An explainable long short-term memory network for surgical site infection identification.

BACKGROUND: Currently, surgical site infection surveillance relies on labor-intensive manual chart review. Recently suggested solutions involve machine learning to identify surgical site infections directly from the medical record. Deep learning is a form of machine learning that has historically performed better than traditional methods while being harder to interpret. We propose a deep learning model, a long short-term memory network, for the identification of surgical site infection from the medical record with an attention layer for explainability. METHODS: We retrieved structured data and clinical notes from the University of Utah Health System's electronic health care record for operative events randomly selected for manual chart review from January 2016 to June 2021. Surgical site infection occurring within 30 days of surgery was determined according to the National Surgical Quality Improvement Program definition. We trained the long short-term memory model along with traditional machine learning models for comparison. We calculated several performance metrics from a holdout test set and performed additional analyses to understand the performance of the long short-term memory, including an explainability analysis. RESULTS: Surgical site infection was present in 4.7% of the total 9,185 operative events. The area under the receiver operating characteristic curve and sensitivity of the long short-term memory was higher (area under the receiver operating characteristic curve: 0.954, sensitivity: 0.920) compared to the top traditional model (area under the receiver operating characteristic curve: 0.937, sensitivity: 0.736). The top 5 features of the long short-term memory included 2 procedure codes and 3 laboratory values. CONCLUSION: Surgical site infection surveillance is vital for the reduction of surgical site infection rates. Our explainable long short-term memory achieved a comparable area under the receiver operating characteristic curve and greater sensitivity when compared to traditional machine learning methods. With explainable deep learning, automated surgical site infection surveillance could replace burdensome manual chart review processes.

Cost and Late Hospital Care of Publicly Insured Children After Appendectomy.

INTRODUCTION: Immediate complications of appendicitis are common, but the prevalence of long-term complications is uncertain. METHODS: We studied all publicly-insured children in the US with uncomplicated or complicated appendicitis in 2018-2019 using administrative claims. The main outcome was late hospital care defined as hospitalization or abdominal procedure within 180 d of an appendicitis discharge, excluding interval appendectomies. Time to late hospital care was evaluated using Cox regression. We evaluated health-care expenditures arising from appendicitis episodes. RESULTS: Among 95,942 children with appendicitis, 5727 (6.0%) had late hospital care, with 5062 requiring rehospitalization and 2012 (2.1%) surgery. The median time to late hospital care was 10 d (interquartile range 4-33). Age under 5 y (compared with >14 y, hazard ratio [HR] 1.88, 95% confidence interval [CI] 1.70-2.08), complex chronic conditions (HR 2.35, 95% CI 2.13-2.59), and complicated appendicitis (HR 2.81, 95% CI 2.67, 2.96) were each associated with time to late hospital care. Expenditures over 180 d were a median $6553 and $19,589 respectively in those requiring no late hospital care versus those requiring it (P < 0.001). CONCLUSIONS: Late hospital care is uncommon in pediatric appendicitis but is costly. Prevention efforts should be targeted to the youngest, most complex children, and those with complicated appendicitis at presentation.

Emergency Department Utilization for Pediatric Gastrostomy Tubes Across the United States.

INTRODUCTION: Emergency Department (ED) visits for gastrostomy tube complications in children represent a substantial health-care burden, and many ED visits are potentially preventable. The number and nature of ED visits to community hospitals for pediatric gastrostomy tube complications is unknown. METHODS: Using the 2019 Nationwide Emergency Department Sample, we performed a retrospective cross-sectional analysis of pediatric patients (<18 y) with a primary diagnosis of gastrostomy tube complication. Our primary outcome was a potentially preventable ED visit, defined as an encounter that did not result in any imaging, procedures, or an inpatient admission. Univariate and multivariable logistic regression analyses were used to determine the associations between patient factors and our primary outcome. RESULTS: We observed 32,036 ED visits at 535 hospitals and 15,165 (47.3%) were potentially preventable. The median (interquartile range) age was 2 (1, 6) years and 17,707 (55%) were male. Compared to White patients, patients with higher odds of potentially preventable visits were Black (adjusted odds ratio (aOR) [95% confidence interval {CI}]: 1.07 [1.05-1.11], P < 0.001) and Hispanic (aOR [95% CI]: 1.05 [1.02-1.08], P = 0.004). Patients with residential zip codes in the first (aOR [95% CI]: 1.08 [1.04, 1.12], P < 0.001), second (aOR [95% CI]: 1.07 [1.03, 1.11], P < 0.001), and third (aOR [95% CI]: 1.09 [1.05, 1.13], P < 0.001) median household income quartiles had higher odds of potentially preventable visits compared to the highest. CONCLUSIONS: In a nationally representative sample of EDs, 47.3% of visits for pediatric gastrostomy tubes were potentially preventable. Efforts to improve outpatient management are warranted to reduce health-care utilization for these patients.

Disparities in Ultrasound Use for Diagnosing Pediatric Appendicitis Across United States Emergency Departments.

INTRODUCTION: An ultrasound (US)-first approach for evaluating appendicitis is recommended by the American College of Radiology. We sought to assess the access to and utilization of an US-first approach for children with acute appendicitis in United States Emergency Departments. METHODS: Utilizing the 2019 Nationwide Emergency Department Sample, we performed a retrospective cohort study of patients <18 y with a primary diagnosis of acute appendicitis based on International Classification of Disease 10th Edition Diagnosis codes. Our primary outcome was the presentation to a hospital that does not perform US for children with acute appendicitis. Our secondary outcome was the receipt of a US at US-capable hospital. We developed generalized linear models with inverse-probability weighting to determine the association between patient characteristics and outcomes. RESULTS: Of 49,703 total children, 24,102 (48%) received a US evaluation. The odds of presenting at a hospital with no US use were significantly higher for patients aged 11-17 compared to patients <6 y (adjusted odds ratio [aOR] [95% confidence interval (CI)]: 1.59, [1.19- 2.13], P = 0.002); lowest median household income quartile compared to highest (aOR [95% CI]: 2.50, [1.52-4.10], P < 0.001); rural locations compared to metropolitan (aOR [95% CI]: 8.36 [5.54-12.6], P < 0.001), and Hispanic compared to non-Hispanic White (aOR [95% CI]: 0.63 [0.45-0.90], P = 0.01). The odds of receiving a US at US-capable hospitals were significantly lower for patients >6 y, lowest median household income quartiles, and rural locations (P < 0.05). CONCLUSIONS: Rural, older, and poorer children are more likely to present to hospitals that do not utilize US in the diagnosis of acute appendicitis and are less likely to undergo US at US-capable hospitals.

Developing an LSTM Model to Identify Surgical Site Infections using Electronic Healthcare Records.

Recently, hospitals and healthcare providers have made efforts to reduce surgical site infections as they are a major cause of surgical complications, a prominent reason for hospital readmission, and associated with significantly increased healthcare costs. Traditional surveillance methods for SSI rely on manual chart review, which can be laborious and costly. To assist the chart review process, we developed a long short-term memory (LSTM) model using structured electronic health record data to identify SSI. The top LSTM model resulted in an average precision (AP) of 0.570 [95% CI 0.567, 0.573] and area under the receiver operating characteristic curve (AUROC) of 0.905 [95% CI 0.904, 0.906] compared to the top traditional machine learning model, a random forest, which achieved 0.552 [95% CI 0.549, 0.555] AP and 0.899 [95% CI 0.898, 0.900] AUROC. Our LSTM model represents a step toward automated surveillance of SSIs, a critical component of quality improvement mechanisms.

Decreased Incidence of Hirschsprung-Associated Enterocolitis During COVID-19 Across United States Children's Hospitals.

BACKGROUND: The Coronavirus Disease 2019 pandemic provided a natural experiment to study the effect of social distancing on the risk of developing Hirschsprung's Associated Enterocolitis (HAEC). METHODS: Using the Pediatric Health Information System (PHIS), a retrospective cohort study of children (<18 years) with Hirschsprung's Disease (HSCR) across 47 United States children's hospitals was performed. The primary outcome was HAEC admissions per 10,000 patient-days. The exposure (COVID-19) was defined as April 2020-December 2021. The unexposed (historical control) period was April 2018-December 2019. Secondary outcomes included sepsis, bowel perforation, intensive care unit (ICU) admission, mortality, and length of stay. RESULTS: Overall, we included 5707 patients with HSCR during the study period. There were 984 and 834 HAEC admissions during the pre-pandemic and pandemic periods, respectively (2.6 vs. 1.9 HAEC admissions per 10,000 patient-days, incident rate ratio [95% confidence interval]: 0.74 [0.67, 0.81], p < 0.001). Compared to pre-pandemic, those with HAEC during the pandemic were younger (median [IQR]: 566 [162, 1430] days pandemic vs. 746 [259, 1609] days pre-pandemic, p < 0.001) and more likely to live in the lowest quartile of median household income zip codes (24% pandemic vs. 19% pre-pandemic, p = 0.02). There were no significant differences in rates of sepsis (6.1% pandemic vs. 6.1% pre-pandemic, p > 0.9), bowel perforation (1.3% pandemic vs. 1.2% pre-pandemic, p = 0.8), ICU admissions (9.6% pandemic vs. 12% pre-pandemic, p = 0.2), mortality (0.5% pandemic vs. 0.6% pre-pandemic, p = 0.8), or length of stay (median [interquartile range]: 4 [(Pastor et al., 2009; Gosain and Brinkman, 2015) 2,112,11 days pandemic vs. 5 [(Pastor et al., 2009; Tang et al., 2020) 2,102,10 days pre-pandemic, p = 0.4). CONCLUSIONS: The COVID-19 pandemic was associated with significantly decreased incidence of HAEC admissions across US children's hospitals. Possible etiologies such as social distancing should be explored. LEVEL OF EVIDENCE: II.

Concurrent SARS-COV-19 and acute appendicitis: Management and outcomes across United States children's hospitals.

BACKGROUND: Nonoperative management of acute appendicitis is a safe and effective alternative to appendectomy, though rates of treatment failure and disease recurrence are significant. The purpose of this study was to determine whether COVID-19-positive children with acute appendicitis were more likely to undergo nonoperative management when compared to COVID-19-negative peers and to compare clinical outcomes and healthcare use for these groups. METHODS: A retrospective cohort study of children <18 years with acute appendicitis across 45 US Children's Hospitals during the first 12 months of the COVID-19 pandemic was performed. Operative management was defined as appendectomy or percutaneous drain placement, whereas nonoperative management was defined as admission with antibiotics alone. Multivariable hierarchical logistic regression using an exact matched cohort was used to determine the association between COVID-19 positivity and nonoperative management. The secondary outcomes included intensive care unit admission, mechanical ventilation, length of stay, nonoperative management failure rates, and hospital variation in nonoperative management. RESULTS: Of 17,481 children in the cohort, 581 (3.3%) were positive for COVID-19. The odds of nonoperative management was significantly higher in the COVID-19-positive group (adjusted odds ratio [95% confidence interval]: 13.4 [10.7-16.8], P < .001). Patients positive for COVID-19 had increased odds of intensive care unit admission (adjusted odds ratio [95% confidence interval]: 3.78 [2.01-7.12], P < .001) and longer length of stay (median 2 days vs 1 day, P < .001). Hospital rates of nonoperative management ranged from 0% to 100% for COVID-19-positive patients and 0% to 42% for COVID-19-negative patients. CONCLUSION: Children with concurrent acute appendicitis and COVID-19 positivity are significantly more likely to undergo nonoperative management. Both groups experience infrequent nonoperative management failure rates and rare intensive care unit admissions. Marked hospital variability in nonoperative management practices was demonstrated.

Postdischarge Racial and Ethnic Disparities in Pediatric Appendicitis: A Mediation Analysis.

INTRODUCTION: Significant racial and ethnic disparities exist for children presenting with acute appendicitis; however, it is unknown if disparities persist after initial management and hospital discharge. MATERIALS AND METHODS: We performed a retrospective cohort study of children (aged < 18 y) who underwent treatment for acute appendicitis in 47 U.S. Children's Hospitals between 2017 and 2019. Primary outcomes were 30-d emergency department (ED) visits and 30-d inpatient readmission. Hierarchical multivariable logistic regression models were developed to determine the association of race and ethnicity on the primary outcomes. Inverse odds-weighted mediation analyses were used to estimate the degree to which complicated disease, insurance status, urbanicity, and residential socioeconomic status- mediated disparate outcomes. RESULTS: A total of 67,303 patients were included. Compared with Non-Hispanic White children, Non-Hispanic Black (NHB) (odds ratio [OR] 1.40, 95% confidence interval [CI] 1.23-1.59) and Hispanic/Latinx (HL) children (OR 1.55, 95% CI 1.44-1.67) had higher odds of ED visits. Only NHB children had higher odds of readmission (OR 1.43, 95% CI 1.30-1.57). On a multivariable analysis, NHB (adjusted OR 1.19, 95% CI 1.04-1.36) and HL (adjusted OR 1.19, 95% CI 1.09-1.31) children had higher odds of ED visits. Insurance, disease severity, socioeconomic status, and urbanicity mediated 61.6% (95% CI 29.7-100%) and 66.3% (95% CI 46.9-89.3%) of disparities for NHB and HL children, respectively. CONCLUSIONS: Children of racial and ethnic minorities are more likely to visit the ED after treatment for acute appendicitis, but HL patients did not have a corresponding increase in readmission. These differences were mediated mainly by insurance status and urban residence. A lack of appropriate postdischarge education and follow-up may drive disparities in healthcare utilization after pediatric appendicitis.

Surge in Pediatric Firearm Injuries Presenting to US Children's Hospitals During the COVID-19 Pandemic.

This cohort study uses administrative health data to evaluate trends in pediatric firearm injuries before and during the COVID-19 pandemic.

Racial and ethnic disparities in interhospital transfer for complex emergency general surgical disease across the United States.

BACKGROUND: Differential access to specialty surgical care can drive health care disparities, and interhospital transfer (IHT) is one mechanism through which access barriers can be realized for vulnerable populations. The association between race/ethnicity and IHT for patients presenting with complex emergency general surgery (EGS) disease is understudied. METHODS: Using the 2019 Nationwide Emergency Department Sample, we identified patients 18 years and older with 1 of 13 complex EGS diseases based on International Classification of Diseases, Tenth Revision , diagnosis codes. The primary outcome was IHT. A series of weighted logistic regression models was created to determine the association of race/ethnicity with the primary outcome while controlling for patient and hospital characteristics. RESULTS: Of 387,610 weighted patient encounters from 989 hospitals, 59,395 patients (15.3%) underwent IHT. Compared with non-Hispanic White patients, rates of IHT were significantly lower for non-Hispanic Black (15% vs. 17%; unadjusted odds ratio (uOR) [95% confidence interval (CI)], 0.58 [0.49-0.68]; p < 0.001), Hispanic/Latinx (HL) (9.0% vs. 17%; uOR [95% CI], 0.48 [0.43-0.54]; p < 0.001), Asian/Pacific Islander (Asian/PI) (11% vs. 17%; uOR [95% CI], 0.84 [0.78-0.91]; p < 0.001), and other race/ethnicity (12% vs. 17%; uOR [95% CI], 0.68 [0.57-0.81]; p < 0.001) patients. In multivariable models, the adjusted odds of IHT remained significantly lower for HL (adjusted odds ratio [95% CI], 0.76 [0.72-0.83]; p < 0.001) and Asian/PI patients (adjusted odds ratio [95% CI], 0.73 [0.62-0.86]; p < 0.001) but not for non-Hispanic Black and other race/ethnicity patients ( p > 0.05). CONCLUSION: In a nationally representative sample of emergency departments across the United States, patients of minority race/ethnicity presenting with complex EGS disease were less likely to undergo IHT when compared with non-Hispanic White patients. Disparities persisted for HL and Asian/PI patients when controlling for comorbid conditions, hospital and residential geography, neighborhood socioeconomic status, and insurance; these patients may face unique barriers in accessing surgical care. LEVEL OF EVIDENCE: Prognostic and Epidemiologic; Level III.

Caregiver health-related quality of life 1 year following pediatric gastrostomy tube placement.

Introduction: Children undergoing gastrostomy tube placement often have complex medical conditions that can increase caregiver burden and decrease caregiver health-related quality of life. Our goal was to identify changes in health-related quality of life over a 12-month period in the caregivers of these patients. Methods: We performed a prospective cohort study of pediatric patients undergoing gastrostomy tube placement. Using the PedsQL 2.0 Family Impact Module, we surveyed the caregivers of these patients at baseline (prior to gastrostomy tube placement) and 1 month, 3 months, 6 months, and 12 months following gastrostomy tube placement. We administered the same survey, at baseline only, to a control group composed of caregivers of pediatric patients undergoing elective hernia repair. Results: We enrolled 130 patients undergoing gastrostomy tube placement and 35 patients undergoing hernia repair. At baseline, these caregivers had significantly lower total health-related quality of life compared to caregivers of children undergoing hernia repair (47.4 ±â€¯16.1 vs 86.0 ±â€¯15.6, P < .001). In the first year after gastrostomy tube placement, caregivers had significant increases in total health-related quality of life (P < .01) and the physical functioning (P < .05), communication (P < .05), worry (P < .05), and daily activities (P < .001) subdomains. A within-subjects analysis comparing scores at baseline and 12 months revealed a significant increase in the total health-related quality of life score over this 1-year period (60.7 ±â€¯18.9 vs 47.4 ±â€¯16.1, P < .01). The total health-related quality of life at 12 months, however, remained significantly lower compared to controls (60.7 ±â€¯18.9 vs 86.0 ±â€¯15.6, P < .001). Conclusion: Our findings highlight the ongoing caregiver burden associated with children undergoing gastrostomy tube placement and provide evidence of the need for efforts directed at caregiver support post gastrostomy tube placement.

Standard Vocabularies to Improve Machine Learning Model Transferability With Electronic Health Record Data: Retrospective Cohort Study Using Health Care-Associated Infection.

BACKGROUND: With the widespread adoption of electronic healthcare records (EHRs) by US hospitals, there is an opportunity to leverage this data for the development of predictive algorithms to improve clinical care. A key barrier in model development and implementation includes the external validation of model discrimination, which is rare and often results in worse performance. One reason why machine learning models are not externally generalizable is data heterogeneity. A potential solution to address the substantial data heterogeneity between health care systems is to use standard vocabularies to map EHR data elements. The advantage of these vocabularies is a hierarchical relationship between elements, which allows the aggregation of specific clinical features to more general grouped concepts. OBJECTIVE: This study aimed to evaluate grouping EHR data using standard vocabularies to improve the transferability of machine learning models for the detection of postoperative health care-associated infections across institutions with different EHR systems. METHODS: Patients who underwent surgery from the University of Utah Health and Intermountain Healthcare from July 2014 to August 2017 with complete follow-up data were included. The primary outcome was a health care-associated infection within 30 days of the procedure. EHR data from 0-30 days after the operation were mapped to standard vocabularies and grouped using the hierarchical relationships of the vocabularies. Model performance was measured using the area under the receiver operating characteristic curve (AUC) and F1-score in internal and external validations. To evaluate model transferability, a difference-in-difference metric was defined as the difference in performance drop between internal and external validations for the baseline and grouped models. RESULTS: A total of 5775 patients from the University of Utah and 15,434 patients from Intermountain Healthcare were included. The prevalence of selected outcomes was from 4.9% (761/15,434) to 5% (291/5775) for surgical site infections, from 0.8% (44/5775) to 1.1% (171/15,434) for pneumonia, from 2.6% (400/15,434) to 3% (175/5775) for sepsis, and from 0.8% (125/15,434) to 0.9% (50/5775) for urinary tract infections. In all outcomes, the grouping of data using standard vocabularies resulted in a reduced drop in AUC and F1-score in external validation compared to baseline features (all P<.001, except urinary tract infection AUC: P=.002). The difference-in-difference metrics ranged from 0.005 to 0.248 for AUC and from 0.075 to 0.216 for F1-score. CONCLUSIONS: We demonstrated that grouping machine learning model features based on standard vocabularies improved model transferability between data sets across 2 institutions. Improving model transferability using standard vocabularies has the potential to improve the generalization of clinical prediction models across the health care system.

Geographic Proximity of Family Members and Healthcare Utilization After Complex Surgical Procedures.

OBJECTIVE: We sought to determine the relationship between a patient's proximal familial social support, defined as the geographic proximity of family members, and healthcare utilization after complex cardiovascular and oncologic procedures. BACKGROUND: Social support mechanisms are increasingly identified as modifiable risk factors for healthcare utilization. METHODS: We performed a retrospective cohort study of 60,895 patients undergoing complex cardiovascular procedures or oncologic procedures. We defined healthcare utilization outcomes as 30-day all-cause readmission unplanned readmission, nonindex hospital readmission, index hospital length of stay, and home discharge disposition. For each patient, we aggregated the number of first-degree relatives (FDR) living within 30 miles of the patient's home address at the time of the surgical procedure into the following categories: 0 to 1, 2 to 3, 4 to 5, 6+ FDRs. We developed hierarchical multivariable regression models to determine the relationship between the number of FDR living within 30 miles of the patient and the healthcare utilization outcomes. RESULTS: Compared with patients with 0 to 1 FDRs, patients with 6+ FDRs living in close proximity had significantly lower rates of all-cause readmission (12.1% vs 13.5%, P <0.001), unplanned readmission (10.9% vs 12.0%, P =0.001), nonindex readmission (2.6% vs 3.2%, P =0.003); higher rates of home discharge (88.0% vs 85.3%, P <0.001); and shorter length of stay (7.3 vs 7.5 days, P =0.02). After multivariable adjustment, a larger number of FDRs living within 30 miles of the patient was significantly associated with a lower likelihood of all-cause readmission ( P <0.001 for trend), 30-day unplanned readmission ( P <0.001), nonindex readmission ( P <0.001); higher likelihood of home discharge ( P <0.001); and shorter index length of stay ( P <0.001). CONCLUSIONS: The geographic proximity of family members is significantly associated with decreased healthcare utilization after complex cardiovascular and oncologic surgical procedures.

Pediatric surgery milestones 2.0: A primer.

More than twenty years ago, the Accreditation Council for Graduate Medical Education and the American Board of Medical Specialties began the conversion of graduate medical education from a structure- and process-based model to a competency-based framework. The educational outcomes assessment tool, known as the Milestones, was introduced in 2013 for seven specialties and by 2015 for the remaining specialties, including pediatric surgery. Designed to be an iterative process with improvements over time based on feedback and evidence-based literature, the Milestones started the evolution from 1.0 to 2.0 in 2016. The formation of Pediatric Surgery Milestones 2.0 began in 2019 and was finalized in 2021 for implementation in the 2022-2023 academic year. Milestones 2.0 are fewer in number and are stated in more straightforward language. It incorporated the harmonized milestones, subcompetencies for non-patient care and non-medical knowledge that are consistent across all medical and surgical specialties. There is a new Supplemental Guide that lists examples, references and links to other assessment tools and resources for each subcompetency. Milestones 2.0 represents a continuous process of feedback, literature review and revision with goals of improving patient care and maintaining public trust in graduate medical education's ability to self-regulate. LEVEL OF EVIDENCE: V.

Healthcare Burden and Cost in Children with Anorectal Malformation During the First 5 Years of Life.

OBJECTIVE: To identify cumulative 5-year healthcare costs and healthcare days in children with anorectal malformation (ARM) and to compare the cumulative 5-year healthcare costs and healthcare days in children with ARM with 3 control cohorts: healthy, premature, and congenital heart disease (CHD). STUDY DESIGN: We performed a retrospective case-control study using the Truven MarketScan database of commercial claims encounters between 2008 and 2017. The ARM, CHD, and premature cohorts were identified using a targeted list of International Classification of Diseases 9th or 10th Revision diagnosis and Current Procedural Terminology codes. The healthy cohort included patients without ARM, preterm birth, or CHD. RESULTS: We identified 664 children with ARM, 3356 children with heart disease, 63 190 children who were born preterm, and 2947 healthy patients. At 5 years, the total healthcare costs of children with ARM ($273K, 95% CI $168K-$378K) were similar to the premature cohort ($246K, 95% CI $237K-$255K) and lower than the CHD cohort ($466K, 95% CI $401K-$530K, P < .001). Total healthcare days were similar in children with ARM (158 days, 95% CI 117-198) and prematurity (141 days, 95% CI 137-144) but lower than CHD (223 days, 95% CI 197-250, P = .02). In ARM, outpatient care (126 days, 95% CI 93-159) represented the largest contribution to total healthcare days. CONCLUSIONS: Children with ARM accumulate similar healthcare costs to children with prematurity and comparable healthcare days to children with CHD and prematurity in the first 5 years of life. Outpatient care represents the majority of healthcare days in children with ARM, identifying this as a target for quality improvement and demonstrating the long-term impact of this condition.

Natural language processing for the surveillance of postoperative venous thromboembolism.

BACKGROUND: The objective of this study was to develop a portal natural language processing approach to aid in the identification of postoperative venous thromboembolism events from free-text clinical notes. METHODS: We abstracted clinical notes from 25,494 operative events from 2 independent health care systems. A venous thromboembolism detected as part of the American College of Surgeons National Surgical Quality Improvement Program (ACS NSQIP) was used as the reference standard. A natural language processing engine, easy clinical information extractor-pulmonary embolism/deep vein thrombosis (EasyCIE-PEDVT), was trained to detect pulmonary embolism and deep vein thrombosis from clinical notes. International Classification of Diseases (ICD) discharge diagnosis codes for venous thromboembolism were used as baseline comparators. The classification performance of EasyCIE-PEDVT was compared with International Classification of Diseases codes using sensitivity, specificity, area under the receiver operating characteristic curve, using an internal and external validation cohort. RESULTS: To detect pulmonary embolism, EasyCIE-PEDVT had a sensitivity of 0.714 and 0.815 in internal and external validation, respectively. To detect deep vein thrombosis, EasyCIE-PEDVT had a sensitivity of 0.846 and 0.849 in internal and external validation, respectively. EasyCIE-PEDVT had significantly higher discrimination for deep vein thrombosis compared with International Classification of Diseases codes in internal validation (area under the receiver operating characteristic curve: 0.920 vs 0.761; P < .001) and external validation (area under the receiver operating characteristic curve: 0.921 vs 0.794; P < .001). There was no significant difference in the discrimination for pulmonary embolism between EasyCIE-PEDVT and ICD codes. CONCLUSION: Accurate surveillance of postoperative venous thromboembolism may be achieved using natural language processing on clinical notes in 2 independent health care systems. These findings suggest natural language processing may augment manual chart abstraction for large registries such as NSQIP.

Impact of Insurance Expansion on Disposition for Pediatric Trauma Patients: A National Trauma Data Bank Cohort Study.

BACKGROUND: The insurance status of pediatric trauma patients is associated with access to post-discharge resources, including inpatient rehabilitation. Our goal was to understand the impact of changes in insurance coverage on access to post-acute care resources for pediatric trauma patients. MATERIALS AND METHODS: We utilized the National Trauma Data Bank from 2012 to 2016 for all pediatric trauma patients with a highest body region abbreviated injury score >2. Our primary outcome was hospital discharge disposition location. We used multivariable regression to adjust for salient patient and trauma center characteristics. Additionally, we performed a sensitivity analysis including only high-volume hospitals to examine the relationship between the magnitude of facility level expansion in coverage and changes in patient disposition. RESULTS: We identified 195,649 pediatric trauma patients meeting inclusion criteria. From 2012 to 2016 the proportion of patients with Medicaid (35% versus 39%, P < 0.001) and private insurance (39% versus 45%, P < 0.001) increased. Increased discharges with home health (HH) (+0.6% adjusted risk difference 2012 to 2016, P < 0.001) and decreased discharges to inpatient rehabilitation and skilled nursing facilities (-0.6% adjusted risk difference 2012 to 2016, P = 0.01) were associated with changes in insurance coverage. After stratifying high volume facilities by magnitude of increase in Medicaid and private insurance, we found that, for all groups, discharge to inpatient rehabilitation and skilled nursing facilities decreased (P < 0.01). CONCLUSIONS: Expanded insurance coverage for children was paradoxically associated with decreased discharge to inpatient rehabilitation and skilled nursing facilities after severe traumatic injury. These findings suggest that additional barriers may drive limitations in access to pediatric post-discharge services.

Assessment of Bone Mineral Density From a Computed Tomography Topogram of Photon-Counting Detector Computed Tomography-Effect of Phantom Size and Tube Voltage.

PURPOSE: The aim of this study was to assess the accuracy and impact of different sizes and tube voltages on bone mineral density (BMD) assessment using a computed tomography (CT) topogram acquired with photon-counting detector CT in an osteopenic ex vivo animal spine. MATERIALS AND METHODS: The lumbar back of a piglet was used to simulate osteopenia of the lumbar spine. Five fat layers (each with a thickness of 3 cm) were consecutively placed on top of the excised spine to emulate a total of 5 different sizes. Each size was repeatedly imaged on (A) a conventional dual-energy x-ray absorptiometry scanner as the reference standard, (B) a prototype photon-counting detector CT system at 120 kVp with energy thresholds at 20 and 70 keV, and (C) the same prototype system at 140 kVp with thresholds at 20 and 75 keV. Material-specific data were reconstructed from spectral topograms for B and C. Bone mineral density was measured for 3 lumbar vertebrae (L2-L4). A linear mixed-effects model was used to estimate the impact of vertebra, imaging setup, size, and their interaction term on BMD. RESULTS: The BMD of the lumbar spine corresponded to a T score in humans between -4.2 and -4.8, which is seen in osteoporosis. Averaged across the 3 vertebrae and 5 sizes, mean BMD was 0.56 ± 0.03, 0.55 ± 0.02, and 0.55 ± 0.02 g/cm2 for setup A, B, and C, respectively. There was no significant influence of imaging setup (P = 0.7), simulated size (P = 0.67), and their interaction term (both P > 0.2) on BMD. Bone mineral density decreased significantly from L2 to L4 for all 3 setups (all P < 0.0001). Bone mineral density was 0.59 ± 0.01, 0.57 ± 0.01, and 0.52 ± 0.02 g/cm2 for L2, L3, and L4, respectively, for setup A; 0.57 ± 0.02, 0.55 ± 0.01, and 0.53 ± 0.01 g/cm2 for setup B; and 0.57 ± 0.01, 0.55 ± 0.01, and 0.53 ± 0.01 g/cm2 for setup C. CONCLUSION: A single CT topogram acquired on photon-counting detector CT with 2 energy thresholds enabled BMD quantification with similar accuracy compared with dual-energy x-ray absorptiometry over a range of simulated sizes and tube voltages in an osteopenic ex vivo animal spine.