID-19 patients at risk of receiving ECMO to guide patient triage and resource allocation. MATERIAL AND METHODS: We included COVID-19 patients admitted to intensive care units for >24 h from March 2020 to October 2021, divided into training and testing development and testing-only holdout cohorts. We developed ECMO deployment timely prediction model ForecastECMO using Gradient Boosting Tree (GBT), with pre-ECMO prediction horizons from 0 to 48 h, compared to PaO2/FiO2 ratio, Sequential Organ Failure Assessment score, PREdiction of Survival on ECMO Therapy score, logistic regression, and 30 pre-selected clinical variables GBT Clinical GBT models, with area under the receiver operator curve (AUROC) and precision recall curve (AUPRC) metrics. RESULTS: ECMO prevalence was 2.89% and 1.73% in development and holdout cohorts. ForecastECMO had the best performance in both cohorts. At the 18-h prediction horizon, a potentially clinically actionable pre-ECMO window, ForecastECMO, had the highest AUROC (0.94 and 0.95) and AUPRC (0.54 and 0.37) in development and holdout cohorts in identifying ECMO patients without data 18 h prior to ECMO. DISCUSSION AND CONCLUSIONS: We developed a multi-horizon model, ForecastECMO, with high performance in identifying patients receiving ECMO at various prediction horizons. This model has potential to be used as early alert tool to guide ECMO resource allocation for COVID-19 patients. Future prospective multicenter validation would provide evidence for generalizability and real-world application of such models to improve patient outcomes.