On the Use of Minimum Penalties in Statistical Learning.

Modern multivariate machine learning and statistical methodologies estimate parameters of interest while leveraging prior knowledge of the association between outcome variables. The methods that do allow for estimation of relationships do so typically through an error covariance matrix in multivariate regression which does not generalize to other types of models. In this article we proposed the MinPen framework to simultaneously estimate regression coefficients associated with the multivariate regression model and the relationships between outcome variables using common assumptions. The MinPen framework utilizes a novel penalty based on the minimum function to simultaneously detect and exploit relationships between responses. An iterative algorithm is proposed as a solution to the non-convex optimization. Theoretical results such as high dimensional convergence rates, model selection consistency, and a framework for post selection inference are provided. We extend the proposed MinPen framework to other exponential family loss functions, with a specific focus on multiple binomial responses. Tuning parameter selection is also addressed. Finally, simulations and two data examples are presented to show the finite sample properties of this framework. Supplemental material providing proofs, additional simulations, code, and data sets are available online.

A retrospective, multi-agency 'target trial emulation' for the comparison of post-resuscitation epinephrine to norepinephrine.

INTRODUCTION: Epinephrine and norepinephrine are the two most commonly used prehospital vasopressors in the United States. Prior studies have suggested that use of a post-ROSC epinephrine infusion may be associated with increased rearrest and mortality in comparison to use of norepinephrine. We used target trial emulation methodology to compare the rates of rearrest and mortality between the groups of OHCA patients receiving these vasopressors in the prehospital setting. METHODS: Adult (18-80 years of age) non-traumatic OHCA patients in the 2018-2022 ESO Data Collaborative datasets with a documented post-ROSC norepinephrine or epinephrine infusion were included in this study. Logistic regression modeling was used to evaluate the association between vasopressor agent and outcome using two sets of covariables. The first set of covariables included standard Utstein factors, the dispatch to ROSC interval, the ROSC to vasopressor interval, and the follow-up interval. The second set added prehospital systolic blood pressure and SpO2 values. Kaplan-Meier time-to-event analysis was also conducted and the vasopressor groups were compared using a multivariable Cox regression model. RESULTS: Overall, 1,893 patients treated by 309 EMS agencies were eligible for analysis. 1,010 (53.4%) received an epinephrine infusion and 883 (46.7%) received a norepinephrine infusion as their initial vasopressor. Adjusted analyses did not discover an association between vasopressor agent and rearrest (aOR: 0.93 [0.72, 1.21]) or mortality (aOR: 1.00 [0.59, 1.69]). CONCLUSIONS: In this multi-agency target trial emulation, the use of a post-resuscitation epinephrine infusion was not associated with increased odds of rearrest in comparison to the use of a norepinephrine infusion.

Enhanced SARS-CoV-2 case prediction using public health data and machine learning models.

Objectives: The goal of this study is to propose and test a scalable framework for machine learning (ML) algorithms to predict near-term severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) cases by incorporating and evaluating the impact of real-time dynamic public health data. Materials and Methods: Data used in this study include patient-level results, procurement, and location information of all SARS-CoV-2 tests reported in West Virginia as part of their mandatory reporting system from January 2021 to March 2022. We propose a method for incorporating and comparing widely available public health metrics inside of a ML framework, specifically a long-short-term memory network, to forecast SARS-CoV-2 cases across various feature sets. Results: Our approach provides better prediction of localized case counts and indicates the impact of the dynamic elements of the pandemic on predictions, such as the influence of the mixture of viral variants in the population and variable testing and vaccination rates during various eras of the pandemic. Discussion: Utilizing real-time public health metrics, including estimated Rt from multiple SARS-CoV-2 variants, vaccination rates, and testing information, provided a significant increase in the accuracy of the model during the Omicron and Delta period, thus providing more precise forecasting of daily case counts at the county level. This work provides insights on the influence of various features on predictive performance in rural and non-rural areas. Conclusion: Our proposed framework incorporates available public health metrics with operational data on the impact of testing, vaccination, and current viral variant mixtures in the population to provide a foundation for combining dynamic public health metrics and ML models to deliver forecasting and insights in healthcare domains. It also shows the importance of developing and deploying ML frameworks in rural settings.

Reverse shock index multiplied by the motor component of the Glasgow Coma Scale predicts mortality and need for intervention in pediatric trauma patients.

BACKGROUND: Timely identification of high-risk pediatric trauma patients and appropriate resource mobilization may lead to improved outcomes. We hypothesized that reverse shock index times the motor component of the Glasgow Coma Scale (rSIM) would perform equivalently to reverse shock index times the total Glasgow Coma Scale (rSIG) in the prediction of mortality and the need for intervention following pediatric trauma. METHODS: The 2017-2020 National Trauma Data Bank datasets were used. We included all patients <16 years of age that had a documented prehospital and trauma bay systolic blood pressure, heart rate, and total GCS. We excluded all patients who arrived at the trauma center without vital signs and interfacility transport patients. Receiver operating characteristic (ROC) curves were used to model the performance of each metric as a classifier with respect to our primary and secondary outcomes, and the area under the ROC curve (AUC) was used for comparison. Our primary outcome was mortality prior to hospital discharge. Secondary outcomes included blood product administration or hemorrhage control intervention (surgery or angiography) < 4 hours following hospital arrival and ICU admission. RESULTS: After application of exclusion criteria, 77,996 patients were included in our analysis. rSIM and rSIG performed equivalently as predictors of mortality in the 1-2 (p = 0.05) and 3-5 (p = 0.28) year categories, but rSIM was statistically outperformed by rSIG in the 6-12 (AUC: 0.96 vs. 0.95, p = 0.04) and 13-16 (AUC: 0.96 vs. 0.95, p < 0.01) year-old age categories. rSIM and rSIG also performed similarly with respect to prediction of secondary outcomes. CONCLUSION: rSIG and rSIM are both outstanding predictors of mortality following pediatric trauma. Statistically significant differences in favor of rSIG were noted in some age groups. Because of the simplicity of calculation, rSIM may be a useful tool for pediatric trauma triage. LEVEL OF EVIDENCE: III, Diagnostic Tests or Criteria.

Lessons from West Virginia's Pandemic Response.

In this editorial discussion we describe our experience developing and implementing predictive models during the pandemic response in the state of West Virginia. We provide insights the on the importance of communication and the dynamic environment that exists that impacts predictive modeling in situations such as those that we faced. It is our hope that this work brings insight to those who may experience similar challenges while working in public health policy.

Impact of stress and stress mindset on prevalence of cardiovascular disease risk factors among first responders.

BACKGROUND: Psychological stress is recognized as an important modifiable risk factor for cardiovascular disease (CVD). Despite its potential significance, few to no studies have evaluated the association between stress, stress mindset, and CVD risk factors among rural first responders. The objectives of this study were to identify relationships between general stress, stress mindset, and CVD risk factors. METHODS: The study sample (n = 148) included those 18 years or older and who currently serve as a first responder, defined as either EMS, firefighter, or law enforcement. Questionnaires captured information on demographics, years of work experience as a first responder, multiple first responder occupations, general stress, stress mindset, and self-reported CVD risk factors. Data were analyzed using regression analyses. RESULTS: Findings suggest that first responders with a stress-is-negative mindset have significantly higher general stress levels (ß = 2.20, p = 0.01). Of note, general stress was not a significant predictor of CVD risk factors (AOR = 1.00, 95%CI = 0.93, 1.08) included in our study. However, a negative stress mindset was statistically significant predictor of CVD risk factors (AOR = 2.82, 95%CI = 1.29, 6.41), after adjusting for general stress and other potential confounders. CONCLUSIONS: Findings suggest that stress mindset is an independent predictor of stress and CVD risk factors among rural first responders. These results have the potential to inform educational and organization level interventions targeting stress appraisal for this vulnerable sub population of workers.

Stay and play or load and go? The association of on-scene advanced life support interventions with return of spontaneous circulation following traumatic cardiac arrest.

INTRODUCTION: Traumatic out-of-hospital cardiac arrest (tOHCA) has a mortality rate over 95%. Many current protocols dictate rapid intra-arrest transport of these patients. We hypothesized that on-scene advanced life support (ALS) would increase the odds of arriving at the emergency department with ROSC (ROSC at ED) in comparison to performance of no ALS or ALS en route. METHODS: We utilized the 2018-2021 ESO Research Collaborative public use datasets for this study, which contain patient care records from ~2000 EMS agencies across the US. All OHCA patients with an etiology of "trauma" or "exsanguination" were screened (n=15,691). The time of advanced airway management, vascular access, and chest decompression was determined for each patient. Logistic regression modeling was used to evaluate the association of ALS intervention timing with ROSC at ED. RESULTS: 4942 patients met inclusion criteria. 14.6% of patients had ROSC at ED. In comparison to no vascular access, on-scene (aOR: 2.14 [1.31, 3.49]) but not en route vascular access was associated with increased odds of having ROSC at ED arrival. In comparison to no chest decompression, neither en route nor on-scene chest decompression were associated with ROSC at ED arrival. Similarly, in comparison to no advanced airway management, neither en route nor on-scene advanced airway management were associated with ROSC at ED arrival. The odds of ROSC at ED decreased by 3% (aOR: 0.97 [0.94, 0.99]) for every 1-minute increase in time to vascular access and decreased by 5% (aOR: 0.95 [0.94, 0.99]) for every 1-minute increase in time to epinephrine. CONCLUSION: On-scene ALS interventions were associated with increased ROSC at ED in our study. These data suggest that initiating ALS prior to rapid transport to definitive care in the setting of tOHCA may increase the number of patients with a palpable pulse at ED arrival.

Assessing the effects of therapeutic combinations on SARS-CoV-2 infected patient outcomes: A big data approach.

BACKGROUND: The COVID-19 pandemic has demonstrated the need for efficient and comprehensive, simultaneous assessment of multiple combined novel therapies for viral infection across the range of illness severity. Randomized Controlled Trials (RCT) are the gold standard by which efficacy of therapeutic agents is demonstrated. However, they rarely are designed to assess treatment combinations across all relevant subgroups. A big data approach to analyzing real-world impacts of therapies may confirm or supplement RCT evidence to further assess effectiveness of therapeutic options for rapidly evolving diseases such as COVID-19. METHODS: Gradient Boosted Decision Tree, Deep and Convolutional Neural Network classifiers were implemented and trained on the National COVID Cohort Collaborative (N3C) data repository to predict the patients' outcome of death or discharge. Models leveraged the patients' characteristics, the severity of COVID-19 at diagnosis, and the calculated proportion of days on different treatment combinations after diagnosis as features to predict the outcome. Then, the most accurate model is utilized by eXplainable Artificial Intelligence (XAI) algorithms to provide insights about the learned treatment combination impacts on the model's final outcome prediction. RESULTS: Gradient Boosted Decision Tree classifiers present the highest prediction accuracy in identifying patient outcomes with area under the receiver operator characteristic curve of 0.90 and accuracy of 0.81 for the outcomes of death or sufficient improvement to be discharged. The resulting model predicts the treatment combinations of anticoagulants and steroids are associated with the highest probability of improvement, followed by combined anticoagulants and targeted antivirals. In contrast, monotherapies of single drugs, including use of anticoagulants without steroid or antivirals are associated with poorer outcomes. CONCLUSIONS: This machine learning model by accurately predicting the mortality provides insights about the treatment combinations associated with clinical improvement in COVID-19 patients. Analysis of the model's components suggests benefit to treatment with combination of steroids, antivirals, and anticoagulant medication. The approach also provides a framework for simultaneously evaluating multiple real-world therapeutic combinations in future research studies.

Prehospital shock index predicts outcomes after prolonged transport: A multicenter rural study.

BACKGROUND: Shock index (SI) predicts outcomes after trauma. Prior single-center work demonstrated that emergency medical services (EMSs) initial SI was the most accurate predictor of hospital outcomes in a rural environment. This study aimed to evaluate the predictive ability of SI in multiple rural trauma systems with prolonged transport times to a definitive care facility. METHODS: This retrospective review was performed at four American College of Surgeons-verified level 1 trauma centers with large rural catchment basins. Adult trauma patients who were transferred and arrived >60 minutes from scene during 2018 were included. Patients who sustained blunt chest or abdominal trauma were analyzed. Subjects with missing data or severe head trauma (Abbreviated Injury Scale score, >2) were excluded. Poisson and binomial logistic regression were used to study the effect of SI and delta shock index (∆SI) on outcomes. RESULTS: After applying the criteria, 789 patients were considered for analysis (502 scene patients and 287 transfers). The mean Injury Severity Score was 8 (interquartile range, 6) for scene and 8.9 (interquartile range, 5) for transfers. Initial EMSs SI was a significant predictor of the need for blood transfusion and intensive care unit care in both scene and transferred patients. An increase in ∆SI was predictive of the need for operative intervention ( p < 0.05). There were increased odds for mortality for every 0.1 change in EMSs SI; those changes were not deemed significant among both scene and transfer patients ( p < 0.1). CONCLUSION: Providers must maintain a high level of clinical suspicion for patients who had an initially elevated SI. Emergency medical services SI is a significant predictor for use of blood and intensive care unit care, as well as mortality for scene patients. This highlights the importance of SI and ∆SI in rural trauma care. LEVEL OF EVIDENCE: Prognostic and Epidemiological; Level IV.

Community risks for SARS-CoV-2 infection among fully vaccinated US adults by rurality: A retrospective cohort study from the National COVID Cohort Collaborative.

BACKGROUND: While COVID-19 vaccines reduce adverse outcomes, post-vaccination SARS-CoV-2 infection remains problematic. We sought to identify community factors impacting risk for breakthrough infections (BTI) among fully vaccinated persons by rurality. METHODS: We conducted a retrospective cohort study of US adults sampled between January 1 and December 20, 2021, from the National COVID Cohort Collaborative (N3C). Using Kaplan-Meier and Cox-Proportional Hazards models adjusted for demographic differences and comorbid conditions, we assessed impact of rurality, county vaccine hesitancy, and county vaccination rates on risk of BTI over 180 days following two mRNA COVID-19 vaccinations between January 1 and September 21, 2021. Additionally, Cox Proportional Hazards models assessed the risk of infection among adults without documented vaccinations. We secondarily assessed the odds of hospitalization and adverse COVID-19 events based on vaccination status using multivariable logistic regression during the study period. RESULTS: Our study population included 566,128 vaccinated and 1,724,546 adults without documented vaccination. Among vaccinated persons, rurality was associated with an increased risk of BTI (adjusted hazard ratio [aHR] 1.53, 95% confidence interval [CI] 1.42-1.64, for urban-adjacent rural and 1.65, 1.42-1.91, for nonurban-adjacent rural) compared to urban dwellers. Compared to low vaccine-hesitant counties, higher risks of BTI were associated with medium (1.07, 1.02-1.12) and high (1.33, 1.23-1.43) vaccine-hesitant counties. Compared to counties with high vaccination rates, a higher risk of BTI was associated with dwelling in counties with low vaccination rates (1.34, 1.27-1.43) but not medium vaccination rates (1.00, 0.95-1.07). Community factors were also associated with higher odds of SARS-CoV-2 infection among persons without a documented vaccination. Vaccinated persons with SARS-CoV-2 infection during the study period had significantly lower odds of hospitalization and adverse events across all geographic areas and community exposures. CONCLUSIONS: Our findings suggest that community factors are associated with an increased risk of BTI, particularly in rural areas and counties with high vaccine hesitancy. Communities, such as those in rural and disproportionately vaccine hesitant areas, and certain groups at high risk for adverse breakthrough events, including immunosuppressed/compromised persons, should continue to receive public health focus, targeted interventions, and consistent guidance to help manage community spread as vaccination protection wanes.

An ordinal severity scale for COVID-19 retrospective studies using Electronic Health Record data.

Objectives: Although the World Health Organization (WHO) Clinical Progression Scale for COVID-19 is useful in prospective clinical trials, it cannot be effectively used with retrospective Electronic Health Record (EHR) datasets. Modifying the existing WHO Clinical Progression Scale, we developed an ordinal severity scale (OS) and assessed its usefulness in the analyses of COVID-19 patient outcomes using retrospective EHR data. Materials and Methods: An OS was developed to assign COVID-19 disease severity using the Observational Medical Outcomes Partnership common data model within the National COVID Cohort Collaborative (N3C) data enclave. We then evaluated usefulness of the developed OS using heterogenous EHR data from January 2020 to October 2021 submitted to N3C by 63 healthcare organizations across the United States. Principal component analysis (PCA) was employed to characterize changes in disease severity among patients during the 28-day period following COVID-19 diagnosis. Results: The data set used in this analysis consists of 2 880 456 patients. PCA of the day-to-day variation in OS levels over the totality of the 28-day period revealed contrasting patterns of variation in disease severity within the first and second 14 days and illustrated the importance of evaluation over the full 28-day period. Discussion: An OS with well-defined, robust features, based on discrete EHR data elements, is useful for assessments of COVID-19 patient outcomes, providing insights on the progression of COVID-19 disease severity over time. Conclusions: The OS provides a framework that can facilitate better understanding of the course of acute COVID-19, informing clinical decision-making and resource allocation.

Emergency medical services shock index is the most accurate predictor of patient outcomes after blunt torso trauma.

INTRODUCTION: Shock index (SI) and delta shock index (∆SI) predict mortality and blood transfusion in trauma patients. This study aimed to evaluate the predictive ability of SI and ∆SI in a rural environment with prolonged transport times and transfers from critical access hospitals or level IV trauma centers. METHODS: We completed a retrospective database review at an American College of Surgeons verified level 1 trauma center for 2 years. Adult subjects analyzed sustained torso trauma. Subjects with missing data or severe head trauma were excluded. For analysis, poisson regression and binomial logistic regression were used to study the effect of time in transport and SI/∆SI on resource utilization and outcomes. p < 0.05 was considered significant. RESULTS: Complete data were available on 549 scene patients and 127 transfers. Mean Injury Severity Score was 11 (interquartile range, 9.0) for scene and 13 (interquartile range, 6.5) for transfers. Initial emergency medical services SI was the most significant predictor for blood transfusion and intensive care unit care in both scene and transferred patients (p < 0.0001) compared with trauma center arrival SI or transferring center SI. A negative ∆SI was significantly associated with the need for transfusion and the number of units transfused. Longer transport time also had a significant relationship with increasing intensive care unit length of stay. Cohorts were analyzed separately. CONCLUSION: Providers must maintain a high level of clinical suspicion for patients who had an initially elevated SI. Emergency medical services SI was the greatest predictor of injury and need for resources. Enroute SI and ∆SI were less predictive as time from injury increased. This highlights the improvements in en route care but does not eliminate the need for high-level trauma intervention. LEVEL OF EVIDENCE: Therapeutic/care management, level IV.

Predicting increases in COVID-19 incidence to identify locations for targeted testing in West Virginia: A machine learning enhanced approach.

During the COVID-19 pandemic, West Virginia developed an aggressive SARS-CoV-2 testing strategy which included utilizing pop-up mobile testing in locations anticipated to have near-term increases in SARS-CoV-2 infections. This study describes and compares two methods for predicting near-term SARS-CoV-2 incidence in West Virginia counties. The first method, Rt Only, is solely based on producing forecasts for each county using the daily instantaneous reproductive numbers, Rt. The second method, ML+Rt, is a machine learning approach that uses a Long Short-Term Memory network to predict the near-term number of cases for each county using epidemiological statistics such as Rt, county population information, and time series trends including information on major holidays, as well as leveraging statewide COVID-19 trends across counties and county population size. Both approaches used daily county-level SARS-CoV-2 incidence data provided by the West Virginia Department Health and Human Resources beginning April 2020. The methods are compared on the accuracy of near-term SARS-CoV-2 increases predictions by county over 17 weeks from January 1, 2021- April 30, 2021. Both methods performed well (correlation between forecasted number of cases and the actual number of cases week over week is 0.872 for the ML+Rt method and 0.867 for the Rt Only method) but differ in performance at various time points. Over the 17-week assessment period, the ML+Rt method outperforms the Rt Only method in identifying larger spikes. Results show that both methods perform adequately in both rural and non-rural predictions. Finally, a detailed discussion on practical issues regarding implementing forecasting models for public health action based on Rt is provided, and the potential for further development of machine learning methods that are enhanced by Rt.

Predicting increases in COVID-19 incidence to identify locations for targeted testing in West Virginia: A machine learning enhanced approach.

During the COVID-19 pandemic, West Virginia developed an aggressive SARS-CoV-2 testing strategy which included utilizing pop-up mobile testing in locations anticipated to have near-term increases in SARS-CXoV-2 infections. In this study, we describe and compare two methods for predicting near-term SARS-CoV-2 incidence in West Virginia counties. The first method, R t Only, is solely based on producing forecasts for each county using the daily instantaneous reproductive numbers, R t. The second method, ML+ R t , is a machine learning approach that uses a Long Short-Term Memory network to predict the near-term number of cases for each county using epidemiological statistics such as Rt, county population information, and time series trends including information on major holidays, as well as leveraging statewide COVID-19 trends across counties and county population size. Both approaches used daily county-level SARS-CoV-2 incidence data provided by the West Virginia Department Health and Human Resources beginning April 2020. The methods are compared on the accuracy of near-term SARS-CoV-2 increases predictions by county over 17 weeks from January 1, 2021-April 30, 2021. Both methods performed well (correlation between forecasted number of cases and the actual number of cases week over week is 0.872 for the ML+R t method and 0.867 for the R t Only method) but differ in performance at various time points. Over the 17-week assessment period, the ML+R t method outperforms the R t Only method in identifying larger spikes. We also find that both methods perform adequately in both rural and non-rural predictions. Finally, we provide a detailed discussion on practical issues regarding implementing forecasting models for public health action based on R t , and the potential for further development of machine learning methods that are enhanced by R t.

Forensic intelligence: Data analytics as the bridge between forensic science and investigation.

Scientists should not play a role in investigations nor should investigators play a role in the scientific analyses. One way to bridge the relationship between the forensic scientist and the police investigator is through an Intelligence Analyst (IA) who is part of the forensic services operation. The IA offers the ability to walk between the role of scientist and law enforcement, receiving data after completion of scientific analyses and translating the information into actionable intelligence. The additional bridging and translating services represent a paradigm shift with increased emphasis on investigative contributions from forensic analysis. Forensic intelligence incorporates forensic data early in an investigation in a holistic case approach that incorporates possible datasets and information that could be relevant to the investigation. We present a brief review of the value added when an IA provides the bridge between the forensic laboratory and police investigators to enhance the use of forensic evidence.