Intensive care units (ICUs) provide care for critically-ill patients who require constant monitoring and the availability of specialized equipment and personnel. In this environment, a high volume of information and a high degree of uncertainty present a burden to clinicians. In specialized cohorts, such as pediatric patients with congenital heart defects (CHDs), this burden is exacerbated by increased complexity, the inadequacy of existing decision support aids, and the limited and decreasing availability of highly-specialized clinicians. Among CHD patients, infants with single ventricle (SV) physiology are one of the most complex and severely-ill sub-populations. While SV mortality rates have dropped, patient deterioration may happen unexpectedly in the period before patients undergo stage-2 palliative surgery. Even in expert hands, critical and potentially catastrophic events (CEs), such as cardiopulmonary resuscitation (CPR), emergent endotracheal intubation (EEI), or extracorporeal membrane oxygenation (ECMO) are common in SV patients, and may negatively impact morbidity, mortality, and hospital length of stay. There is a clinical need of predictive tools that help intensivists assess and forecast the advent of CEs in SV infants. Although ubiquitous, widely adopted ICU severity-of-illness scores or early warning systems (EWS), e.g., PRISM and PIM, have not met this need. They are often v developed for general ICU use and do not generalize well to specialized populations. Furthermore, most EWS are developed for prediction of patient mortality. Among SV patients, however, death is semi-elective. On the other hand, prediction of CEs may help clinicians improve patient care by anticipating the advent of patient deterioration. In this dissertation, we aimed to develop and validate predictive models that achieve early and accurate prediction of CEs in infants with SV physiology. Such models may provide early and actionable information to clinicians and may be used to perform clinical interventions aimed at preventing CEs, and to reducing morbidity, mortality, and healthcare costs. We assert that our work is significant in that it addresses an unmet clinical need by achieving state-of-the-art, early prediction of patient deterioration in a challenging and vulnerable population. vi TABLE OF CONTENTS

Objective Limited research exists in predicting first-time suicide attempts that account for two-thirds of suicide decedents. We aimed to predict first-time suicide attempts using a large data-driven approach that applies natural language processing (NLP) and machine learning (ML) to unstructured (narrative) clinical notes and structured electronic health record (EHR) data. Methods This case-control study included patients aged 10–75 years who were seen between 2007 and 2016 from emergency departments and inpatient units. Cases were first-time suicide attempts from coded diagnosis; controls were randomly selected without suicide attempts regardless of demographics, following a ratio of nine controls per case. Four data-driven ML models were evaluated using 2-year historical EHR data prior to suicide attempt or control index visits, with prediction windows from 7 to 730 days. Patients without any historical notes were excluded. Model evaluation on accuracy and robustness was performed on a blind dataset (30% cohort). Results The study cohort included 45 238 patients (5099 cases, 40 139 controls) comprising 54 651 variables from 5.7 million structured records and 798 665 notes. Using both unstructured and structured data resulted in significantly greater accuracy compared to structured data alone (area-under-the-curve [AUC]: 0.932 vs. 0.901 P < .001). The best-predicting model utilized 1726 variables with AUC = 0.932 (95% CI, 0.922–0.941). The model was robust across multiple prediction windows and subgroups by demographics, points of historical most recent clinical contact, and depression diagnosis history. Conclusions Our large data-driven approach using both structured and unstructured EHR data demonstrated accurate and robust first-time suicide attempt prediction, and has the potential to be deployed across various populations and clinical settings.

ObjectivesThis study evaluates the accuracy and transferability of Bayesian case detection systems (BCD) that use clinical notes from emergency department (ED) to detect influenza cases.MethodsA BCD uses natural language processing (NLP) to infer the presence or absence of clinical findings from ED notes, which are fed into a Bayesain network classifier (BN) to infer patients’ diagnoses. We developed BCDs at the University of Pittsburgh Medical Center (BCDUPMC) and Intermountain Healthcare in Utah (BCDIH). At each site, we manually built a rule-based NLP and trained a Bayesain network classifier from over 40,000 ED encounters between Jan. 2008 and May. 2010 using feature selection, machine learning, and expert debiasing approach. Transferability of a BCD in this study may be impacted by seven factors: development (source) institution, development parser, application (target) institution, application parser, NLP transfer, BN transfer, and classification task. We employed an ANOVA analysis to study their impacts on BCD performance.ResultsBoth BCDs discriminated well between influenza and non-influenza on local test cases (AUCs > 0.92). When tested for transferability using the other institution’s cases, BCDUPMC discriminations declined minimally (AUC decreased from 0.95 to 0.94, p<0.01), and BCDIH discriminations declined more (from 0.93 to 0.87, p<0.0001). We attributed the BCDIH decline to the lower recall of the IH parser on UPMC notes. The ANOVA analysis showed five significant factors: development parser, application institution, application parser, BN transfer, and classification task.ConclusionWe demonstrated high influenza case detection performance in two large healthcare systems in two geographically separated regions, providing evidentiary support for the use of automated case detection from routinely collected electronic clinical notes in national influenza surveillance. The transferability could be improved by training Bayesian network classifier locally and increasing the accuracy of the NLP parser.

No abstract