The Potential For Bias In Machine Learning And Opportunities For Health Insurers To Address It.

As the use of machine learning algorithms in health care continues to expand, there are growing concerns about equity, fairness, and bias in the ways in which machine learning models are developed and used in clinical and business decisions. We present a guide to the data ecosystem used by health insurers to highlight where bias can arise along machine learning pipelines. We suggest mechanisms for identifying and dealing with bias and discuss challenges and opportunities to increase fairness through analytics in the health insurance industry.

Telemedicine catches on: changes in the utilization of telemedicine services during the COVID-19 pandemic.

OBJECTIVES: To determine the degree of telemedicine expansion overall and across patient subpopulations and diagnoses. We hypothesized that telemedicine visits would increase substantially due to the need for continuity of care despite the disruptive effects of COVID-19. STUDY DESIGN: A retrospective study of health insurance claims for telemedicine visits from January 1, 2018, through March 10, 2020 (prepandemic period), and March 11, 2020, through October 31, 2020 (pandemic period). METHODS: We analyzed claims from 1,589,777 telemedicine visits that were submitted to Independence Blue Cross (Independence) from telemedicine-only providers and providers who traditionally deliver care in person. The primary exposure was the combination of individual behavior changes, state stay-at-home orders, and the Independence expansion of billing policies for telemedicine. The comparison population consisted of telemedicine visits in the prepandemic period. RESULTS: Telemedicine increased rapidly from a mean (SD) of 773 (155) weekly visits in prepandemic 2020 to 45,632 (19,937) weekly visits in the pandemic period. During the pandemic period, a greater proportion of telemedicine users were older, had Medicare Advantage insurance plans, had existing chronic conditions, or resided in predominantly non-Hispanic Black or African American Census tracts compared with during the prepandemic period. A significant increase in telemedicine claims containing a mental health-related diagnosis was observed. CONCLUSIONS: Telemedicine expanded rapidly during the COVID-19 pandemic across a broad range of clinical conditions and demographics. Although levels declined later in 2020, telemedicine utilization remained markedly higher than 2019 and 2018 levels. Trends suggest that telemedicine will likely play a key role in postpandemic care delivery.

The Lyme Disease Biobank: Characterization of 550 Patient and Control Samples from the East Coast and Upper Midwest of the United States.

Lyme disease (LD) is an increasing public health problem. Current laboratory testing is insensitive in early infection, the stage at which appropriate treatment is most effective in preventing disease sequelae. The Lyme Disease Biobank (LDB) collects samples from individuals with symptoms consistent with early LD presenting with or without erythema migrans (EM) or an annular, expanding skin lesion and uninfected individuals from areas of endemicity. Samples were collected from 550 participants (298 cases and 252 controls) according to institutional review board-approved protocols and shipped to a centralized biorepository. Testing was performed to confirm the presence of tick-borne pathogens by real-time PCR, and a subset of samples was tested for Borrelia burgdorferi by culture. Serology was performed on all samples using the CDC's standard two-tiered testing algorithm (STTTA) for LD. LD diagnosis was supported by laboratory testing in 82 cases, including positive results by use of the STTTA, PCR, or culture or positive results by two enzyme-linked immunosorbent assays for cases presenting with EM lesion sizes of >5 cm. The remaining 216 cases had negative laboratory testing results. For the controls, 43 were positive by at least one of the tiers and 6 were positive by use of the STTTA. The results obtained with this collection highlight and reinforce the known limitations of serologic testing in early LD, with only 29% of individuals presenting with EM lesion sizes of >5 cm yielding a positive result using the STTTA. Aliquots of whole blood, serum, and urine from clinically characterized patients with and without LD are available to investigators in academia and industry for evaluation or development of novel diagnostic assays for LD, to continue to improve upon currently available methods.

How Individual Variation in Host Tolerance Affects Competence to Transmit Parasites.

Tolerance, or the maintenance of host health or fitness at a given parasite burden, has often been studied in evolutionary and medical contexts, particularly with respect to effects on the evolution of parasite virulence and individual patient outcomes. These bodies of work have provided insight about tolerance for evolutionary phenomena (e.g., virulence) and individual health (e.g., recovering from an infection). However, due to the specific motivations of that work, few studies have considered the ecological ramifications of variation in tolerance, namely, how variation in forms of tolerance could mediate parasite movement through populations and even community-level disease dynamics. Tolerance is most commonly regarded as the relationship between host fitness and parasite burden. However, few if any studies have actually quantified host fitness, instead utilizing proxies of fitness as the response variables to be regressed against parasite burden. Here, we address how attention to the effects of parasite burden on traits that are relevant to host competence (i.e., the ability to amplify parasites to levels transmissible to other hosts/vectors) will enhance our understanding of disease dynamics in nature. We also provide several forms of guidance for how to overcome the challenges of quantifying tolerance in wild organisms.

Sharing an environment with sick conspecifics alters odors of healthy animals.

Body odors change with health status and the odors of sick animals can induce avoidance behaviors in healthy conspecifics. Exposure to sickness odors might also alter the physiology of healthy conspecifics and modify the odors they produce. We hypothesized that exposure to odors of sick (but non-infectious) animals would alter the odors of healthy cagemates. To induce sickness, we injected mice with a bacterial endotoxin, lipopolysaccharide. We used behavioral odor discrimination assays and analytical chemistry techniques followed by predictive classification modeling to ask about differences in volatile odorants produced by two types of healthy mice: those cohoused with healthy conspecifics and those cohoused with sick conspecifics. Mice trained in Y-maze behavioral assays to discriminate between the odors of healthy versus sick mice also discriminated between the odors of healthy mice cohoused with sick conspecifics and odors of healthy mice cohoused with healthy conspecifics. Chemical analyses paired with statistical modeling revealed a parallel phenomenon. Urine volatiles of healthy mice cohoused with sick partners were more likely to be classified as those of sick rather than healthy mice based on discriminant model predictions. Sickness-related odors could have cascading effects on neuroendocrine or immune responses of healthy conspecifics, and could affect individual behaviors, social dynamics, and pathogen spread.

Parasite Tolerance and Host Competence in Avian Host Defense to West Nile Virus.

Competence, or the propensity of a host to transmit parasites, is partly underlain by host strategies to cope with infection (e.g., resistance and tolerance). Resistance represents the ability of hosts to prevent or clear infections, whereas tolerance captures the ability of individuals to cope with a given parasite burden. Here, we investigated (1) whether one easy-to-measure form of tolerance described well the dynamic relationships between host health and parasite burden, and (2) whether individual resistance and tolerance to West Nile virus (WNV) were predictable from single cytokine measures. We exposed house sparrows (HOSP) to WNV and measured subsequent changes in host performance, viral burden, and cytokine expression. We then used two novel approaches (one complex, one simpler) to estimate tolerance within-individual HOSP using four separate host performance traits. We lastly investigated changes in the expression of pro-inflammatory cytokine interferon-γ (IFN-γ) and anti-inflammatory cytokine interleukin-10 (IL-10). Both approaches to estimating tolerance were equivalent among WNV-infected HOSP; thus, an easy-to-measure tolerance estimation may be successfully applied in field studies. Constitutive expression of IFN-γ and IL-10 were predictive of resistance and tolerance to WNV, implicating these cytokines as viable biomarkers of host competence to WNV.

Phylogenetic patterns of trait and trait plasticity evolution: Insights from amphibian embryos.

Environmental variation favors the evolution of phenotypic plasticity. For many species, we understand the costs and benefits of different phenotypes, but we lack a broad understanding of how plastic traits evolve across large clades. Using identical experiments conducted across North America, we examined prey responses to predator cues. We quantified five life-history traits and the magnitude of their plasticity for 23 amphibian species/populations (spanning three families and five genera) when exposed to no cues, crushed-egg cues, and predatory crayfish cues. Embryonic responses varied considerably among species and phylogenetic signal was common among the traits, whereas phylogenetic signal was rare for trait plasticities. Among trait-evolution models, the Ornstein-Uhlenbeck (OU) model provided the best fit or was essentially tied with Brownian motion. Using the best fitting model, evolutionary rates for plasticities were higher than traits for three life-history traits and lower for two. These data suggest that the evolution of life-history traits in amphibian embryos is more constrained by a species' position in the phylogeny than is the evolution of life history plasticities. The fact that an OU model of trait evolution was often a good fit to patterns of trait variation may indicate adaptive optima for traits and their plasticities.

Stress hormones predict a host superspreader phenotype in the West Nile virus system.

Glucocorticoid stress hormones, such as corticosterone (CORT), have profound effects on the behaviour and physiology of organisms, and thus have the potential to alter host competence and the contributions of individuals to population- and community-level pathogen dynamics. For example, CORT could alter the rate of contacts among hosts, pathogens and vectors through its widespread effects on host metabolism and activity levels. CORT could also affect the intensity and duration of pathogen shedding and risk of host mortality during infection. We experimentally manipulated songbird CORT, asking how CORT affected behavioural and physiological responses to a standardized West Nile virus (WNV) challenge. Although all birds became infected after exposure to the virus, only birds with elevated CORT had viral loads at or above the infectious threshold. Moreover, though the rate of mortality was faster in birds with elevated CORT compared with controls, most hosts with elevated CORT survived past the day of peak infectiousness. CORT concentrations just prior to inoculation with WNV and anti-inflammatory cytokine concentrations following viral exposure were predictive of individual duration of infectiousness and the ability to maintain physical performance during infection (i.e. tolerance), revealing putative biomarkers of competence. Collectively, our results suggest that glucocorticoid stress hormones could directly and indirectly mediate the spread of pathogens.

Linking Ecology and Epidemiology to Understand Predictors of Multi-Host Responses to an Emerging Pathogen, the Amphibian Chytrid Fungus.

Variation in host responses to pathogens can have cascading effects on populations and communities when some individuals or groups of individuals display disproportionate vulnerability to infection or differ in their competence to transmit infection. The fungal pathogen, Batrachochytrium dendrobatidis (Bd) has been detected in almost 700 different amphibian species and is implicated in numerous global amphibian population declines. Identifying key hosts in the amphibian-Bd system-those who are at greatest risk or who pose the greatest risk for others-is challenging due in part to many extrinsic environmental factors driving spatiotemporal Bd distribution and context-dependent host responses to Bd in the wild. One way to improve predictive risk models and generate testable mechanistic hypotheses about vulnerability is to complement what we know about the spatial epidemiology of Bd with data collected through comparative experimental studies. We used standardized pathogen challenges to quantify amphibian survival and infection trajectories across 20 post-metamorphic North American species raised from eggs. We then incorporated trait-based models to investigate the predictive power of phylogenetic history, habitat use, and ecological and life history traits in explaining responses to Bd. True frogs (Ranidae) displayed the lowest infection intensities, whereas toads (Bufonidae) generally displayed the greatest levels of mortality after Bd exposure. Affiliation with ephemeral aquatic habitat and breadth of habitat use were strong predictors of vulnerability to and intensity of infection and several other traits including body size, lifespan, age at sexual maturity, and geographic range also appeared in top models explaining host responses to Bd. Several of the species examined are highly understudied with respect to Bd such that this study represents the first experimental susceptibility data. Combining insights gained from experimental studies with observations of landscape-level disease prevalence may help explain current and predict future pathogen dynamics in the Bd system.

Host Competence: An Organismal Trait to Integrate Immunology and Epidemiology.

The new fields of ecological immunology and disease ecology have begun to merge, and the classic fields of immunology and epidemiology are beginning to blend with them. This merger is occurring because the integrative study of host-parasite interactions is providing insights into disease in ways that traditional methods have not. With the advent of new tools, mathematical and technological, we could be on the verge of developing a unified theory of infectious disease, one that supersedes the barriers of jargon and tradition. Here we argue that a cornerstone of any such synthesis will be host competence, the propensity of an individual host to generate new infections in other susceptible hosts. In the last few years, the emergence of systems immunology has led to novel insight into how hosts control or eliminate pathogens. Most such efforts have stopped short of considering transmission and the requisite behaviors of infected individuals that mediate it, and few have explicitly incorporated ecological and evolutionary principles. Ultimately though, we expect that the use of a systems immunology perspective will help link suborganismal processes (i.e., health of hosts and selection on genes) to superorganismal outcomes (i.e., community-level disease dynamics and host-parasite coevolution). Recently, physiological regulatory networks (PRNs) were cast as whole-organism regulatory systems that mediate homeostasis and hence link suborganismal processes with the fitness of individuals. Here, we use the PRN construct to develop a roadmap for studying host competence, taking guidance from systems immunology and evolutionary ecology research. We argue that PRN variation underlies heterogeneity in individual host competence and hence host-parasite dynamics.

Host stress hormones alter vector feeding preferences, success, and productivity.

Stress hormones might represent a key link between individual-level infection outcome, population-level parasite transmission, and zoonotic disease risk. Although the effects of stress on immunity are well known, stress hormones could also affect host-vector interactions via modification of host behaviours or vector-feeding patterns and subsequent reproductive success. Here, we experimentally manipulated songbird stress hormones and examined subsequent feeding preferences, feeding success, and productivity of mosquito vectors in addition to defensive behaviours of hosts. Despite being more defensive, birds with elevated stress hormone concentrations were approximately twice as likely to be fed on by mosquitoes compared to control birds. Moreover, stress hormones altered the relationship between the timing of laying and clutch size in blood-fed mosquitoes. Our results suggest that host stress could affect the transmission dynamics of vector-borne parasites via multiple pathways.

Differences in sensitivity to the fungal pathogen Batrachochytrium dendrobatidis among amphibian populations.

Contributing to the worldwide biodiversity crisis are emerging infectious diseases, which can lead to extirpations and extinctions of hosts. For example, the infectious fungal pathogen Batrachochytrium dendrobatidis (Bd) is associated with worldwide amphibian population declines and extinctions. Sensitivity to Bd varies with species, season, and life stage. However, there is little information on whether sensitivity to Bd differs among populations, which is essential for understanding Bd-infection dynamics and for formulating conservation strategies. We experimentally investigated intraspecific differences in host sensitivity to Bd across 10 populations of wood frogs (Lithobates sylvaticus) raised from eggs to metamorphosis. We exposed the post-metamorphic wood frogs to Bd and monitored survival for 30 days under controlled laboratory conditions. Populations differed in overall survival and mortality rate. Infection load also differed among populations but was not correlated with population differences in risk of mortality. Such population-level variation in sensitivity to Bd may result in reservoir populations that may be a source for the transmission of Bd to other sensitive populations or species. Alternatively, remnant populations that are less sensitive to Bd could serve as sources for recolonization after epidemic events.

The context of host competence: a role for plasticity in host-parasite dynamics.

Even apparently similar hosts can respond differently to the same parasites. Some individuals or specific groups of individuals disproportionately affect disease dynamics. Understanding the sources of among-host heterogeneity in the ability to transmit parasites would improve disease management. A major source of host variation might be phenotypic plasticity - the tendency for phenotypes to change across different environments. Plasticity might be as important as, or even more important than, genetic change, especially in light of human modifications of the environment, because it can occur on a more rapid timescale than evolution. We argue that variation in phenotypic plasticity among and within species strongly contributes to epidemiological dynamics when parasites are shared among multiple hosts, which is often the case.

Carotenoids and amphibians: effects on life history and susceptibility to the infectious pathogen, Batrachochytrium dendrobatidis.

Carotenoids are considered beneficial nutrients because they provide increased immune capacity. Although carotenoid research has been conducted in many vertebrates, little research has been done in amphibians, a group that is experiencing global population declines from numerous causes, including disease. We raised two amphibian species through metamorphosis on three carotenoid diets to quantify the effects on life-history traits and post-metamorphic susceptibility to a fungal pathogen (Batrachochytrium dendrobatidis; Bd). Increased carotenoids had no effect on survival to metamorphosis in gray treefrogs (Hyla versicolor) but caused lower survival to metamorphosis in wood frogs [Lithobates sylvaticus (Rana sylvatica)]. Increased carotenoids caused both species to experience slower development and growth. When exposed to Bd after metamorphosis, wood frogs experienced high mortality, and the carotenoid diets had no mitigating effects. Gray treefrogs were less susceptible to Bd, which prevented an assessment of whether carotenoids could mitigate the effects of Bd. Moreover, carotenoids had no effect on pathogen load. As one of only a few studies examining the effects of carotenoids on amphibians and the first to examine potential interactions with Bd, our results suggest that carotenoids do not always serve amphibians in the many positive ways that have become the paradigm in other vertebrates.

Larval exposure to predator cues alters immune function and response to a fungal pathogen in post-metamorphic wood frogs.

For the past several decades, amphibian populations have been decreasing around the globe at an unprecedented rate. Batrachochytrium dendrobatidis (Bd), the fungal pathogen that causes chytridiomycosis in amphibians, is contributing to amphibian declines. Natural and anthropogenic environmental factors are hypothesized to contribute to these declines by reducing the immunocompetence of amphibian hosts, making them more susceptible to infection. Antimicrobial peptides (AMPs) produced in the granular glands of a frog's skin are thought to be a key defense against Bd infection. These peptides may be a critical immune defense during metamorphosis because many acquired immune functions are suppressed during this time. To test if stressors alter AMP production and survival of frogs exposed to Bd, we exposed wood frog (Lithobates sylvaticus) tadpoles to the presence or absence of dragonfly predator cues crossed with a single exposure to three nominal concentrations of the insecticide malathion (0, 10, or 100 parts per billion [ppb]). We then exposed a subset of post-metamorphic frogs to the presence or absence of Bd zoospores and measured frog survival. Although predator cues and malathion had no effect on survival or size at metamorphosis, predator cues increased the time to metamorphosis by 1.5 days and caused a trend of a 20% decrease in hydrophobic skin peptides. Despite this decrease in peptides determined shortly after metamorphosis, previous exposure to predator cues increased survival in both Bd-exposed and unexposed frogs several weeks after metamorphosis. These results suggest that exposing tadpoles to predator cues confers fitness benefits later in life.

Experimental evidence for American bullfrog (Lithobates catesbeianus) susceptibility to chytrid fungus (Batrachochytrium dendrobatidis).

The emerging fungal pathogen, Batrachochytrium dendrobatidis (Bd), has been associated with global amphibian population declines and extinctions. American bullfrogs (Lithobates catesbeianus) are widely reported to be a tolerant host and a carrier of Bd that spreads the pathogen to less tolerant hosts. Here, we examined whether bullfrogs raised from eggs to metamorphosis in outdoor mesocosms were susceptible to Bd. We experimentally exposed metamorphic juveniles to Bd in the laboratory and compared mortality rates of pathogen-exposed animals to controls (non-exposed) in two separate experiments; one using a Bd strain isolated from a Western toad and another using a strain isolated from an American bullfrog. We wanted to examine whether metamorphic bullfrogs were susceptible to either of these strains. We show that bullfrogs were susceptible to one strain of Bd and not the other. In both experiments, infection load detected in the skin decreased over time, suggesting that metamorphic bullfrogs from some populations may be inefficient long-term carriers of Bd.

Host identity matters in the amphibian-Batrachochytrium dendrobatidis system: fine-scale patterns of variation in responses to a multi-host pathogen.

Species composition within ecological assemblages can drive disease dynamics including pathogen invasion, spread, and persistence. In multi-host pathogen systems, interspecific variation in responses to infection creates important context dependency when predicting the outcome of disease. Here, we examine the responses of three sympatric host species to a single fungal pathogen, Batrachochytrium dendrobatidis, which is associated with worldwide amphibian population declines and extinctions. Using an experimental approach, we show that amphibian species from three different genera display significant differences in patterns of pathgen-induced mortality as well as the magnitude and temporal dynamics of infection load. We exposed amphibians to one of four inoculation dose treatments at both larval and post- metamorphic stages and quantified infection load on day 8 and day 15 post-inoculation. Of the three species examined, only one (the Pacific treefrog; Pseudacris regilla) displayed "dose-dependent" responses; survival was reduced and infection load was elevated as inoculation dose was increased. We observed a reduction in survival but no differences in infection load across pathogen treatments in Cascades frogs (Rana cascadae). Western toads (Anaxyrus boreas) displayed differences in infection load but no differences in survival across pathogen treatments. Within species, responses to the pathogen varied with life history stage, and the most heavily infected species at the larval stage was different from the most heavily infected species at the post-metamorphic stage. Temporal changes in infection load were species and life history stage-specific. We show that variation in susceptibility to this multi-host pathogen is complex when viewed at a fine-scale and may be mediated through intrinsic host traits.

Ecophysiology meets conservation: understanding the role of disease in amphibian population declines.

Infectious diseases are intimately associated with the dynamics of biodiversity. However, the role that infectious disease plays within ecological communities is complex. The complex effects of infectious disease at the scale of communities and ecosystems are driven by the interaction between host and pathogen. Whether or not a given host-pathogen interaction results in progression from infection to disease is largely dependent on the physiological characteristics of the host within the context of the external environment. Here, we highlight the importance of understanding the outcome of infection and disease in the context of host ecophysiology using amphibians as a model system. Amphibians are ideal for such a discussion because many of their populations are experiencing declines and extinctions, with disease as an important factor implicated in many declines and extinctions. Exposure to pathogens and the host's responses to infection can be influenced by many factors related to physiology such as host life history, immunology, endocrinology, resource acquisition, behaviour and changing climates. In our review, we discuss the relationship between disease and biodiversity. We highlight the dynamics of three amphibian host-pathogen systems that induce different effects on hosts and life stages and illustrate the complexity of amphibian-host-parasite systems. We then review links between environmental stress, endocrine-immune interactions, disease and climate change.

The complexity of amphibian population declines: understanding the role of cofactors in driving amphibian losses.

Population losses and extinctions of species are occurring at unprecedented rates, as exemplified by declines and extinctions of amphibians worldwide. However, studies of amphibian population declines generally do not address the complexity of the phenomenon or its implications for ecological communities, focusing instead on single factors affecting particular amphibian species. We argue that the causes for amphibian population declines are complex; may differ among species, populations, and life stages within a population; and are context dependent with multiple stressors interacting to drive declines. Because amphibians are key components of communities, we emphasize the importance of investigating amphibian declines at the community level. Selection pressures over evolutionary time have molded amphibian life history characteristics, such that they may remain static even in the face of strong, recent human-induced selection pressures.