Browsing by Subject "Healthcare"

Between 1890 and 1965, the ideology of government responsibility for maternal and child health represented a continuous and central goal that fueled programs and institutional networks of progressive and liberal social policy advocates. Beginning in the settlement houses of the 1890s, a cadre of female bureaucrats, social reformers, and their political allies developed an array of federally based programs. Conservative stakeholders--among them anti-feminists, representatives of the medical industry, anti-communists, and white supremacists--strenuously opposed this vision of health justice, arguing that health was a personal responsibility in which government should play no part. Despite the achievements of government-based progressive reformers in instituting their vision in urban settlement houses, under the Sheppard-Towner Act of the mid-1920s and during the years of the New Deal and World War II, the Cold War's approach to domestic social policy after 1947 clamped down on their vision. After this conservative turn against social democratic solutions to welfare needs, these progressive advocates shifted their attention to the international health rights movement and to community-based maternal and child health activities.

My dissertation introduces the concept of health justice as an interpretive lens to trace the history of health policy progressives and their institutional networks. On the one hand, health justice reflects the communitarian premise that the health of all members of society is essential for the common good. On the other hand, health justice implies that health and health care are individual rights that government ought to protect. While communitarian arguments were often on the tip of the tongues of social reformers, a passionate belief in citizenship-based rights and redistributive and humanitarian ideas of social justice undergirded their policy ideas and became a more explicitly stated position during the New Deal and World War II. This justice-based approach to maternal and child health policy was consistently undermined by the prevailing counter-ideologies of individual responsibility for health, local control of public services, racial segregation in health services, and the commodification of health care.

My work relies on primary evidence collected from the personal papers of key protagonists, the administrative records of the Children's Bureau housed at the National Archives, oral histories, and the presidential papers of Harry S. Truman. Published primary materials have been culled from memoirs, professional public health and medical journals, as well as the popular press. I also draw from a body of historical and political science scholarship of the past twenty-five years to contextualize the narrative.

Though families come in all shapes and sizes, many people recognize the birth of their first child as the start of their new family. The transition to parenthood that expectant parents experience has important implications for their future health and the health of their children. This dissertation investigates the experiences of new and expectant parents as they develop their new roles. The findings draw on publicly-available conversations from parenting forums. Investigative phenomenology, descriptive phenomenology, and quantitative analysis are used to explore three research questions: 1) How do people experience perinatal loss? 2) What are parents’ experiences of working with nurses when their infant is in a neonatal critical care unit? 3) What emotions do men experience on their journey to fatherhood? The findings of this dissertation indicate that the transition to parenthood is a time of ambiguity, stress, and potentially, great joy for new parents. During this transition, people take on new identities, perform new roles, experience a broad range of emotions, and develop new relationships. The impacts of this transition are lifelong, so support is vital to promoting the formation of healthy, well-adjusted families. For healthcare providers and researchers, there is a great deal that can be done to help new and expectant parents feel supported and respected. A few of the many potential tools providers and researchers can use include mindfulness, non-judgement, and therapeutic communication.

Clinical prediction models offer the ability to help physicians make better data-driven decisions that can improve patient outcomes. Given the wealth of data available with the widespread adoption of electronic health records, more flexible statistical models are required that can account for the messiness and complexity of this data. In this dissertation we focus on developing models for clinical time series, as most data within healthcare is collected longitudinally and it is important to take this structure into account. Models built off of Gaussian processes are natural in this setting of irregularly sampled, noisy time series with many missing values. In addition, they have the added benefit of accounting for and quantifying uncertainty, which can be extremely useful in medical decision making. In this dissertation, we develop new Gaussian process-based models for medical time series along with associated algorithms for efficient inference on large-scale electronic health records data. We apply these models to several real healthcare applications, using local data obtained from the Duke University healthcare system.

In Chapter 1 we give a brief overview of clinical prediction models, electronic health records, and Gaussian processes. In Chapter 2, we develop several Gaussian process models for clinical time series in the context of chronic kidney disease management. We show how our proposed joint model for longitudinal and time-to-event data and model for multivariate time series can make accurate predictions about a patient's future disease trajectory. In Chapter 3, we combine multi-output Gaussian processes with a downstream black-box deep recurrent neural network model from deep learning. We apply this modeling framework to clinical time series to improve early detection of sepsis among patients in the hospital, and show that the Gaussian process preprocessing layer both allows for uncertainty quantification and acts as a form of data augmentation to reduce overfitting. In Chapter 4, we again use multi-output Gaussian processes as a preprocessing layer in model-free deep reinforcement learning. Here the goal is to learn optimal treatments for sepsis given clinical time series and historical treatment decisions taken by clinicians, and we show that the Gaussian process preprocessing layer and use of a recurrent architecture offers improvements over standard deep reinforcement learning methods. We conclude in Chapter 5 with a summary of future areas for work, and a discussion on practical considerations and challenges involved in deploying machine learning models into actual clinical practice.

Health care costs for U.S. employers have tripled in the past twenty years. By constructing a novel dataset with firm-specific health care expenses, I show that firms negatively adjust both capital expenditures and R&D expenses in response to changes in health care costs. I estimate that, on average, a 1% increase in health care costs is associated with a 0.7% decrease in total investment. The effects are stronger for financially constrained firms, firms employing more high-skilled workers, and firms working with fewer insurers. Additional tests confirm that hiring fewer workers and reducing wages do not offset rising health costs enough to counteract this lower investment channel. Overall, my findings suggest that rising health care costs limit firms’ ability to expand either physically or via innovation.

Introduction: Hypertension, the primary risk factor for the leading cause of mortality, cardiovascular diseases, is increasing in all parts of the world, including low-income countries. This thesis explores the prevalence and predictors of elevated blood pressure in rural farmers in northeastern Madagascar. The prevalence of hypertensive blood pressure was compared as defined by two commonly used hypertension standards, the World Health Organization (WHO), and the American College of Cardiology & the American Heart Association (ACC/AHA). The hypothesized drivers of systolic and diastolic blood pressure were investigated, focusing specifically on age, gender, body mass index (BMI), salt intake, and stress. Methods: Over 1,200 surveys were conducted from 2019 to 2022 in four villages in the Sava Region of Madagascar, where blood pressure readings were obtained on each participant. A subset of the sample in the last field season, participated in additional questions related to salt consumption and a measurement of salt taken over the course of seven days. A subsample of forty-eight participants completed an additional survey regarding stress. Generalized linear mixed models were run to assess associations involving blood pressure, salt intake and stress, and other predictions involving age, gender, and BMI. Results: The different standards of hypertension, as defined by the WHO and the ACC/AHA, changed the respective prevalence of elevated blood pressure for this population markedly, nearly doubling the prevalence of hypertension if the more stringent American standard was used (WHO=29.7%, ACC/AHA=59.6%). Age was found to be a strong predictor of blood pressure as well as BMI, while salt intake, stress, and gender had weaker associations. Conclusion: The prevalence of elevated blood pressure in this population was markedly higher than that found in other studies conducted in this region and globally. Findings regarding lifestyle factors, salt intake and stress, as individual drivers of blood pressure were found to be inconclusive. Further research should be performed in this population to determine the predictors and lifestyle factors associated with blood pressure and the extent of their influence in this region. Consideration for standards of practice and diagnosis should be carefully considered in this population, as the burden of hypertension would potentially increase with a change in practice standard.

Healthcare stands to benefit from the advent of deep learning on account of (i) the massive amounts of data generated by the health system and (ii) the ability of deep models to make predictions from complex inputs. This dissertation centers on two applications of deep learning to challenging problems in healthcare.

First, we discuss deep learning for treatment effect/counterfactual estimation in the observational setting, i.e., where the treatment assignment is not randomized (Chapters 2 and 3). For example, we may want to know the causal effect of a drug on a patient's blood pressure. We combine deep learning with classical weighting techniques to estimate average and conditional average treatment effects from observational data. We show theoretical properties of our method, including guarantees about when "balance" can be achieved between treatment groups. We then weaken the typical "ignorability" assumption and generate treatment effect intervals (instead of point-estimates).

Second, we explore the use of deep learning applied to a difficult problem in medical imaging: classifying malignancy from thyroid cytopathology slides (Chapters 4, 5, and 6). The difficulty of this problem arises from the image size, which is typically on the order of tens of gigabytes (i.e., around 3 to 4 orders of magnitude larger than image sizes in popular deep learning architectures). Our approach is a two-step process: (i) automatically finding image regions containing follicular cell groups, (ii) classifying each region and aggregating the predictions. We show that our system works well for mobile phone images of thyroid biopsy slides, and that our system compares favorably with state-of-the-art genetic testing for malignancy.

Finally, after my Ph.D. I plan to enter a career in autonomous driving. As an "epilogue" of this dissertation (Chapter 7), we present a method to make deep learning point-cloud models for autonomous driving which are invariant (or equivariant) to rotations. Intuitively, this is an important requirement -- a rotated bicycle should still be classified as a bicycle, and driving behavior should be independent of direction of travel. However, most deep learning models used in autonomous driving today do not satisfy these properties exactly. We propose a practical model (based on the Transformer architecture) to address this pitfall, and we showcase its performance on point-cloud classification and trajectory forecasting tasks.

We present a scalable end-to-end classifier that uses streaming physiological and medication data to accurately predict the onset of sepsis, a life-threatening complication from infections that has high mortality and morbidity. Our proposed framework models the multivariate trajectories of continuous-valued physiological time series using multitask Gaussian processes, seamlessly accounting for the high uncertainty, frequent missingness, and irregular sampling rates typically associated with real clinical data. The Gaussian process is directly connected to a black-box classifier that predicts whether a patient encounter will become septic, chosen in our case to be a recurrent neural network to account for the extreme variability in the length of patient encounters. We show how several approximations scale the computations associated with the Gaussian process in a manner so that the entire system can be trained discriminatively end-to-end using backpropagation. In a large cohort of heterogeneous inpatient encounters at our university health system we find that it outperforms several baselines at predicting sepsis, and yields 33\% and 195\% improved areas under the Receiver Operating Characteristic and Precision Recall curves as compared to the NEWS score currently in use on our own hospital wards.

This dissertation consists of three essays that analyze healthcare and health insurance markets in relation to healthcare reform, and particularly in the context of the Affordable Care Act (ACA). The first essay uses a nationwide datasets of plan offerings, premiums, and network sizes for the ACA Health Insurance Marketplaces in 2014 to document patterns relating to the effects of competition on premiums and plan network characteristics. The results suggest that greater competition is associated with lower premiums, and that narrow network plans do offer lower premiums. This study also documents heterogeneity along these dimensions across types of insurance plans by ownership structure (not-for-profit, for-profit, and CO-OP). This heterogeneity suggests that a market's overall welfare may depend on the equilibrium market shares and ownership types of the competing firms.

The second and third essays use the State of Colorado's new All-Payer Claims Database (APCD) to examine the welfare consequences in the state's non-group health insurance market, which includes the ACA Marketplace. In the second essay, I test for adverse selection into the ACA Marketplace, and evaluate policies that may help to ameliorate the welfare loss due to adverse selection. Specifically, I use plausibly exogenous premium variation generated by geographic discontinuities to provide evidence of adverse selection, whereby low-cost individuals exit the market in response to rising premiums. Specifically, a 1% increase in premiums causes a 0.8% increase in medical expenditures of the insured population. The estimates indicate that additional premium subsidies, and especially age-targeted subsidies, would enhance welfare. These results offer the first quasi-experimental evidence of selection in the ACA Exchanges.

In the third essay, my co-author Eli Liebman and I extend this analysis to take into account imperfect competition in both health insurance and hospital markets. We bring together the literatures on insurer-hospital bargaining and selection in imperfectly competitive insurance markets to propose a model that captures features salient to the health insurance marketplaces. In particular, although insurance markets tend to be concentrated, the ACA aimed to foster competitive marketplaces, highlighting the importance for understanding the interaction between imperfect competition and selection. The degree of competition among insurance plans affects both selection across plans and on the extensive margin, as well as simultaneously affecting the prices negotiated with providers. We show theoretically that provider market power and adverse selection can interact to amplify the welfare loss due to either one of these two channels individually. We also show why ignoring adverse selection will lead to biased estimates of bargaining parameters in the standard model of hospital-insurer bargaining. Finally, we use medical claims from the State of Colorado, to quantify the welfare consequences for that market. These considerations are relevant for evaluating the effects of policy interventions in the ACA's health insurance marketplaces that affect insurer entry/exit and premium setting.