Automated, machine learning-based alerts increase epilepsy surgery referrals: A randomized controlled trial.

OBJECTIVE: To determine whether automated, electronic alerts increased referrals for epilepsy surgery. METHODS: We conducted a prospective, randomized controlled trial of a natural language processing-based clinical decision support system embedded in the electronic health record (EHR) at 14 pediatric neurology outpatient clinic sites. Children with epilepsy and at least two prior neurology visits were screened by the system prior to their scheduled visit. Patients classified as a potential surgical candidate were randomized 2:1 for their provider to receive an alert or standard of care (no alert). The primary outcome was referral for a neurosurgical evaluation. The likelihood of referral was estimated using a Cox proportional hazards regression model. RESULTS: Between April 2017 and April 2019, at total of 4858 children were screened by the system, and 284 (5.8%) were identified as potential surgical candidates. Two hundred four patients received an alert, and 96 patients received standard care. Median follow-up time was 24 months (range: 12-36 months). Compared to the control group, patients whose provider received an alert were more likely to be referred for a presurgical evaluation (3.1% vs 9.8%; adjusted hazard ratio [HR] = 3.21, 95% confidence interval [CI]: 0.95-10.8; one-sided p = .03). Nine patients (4.4%) in the alert group underwent epilepsy surgery, compared to none (0%) in the control group (one-sided p = .03). SIGNIFICANCE: Machine learning-based automated alerts may improve the utilization of referrals for epilepsy surgery evaluations.

The Risk of Coding Racism into Pediatric Sepsis Care: The Necessity of Antiracism in Machine Learning.

Machine learning holds the possibility of improving racial health inequalities by compensating for human bias and structural racism. However, unanticipated racial biases may enter during model design, training, or implementation and perpetuate or worsen racial inequalities if ignored. Pre-existing racial health inequalities could be codified into medical care by machine learning without clinicians being aware. To illustrate the importance of a commitment to antiracism at all stages of machine learning, we examine machine learning in predicting severe sepsis in Black children, focusing on the impacts of structural racism that may be perpetuated by machine learning and difficult to discover. To move toward antiracist machine learning, we recommend partnering with ethicists and experts in model development, enrolling representative samples for training, including socioeconomic inputs with proximate causal associations to racial inequalities, reporting outcomes by race, and committing to equitable models that narrow inequality gaps or at least have equal benefit.

Aligning Provider Prescribing With Guidelines for Soft Tissue Infections.

OBJECTIVE: Despite evidence-based guidelines, antibiotics prescribed for uncomplicated skin and soft tissue infections can involve inappropriate microbial coverage. Our aim was to evaluate the appropriateness of antibiotic prescribing practices for mild nonpurulent cellulitis in a pediatric tertiary academic medical center over a 1-year period. METHODS: Eligible patients treated in the emergency department or urgent care settings for mild nonpurulent cellulitis from January 2017 to December 2017 were identified by an International Classification of Diseases, Tenth Revision, code for cellulitis. The primary outcome was appropriateness of prescribed antibiotics as delineated by adherence with the Infectious Diseases Society of America guidelines. Secondary outcomes include reutilization rate as defined by revisit to the emergency department/urgent cares within 14 days of the initial encounter. RESULTS: A total of 967 encounters were evaluated with 60.0% overall having guideline-adherent care. Common reasons for nonadherence included inappropriate coverage of MRSA with clindamycin (n = 217, 56.1%) and single-agent coverage with sulfamethoxazole-trimethoprim (n = 129, 33.3%). There were 29 revisits within 14 days of initial patient encounters or a reutilization rate of 3.0%, which was not significantly associated with the Infectious Diseases Society of America adherence. CONCLUSIONS: Our data show antibiotic prescription for nonpurulent cellulitis as a potential area of standardization and optimization of care at our center.

Early identification of epilepsy surgery candidates: A multicenter, machine learning study.

OBJECTIVES: Epilepsy surgery is underutilized. Automating the identification of potential surgical candidates may facilitate earlier intervention. Our objective was to develop site-specific machine learning (ML) algorithms to identify candidates before they undergo surgery. MATERIALS & METHODS: In this multicenter, retrospective, longitudinal cohort study, ML algorithms were trained on n-grams extracted from free-text neurology notes, EEG and MRI reports, visit codes, medications, procedures, laboratories, and demographic information. Site-specific algorithms were developed at two epilepsy centers: one pediatric and one adult. Cases were defined as patients who underwent resective epilepsy surgery, and controls were patients with epilepsy with no history of surgery. The output of the ML algorithms was the estimated likelihood of candidacy for resective epilepsy surgery. Model performance was assessed using 10-fold cross-validation. RESULTS: There were 5880 children (n = 137 had surgery [2.3%]) and 7604 adults with epilepsy (n = 56 had surgery [0.7%]) included in the study. Pediatric surgical patients could be identified 2.0 years (range: 0-8.6 years) before beginning their presurgical evaluation with AUC =0.76 (95% CI: 0.70-0.82) and PR-AUC =0.13 (95% CI: 0.07-0.18). Adult surgical patients could be identified 1.0 year (range: 0-5.4 years) before beginning their presurgical evaluation with AUC =0.85 (95% CI: 0.78-0.93) and PR-AUC =0.31 (95% CI: 0.14-0.48). By the time patients began their presurgical evaluation, the ML algorithms identified pediatric and adult surgical patients with AUC =0.93 and 0.95, respectively. The mean squared error of the predicted probability of surgical candidacy (Brier scores) was 0.018 in pediatrics and 0.006 in adults. CONCLUSIONS: Site-specific machine learning algorithms can identify candidates for epilepsy surgery early in the disease course in diverse practice settings.

Prospective validation of a machine learning model that uses provider notes to identify candidates for resective epilepsy surgery.

OBJECTIVE: Delay to resective epilepsy surgery results in avoidable disease burden and increased risk of mortality. The objective was to prospectively validate a natural language processing (NLP) application that uses provider notes to assign epilepsy surgery candidacy scores. METHODS: The application was trained on notes from (1) patients with a diagnosis of epilepsy and a history of resective epilepsy surgery and (2) patients who were seizure-free without surgery. The testing set included all patients with unknown surgical candidacy status and an upcoming neurology visit. Training and testing sets were updated weekly for 1 year. One- to three-word phrases contained in patients' notes were used as features. Patients prospectively identified by the application as candidates for surgery were manually reviewed by two epileptologists. Performance metrics were defined by comparing NLP-derived surgical candidacy scores with surgical candidacy status from expert chart review. RESULTS: The training set was updated weekly and included notes from a mean of 519 ± 67 patients. The area under the receiver operating characteristic curve (AUC) from 10-fold cross-validation was 0.90 ± 0.04 (range = 0.83-0.96) and improved by 0.002 per week (P < .001) as new patients were added to the training set. Of the 6395 patients who visited the neurology clinic, 4211 (67%) were evaluated by the model. The prospective AUC on this test set was 0.79 (95% confidence interval [CI] = 0.62-0.96). Using the optimal surgical candidacy score threshold, sensitivity was 0.80 (95% CI = 0.29-0.99), specificity was 0.77 (95% CI = 0.64-0.88), positive predictive value was 0.25 (95% CI = 0.07-0.52), and negative predictive value was 0.98 (95% CI = 0.87-1.00). The number needed to screen was 5.6. SIGNIFICANCE: An electronic health record-integrated NLP application can accurately assign surgical candidacy scores to patients in a clinical setting.

Feedback at the Point of Care to Decrease Medication Alert Rates in an Electronic Health Record.

Frequently overridden alerts in the electronic health record can highlight alerts that may need revision. This method is a way of fine-tuning clinical decision support. We evaluated the feasibility of a complementary, yet different method that directly involved pediatric emergency department (PED) providers in identifying additional medication alerts that were potentially incorrect or intrusive. We then evaluated the effect subsequent resulting modifications had on alert salience. METHODS: We performed a prospective, interventional study over 34 months (March 6, 2014, to December 31, 2016) in the PED. We implemented a passive alert feedback mechanism by enhancing the native electronic health record functionality on alert reviews. End-users flagged potentially incorrect/bothersome alerts for review by the study's team. The alerts were updated when clinically appropriate and trends of the impact were evaluated. RESULTS: More than 200 alerts were reported from both inside and outside the PED, suggesting an intuitive approach. On average, we processed 4 reviews per week from the PED, with attending physicians as major contributors. The general trend of the impact of these changes seems favorable. DISCUSSION: The implementation of the review mechanism for user-selected alerts was intuitive and sustainable and seems to be able to detect alerts that are bothersome to the end-users. The method should be run in parallel with the traditional data-driven approach to support capturing of inaccurate alerts. CONCLUSIONS: User-centered, context-specific alert feedback can be used for selecting suboptimal, interruptive medication alerts.

Utilization of a Clinical Decision Support Tool to Reduce Child Tobacco Smoke Exposure in the Urgent Care Setting.

BACKGROUND: Clinical decision support systems (CDSS) may facilitate caregiver tobacco screening and counseling by pediatric urgent care (UC) nurses. OBJECTIVE: This study aimed to assess the feasibility of a CDSS to address caregivers' tobacco use and child tobacco smoke exposure (TSE). METHODS: We conducted a 3-month prospective study on caregivers screened using a CDSS. Nurses used the CDSS to advise, assess, and assist caregivers to quit. We assessed caregiver sociodemographics, smoking habits, and child TSE. RESULTS: We screened 185 caregivers whose children were exposed to TSE for study inclusion; 155 (84%) met the eligibility criteria, and 149 (80.5%) were included in the study. Study nurses advised 35.2% of the caregivers to quit, assessed 35.9% for readiness to quit, and assisted 32.4%. Of the 149 participants, 83.1% were female; 47.0% were white and 45.6% African American; 84.6% had public insurance or were self-pay; 71.1% were highly nicotine dependent; 50.0% and 50.7% allowed smoking in the home and car, respectively; and 81.3% of children were biochemically confirmed to be exposed to tobacco smoke. At follow-up (86.6% retention), 58.9% reported quit attempts at 3 months. There was a significant decrease in nicotine dependence and a significant increase in motivation to quit. Self-reported quit rate was 7.8% at 3 months. CONCLUSIONS: An electronic health record-embedded CDSS was feasible to incorporate into busy UC nurses' workloads and was associated with encouraging changes in the smoking behavior of caregivers. More research on the use of CDSS to screen and counsel caregivers who smoke in the UC and other acute care settings is warranted.

Investigation of bias in an epilepsy machine learning algorithm trained on physician notes.

Racial disparities in the utilization of epilepsy surgery are well documented, but it is unknown whether a natural language processing (NLP) algorithm trained on physician notes would produce biased recommendations for epilepsy presurgical evaluations. To assess this, an NLP algorithm was trained to identify potential surgical candidates using 1097 notes from 175 epilepsy patients with a history of resective epilepsy surgery and 268 patients who achieved seizure freedom without surgery (total N = 443 patients). The model was tested on 8340 notes from 3776 patients with epilepsy whose surgical candidacy status was unknown (2029 male, 1747 female, median age = 9 years; age range = 0-60 years). Multiple linear regression using demographic variables as covariates was used to test for correlations between patient race and surgical candidacy scores. After accounting for other demographic and socioeconomic variables, patient race, gender, and primary language did not influence surgical candidacy scores (P > .35 for all). Higher scores were given to patients >18 years old who traveled farther to receive care, and those who had a higher family income and public insurance (P < .001, .001, .001, and .01, respectively). Demographic effects on surgical candidacy scores appeared to reflect patterns in patient referrals.

A Time-and-Motion Study of Clinical Trial Eligibility Screening in a Pediatric Emergency Department.

OBJECTIVE: Challenges with efficient patient recruitment including sociotechnical barriers for clinical trials are major barriers to the timely and efficacious conduct of translational studies. We conducted a time-and-motion study to investigate the workflow of clinical trial enrollment in a pediatric emergency department. METHODS: We observed clinical research coordinators during 3 clinically staffed shifts. One clinical research coordinator was shadowed at a time. Tasks were marked in 30-second intervals and annotated to include patient screening, patient contact, performing procedures, and physician contact. Statistical analysis was conducted on the patient enrollment activities. RESULTS: We conducted fifteen 120-minute observations from December 12, 2013, to January 3, 2014 and shadowed 8 clinical research coordinators. Patient screening took 31.62% of their time, patient contact took 18.67%, performing procedures took 17.6%, physician contact was 1%, and other activities took 31.0%. CONCLUSIONS: Screening patients for eligibility constituted the most time. Automated screening methods could help reduce this time. The findings suggest improvement areas in recruitment planning to increase the efficiency of clinical trial enrollment.

The Reliability of Computerized Physician Order Entry Data for Research Studies.

OBJECTIVES: In the United States, adolescents account for nearly half of the newly diagnosed sexually transmitted infections annually, and many of these infections are asymptomatic. Adolescents often seek care in pediatric emergency departments; thus, the emergency department is an important setting to implement adolescent sexually transmitted infection screening. Before implementation, baseline data reflecting current screening rates of symptomatic and asymptomatic patients were needed. This study aimed to evaluate the accuracy of provider-reported rates of symptomatic and asymptomatic chlamydia (CT) and gonorrhea (GC) testing in adolescents overall and pre-electronic health record (EHR) and post-EHR order modification in preparation for a research intervention. METHODS: This was a 1-year prospective, observational study. Provider reason for CT/GC testing was added to the existing EHR order. Chart reviews were performed to ensure the accuracy of clinician CT/GC testing choices (symptomatic vs asymptomatic). Frequencies of testing choices were obtained. Order modifications were made to further clarify the definitions. A Student t test was used to compare data preorder and postorder modification. RESULTS: When relying on providers to report reasons for CT/GC testing (symptomatic vs asymptomatic), many patients were misclassified based on a priori defined testing reasons. After order modification, rates of provider-reported symptomatic testing remained unchanged (P = 0.16). Provider-reported asymptomatic testing significantly declined (P = 0.004); however, 23.2% of those tested continued to be misclassified. CONCLUSIONS: Provider-entered EHR data are increasingly being used in research studies; thus, it is important to ensure its accuracy and reliability before study implementation.

Evaluation of Early Childhood Home Visiting to Prevent Medically Attended Unintentional Injury.

STUDY OBJECTIVE: We evaluated the influence of home visiting on the risk for medically attended unintentional injury during home visiting (0 to 3 years) and subsequent to home visiting (3 to 5 years). METHODS: A retrospective, quasi-experimental study was conducted in a cohort of mother-child pairs in Hamilton County, OH. The birth cohort (2006 to 2012) was linked to administrative home visiting records and data from a population-based injury surveillance system containing records of emergency department (ED) visits and hospitalizations. Cox proportional-hazard regression was used to compare medically attended unintentional injury risk (0 to 2, 0 to 3, and 3 to 5 years) in a home-visited group versus a propensity score-matched comparison group. The study population was composed of 2,729 mother-child pairs who received home visiting and 2,729 matched mother-child pairs in a comparison group. RESULTS: From birth to 2 years, 17.2% of the study population had at least one medically attended unintentional injury. The risk for medically attended unintentional injury from aged 0 to 2 and 0 to 3 years was significantly higher in the home-visited group relative to the comparison group (hazard ratio 1.17, 95% confidence interval 1.01 to 1.35; hazard ratio 1.15, 95% confidence interval 1.00 to 1.31, respectively). Additional injuries in the home-visited group were superficial, and the increased risk for medically attended unintentional injury was observed for ED visits and not hospitalizations. CONCLUSION: Home-visited children were more likely to have a medically attended unintentional injury from birth to aged 3 years. This finding may be partially attributed to home visitor surveillance of injuries or greater health care-seeking behavior. Implications and alternative explanations are discussed.

Adolescents with Mild Traumatic Brain Injury Get SMART: An Analysis of a Novel Web-Based Intervention.

BACKGROUND: There have been few trials of interventions to facilitate recovery following mild traumatic brain injury (mTBI) in adolescence. To address this gap, we developed and piloted a novel Web-based intervention, entitled Self-Management Activity Restriction and Relaxation Training (SMART), and examined its impact on symptom burden, functional disability, and executive functioning during the month following mTBI in adolescents. MATERIALS AND METHODS: Open-label, single arm study. Adolescents with recent mTBI and a parent were recruited from the emergency department and provided access upon discharge to SMART-a Web-based program designed to facilitate recovery via self-management and education about symptoms and sequelae associated with mTBI. Symptom burden, functional disability, and executive functioning were rated by both the adolescent and the parent initially and at assessments at 1-, 2- and 4-weeks postinjury. Mixed models analyses were used to examine trajectories on these outcomes. RESULTS: Of the 21 adolescent/parent dyads enrolled, 13 engaged in the program and reported significant improvement in symptoms over the 4-week program (adolescent, p = 0.0005; parent, p = 0.004). Adolescents spent a median of 35.5 min (range 1.1-107.6) using the program. Parent ratings of the adolescent's functional disability and executive functioning significantly improved over the 4-week period from baseline (p = 0.009 and p = 0.03, respectively), whereas adolescents themselves did not report significant changes in either outcome. All participants improved and there were no adverse outcomes. CONCLUSION: The SMART program, a novel Web-based intervention, may serve as a self-management tool for adolescents and their parents to assist with the recovery following a recent mTBI.

Feasibility and Potential Benefits of a Web-Based Intervention Delivered Acutely After Mild Traumatic Brain Injury in Adolescents: A Pilot Study.

BACKGROUND: There is a paucity of evidence-based interventions for mild traumatic brain injury (mTBI). OBJECTIVE: To evaluate the feasibility and potential benefits of an interactive, Web-based intervention for mTBI. SETTING: Emergency department and outpatient settings. PARTICIPANTS: Of the 21 adolescents aged 11 to 18 years with mTBI recruited from November 2013 to June 2014 within 96 hours of injury, 13 completed the program. DESIGN: Prospective, open pilot. INTERVENTION: The Web-based Self-Management Activity-restriction and Relaxation Training (SMART) program incorporates anticipatory guidance and psychoeducation, self-management and pacing of cognitive and physical activities, and cognitive-behavioral principles for early management of mTBI in adolescents. MAIN MEASURES: Primary: Daily Post-Concussion Symptom Scale (PCSS). Secondary: Daily self-reported ratings of activities and satisfaction survey. RESULTS: Average time from injury to baseline testing was 14.0 (standard deviation = 16.7) hours. Baseline PCSS was 23.6 (range: 0-46), and daily activity was 1.8 (range: 0-5.75) hours. Repeated-measures, generalized linear mixed-effects model analysis demonstrated a significant decrease of PCSS at a rate of 2.0 points per day that stabilized after about 2 weeks. Daily activities, screen time, and physical activity increased by 0.06 (standard error [SE] = 0.04, P = .09), 0.04 (SE = 0.02, P = .15), and 0.03 (SE = 0.02, P = .05) hours per day, respectively, over the 4-week follow-up. Satisfaction was rated highly by parents and youth. CONCLUSIONS: Self-Management Activity-restriction and Relaxation Training is feasible and reported to be helpful and enjoyable by participants. Future research will need to determine the comparative benefits of SMART and ideal target population.

Development of a Tobacco Cessation Clinical Decision Support System for Pediatric Emergency Nurses.

Almost 50% of children who visit the pediatric emergency department are exposed to tobacco smoke. However, pediatric emergency nurses do not routinely address this issue. The incorporation of a clinical decision support system into the electronic health record may improve the rates of tobacco exposure screening and interventions. We used a mixed-methods design to develop, refine, and implement an evidence-based clinical decision support system to help nurses screen, educate, and assist caregivers to quit smoking. We included an advisory panel of emergency department experts and leaders and focus and user groups of nurses. The prompts include the following: (1) "Ask" about child smoke exposure and caregiver smoking; (2) "Advise" caregivers to reduce their child's smoke exposure by quitting smoking; (3) "Assess" interest; and (4) "Assist" caregivers to quit. The clinical decision support system was created to reflect nurses' suggestions and was implemented in five busy urgent care settings with 38 nurses. The nurses reported that the system was easy to use and helped them to address caregiver smoking. The use of this innovative tool may create a sustainable and disseminable model for prompting nurses to provide evidence-based tobacco cessation treatment.

Preparing for International Classification of Diseases, 10th Revision, Clinical Modification implementation: strategies for maintaining an efficient workflow.

The International Classification of Diseases, 10th Revision, is required to be used by the Centers for Medicare and Medicaid Services health care billing data starting in October 2015 in the United States. The International Classification of Diseases, 10th Revision, is an update to the International Classification of Diseases, Ninth Revision, and contains approximately 70,000 codes compared with 14,000 codes. We aimed to discuss how our institution is updating the coding system in a manner that alleviates the possible burden placed on providers including more coding information required and longer load times. We performed a simulation test including testing the diagnosis calculator, timing, and how well the new and old codes mapped. We conducted a gap analysis to ensure that coding could begin in October of 2015 with minimal service interruptions. We will describe strategies and procedures to transition between systems while maintaining efficiency and helping to improve classification.

A systematic review of the implementation and impact of asthma protocols.

BACKGROUND: Asthma is one of the most common childhood illnesses. Guideline-driven clinical care positively affects patient outcomes for care. There are several asthma guidelines and reminder methods for implementation to help integrate them into clinical workflow. Our goal is to determine the most prevalent method of guideline implementation; establish which methods significantly improved clinical care; and identify the factors most commonly associated with a successful and sustainable implementation. METHODS: PUBMED (MEDLINE), OVID CINAHL, ISI Web of Science, and EMBASE. STUDY SELECTION: Studies were included if they evaluated an asthma protocol or prompt, evaluated an intervention, a clinical trial of a protocol implementation, and qualitative studies as part of a protocol intervention. Studies were excluded if they had non-human subjects, were studies on efficacy and effectiveness of drugs, did not include an evaluation component, studied an educational intervention only, or were a case report, survey, editorial, letter to the editor. RESULTS: From 14,478 abstracts, we included 101 full-text articles in the analysis. The most frequent study design was pre-post, followed by prospective, population based case series or consecutive case series, and randomized trials. Paper-based reminders were the most frequent with fully computerized, then computer generated, and other modalities. No study reported a decrease in health care practitioner performance or declining patient outcomes. The most common primary outcome measure was compliance with provided or prescribing guidelines, key clinical indicators such as patient outcomes or quality of life, and length of stay. CONCLUSIONS: Paper-based implementations are by far the most popular approach to implement a guideline or protocol. The number of publications on asthma protocol reminder systems is increasing. The number of computerized and computer-generated studies is also increasing. Asthma guidelines generally improved patient care and practitioner performance regardless of the implementation method.

Machine Learning-Based Prediction of Escitalopram and Sertraline Side Effects With Pharmacokinetic Data in Children and Adolescents.

Selective serotonin reuptake inhibitors (SSRI) are the first-line pharmacologic treatment for anxiety and depressive disorders in children and adolescents. Many patients experience side effects that are difficult to predict, are associated with significant morbidity, and can lead to treatment discontinuation. Variation in SSRI pharmacokinetics could explain differences in treatment outcomes, but this is often overlooked as a contributing factor to SSRI tolerability. This study evaluated data from 288 escitalopram-treated and 255 sertraline-treated patients ≤ 18 years old to develop machine learning models to predict side effects using electronic health record data and Bayesian estimated pharmacokinetic parameters. Trained on a combined cohort of escitalopram- and sertraline-treated patients, a penalized logistic regression model achieved an area under the receiver operating characteristic curve (AUROC) of 0.77 (95% confidence interval (CI): 0.66-0.88), with 0.69 sensitivity (95% CI: 0.54-0.86), and 0.82 specificity (95% CI: 0.72-0.87). Medication exposure, clearance, and time since the last dose increase were among the top features. Individual escitalopram and sertraline models yielded an AUROC of 0.73 (95% CI: 0.65-0.81) and 0.64 (95% CI: 0.55-0.73), respectively. Post hoc analysis showed sertraline-treated patients with activation side effects had slower clearance (P = 0.01), which attenuated after accounting for age (P = 0.055). These findings raise the possibility that a machine learning approach leveraging pharmacokinetic data can predict escitalopram- and sertraline-related side effects. Clinicians may consider differences in medication pharmacokinetics, especially during dose titration and as opposed to relying on dose, when managing side effects. With further validation, application of this model to predict side effects may enhance SSRI precision dosing strategies in youth.

Early Identification of Candidates for Epilepsy Surgery: A Multicenter, Machine Learning, Prospective Validation Study.

BACKGROUND AND OBJECTIVES: Epilepsy surgery is often delayed. We previously developed machine learning (ML) models to identify candidates for resective epilepsy surgery earlier in the disease course. In this study, we report the prospective validation. METHODS: In this multicenter, prospective, longitudinal cohort study, random forest models were validated at a pediatric epilepsy center consisting of 2 hospitals and 14 outpatient neurology clinic sites and an adult epilepsy center with 2 hospitals and 27 outpatient neurology clinic sites. The models used neurology visit notes, EEG and MRI reports, visit patterns, hospitalizations, and medication, laboratory, and procedure orders to identify candidates for surgery. The models were trained on historical data up to May 10, 2019. Patients with an ICD-10 diagnosis of epilepsy who visited from May 11, 2019, to May 10, 2020, were screened by the algorithm and assigned surgical candidacy scores. The primary outcome was area under the curve (AUC), which was calculated by comparing scores from patients who underwent epilepsy surgery before November 10, 2020, against scores from nonsurgical patients. Nonsurgical patients' charts were reviewed to determine whether patients with high scores were more likely to be missed surgical candidates. Delay to surgery was defined as the time between the first visit that a surgical candidate was identified by the algorithm and the date of the surgery. RESULTS: A total of 5,285 pediatric and 5,782 adult patients were included to train the ML algorithms. During the study period, 41 children and 23 adults underwent resective epilepsy surgery. In the pediatric cohort, AUC was 0.91 (95% CI 0.87-0.94), positive predictive value (PPV) was 0.08 (0.05-0.10), and negative predictive value (NPV) was 1.00 (0.99-1.00). In the adult cohort, AUC was 0.91 (0.86-0.97), PPV was 0.07 (0.04-0.11), and NPV was 1.00 (0.99-1.00). The models first identified patients at a median of 2.1 years (interquartile range [IQR]: 1.2-4.9 years, maximum: 11.1 years) before their surgery and 1.3 years (IQR: 0.3-4.0 years, maximum: 10.1 years) before their presurgical evaluations. DISCUSSION: ML algorithms can identify surgical candidates earlier in the disease course. Even at specialized epilepsy centers, there is room to shorten the time to surgery. CLASSIFICATION OF EVIDENCE: This study provides Class II evidence that a machine learning algorithm can accurately distinguish patients with epilepsy who require resective surgery from those who do not.

Modifications and integration of the electronic tracking board in a pediatric emergency department.

Electronic health records (EHRs) are used for data storage; provider, laboratory, and patient communication; clinical decision support; procedure and medication orders; and decision support alerts. Clinical decision support is part of any EHR and is designed to help providers make better decisions. The emergency department (ED) poses a unique environment to the use of EHRs and clinical decision support. Used effectively, computerized tracking boards can help improve flow, communication, and the dissemination of pertinent visit information between providers and other departments in a busy ED. We discuss the unique modifications and decisions made in the implementation of an EHR and computerized tracking board in a pediatric ED. We discuss the changing views based on provider roles, customization to the user interface including the layout and colors, decision support, tracking board best practices collected from other institutions and colleagues, and a case study of using reminders on the electronic tracking board to drive pain reassessments.