A Pipeline for Evaluation of Machine Learning/Artificial Intelligence Models to Quantify Programmed Death Ligand 1 Immunohistochemistry.

Immunohistochemistry (IHC) is used to guide treatment decisions in multiple cancer types. For treatment with checkpoint inhibitors, programmed death ligand 1 (PD-L1) IHC is used as a companion diagnostic. However, the scoring of PD-L1 is complicated by its expression in cancer and immune cells. Separation of cancer and noncancer regions is needed to calculate tumor proportion scores (TPS) of PD-L1, which is based on the percentage of PD-L1-positive cancer cells. Evaluation of PD-L1 expression requires highly experienced pathologists and is often challenging and time-consuming. Here, we used a multi-institutional cohort of 77 lung cancer cases stained centrally with the PD-L1 22C3 clone. We developed a 4-step pipeline for measuring TPS that includes the coregistration of hematoxylin and eosin, PD-L1, and negative control (NC) digital slides for exclusion of necrosis, segmentation of cancer regions, and quantification of PD-L1+ cells. As cancer segmentation is a challenging step for TPS generation, we trained DeepLab V3 in the Visiopharm software package to outline cancer regions in PD-L1 and NC images and evaluated the model performance by mean intersection over union (mIoU) against manual outlines. Only 14 cases were required to accomplish a mIoU of 0.82 for cancer segmentation in hematoxylin-stained NC cases. For PD-L1-stained slides, a model trained on PD-L1 tiles augmented by registered NC tiles achieved a mIoU of 0.79. In segmented cancer regions from whole slide images, the digital TPS achieved an accuracy of 75% against the manual TPS scores from the pathology report. Major reasons for algorithmic inaccuracies include the inclusion of immune cells in cancer outlines and poor nuclear segmentation of cancer cells. Our transparent and stepwise approach and performance metrics can be applied to any IHC assay to provide pathologists with important insights on when to apply and how to evaluate commercial automated IHC scoring systems.

Skin Color Representation in Teaching Photographs: One Institution's Approach to Addressing Visual Racism in Medical Education.

Problem: Visual racism refers to both the underrepresentation and inappropriate representation of darker skin types in medical education. By not teaching medical students and resident physicians to recognize common conditions in darker skin, it perpetuates biases that contribute to healthcare disparities for racial and ethnic minoritized groups. In this paper we describe our efforts to engage in institutional anti-racism work by addressing imbalances in representation of darker skin types in visual teaching images within our institution's curriculum. Intervention: We initially surveyed preclinical medical students regarding their perceptions of skin color representation in two courses. Researchers recorded the skin types of all teaching photographs in these courses in 2020. We then provided feedback and education to faculty, proposing that they increase brown and black skin color representation in educational content. During 2021, we reviewed the same courses and surveyed students again to ascertain the implementation and impact of our proposal. Context: We applied our intervention to two courses, Host & Defense (H&D) and Skin, Muscle, Bone, and Joint (SMBJ) since both courses utilize a large number of teaching images. Impact: From 2020 to 2021, both H&D and SMBJ significantly increased the proportion of visual teaching images that included darker skin types, with an increase from 28% to 42% in H&D and 20% to 30% in SMBJ. Significantly more students in the courses' 2021 iterations (73% in H&D, 93% in SMBJ) felt that lectures had appropriate representations of darker skin types when compared to students who took the course in 2020 (8% in H&D, 51% in SMBJ). Students in 2021 felt more confident in recognizing dermatological signs and symptoms in patients with darker skin than students in 2020. The majority of students in both 2020 and 2021 reported wanting to see a gradient of skin types for every dermatological condition discussed. Lessons learned: Our work suggests that addressing visual racism can be achieved partly by setting expectations for increased visual representation, collaborating across educational departments, and establishing clear metrics for assessing implementation. Future interventions will require a continual feedback loop of monitoring learning material, assessing faculty and student perception, refining resources, and recommending revisions to improve visual representation across the entire curriculum.

Analysis of Readmissions and Reoperations in Pediatric Microvascular Reconstruction.

BACKGROUND: Free tissue transfer is utilized as a reconstructive option for various anatomic defects. While it has long been performed in adults, reconstructive surgeons have used free tissue transfer to a lesser degree in children. As such, there are few analyses of factors associated with complications in free tissue transfer within this population. The aim of this study is to assess factors associated with readmission and reoperation in pediatric free flap patients utilizing the pediatric National Surgical Quality Improvement Program database. METHODS: Pediatric patients who underwent microvascular reconstruction between 2015 and 2020 were included. Patients were identified by five microvascular reconstruction Current Procedural Terminology codes and were then stratified by flap site (head and neck, extremities, trunk) and defect etiology (congenital, trauma, infection, neoplasm). Multivariate logistic regression was performed to identify factors associated with readmissions and reoperations. RESULTS: The study cohort consisted of 258 patients. The average age was 10.0 ± 4.7 years and the majority of patients were male (n = 149, 57.8%), were of white race (n = 164, 63.6%), and had a normal body mass index. Twenty-two patients (8.5%) experienced an unplanned readmission within 30 days of the initial operation, most commonly for wound disruption (31.8% of readmissions). The overall rate of unplanned reoperation within 30 days was 11.6% (n = 30) for all patients, with an average of 8.9 ± 7.5 days to reoperation. On multivariate regression analysis, each hour increase in operative time was associated with an increased odds of reoperation (odds ratio [OR]: 1.27; 95% confidence interval [CI]: 1.12, 1.45) and readmission (OR: 1.16; 95% CI: 1.02, 1.34). CONCLUSION: In pediatric patients undergoing free tissue transfer, higher readmission and reoperation risk was associated with longer operative duration. Overall, free tissue transfer is safe in the pediatric population with relatively low rates of readmission and reoperation.

Incidence and severity of asymptomatic ocular injury in adult and pediatric orbital fractures.

PURPOSE: To evaluate the incidence of severe ocular injury requiring emergent ophthalmic evaluation in visually asymptomatic patients presenting with orbital fractures. METHODS: We performed a retrospective chart review of all adult and pediatric orbital fractures between 2012-2022 at a level 1 trauma center. Ocular injuries were categorized into severe, moderate, and mild. We evaluated symptoms, mechanism of injury, visual acuity (VA), and severity of injuries using the Cochran-Armitage and linear-by-linear tests. RESULTS: Of the 2495 cases, 1534 had ophthalmology evaluation. The mean ± standard deviation age was 40.4 ± 20.4 years. Most patients were male (73.1%) and Caucasian (75.9%). The mean time to evaluation was 0.6 ± 2.5 days. 486 (31.7%) were visually symptomatic, 760 (49.5%) were asymptomatic, and 288 (18.8%) were unknown. Of the symptomatic, 135 (27.8%) had severe injuries, 108 (22.2%) had moderate injuries, 216 (44.4%) had mild injuries, and 27 (5.6%) had no injuries. Of the asymptomatic, 67 (8.8%) had severe injuries, 183 (24.1%) had moderate injuries, 468 (61.6%) had mild injuries, and 42 (5.5%) had no injuries. Symptoms correlated with injury severity (p-value <.001). The most common mechanism of injury were falls (24.0%), assaults (22.2%), and motor vehicle accidents (14.8%). CONCLUSIONS: Visually asymptomatic orbital fractures were less likely to have severe ocular injuries; however, many patients were unable to express symptoms. Emergent ophthalmology evaluation should be considered in all patients presenting with orbital fractures, especially patients with visual symptoms or are unable to report symptoms.

Combining antibody markers for serosurveillance of SARS-CoV-2 to estimate seroprevalence and time-since-infection.

Serosurveillance is an important epidemiologic tool for severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), used to estimate infection rates and the degree of population immunity. There is no general agreement on which antibody biomarker(s) should be used, especially with the rollout of vaccines globally. Here, we used random forest models to demonstrate that a single spike or receptor-binding domain (RBD) antibody was adequate for classifying prior infection, while a combination of two antibody biomarkers performed better than any single marker for estimating time-since-infection. Nucleocapsid antibodies performed worse than spike or RBD antibodies for classification, but can be useful for estimating time-since-infection, and in distinguishing infection-induced from vaccine-induced responses. Our analysis has the potential to inform the design of serosurveys for SARS-CoV-2, including decisions regarding a number of antibody biomarkers measured.

Characterizing the volume of surgery and post-operative complications during the COVID-19 pandemic.

PURPOSE: The COVID-19 pandemic led to unprecedented changes in volume and quality of surgery. Utilizing the American College of Surgeons National Surgical Quality Improvement Program (NSQIP) database, the current study assesses the impact of COVID-19 on surgical volume during each quarter of 2020 in comparison to 2019. Quality of surgical care during 2020 was also investigated by assessing postoperative complications, readmissions, and reoperations during 2020 in comparison to the previous 5 years. MATERIALS AND METHODS: The NSQIP database was queried from 2015 to 2020. Descriptive statistics and a chi-squared test were utilized to compare demographic variables. A seasonal autoregressive integrated moving average time-series model was fit to assess the trend and seasonality of complications from 2015 to 2019 and was used to forecast the proportion of complications in the year 2020 and compared the forecast with the actual proportions graphically. RESULTS: There were fewer patients operated on in 2020 compared to 2019, with the most dramatic drop in Q2 with a nearly 27% decrease. Patients with ASA class 3 or greater were operated on at a greater proportion in every quarter of 2020. Q2 of 2020 represented the highest proportion of any operative complications since 2015 at ~13%. Q4 of 2020 demonstrated a return to 2020 Q1 complication proportions. CONCLUSION: Surgical volume was heavily affected in 2020, particularly in Q2. Patients during Q2 of 2020 were generally of a higher ASA class and had increased operative complications. Operative volume and overall surgical complication rate normalized over the next two quarters.

Sharp Edge Eye Syndrome: A Case Report and Survey of Self-Identified Individuals.

BACKGROUND: Sharp edge eye syndrome (SEES), sometimes known as visual looming syndrome, is a condition in which the patient experiences ocular pain or discomfort when viewing or mentally picturing sharp objects and edges. Patients may present for medical care because they perceive the condition to represent an ophthalmic problem or a sign of a more serious underlying condition. An individual case report of SEES is included to aid in illustrating syndrome characteristics. Our aim is to describe the syndrome, vision-related quality of life (VRQOL), and psychosocial characteristics in patients with self-identified SEES. METHODS: A cross-sectional web-based survey was made available on social media webpages dedicated to SEES. The study included 22 questions developed by the research team, demographic questions, and 4 standardized questionnaires [ID Migraine, the National Eye Institute's Visual Function Questionnaire (NEI-VFQ-25), General Anxiety Disorder-2 (GAD-2), and Patient Health Questionnaire (depression) Scale-2]. RESULTS: Seventy-seven respondents had an average age of 29 and were 57% male. 92% reported symptoms before age 18. The main site of pain or discomfort was the eyes, with onset resulting from viewing or thinking of sharp objects and edges. Symptoms lasted from seconds to hours and could be prolonged even after closing eyes or avoiding viewing the trigger. The composite and subscale scores on the NEI-VFQ-25 were low, with a mean composite score of 78 and selected subscores of general health (61), general vision (73), ocular pain (68), driving (79), mental health (61), and role difficulties (72). Anxiety was reported in 58% of participants, and depression in 57%. Migraine or headache was reported in 46% of participants. Participants reported Alice in Wonderland syndrome, visual snow, obsessive-compulsive disorder, attention deficit hyperactivity disorder, stripe-induced visual discomfort, and synesthesia. CONCLUSION: From this survey, we have the beginnings of an understanding of the characteristics of SEES, as well as VRQOL impacts. These survey responses lead us to postulate that SEES may be a distinct visual phenomenon and to propose SEES criteria. Systematic studies of this condition's clinical features and treatment responses will be additional steps toward improving patient care.

Staphylococcal scalded skin syndrome: Clinical features, ancillary testing, and patient management.

BACKGROUND/OBJECTIVES: The utility of ancillary testing in improving diagnostic precision or improving patient outcomes in staphylococcal scalded skin syndrome (SSSS) is unclear. Similarly, an optimal antibiotic regimen has yet to be established. Our goal was to describe clinical characteristics and ancillary work-up of SSSS, report bacterial resistance patterns, and examine patient outcomes under varying therapeutic strategies with the aim of developing an evidence-based management algorithm. METHODS: We performed a retrospective review of pediatric patients diagnosed with SSSS at Intermountain Healthcare facilities between 2010 and 2021. A Kruskal-Wallis rank sum test was used to assess median length of stay between different antibiotic regimens. RESULTS: Eighty-five cases were identified. The most common ancillary tests obtained were a complete blood count (88%), followed by chemistry analysis (80%). Blood cultures were collected in more patients (79%) compared to aerobic cultures (60%). No blood culture was positive for Staphylococcus aureus. All S. aureus isolates were methicillin-sensitive. Of those found resistant to clindamycin (36%), all demonstrated macrolide-induced clindamycin resistance. None were constitutively resistant to clindamycin. There was no statistical difference between antibiotic regimen and length of stay (p = .691). Receiving opiate medications was the only risk factor associated with prolonged hospitalization (p = .001). CONCLUSIONS: Ancillary testing does not improve diagnostic precision and can be reduced. Clindamycin does not improve patient outcomes, suggesting beta-lactams should be considered first line. Susceptibility patterns in our cohort demonstrate inducible clindamycin resistance as opposed to constitutive.

Cognitive enhancing supplements and medications in United States Resident Physicians.

BACKGROUND: The use of cognitive-enhancing medications and supplements among healthy adults continues to rise. Limited data exists on their use among resident physicians. Given their highly competitive and stressful lifestyle, we sought to evaluate the prevalence, motivations, and side effects of using cognitive-enhancing supplements and medications among resident physicians at a large United States academic institution. METHODS: An anonymous web-based survey was circulated to resident physicians inquiring about using cognitive-enhancing supplements and medications, as well as personal characteristics such as gender, marital and parental status, medical diagnoses, and medical specialty. Before circulation, we performed a pilot study. Weighted logistic regression analyses estimated the impact of personal characteristics on the probability of using both supplements and medications. RESULTS: Survey response rate was 46.4%. Of respondents, 48.6% were female, 45.9% were married, 70.9% were without children, and 67.2% were in a non-surgical medical specialty. Few respondents had a related medical diagnosis, with attention deficit hyperactivity disorder being the most common (7.1%). Male, non-married, surgical residents were more likely to take supplements (odds ratio (OR) = 1.06, 1.05, and 1.05). Males, without children, and those who felt pressure to perform well, were afraid of being left behind, felt pressure because colleagues take them, or felt they could not reach their current level of training without medications were more likely to take medications (OR = 1.11, 1.04, 1.05, and 1.08). Adverse effects with medications were common. CONCLUSION: Supplement and medication use for cognitive enhancement was high among resident physicians at a single institution despite few having a related medical diagnosis. This study raises awareness of the growing pressure in competitive residency environments to use cognitive enhancement regardless of the potential side effects.

The Effects of Using a Clinical Prediction Rule to Prioritize Diagnostic Testing on Transmission and Hospital Burden: A Modeling Example of Early Severe Acute Respiratory Syndrome Coronavirus 2.

BACKGROUND: Prompt identification of infections is critical for slowing the spread of infectious diseases. However, diagnostic testing shortages are common in emerging diseases, low resource settings, and during outbreaks. This forces difficult decisions regarding who receives a test, often without knowing the implications of those decisions on population-level transmission dynamics. Clinical prediction rules (CPRs) are commonly used tools to guide clinical decisions. METHODS: Using early severe acute respiratory syndrome coronavirus disease 2 (SARS-CoV-2) as an example, we used data from electronic health records to develop a parsimonious 5-variable CPR to identify those who are most likely to test positive. To consider the implications of gains in daily case detection at the population level, we incorporated testing using the CPR into a compartmentalized model of SARS-CoV-2. RESULTS: We found that applying this CPR (area under the curve, 0.69; 95% confidence interval, .68-.70) to prioritize testing increased the proportion of those testing positive in settings of limited testing capacity. We found that prioritized testing led to a delayed and lowered infection peak (ie, "flattens the curve"), with the greatest impact at lower values of the effective reproductive number (such as with concurrent community mitigation efforts), and when higher proportions of infectious persons seek testing. In addition, prioritized testing resulted in reductions in overall infections as well as hospital and intensive care unit burden. CONCLUSION: We highlight the population-level benefits of evidence-based allocation of limited diagnostic capacity.SummaryWhen the demand for diagnostic tests exceeds capacity, the use of a clinical prediction rule to prioritize diagnostic testing can have meaningful impact on population-level outcomes, including delaying and lowering the infection peak, and reducing healthcare burden.

Etiologies of bloody diarrhea in children presenting with acute gastroenteritis to US emergency departments.

Among 111 children presenting with bloody diarrhea in a multicenter study of molecular testing in US emergency departments, we found viral pathogens in 18%, bacteria in 48%, protozoa in 2%, and no pathogens detected in 38%.

Integration of population-level data sources into an individual-level clinical prediction model for dengue virus test positivity.

The differentiation of dengue virus (DENV) infection, a major cause of acute febrile illness in tropical regions, from other etiologies, may help prioritize laboratory testing and limit the inappropriate use of antibiotics. While traditional clinical prediction models focus on individual patient-level parameters, we hypothesize that for infectious diseases, population-level data sources may improve predictive ability. To create a clinical prediction model that integrates patient-extrinsic data for identifying DENV among febrile patients presenting to a hospital in Thailand, we fit random forest classifiers combining clinical data with climate and population-level epidemiologic data. In cross-validation, compared to a parsimonious model with the top clinical predictors, a model with the addition of climate data, reconstructed susceptibility estimates, force of infection estimates, and a recent case clustering metric significantly improved model performance.

Integrating molecular and clinical variables to predict myocardial recovery.

Mechanical unloading and circulatory support with left ventricular assist devices (LVADs) mediate significant myocardial improvement in a subset of advanced heart failure (HF) patients. The clinical and biological phenomena associated with cardiac recovery are under intensive investigation. Left ventricular (LV) apical tissue, alongside clinical data, were collected from HF patients at the time of LVAD implantation (n=208). RNA was isolated and mRNA transcripts were identified through RNA sequencing and confirmed with RT-qPCR. To our knowledge this is the first study to combine transcriptomic and clinical data to derive predictors of myocardial recovery. We used a bioinformatic approach to integrate 59 clinical variables and 22,373 mRNA transcripts at the time of LVAD implantation for the prediction of post-LVAD myocardial recovery defined as LV ejection fraction (LVEF) ≥40% and LV end-diastolic diameter (LVEDD) ≤5.9cm, as well as functional and structural LV improvement independently by using LVEF and LVEDD as continuous variables, respectively. To substantiate the predicted variables, we used a multi-model approach with logistic and linear regressions. Combining RNA and clinical data resulted in a gradient boosted model with 80 features achieving an AUC of 0.731±0.15 for predicting myocardial recovery. Variables associated with myocardial recovery from a clinical standpoint included HF duration, pre-LVAD LVEF, LVEDD, and HF pharmacologic therapy, and LRRN4CL (ligand binding and programmed cell death) from a biological standpoint. Our findings could have diagnostic, prognostic, and therapeutic implications for advanced HF patients, and inform the care of the broader HF population.

Predicting IHC staining classes of NF1 using features in the hematoxylin channel.

Immunohistochemistry (IHC) highlights specific cell types in tissues and traditionally involves antibody staining together with a hematoxylin counterstain. The intensity and pattern of hematoxylin staining differs between cell types and reveals morphological characteristics of cells. Here, we propose that features in the hematoxylin stain can be used to predict IHC labels, such as Neurofibromin (encoded by the gene NF1). The dataset consists of 7.2 million cells from benign and kidney cancer cores in a tissue microarray. Morphology and hematoxylin (H&M) features defined within QuPath are subjected to a clustering analysis in CytoMap. H&M features are also used to train 4 different XGBoost models to predict high, low, and negative NF1 stain classes in benign renal tubules, clear cell (ccRCC), papillary (PRCC), and chromophobe (ChRCC) renal carcinoma. The prediction accuracies of NF1 staining classes in benign, ccRCC, ChRCC, and PRCC range between 70% and 90% with areas under the precision recall curve PRAUCNF1-high = 0.82+0.12, PRAUCNF1-low = 0.62+0.25, and PRAUCNF1-negative = 0.83+0.16. The most important feature for predicting the NF1 class involves the minimum cellular hematoxylin staining intensity. Together, these results demonstrate the feasibility to predict NF1 expression solely from features in hematoxylin staining using open source software. Since the hematoxylin features can be obtained from regular H&E and IHC slides, the proposed workflow has broad applicability.

Structured Case-Based Ethics Discussion for Trainees and Faculty on Dermatopathology.

Introduction: Ethical and professional dilemmas are part of the day-to-day practice of medicine, including within dermatopathology (e.g., ethical implications of self-referring skin biopsies for pathology interpretation). There is a need for teaching aids that dermatology educators can easily access to help provide ethics education. Methods: We held an hour-long, faculty-facilitated, interactive, virtual discussion about ethical issues in dermatopathology. The session followed a structured, case-based format. We administered anonymous online feedback surveys after the session and used the Wilcoxon signed rank test to compare participants' before and after responses. Results: Seventy-two individuals from two academic institutions participated in the session. We collected 35 total responses (49%) from dermatology residents (n = 15), dermatology faculty (n = 14), medical students (n = 2), and other providers and learners (n = 4). Feedback was largely positive, with 21 attendees (60%) indicating they learned a few things and 11 (31%) indicating they learned a great deal. Additionally, 32 participants (91%) indicated they would recommend the session to a colleague. Our analysis showed that attendees had a greater self-perceived level of achievement for each of our three objectives after the session. Discussion: This dermatoethics session is structured so as to be easily shared, deployed, and built on by other institutions. We hope that other institutions will use our materials and results to improve upon the foundation presented here and that this framework will be used by other medical specialties seeking to foster ethics education in their training programs.

Assessment of Model Estimated and Directly Observed Weather Data for Etiological Prediction of Diarrhea.

Recent advances in clinical prediction for diarrheal etiology in low- and middle-income countries have revealed that addition of weather data improves predictive performance. However, the optimal source of weather data remains unclear. We aim to compare model estimated satellite- and ground-based observational data with weather station directly-observed data for diarrheal prediction. We used clinical and etiological data from a large multi-center study of children with diarrhea to compare these methods. We show that the two sources of weather conditions perform similarly in most locations. We conclude that while model estimated data is a viable, scalable tool for public health interventions and disease prediction, directly observed weather station data approximates the modeled data, and given its ease of access, is likely adequate for prediction of diarrheal etiology in children in low- and middle-income countries.

Association between specific unmet functional needs and desire to institutionalize among caregivers of older veterans.

OBJECTIVES: To evaluate the associations between specific functional needs of older Veterans and the desire to institutionalize (DTI) among their caregivers. METHODS: Cross-sectional multivariable logistic regression analysis of 3579 Hero Care survey responses from caregivers of Veterans at five US sites from July to December 2021. Unmet needs were areas in which the caregiver reported the Veteran needed a little more or a lot more help. Caregiver DTI was defined as the caregiver reporting that they had discussed, considered, or taken steps toward a nursing home or assisted living placement for the Veteran or that they felt the Veteran would be better off in such a setting or they were likely to move the Veteran to another living arrangement. RESULTS: Caregivers were largely white, retired, females with an average age of 71 and with some college education who spent an average of 8-9 h per day 6 days a week caring for a Veteran spouse. There was evidence of associations between the following needs and a DTI: managing incontinence, using the telephone, transportation, and arranging services in the home such as visiting nurses, home care aides, or meals on wheels. Unmet functional needs in other selected domains were not associated with the DTI. CONCLUSION: Among caregivers of older Veterans, a need for more assistance managing incontinence, telephone use, transportation, and arranging in-home services were associated with the DTI. These may represent functional markers of important clinical determinants for institutionalization as well as potential targets for intervention to reduce caregiver DTI, such as programs that provide more caregiver or Veteran support in the home to meet these needs and reduce caregiver burden.

Development of a prediction model for the acquisition of extended spectrum beta-lactam-resistant organisms in U.S. international travellers.

BACKGROUND: Extended spectrum beta-lactamase producing Enterobacterales (ESBL-PE) present a risk to public health by limiting the efficacy of multiple classes of beta-lactam antibiotics against infection. International travellers may acquire these organisms and identifying individuals at high risk of acquisition could help inform clinical treatment or prevention strategies. METHODS: We used data collected from a cohort of 528 international travellers enrolled in a multicentre US-based study to derive a clinical prediction rule (CPR) to identify travellers who developed ESBL-PE colonization, defined as those with new ESBL positivity in stool upon return to the United States. To select candidate features, we used data collected from pre-travel and post-travel questionnaires, alongside destination-specific data from external sources. We utilized LASSO regression for feature selection, followed by random forest or logistic regression modelling, to derive a CPR for ESBL acquisition. RESULTS: A CPR using machine learning and logistic regression on 10 features has an internally cross-validated area under the receiver operating characteristic curve (cvAUC) of 0.70 (95% confidence interval 0.69-0.71). We also demonstrate that a four-feature model performs similarly to the 10-feature model, with a cvAUC of 0.68 (95% confidence interval 0.67-0.69). This model uses traveller's diarrhoea, and antibiotics as treatment, destination country waste management rankings and destination regional probabilities as predictors. CONCLUSIONS: We demonstrate that by integrating traveller characteristics with destination-specific data, we could derive a CPR to identify those at highest risk of acquiring ESBL-PE during international travel.

Derivation and external validation of a clinical prognostic model identifying children at risk of death following presentation for diarrheal care.

Diarrhea continues to be a leading cause of death for children under-five. Amongst children treated for acute diarrhea, mortality risk remains elevated during and after acute medical management. Identification of those at highest risk would enable better targeting of interventions, but available prognostic tools lack validation. We used clinical and demographic data from the Global Enteric Multicenter Study (GEMS) to build predictive models for death (in-treatment, after discharge, or either) in children aged ≤59 months presenting with moderate-to-severe diarrhea (MSD), in Africa and Asia. We screened variables using random forests, and assessed predictive performance with random forest regression and logistic regression using repeated cross-validation. We used data from the Kilifi Health and Demographic Surveillance System (KHDSS) and Kilifi County Hospital (KCH) in Kenya to externally validate our GEMS-derived clinical prognostic model (CPM). Of 8060 MSD cases, 43 (0.5%) children died in treatment and 122 (1.5% of remaining) died after discharge. MUAC at presentation, respiratory rate, age, temperature, number of days with diarrhea at presentation, number of people living in household, number of children <60 months old living in household, and how much the child had been offered to drink since diarrhea started were predictive of death both in treatment and after discharge. Using a parsimonious 2-variable prediction model, we achieve an AUC=0.84 (95% CI: 0.82, 0.86) in the derivation dataset, and an AUC=0.74 (95% CI 0.71, 0.77) in the external dataset. Our findings suggest it is possible to identify children most likely to die after presenting to care for acute diarrhea. This could represent a novel and cost-effective way to target resources for the prevention of childhood mortality.