Surgical scheduling via optimization and machine learning with long-tailed data : Health care management science, in press.

Using data from cardiovascular surgery patients with long and highly variable post-surgical lengths of stay (LOS), we develop a modeling framework to reduce recovery unit congestion. We estimate the LOS and its probability distribution using machine learning models, schedule procedures on a rolling basis using a variety of optimization models, and estimate performance with simulation. The machine learning models achieved only modest LOS prediction accuracy, despite access to a very rich set of patient characteristics. Compared to the current paper-based system used in the hospital, most optimization models failed to reduce congestion without increasing wait times for surgery. A conservative stochastic optimization with sufficient sampling to capture the long tail of the LOS distribution outperformed the current manual process and other stochastic and robust optimization approaches. These results highlight the perils of using oversimplified distributional models of LOS for scheduling procedures and the importance of using optimization methods well-suited to dealing with long-tailed behavior.

Increased dominance of heat-tolerant symbionts creates resilient coral reefs in near-term ocean warming.

Climate change is radically altering coral reef ecosystems, mainly through increasingly frequent and severe bleaching events. Yet, some reefs have exhibited higher thermal tolerance after bleaching severely the first time. To understand changes in thermal tolerance in the eastern tropical Pacific (ETP), we compiled four decades of temperature, coral cover, coral bleaching, and mortality data, including three mass bleaching events during the 1982 to 1983, 1997 to 1998 and 2015 to 2016 El Niño heatwaves. Higher heat resistance in later bleaching events was detected in the dominant framework-building genus, Pocillopora, while other coral taxa exhibited similar susceptibility across events. Genetic analyses of Pocillopora spp. colonies and their algal symbionts (2014 to 2016) revealed that one of two Pocillopora lineages present in the region (Pocillopora "type 1") increased its association with thermotolerant algal symbionts (Durusdinium glynnii) during the 2015 to 2016 heat stress event. This lineage experienced lower bleaching and mortality compared with Pocillopora "type 3", which did not acquire D. glynnii. Under projected thermal stress, ETP reefs may be able to preserve high coral cover through the 2060s or later, mainly composed of Pocillopora colonies that associate with D. glynnii. However, although the low-diversity, high-cover reefs of the ETP could illustrate a potential functional state for some future reefs, this state may only be temporary unless global greenhouse gas emissions and resultant global warming are curtailed.

Cardiovascular Imaging for Systemic Sclerosis Monitoring and Management.

Systemic sclerosis (SSc) is a complex connective tissue disease with multiple clinical and subclinical cardiac manifestations. SSc can affect most structural components of the heart, including the pericardium, myocardium, valves, and conduction system through a damaging cycle of inflammation, ischemia, and fibrosis. While cardiac involvement is the second leading SSc-related cause of death, it is frequently clinically silent in early disease and often missed with routine screening. To facilitate identification of cardiac disease in this susceptible population, we present here a review of cardiac imaging modalities and potential uses in the SSc patient population. We describe well-characterized techniques including electrocardiography and 2D echocardiography with Doppler, but also discuss more advanced imaging approaches, such as speckle-tracking echocardiography, cardiovascular magnetic resonance imaging (CMR), and stress imaging, among others. We also suggest an algorithm for the appropriate application of these modalities in the workup and management of patients with SSc. Finally, we discuss future opportunities for cardiac imaging in SSc research to achieve early detection and to optimize treatment.

Validating Coding for the Identification of Pediatric Treatment Resistant Epilepsy Patients.

The International Classification of Diseases (ICD) system includes sub codes to indicate that an individual with epilepsy is treatment resistant. These codes would be a valuable tool to identify individuals for quality improvement and population health, as well as for recruitment into clinical trials. However, the accuracy of these codes is unclear. We performed a single center cross sectional study to understand the accuracy of ICD codes for treatment resistant epilepsy. We identified 344 individuals, roughly half with treatment resistant epilepsy The ICD code had a sensitivity of 90% (147 of 164) and specificity of 86% (155 of 180). The miscoding of children with refractory epilepsy was attributed to the following reasons: 5 patients had epilepsy surgery, 4 had absence epilepsy, 4 patients were seen by different providers, and 1 patient was most recently seen in movement disorders clinic. ICD codes accurately identify children with treatment resistant epilepsy.

On incorporating forecasts into linear state space model Markov decision processes.

Weather forecast information will very likely find increasing application in the control of future energy systems. In this paper, we introduce an augmented state space model formulation with linear dynamics, within which one can incorporate forecast information that is dynamically revealed alongside the evolution of the underlying state variable. We use the martingale model for forecast evolution (MMFE) to enforce the necessary consistency properties that must govern the joint evolution of forecasts with the underlying state. The formulation also generates jointly Markovian dynamics that give rise to Markov decision processes (MDPs) that remain computationally tractable. This paper is the first to enforce MMFE consistency requirements within an MDP formulation that preserves tractability. This article is part of the theme issue 'The mathematics of energy systems'.

Geographic Variation in Trends and Disparities in Heart Failure Mortality in the United States, 1999 to 2017.

Background Cardiovascular disease mortality related to heart failure (HF) is rising in the United States. It is unknown whether trends in HF mortality are consistent across geographic areas and are associated with state-level variation in cardiovascular health (CVH). The goal of the present study was to assess regional and state-level trends in cardiovascular disease mortality related to HF and their association with variation in state-level CVH. Methods and Results Age-adjusted mortality rates (AAMR) per 100 000 attributable to HF were ascertained using the Centers for Disease Control and Prevention's Wide-Ranging Online Data for Epidemiologic Research from 1999 to 2017. CVH at the state-level was quantified using the Behavioral Risk Factor Surveillance System. Linear regression was used to assess temporal trends in HF AAMR were examined by census region and state and to examine the association between state-level CVH and HF AAMR. AAMR attributable to HF declined from 1999 to 2011 and increased between 2011 and 2017 across all census regions. Annual increases after 2011 were greatest in the Midwest (ß=1.14 [95% CI, 0.75, 1.53]) and South (ß=0.96 [0.66, 1.26]). States in the South and Midwest consistently had the highest HF AAMR in all time periods, with Mississippi having the highest AAMR (109.6 [104.5, 114.6] in 2017). Within race‒sex groups, consistent geographic patterns were observed. The variability in HF AAMR was associated with state-level CVH (P<0.001). Conclusions Wide geographic variation exists in HF mortality, with the highest rates and greatest recent increases observed in the South and Midwest. Higher levels of poor CVH in these states suggest the potential for interventions to promote CVH and reduce the burden of HF.

The development and deployment of a model for hospital-level COVID-19 associated patient demand intervals from consistent estimators (DICE).

Hospitals commonly project demand for their services by combining their historical share of regional demand with forecasts of total regional demand. Hospital-specific forecasts of demand that provide prediction intervals, rather than point estimates, may facilitate better managerial decisions, especially when demand overage and underage are associated with high, asymmetric costs. Regional point forecasts of patient demand are commonly available, e.g., for the number of people requiring hospitalization due to an epidemic such as COVID-19. However, even in this common setting, no probabilistic, consistent, computationally tractable forecast is available for the fraction of patients in a region that a particular institution should expect. We introduce such a forecast, DICE (Demand Intervals from Consistent Estimators). We describe its development and deployment at an academic medical center in California during the 'second wave' of COVID-19 in the Unite States. We show that DICE is consistent under mild assumptions and suitable for use with perfect, biased and unbiased regional forecasts. We evaluate its performance on empirical data from a large academic medical center as well as on synthetic data.

Heart Failure Risk Distribution and Trends in the United States Population, NHANES 1999-2016.

BACKGROUND: Implementation of effective preventive interventions requires identification of high-risk individuals. We sought to define the distribution and trends of heart failure risk in the US population. METHODS: We calculated 10-year predicted heart failure risk among a representative sample of US adults aged 30-79 years, without baseline cardiovascular disease, from the National Health and Nutrition Examination Surveys (NHANES) 1999-2016. We used the published Pooled Cohort Equations to Prevent Heart Failure (PCP-HF) model, which integrates demographic and risk factor data, to estimate 10-year heart failure risk. Participants were stratified by NHANES cycle, sex, age, and race/ethnicity and by 10-year heart failure risk, defined as low (<1%), intermediate (1% to <5%), and high (≥5%). RESULTS: From 1999-2000 to 2015-2016, mean predicted 10-year heart failure risk increased significantly from 2.0% to 3.0% (P < .05) in the population, most notably among non-Hispanic black (2.1% to 3.7%) and non-Hispanic white (2.4% to 3.6%) men. In 2013-2016, 17.6% of the studied population was at high predicted 10-year heart failure risk. The prevalence of high predicted heart failure risk was highest among non-Hispanic black men (23.1%), followed by non-Hispanic white men (19.2%) and non-Hispanic white women (17.9%). DISCUSSION: Mean population risk of heart failure increased significantly from 1999-2016. A substantial proportion of US adults are at high 10-year heart failure risk (≥5%), particularly non-Hispanic black men. These data underscore the importance of identifiying individuals at increased heart failure risk for targeted prevention measures to reduce the future burden of heart failure.

Evaluation of a Care Management Program for Pediatric Epilepsy Patients.

OBJECTIVE: To evaluate the impact of a pediatric epilepsy care management intervention on emergency department visits, hospitalizations, and seizure freedom. METHODS: We conducted a prospective observational study at a single academic medical center. Children with epilepsy with high risk of frequent emergency department use were enrolled in the intervention from January through May 2015, which included a baseline visit and follow-up support from a care management team. Controls selected from the same institution received standard of care. Baseline and follow-up information were collected from electronic health records and surveys (Family Impact Scale, Pediatric Epilepsy Medication Self-Management Questionnaire). Propensity score-weighted logistic regression compared emergency department visits, unplanned hospitalizations, and 3-month seizure freedom after 1 year in the intervention vs control groups. RESULTS: A total of 56 children were enrolled in the intervention and 359 received standard of care. The intervention group was younger and had greater use of health services at baseline. When comparing the intervention to standard of care after 1 year, we found no significant difference in the risk of any emergency department visit (adjusted odds ratio [OR] 2.2, 95% confidence interval [CI] 0.6-8.5) or seizure freedom (adjusted OR 2.5, 95% CI 0.3-21.5). However, the risk of unplanned hospital admissions remained higher in the intervention group (adjusted OR 23.1, 95% CI 5.1-104). CONCLUSION: We did not find that children with epilepsy who received a care management intervention had less use of health services or better clinical outcomes after a year compared with controls. The study is limited by small sample size and nonrandomized study design.

Population dynamics of the reef crisis: Consequences of the growing human population.

An unequivocal link exists between human population density and environmental degradation, both in the near field (local impacts) and far field (impacts due to teleconnections). Human population is most widely predicted to reach 9-11 billion by 2100, when the demographic transition is expected in all but a handful of countries. Strongest population growth is in the tropics, where coral reefs face dense human population and concomitant heavy usage. In most countries, >50% will be urbanized but growth of rural population and need for food in urban centres will not alleviate pressure on reef resources. Aquaculture will alleviate some fishing pressure, but still utilizes reef surface and is also destructive. Denser coastal populations and greater wealth will lead to reef degradation by coastal construction. Denser populations inland will lead to more runoff and siltation. Effects of human perturbations can be explored with metapopulation theory since they translate to increases in patch-mortality and decreases in patch-colonization (=regeneration). All such changes will result in a habitat with overall fewer settled patches, so fewer live reefs. If rescue effects are included, bifurcations in system dynamics will allow for many empty patches and, depending on system state relative to stable and unstable equilibria, a part-empty system may either trend towards stability at higher patch occupancy or extinction. Thus, unless the disturbance history is known, it may be difficult to assess the direction of system trajectory-making management difficult. If habitat is decreased by destruction, rescue effects become even more important as extinction-debt, accumulated by efficient competitors with weaker dispersal ability, is realized. Easily visible trends in human population dynamics combined with well-established and tested ecological theory give a clear, intuitive, yet quantifiable guide to the severity of survival challenges faced by coral reefs. Management challenges and required actions can be clearly shown and, contrary to frequent claims, no scientific ambiguity exists with regards to the serious threat posed to coral reefs by humankind's continued numerical increase.

A tropical eastern Pacific invasive brittle star species (Echinodermata: Ophiuroidea) reaches southeastern Florida.

The invasive brittle star Ophiothela mirabilis (family Ophiotrichidae), a tropical Indo-Pacific endemic species, first reported in Atlantic waters off southern Brazil in 2000, has extended its range northward to the Caribbean Sea, to the Lesser Antilles in 2011, and was first reported in south Florida in January 2019. Its occurrence in southeast Florida extends along nearly 70km of coastline, from near the Port of Miami, Miami-Dade County, northward to Deerfield Beach, Broward County. It occurs abundantly as an epizoite on octocorals, attaining population densities of 25 individuals and more per 10-cm long octocoral stem. The surface texture of octocoral hosts (rough, smooth) did not affect the densities of the ophiuroid epizoites, and there were significantly greater abundances on octocorals during two winter sampling periods than in the summer. Beige and orange-coloured morphs are sometimes present on the same octocoral stem. Gut content analysis supported a suspension feeding mode, revealing essentially identical ingested items in both colour morphs with a preponderance of amorphous detritus and filamentous algae. Molecular genetic evidence (COI & 16s) has established the identity of O. mirabilis and its relationship to invasive Brazilian populations. The orange and beige morphs form two distinct, but closely related lineages that may represent two separate introductions. The orange morph shares haplotypes with Brazilian and Caribbean specimens suggesting a further range expansion of the 'original' invasion. The beige morph, however, shares haplotypes with specimens from the Mexican Pacific and Peru and potentially represents a secondary introduction. Traits promoting dispersal and establishment of this species in new habitats are manifold: vagility and ability to cling tightly to diverse host taxa (e.g. sponges, cnidarians, bryozoans, and echinoderms), frequent asexual reproduction (fissiparity), suspension feeding, including a wide range of dietary items, possession of integument-covered ossicles and arm spines offering protection from predators, and an effective competitive edge over associated microbiota for substrate space.

Predictive Accuracy of Heart Failure-Specific Risk Equations in an Electronic Health Record-Based Cohort.

BACKGROUND: Guidelines recommend identification of individuals at risk for heart failure (HF). However, implementation of risk-based prevention strategies requires validation of HF-specific risk scores in diverse, real-world cohorts. Therefore, our objective was to assess the predictive accuracy of the Pooled Cohort Equations to Prevent HF within a primary prevention cohort derived from the electronic health record. METHODS: We retrospectively identified patients between the ages of 30 to 79 years in a multi-center integrated healthcare system, free of cardiovascular disease, with available data on HF risk factors, and at least 5 years of follow-up. We applied the Pooled Cohort Equations to Prevent HF tool to calculate sex and race-specific 5-year HF risk estimates. Incident HF was defined by the International Classification of Diseases codes. We assessed model discrimination and calibration, comparing predicted and observed rates for incident HF. RESULTS: Among 31 256 eligible adults, mean age was 51.4 years, 57% were women and 11% Black. Incident HF occurred in 568 patients (1.8%) over 5-year follow-up. The modified Pooled Cohort Equations to Prevent HF model for 5-year risk prediction of HF had excellent discrimination in White men (C-statistic 0.82 [95% CI, 0.79-0.86]) and women (0.82 [0.78-0.87]) and adequate discrimination in Black men (0.69 [0.60-0.78]) and women (0.69 [0.52-0.76]). Calibration was fair in all race-sex subgroups (χ2<20). CONCLUSIONS: A novel sex- and race-specific risk score predicts incident HF in a real-world, electronic health record-based cohort. Integration of HF risk into the electronic health record may allow for risk-based discussion, enhanced surveillance, and targeted preventive interventions to reduce the public health burden of HF.

Evaluation of an Activity Tracker to Detect Seizures Using Machine Learning.

Currently, the tracking of seizures is highly subjective, dependent on qualitative information provided by the patient and family instead of quantifiable seizure data. Usage of a seizure detection device to potentially detect seizure events in a population of epilepsy patients has been previously done. Therefore, we chose the Fitbit Charge 2 smart watch to determine if it could detect seizure events in patients when compared to continuous electroencephalographic (EEG) monitoring for those admitted to an epilepsy monitoring unit. A total of 40 patients were enrolled in the study that met the criteria between 2015 and 2016. All seizure types were recorded. Twelve patients had a total of 53 epileptic seizures. The patient-aggregated receiver operating characteristic curve had an area under the curve of 0.58 [0.56, 0.60], indicating that the neural network models were generally able to detect seizure events at an above-chance level. However, the overall low specificity implied a false alarm rate that would likely make the model unsuitable in practice. Overall, the use of the Fitbit Charge 2 activity tracker does not appear well suited in its current form to detect epileptic seizures in patients with seizure activity when compared to data recorded from the continuous EEG.

Measuring the impact of epilepsy on families.

OBJECTIVE: Caring for a child with illness or a child with disability impacts family in various ways. The ability to assess the impact of this care on families is one way to proactively provide the necessary support and resources for impacted families. Accordingly, the goal of the current study was to assess the impact of pediatric epilepsy on individual families in a comprehensive epilepsy clinic using a slightly modified version of the Impact on Families Scale (IFS). METHODS: Families of patients with epilepsy completed the IFS up to three times. The IFS score and the six categories (i.e., total impact, financial impact, general impact, family/social impact, coping, and sibling impact) were assessed using Student's two sample t-test to determine the differences between binary groups and Pearson's correlation to assess the associations with continuous variables. Linear regression modeling was used to develop a model to predict IFS score. RESULTS: Three hundred and forty-one patients completed the scale at one time point, 314 at two time points, and 61 at three time points. The overall impact of epilepsy on families was 109 (95% confidence interval (CI): 106-112) at time point 1, 111 (95% CI: 108-114) at time point 2, and 112 (95% CI: 105-119) at time point 3. There was no statistical difference in IFS score among the three time points. There were no associations with age or gender. Multivariable modeling using stepwise regression indicated that treatment resistance and seizure-free status were associated with IFS score. No interaction effects were identified. CONCLUSIONS: Findings from the current study suggest that the impact of epilepsy is highest for families that have children with active seizures at the time of their clinical visit and for those with children having treatment-resistant epilepsy. Although intuitive, this is the first study, to our knowledge, that has empirically verified these findings.

Contributions of the UK biobank high impact papers in the era of precision medicine.

To review the highest impact studies published from the UK Biobank and assess their contributions to "precision medicine." We reviewed 140 of 689 studies published between 2008 and May 2019 from the UK Biobank deemed to be high impact by citations, alternative metric data, or publication in a high impact journal. We classified studies according to whether they (1) were largely methods papers, (2) largely replicated prior findings or associations, (3) generated novel findings or associations, (4) developed risk prediction models that did not yield clinically significant improvements in risk estimation over prior models or (5) developed models that produced significant improvements in individualized risk assessment, targeted screening, or targeted treatment. This final category represents "precision medicine." We classified 15 articles as category 1, 33 as category 2, 85 as category 3, six as category 4, and one as category 5. In this assessment of the first 7 years of the UK Biobank and first 4 years of genetic data availability, the majority of high impact UK Biobank studies either replicated known associations or generated novel associations without clinically relevant improvements in risk prediction, screening, or treatment. This information may be useful for designers of other cohort studies in terms of input to design and follow-up to facilitate precision medicine research.

Some environmental and biological determinants of coral richness, resilience and reef building in Galápagos (Ecuador).

Throughout the Galápagos, differences in coral reef development and coral population dynamics were evaluated by monitoring populations from 2000-2019, and environmental parameters (sea temperatures, pH, NO3-, PO43-) from 2015-19. The chief goal was to explain apparent coral community differences between the northern (Darwin and Wolf) and southern (Sta. Cruz, Fernandina, San Cristóbal, Española, Isabela) islands. Site coral species richness was highest at Darwin and Wolf. In the three most common coral taxa, a declining North (N)-South (S) trend in colony sizes existed for Porites lobata and Pocillopora spp., but not for Pavona  spp. Frequent coral recruitment was observed in all areas. Algal competition was highest at Darwin, but competition by bioeroding sea urchins and burrowing fauna (polychaete worms, bivalve mollusks) increased from N to S with declining coral skeletal density. A biophysical model suggested strong connectivity among southern islands with weaker connectivity to Wolf and even less to Darwin. Also, strong connectivity was observed between Darwin and Wolf, but from there only intermittently to the south. From prevailing ocean current trajectories, coral larvae from Darwin and Wolf drift primarily towards Malpelo and Cocos Islands, some reaching Costa Rica and Colombia. Mean temperature, pH, and PO43- declined from N to S. Strong thermocline shoaling, especially in the warm season, was observed at most sites. A single environmental factor could not explain the variability in observed coral community characteristics, with minimum temperature, pH and nutrient levels the strongest determinants. Thus, complex environmental determinants combined with larval connectivity patterns may explain why the northern Galápagos Islands (Darwin, Wolf) have higher coral richness and cover and also recover more rapidly than central/southern islands after region-wide disturbances. These northern islands are therefore potentially of critical conservation importance as important reservoirs of regional coral biodiversity and source of larvae.

Seizure Action Plans for Pediatric Patients With Epilepsy: A Randomized Controlled Trial.

OBJECTIVES: Seizure action plans help patients and caregivers better self-manage their epilepsy. We hypothesized that providing pediatric patients and their caregivers with a seizure action plan would reduce unplanned health care utilization and decrease the impact of epilepsy. METHODS: We developed a seizure action plan for use in pediatric epilepsy patients. A prospective cohort was randomly assigned to receive a seizure action plan in addition to standard epilepsy care or to standard epilepsy care alone. All caregivers were surveyed using the Modified Impact on Families (MIF) questionnaire at enrollment, 3 months, and 12 months. Health care utilization measures and Modified Impact on Families questionnaire scores were compared between the 2 groups. RESULTS: Fifty-four patients received a seizure action plan and standard care, whereas 48 received standard care alone. The groups had similar demographics. There was a significantly higher proportion of overall clinic appointment no shows in the standard care group vs the seizure action plan group (P = .04); however, other significant differences in health care utilization were not found. Among patients with low seizure frequency (12 or fewer seizures per year), Seizure comfort scores on the Modified Impact on Families questionnaire were significantly higher at 12 months among the seizure action plan group compared to the standard care group. SIGNIFICANCE: Caregivers for patients with epilepsy receiving a seizure action plan were more comfortable regarding seizure care and missed fewer appointments. However, differences in health care utilization were not present. The seizure action plan appears to have more impact in patients who experience lower seizure frequencies. Further studies evaluating the impact as well as assessing caregivers' perceptions of the seizure action plan using a larger sample are needed.

Safety and Efficacy of Supratherapeutic Doses of Clobazam.

Clobazam is a commonly used long-acting benzodiazepine approved by the US Food and Drug Administration (FDA) to treat seizures associated with Lennox Gastaut syndrome. The FDA approved maximum dosage of clobazam is 1 mg/kg/d or a total of 40 mg a day. Many providers exceed this dosage but there is limited data on the safety, tolerability, and efficacy of supratherapeutic doses. We reviewed retrospective data at our institution and compared patients on supratherapeutic doses to patients on therapeutic doses. A total of 133 patients met inclusion criteria (65 supratherapeutic, 67 therapeutic). There was no statistically significant difference in terms of seizure control, health care utilization, or side effects between patients on supratherapeutic doses and those on therapeutic doses. This study lends further support to the safety and tolerability of supratherapuetic doses of clobazam.